diff options
Diffstat (limited to 'src/gallium')
469 files changed, 18322 insertions, 14807 deletions
diff --git a/src/gallium/Makefile b/src/gallium/Makefile index 875314f6c3..79ca767f7e 100644 --- a/src/gallium/Makefile +++ b/src/gallium/Makefile @@ -1,26 +1,12 @@ +# src/gallium/Makefile TOP = ../.. include $(TOP)/configs/current - SUBDIRS = $(GALLIUM_DIRS) -# Note winsys/ needs to be built after src/mesa - - -default: subdirs - -subdirs: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - - -clean: - rm -f `find . -name \*.[oa]` - rm -f `find . -name depend` - - -# Dummy install target -install: diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 655e949ca2..98487d43bd 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -1,64 +1,66 @@ -# -*-makefile-*- +# src/gallium/Makefile.template - -# We still have a dependency on the "dri" buffer manager. Most likely -# the interface can be reused in non-dri environments, and also as a -# frontend to simpler memory managers. +# Template makefile for gallium libraries. +# +# Usage: +# The minimum that the including makefile needs to define +# is TOP, LIBNAME and one of of the *_SOURCES. # -COMMON_SOURCES = +# Optional defines: +# LIBRARY_INCLUDES are appended to the list of includes directories. +# LIBRARY_DEFINES is not used for makedepend, but for compilation. -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) +### Basic defines ### +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) -### Include directories INCLUDES = \ -I. \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/include \ - $(DRIVER_INCLUDES) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + -I$(GALLIUM)/src/gallium/include \ + -I$(GALLIUM)/src/gallium/auxiliary \ + -I$(GALLIUM)/src/gallium/drivers \ + $(LIBRARY_INCLUDES) ##### TARGETS ##### -default: depend symlinks lib$(LIBNAME).a - +default: depend lib$(LIBNAME).a lib$(LIBNAME).a: $(OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template - $(TOP)/bin/mklib -o $(LIBNAME) -static $(OBJECTS) $(DRIVER_LIBS) - + $(MKLIB) -o $(LIBNAME) -static $(OBJECTS) depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) rm -f depend touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - + $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) 2> /dev/null # Emacs tags tags: etags `find . -name \*.[ch]` `find ../include` - # Remove .o and backup files -clean:: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak +clean: + rm -f $(OBJECTS) lib$(LIBNAME).a depend depend.bak + +# Dummy target +install: + @echo -n "" + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(LIBRARY_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(LIBRARY_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(LIBRARY_DEFINES) $< -o $@ -include depend +sinclude depend diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile index eaa0f2fe4e..5446eb68a9 100644 --- a/src/gallium/auxiliary/Makefile +++ b/src/gallium/auxiliary/Makefile @@ -1,20 +1,12 @@ +# src/gallium/auxiliary/Makefile TOP = ../../.. include $(TOP)/configs/current - SUBDIRS = $(GALLIUM_AUXILIARY_DIRS) - -default: subdirs - - -subdirs: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - - -clean: - rm -f `find . -name \*.[oa]` diff --git a/src/gallium/auxiliary/cso_cache/Makefile b/src/gallium/auxiliary/cso_cache/Makefile index 6bd6602088..8726afcd94 100644 --- a/src/gallium/auxiliary/cso_cache/Makefile +++ b/src/gallium/auxiliary/cso_cache/Makefile @@ -9,6 +9,3 @@ C_SOURCES = \ cso_hash.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 6b1754ea00..0bc77a5728 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -28,7 +28,7 @@ /* Authors: Zack Rusin <zack@tungstengraphics.com> */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index a9157aad71..f388bf5d95 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -37,7 +37,6 @@ #include "pipe/p_state.h" #include "util/u_memory.h" -#include "pipe/p_inlines.h" #include "tgsi/tgsi_parse.h" #include "cso_cache/cso_context.h" diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index 4e7664f9bf..288cef7b6f 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -30,7 +30,7 @@ * Zack Rusin <zack@tungstengraphics.com> */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "cso_hash.h" diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index bdbf5a08ed..5041dcc072 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -43,8 +43,4 @@ C_SOURCES = \ draw_vs_ppc.c \ draw_vs_sse.c - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 7bd4a2e221..a4f1fcddc1 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -127,7 +127,7 @@ void draw_set_rasterizer_state( struct draw_context *draw, draw->rasterizer = raster; draw->bypass_clipping = - ((draw->rasterizer && draw->rasterizer->bypass_clipping) || + ((draw->rasterizer && draw->rasterizer->bypass_vs_clip_and_viewport) || draw->driver.bypass_clipping); } @@ -138,8 +138,9 @@ void draw_set_driver_clipping( struct draw_context *draw, draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->driver.bypass_clipping = bypass_clipping; - draw->bypass_clipping = (draw->rasterizer->bypass_clipping || - draw->driver.bypass_clipping); + draw->bypass_clipping = + ((draw->rasterizer && draw->rasterizer->bypass_vs_clip_and_viewport) || + draw->driver.bypass_clipping); } diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 20841bb5d6..ca69f0b95e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -32,7 +32,6 @@ */ -#include "pipe/p_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" @@ -410,7 +409,7 @@ aaline_create_texture(struct aaline_stage *aaline) * texels which are zero. Special case the 1x1 and 2x2 levels. */ for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) { - struct pipe_surface *surface; + struct pipe_transfer *transfer; const uint size = aaline->texture->width[level]; ubyte *data; uint i, j; @@ -419,9 +418,9 @@ aaline_create_texture(struct aaline_stage *aaline) /* This texture is new, no need to flush. */ - surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0, - PIPE_BUFFER_USAGE_CPU_WRITE); - data = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE); + transfer = screen->get_tex_transfer(screen, aaline->texture, 0, level, 0, + PIPE_TRANSFER_WRITE, 0, 0, size, size); + data = screen->transfer_map(screen, transfer); if (data == NULL) return FALSE; @@ -440,13 +439,13 @@ aaline_create_texture(struct aaline_stage *aaline) else { d = 255; } - data[i * surface->stride + j] = d; + data[i * transfer->stride + j] = d; } } /* unmap */ - screen->surface_unmap(screen, surface); - screen->tex_surface_release(screen, &surface); + screen->transfer_unmap(screen, transfer); + screen->tex_transfer_destroy(transfer); } return TRUE; } @@ -728,7 +727,7 @@ aaline_destroy(struct draw_stage *stage) aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); if (aaline->texture) - pipe_texture_release(&aaline->texture); + pipe_texture_reference(&aaline->texture, NULL); draw_free_temp_verts( stage ); diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 2c1cacbdb4..5008086cf2 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -38,7 +38,6 @@ */ -#include "pipe/p_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index a0f9716dac..04e59152c5 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -34,7 +34,6 @@ */ -#include "pipe/p_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" @@ -372,7 +371,7 @@ pstip_update_texture(struct pstip_stage *pstip) static const uint bit31 = 1 << 31; struct pipe_context *pipe = pstip->pipe; struct pipe_screen *screen = pipe->screen; - struct pipe_surface *surface; + struct pipe_transfer *transfer; const uint *stipple = pstip->state.stipple->stipple; uint i, j; ubyte *data; @@ -381,10 +380,9 @@ pstip_update_texture(struct pstip_stage *pstip) */ pipe->flush( pipe, PIPE_FLUSH_TEXTURE_CACHE, NULL ); - surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0, - PIPE_BUFFER_USAGE_CPU_WRITE); - data = screen->surface_map(screen, surface, - PIPE_BUFFER_USAGE_CPU_WRITE); + transfer = screen->get_tex_transfer(screen, pstip->texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, 0, 0, 32, 32); + data = screen->transfer_map(screen, transfer); /* * Load alpha texture. @@ -396,18 +394,18 @@ pstip_update_texture(struct pstip_stage *pstip) for (j = 0; j < 32; j++) { if (stipple[i] & (bit31 >> j)) { /* fragment "on" */ - data[i * surface->stride + j] = 0; + data[i * transfer->stride + j] = 0; } else { /* fragment "off" */ - data[i * surface->stride + j] = 255; + data[i * transfer->stride + j] = 255; } } } /* unmap */ - screen->surface_unmap(screen, surface); - screen->tex_surface_release(screen, &surface); + screen->transfer_unmap(screen, transfer); + screen->tex_transfer_destroy(transfer); } @@ -573,7 +571,7 @@ pstip_destroy(struct draw_stage *stage) pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso); - pipe_texture_release(&pstip->texture); + pipe_texture_reference(&pstip->texture, NULL); draw_free_temp_verts( stage ); FREE( stage ); diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 9153bc2f86..12325d30d6 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -29,12 +29,12 @@ * \file * Vertex buffer drawing stage. * - * \author José Fonseca <jrfonsec@tungstengraphics.com> + * \author Jose Fonseca <jrfonsec@tungstengraphics.com> * \author Keith Whitwell <keith@tungstengraphics.com> */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -93,7 +93,6 @@ vbuf_stage( struct draw_stage *stage ) } -static void vbuf_flush_indices( struct vbuf_stage *vbuf ); static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); @@ -109,13 +108,12 @@ overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) static INLINE void check_space( struct vbuf_stage *vbuf, unsigned nr ) { - if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { - vbuf_flush_vertices(vbuf); - vbuf_alloc_vertices(vbuf); + if (vbuf->nr_vertices + nr > vbuf->max_vertices || + vbuf->nr_indices + nr > vbuf->max_indices) + { + vbuf_flush_vertices( vbuf ); + vbuf_alloc_vertices( vbuf ); } - - if (vbuf->nr_indices + nr > vbuf->max_indices ) - vbuf_flush_indices(vbuf); } @@ -202,7 +200,7 @@ vbuf_point( struct draw_stage *stage, * will be flushed if needed and a new one allocated. */ static void -vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) +vbuf_start_prim( struct vbuf_stage *vbuf, uint prim ) { struct translate_key hw_key; unsigned dst_offset; @@ -217,11 +215,7 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) * state change. */ vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render); - - if (vbuf->vertex_size != vbuf->vinfo->size * sizeof(float)) { - vbuf_flush_vertices(vbuf); - vbuf->vertex_size = vbuf->vinfo->size * sizeof(float); - } + vbuf->vertex_size = vbuf->vinfo->size * sizeof(float); /* Translate from pipeline vertices to hw vertices. */ @@ -294,8 +288,8 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) /* Allocate new buffer? */ - if (!vbuf->vertices) - vbuf_alloc_vertices(vbuf); + assert(vbuf->vertices == NULL); + vbuf_alloc_vertices(vbuf); } @@ -305,9 +299,9 @@ vbuf_first_tri( struct draw_stage *stage, { struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( vbuf ); + vbuf_flush_vertices( vbuf ); + vbuf_start_prim(vbuf, PIPE_PRIM_TRIANGLES); stage->tri = vbuf_tri; - vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); stage->tri( stage, prim ); } @@ -318,9 +312,9 @@ vbuf_first_line( struct draw_stage *stage, { struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( vbuf ); + vbuf_flush_vertices( vbuf ); + vbuf_start_prim(vbuf, PIPE_PRIM_LINES); stage->line = vbuf_line; - vbuf_set_prim(vbuf, PIPE_PRIM_LINES); stage->line( stage, prim ); } @@ -331,53 +325,42 @@ vbuf_first_point( struct draw_stage *stage, { struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( vbuf ); + vbuf_flush_vertices(vbuf); + vbuf_start_prim(vbuf, PIPE_PRIM_POINTS); stage->point = vbuf_point; - vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); stage->point( stage, prim ); } -static void -vbuf_flush_indices( struct vbuf_stage *vbuf ) -{ - if(!vbuf->nr_indices) - return; - - assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == - vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); - - vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); - - vbuf->nr_indices = 0; -} - /** * Flush existing vertex buffer and allocate a new one. - * - * XXX: We separate flush-on-index-full and flush-on-vb-full, but may - * raise issues uploading vertices if the hardware wants to flush when - * we flush. */ static void vbuf_flush_vertices( struct vbuf_stage *vbuf ) { - if(vbuf->vertices) { - vbuf_flush_indices(vbuf); - + if(vbuf->vertices) { + + vbuf->render->unmap_vertices( vbuf->render, 0, vbuf->nr_vertices - 1 ); + + if (vbuf->nr_indices) + { + vbuf->render->draw(vbuf->render, + vbuf->indices, + vbuf->nr_indices ); + + vbuf->nr_indices = 0; + } + /* Reset temporary vertices ids */ if(vbuf->nr_vertices) draw_reset_vertex_ids( vbuf->stage.draw ); /* Free the vertex buffer */ - vbuf->render->release_vertices(vbuf->render, - vbuf->vertices, - vbuf->vertex_size, - vbuf->nr_vertices); + vbuf->render->release_vertices( vbuf->render ); + vbuf->max_vertices = vbuf->nr_vertices = 0; vbuf->vertex_ptr = vbuf->vertices = NULL; - } } @@ -402,9 +385,12 @@ vbuf_alloc_vertices( struct vbuf_stage *vbuf ) * and it will flush itself if necessary to do so. If this does * fail, we are basically without usable hardware. */ - vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, - (ushort) vbuf->vertex_size, - (ushort) vbuf->max_vertices); + vbuf->render->allocate_vertices(vbuf->render, + (ushort) vbuf->vertex_size, + (ushort) vbuf->max_vertices); + + vbuf->vertices = (uint *) vbuf->render->map_vertices( vbuf->render ); + vbuf->vertex_ptr = vbuf->vertices; } @@ -415,14 +401,11 @@ vbuf_flush( struct draw_stage *stage, unsigned flags ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( vbuf ); + vbuf_flush_vertices( vbuf ); stage->point = vbuf_first_point; stage->line = vbuf_first_line; stage->tri = vbuf_first_tri; - - if (flags & DRAW_FLUSH_BACKEND) - vbuf_flush_vertices( vbuf ); } diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 4e5ffa0930..9ea0cbe599 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -87,7 +87,7 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_CLIPTEST; } - if (!draw->rasterizer->bypass_vs) { + if (!draw->rasterizer->bypass_vs_clip_and_viewport) { opt |= PT_SHADE; } } diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index c02f229110..6f3e1e0289 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -173,9 +173,7 @@ void draw_pt_emit( struct pt_emit *emit, void draw_pt_emit_linear( struct pt_emit *emit, const float (*vertex_data)[4], - unsigned vertex_count, unsigned stride, - unsigned start, unsigned count ); void draw_pt_emit_destroy( struct pt_emit *emit ); @@ -217,7 +215,7 @@ boolean draw_pt_post_vs_run( struct pt_post_vs *pvs, void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, boolean bypass_clipping, - boolean identity_viewport, + boolean bypass_viewport, boolean opengl ); struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 232dfdaed2..064e16c295 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -165,6 +165,9 @@ void draw_pt_emit( struct pt_emit *emit, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + if (vertex_count == 0) + return; + if (vertex_count >= UNDEFINED_VERTEX_ID) { assert(0); return; @@ -178,9 +181,11 @@ void draw_pt_emit( struct pt_emit *emit, return; } - hw_verts = render->allocate_vertices(render, - (ushort)translate->key.output_stride, - (ushort)vertex_count); + render->allocate_vertices(render, + (ushort)translate->key.output_stride, + (ushort)vertex_count); + + hw_verts = render->map_vertices( render ); if (!hw_verts) { assert(0); return; @@ -201,22 +206,21 @@ void draw_pt_emit( struct pt_emit *emit, vertex_count, hw_verts ); + render->unmap_vertices( render, + 0, + vertex_count - 1 ); + render->draw(render, elts, count); - render->release_vertices(render, - hw_verts, - translate->key.output_stride, - vertex_count); + render->release_vertices(render); } void draw_pt_emit_linear(struct pt_emit *emit, const float (*vertex_data)[4], - unsigned vertex_count, unsigned stride, - unsigned start, unsigned count) { struct draw_context *draw = emit->draw; @@ -231,26 +235,23 @@ void draw_pt_emit_linear(struct pt_emit *emit, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - if (count >= UNDEFINED_VERTEX_ID) { - assert(0); - return; - } + if (count >= UNDEFINED_VERTEX_ID) + goto fail; /* XXX: and work out some way to coordinate the render primitive * between vbuf.c and here... */ - if (!draw->render->set_primitive(draw->render, emit->prim)) { - assert(0); - return; - } + if (!draw->render->set_primitive(draw->render, emit->prim)) + goto fail; - hw_verts = render->allocate_vertices(render, - (ushort)translate->key.output_stride, - (ushort)count); - if (!hw_verts) { - assert(0); - return; - } + if (!render->allocate_vertices(render, + (ushort)translate->key.output_stride, + (ushort)count)) + goto fail; + + hw_verts = render->map_vertices( render ); + if (!hw_verts) + goto fail; translate->set_buffer(translate, 0, vertex_data, stride); @@ -261,12 +262,12 @@ void draw_pt_emit_linear(struct pt_emit *emit, translate->run(translate, 0, - vertex_count, + count, hw_verts); if (0) { unsigned i; - for (i = 0; i < vertex_count; i++) { + for (i = 0; i < count; i++) { debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i); draw_dump_emitted_vertex( emit->vinfo, (const uint8_t *)hw_verts + @@ -274,13 +275,17 @@ void draw_pt_emit_linear(struct pt_emit *emit, } } + render->unmap_vertices( render, 0, count - 1 ); + + render->draw_arrays(render, 0, count); + + render->release_vertices(render); - render->draw_arrays(render, start, count); + return; - render->release_vertices(render, - hw_verts, - translate->key.output_stride, - vertex_count); +fail: + assert(0); + return; } struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index dcb7744b17..6b7d02a19b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -234,9 +234,11 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, return; } - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)feme->translate->key.output_stride, - (ushort)fetch_count ); + draw->render->allocate_vertices( draw->render, + (ushort)feme->translate->key.output_stride, + (ushort)fetch_count ); + + hw_verts = draw->render->map_vertices( draw->render ); if (!hw_verts) { assert(0); return; @@ -259,6 +261,10 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, } } + draw->render->unmap_vertices( draw->render, + 0, + (ushort)(fetch_count - 1) ); + /* XXX: Draw arrays path to avoid re-emitting index list again and * again. */ @@ -268,10 +274,7 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, /* Done -- that was easy, wasn't it: */ - draw->render->release_vertices( draw->render, - hw_verts, - feme->translate->key.output_stride, - fetch_count ); + draw->render->release_vertices( draw->render ); } @@ -288,18 +291,17 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - if (count >= UNDEFINED_VERTEX_ID) { - assert(0); - return; - } + if (count >= UNDEFINED_VERTEX_ID) + goto fail; - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)feme->translate->key.output_stride, - (ushort)count ); - if (!hw_verts) { - assert(0); - return; - } + if (!draw->render->allocate_vertices( draw->render, + (ushort)feme->translate->key.output_stride, + (ushort)count )) + goto fail; + + hw_verts = draw->render->map_vertices( draw->render ); + if (!hw_verts) + goto fail; /* Single routine to fetch vertices and emit HW verts. */ @@ -317,20 +319,21 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, } } + draw->render->unmap_vertices( draw->render, 0, count - 1 ); + /* XXX: Draw arrays path to avoid re-emitting index list again and * again. */ - draw->render->draw_arrays( draw->render, - 0, /*start*/ - count ); + draw->render->draw_arrays( draw->render, 0, count ); /* Done -- that was easy, wasn't it: */ - draw->render->release_vertices( draw->render, - hw_verts, - feme->translate->key.output_stride, - count ); + draw->render->release_vertices( draw->render ); + return; +fail: + assert(0); + return; } @@ -351,9 +354,12 @@ static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, if (count >= UNDEFINED_VERTEX_ID) return FALSE; - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)feme->translate->key.output_stride, - (ushort)count ); + if (!draw->render->allocate_vertices( draw->render, + (ushort)feme->translate->key.output_stride, + (ushort)count )) + return FALSE; + + hw_verts = draw->render->map_vertices( draw->render ); if (!hw_verts) return FALSE; @@ -364,6 +370,8 @@ static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, count, hw_verts ); + draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) ); + /* XXX: Draw arrays path to avoid re-emitting index list again and * again. */ @@ -373,10 +381,7 @@ static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, /* Done -- that was easy, wasn't it: */ - draw->render->release_vertices( draw->render, - hw_verts, - feme->translate->key.output_stride, - count ); + draw->render->release_vertices( draw->render ); return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 84ffe3296a..44147aed9b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -101,7 +101,8 @@ static void fse_prepare( struct draw_pt_middle_end *middle, fse->key.nr_elements = MAX2(fse->key.nr_outputs, /* outputs - translate to hw format */ fse->key.nr_inputs); /* inputs - fetch from api format */ - fse->key.viewport = !draw->identity_viewport; + fse->key.viewport = (!draw->rasterizer->bypass_vs_clip_and_viewport && + !draw->identity_viewport); fse->key.clip = !draw->bypass_clipping; fse->key.const_vbuffers = 0; @@ -234,19 +235,17 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - if (count >= UNDEFINED_VERTEX_ID) { - assert(0); - return; - } + if (count >= UNDEFINED_VERTEX_ID) + goto fail; - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)fse->key.output_stride, - (ushort)count ); + if (!draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)count )) + goto fail; - if (!hw_verts) { - assert(0); - return; - } + hw_verts = draw->render->map_vertices( draw->render ); + if (!hw_verts) + goto fail; /* Single routine to fetch vertices, run shader and emit HW verts. * Clipping is done elsewhere -- either by the API or on hardware, @@ -256,13 +255,7 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, start, count, hw_verts ); - /* Draw arrays path to avoid re-emitting index list again and - * again. - */ - draw->render->draw_arrays( draw->render, - 0, - count ); - + if (0) { unsigned i; for (i = 0; i < count; i++) { @@ -274,12 +267,24 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, (const uint8_t *)hw_verts + fse->key.output_stride * i ); } } + + draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) ); + /* Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw_arrays( draw->render, + 0, + count ); + + + draw->render->release_vertices( draw->render ); + + return; - draw->render->release_vertices( draw->render, - hw_verts, - fse->key.output_stride, - count ); +fail: + assert(0); + return; } @@ -298,18 +303,17 @@ fse_run(struct draw_pt_middle_end *middle, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - if (fetch_count >= UNDEFINED_VERTEX_ID) { - assert(0); - return; - } + if (fetch_count >= UNDEFINED_VERTEX_ID) + goto fail; - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)fse->key.output_stride, - (ushort)fetch_count ); - if (!hw_verts) { - assert(0); - return; - } + if (!draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)fetch_count )) + goto fail; + + hw_verts = draw->render->map_vertices( draw->render ); + if (!hw_verts) + goto fail; /* Single routine to fetch vertices, run shader and emit HW verts. @@ -319,9 +323,6 @@ fse_run(struct draw_pt_middle_end *middle, fetch_count, hw_verts ); - draw->render->draw( draw->render, - draw_elts, - draw_count ); if (0) { unsigned i; @@ -333,12 +334,19 @@ fse_run(struct draw_pt_middle_end *middle, } } + draw->render->unmap_vertices( draw->render, 0, (ushort)(fetch_count - 1) ); + + draw->render->draw( draw->render, + draw_elts, + draw_count ); + - draw->render->release_vertices( draw->render, - hw_verts, - fse->key.output_stride, - fetch_count ); + draw->render->release_vertices( draw->render ); + return; +fail: + assert(0); + return; } @@ -360,13 +368,14 @@ static boolean fse_run_linear_elts( struct draw_pt_middle_end *middle, if (count >= UNDEFINED_VERTEX_ID) return FALSE; - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)fse->key.output_stride, - (ushort)count ); + if (!draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)count )) + return FALSE; - if (!hw_verts) { + hw_verts = draw->render->map_vertices( draw->render ); + if (!hw_verts) return FALSE; - } /* Single routine to fetch vertices, run shader and emit HW verts. * Clipping is done elsewhere -- either by the API or on hardware, @@ -382,11 +391,9 @@ static boolean fse_run_linear_elts( struct draw_pt_middle_end *middle, draw_count ); + draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) ); - draw->render->release_vertices( draw->render, - hw_verts, - fse->key.output_stride, - count ); + draw->render->release_vertices( draw->render ); return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index ec3b41c320..11ac90fc56 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -84,7 +84,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, */ draw_pt_post_vs_prepare( fpme->post_vs, (boolean)draw->bypass_clipping, - (boolean)draw->identity_viewport, + (boolean)(draw->identity_viewport || + draw->rasterizer->bypass_vs_clip_and_viewport), (boolean)draw->rasterizer->gl_rasterization_rules ); @@ -140,9 +141,9 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, (char *)pipeline_verts ); /* Run the shader, note that this overwrites the data[] parts of - * the pipeline verts. If there is no shader, ie a bypass shader, - * then the inputs == outputs, and are already in the correct - * place. + * the pipeline verts. If there is no shader, eg if + * bypass_vs_clip_and_viewport, then the inputs == outputs, and are + * already in the correct place. */ if (opt & PT_SHADE) { @@ -216,9 +217,9 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, (char *)pipeline_verts ); /* Run the shader, note that this overwrites the data[] parts of - * the pipeline verts. If there is no shader, ie a bypass shader, - * then the inputs == outputs, and are already in the correct - * place. + * the pipeline verts. If there is no shader, ie if + * bypass_vs_clip_and_viewport, then the inputs == outputs, and are + * already in the correct place. */ if (opt & PT_SHADE) { @@ -251,9 +252,7 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, else { draw_pt_emit_linear( fpme->emit, (const float (*)[4])pipeline_verts->data, - count, fpme->vertex_size, - 0, /*start*/ count ); } @@ -288,9 +287,9 @@ static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle (char *)pipeline_verts ); /* Run the shader, note that this overwrites the data[] parts of - * the pipeline verts. If there is no shader, ie a bypass shader, - * then the inputs == outputs, and are already in the correct - * place. + * the pipeline verts. If there is no shader, ie if + * bypass_vs_clip_and_viewport, then the inputs == outputs, and are + * already in the correct place. */ if (opt & PT_SHADE) { diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 96dc706b99..00d7197b13 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -200,11 +200,11 @@ boolean draw_pt_post_vs_run( struct pt_post_vs *pvs, void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, boolean bypass_clipping, - boolean identity_viewport, + boolean bypass_viewport, boolean opengl ) { if (bypass_clipping) { - if (identity_viewport) + if (bypass_viewport) pvs->run = post_vs_none; else pvs->run = post_vs_viewport; diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h index a1c4c14445..cccd3bf435 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.h +++ b/src/gallium/auxiliary/draw/draw_vbuf.h @@ -80,9 +80,14 @@ struct vbuf_render { * Hardware renderers will use ttm memory, others will just malloc * something. */ - void *(*allocate_vertices)( struct vbuf_render *, - ushort vertex_size, - ushort nr_vertices ); + boolean (*allocate_vertices)( struct vbuf_render *, + ushort vertex_size, + ushort nr_vertices ); + + void *(*map_vertices)( struct vbuf_render * ); + void (*unmap_vertices)( struct vbuf_render *, + ushort min_index, + ushort max_index ); /** * Notify the renderer of the current primitive when it changes. @@ -109,10 +114,7 @@ struct vbuf_render { /** * Called when vbuf is done with this set of vertices: */ - void (*release_vertices)( struct vbuf_render *, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ); + void (*release_vertices)( struct vbuf_render * ); void (*destroy)( struct vbuf_render * ); }; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 0c693a4a65..9e37a26c1e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -32,7 +32,7 @@ #include "util/u_memory.h" #include "util/u_math.h" #include "pipe/p_shader_tokens.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi/tgsi_exec.h" @@ -143,7 +143,7 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, return x86_make_disp(aos_get_x86(cp, 1, X86_CONSTANTS), idx * 4 * sizeof(float)); default: - ERROR(cp, "unknown reg file"); + AOS_ERROR(cp, "unknown reg file"); return x86_make_reg(0,0); } } @@ -177,7 +177,7 @@ static void spill( struct aos_compilation *cp, unsigned idx ) (cp->xmm[idx].file != TGSI_FILE_INPUT && /* inputs are fetched into xmm & set dirty */ cp->xmm[idx].file != TGSI_FILE_OUTPUT && cp->xmm[idx].file != TGSI_FILE_TEMPORARY)) { - ERROR(cp, "invalid spill"); + AOS_ERROR(cp, "invalid spill"); return; } else { @@ -284,6 +284,15 @@ void aos_release_xmm_reg( struct aos_compilation *cp, } +static void aos_soft_release_xmm( struct aos_compilation *cp, + struct x86_reg reg ) +{ + if (reg.file == file_XMM) { + assert(cp->xmm[reg.idx].last_used == cp->insn_counter); + cp->xmm[reg.idx].last_used = cp->insn_counter - 1; + } +} + /* Mark an xmm reg as holding the current copy of a shader reg. @@ -534,7 +543,7 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, switch (swizzle) { case TGSI_EXTSWIZZLE_ZERO: case TGSI_EXTSWIZZLE_ONE: - ERROR(cp, "not supporting full swizzles yet in tgsi_aos_sse2"); + AOS_ERROR(cp, "not supporting full swizzles yet in tgsi_aos_sse2"); break; default: @@ -555,7 +564,7 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, break; default: - ERROR(cp, "unsupported sign-mode"); + AOS_ERROR(cp, "unsupported sign-mode"); break; } } @@ -584,15 +593,17 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, sse_mulps(cp->func, dst, tmp); aos_release_xmm_reg(cp, tmp.idx); + aos_soft_release_xmm(cp, imm_swz); } else if (negs) { struct x86_reg imm_negs = aos_get_internal_xmm(cp, IMM_NEGS); sse_mulps(cp->func, dst, imm_negs); + aos_soft_release_xmm(cp, imm_negs); } if (abs && abs != 0xf) { - ERROR(cp, "unsupported partial abs"); + AOS_ERROR(cp, "unsupported partial abs"); } else if (abs) { struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); @@ -603,8 +614,10 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, sse_maxps(cp->func, dst, tmp); aos_release_xmm_reg(cp, tmp.idx); + aos_soft_release_xmm(cp, neg); } + aos_soft_release_xmm(cp, arg0); return dst; } @@ -657,7 +670,7 @@ static void x87_fld_src( struct aos_compilation *cp, break; default: - ERROR(cp, "unsupported sign-mode"); + AOS_ERROR(cp, "unsupported sign-mode"); break; } } @@ -1559,7 +1572,6 @@ static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_inst */ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { - if (0) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg r = aos_get_xmm_reg(cp); @@ -1568,21 +1580,30 @@ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } else { - struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); - struct x86_reg r = aos_get_xmm_reg(cp); + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg r = aos_get_xmm_reg(cp); struct x86_reg neg_half = get_reg_ptr( cp, AOS_FILE_INTERNAL, IMM_RSQ ); struct x86_reg one_point_five = x86_make_disp( neg_half, 4 ); struct x86_reg src = get_xmm_writable( cp, arg0 ); - - sse_rsqrtss( cp->func, r, src ); /* rsqrtss(a) */ - sse_mulss( cp->func, src, neg_half ); /* -.5 * a */ - sse_mulss( cp->func, src, r ); /* -.5 * a * r */ - sse_mulss( cp->func, src, r ); /* -.5 * a * r * r */ - sse_addss( cp->func, src, one_point_five ); /* 1.5 - .5 * a * r * r */ - sse_mulss( cp->func, r, src ); /* r * (1.5 - .5 * a * r * r) */ + struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); + struct x86_reg tmp = aos_get_xmm_reg(cp); + + sse_movaps(cp->func, tmp, src); + sse_mulps(cp->func, tmp, neg); + sse_maxps(cp->func, tmp, src); + + sse_rsqrtss( cp->func, r, tmp ); /* rsqrtss(a) */ + sse_mulss( cp->func, tmp, neg_half ); /* -.5 * a */ + sse_mulss( cp->func, tmp, r ); /* -.5 * a * r */ + sse_mulss( cp->func, tmp, r ); /* -.5 * a * r * r */ + sse_addss( cp->func, tmp, one_point_five ); /* 1.5 - .5 * a * r * r */ + sse_mulss( cp->func, r, tmp ); /* r * (1.5 - .5 * a * r * r) */ store_scalar_dest(cp, &op->FullDstRegisters[0], r); + + aos_release_xmm_reg(cp, tmp.idx); + return TRUE; } } diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 264387517b..2cf72ddf7b 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -204,7 +204,7 @@ struct x86_reg aos_get_internal_xmm( struct aos_compilation *cp, unsigned imm ); -#define ERROR(cp, msg) \ +#define AOS_ERROR(cp, msg) \ do { \ if (0) debug_printf("%s: x86 translation failed: %s\n", __FUNCTION__, msg); \ cp->error = 1; \ diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 39f75b50b7..a6eb37d128 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -199,7 +199,7 @@ static boolean load_input( struct aos_compilation *cp, emit_load_R8G8B8A8_UNORM(cp, dataXMM, src); break; default: - ERROR(cp, "unhandled input format"); + AOS_ERROR(cp, "unhandled input format"); return FALSE; } @@ -410,7 +410,7 @@ static boolean emit_output( struct aos_compilation *cp, } break; default: - ERROR(cp, "unhandled output format"); + AOS_ERROR(cp, "unhandled output format"); return FALSE; } diff --git a/src/gallium/auxiliary/gallivm/storagesoa.cpp b/src/gallium/auxiliary/gallivm/storagesoa.cpp index e1e5cabcf5..4984ce985c 100644 --- a/src/gallium/auxiliary/gallivm/storagesoa.cpp +++ b/src/gallium/auxiliary/gallivm/storagesoa.cpp @@ -30,7 +30,7 @@ #include "gallivm_p.h" #include "pipe/p_shader_tokens.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include <llvm/BasicBlock.h> #include <llvm/Module.h> diff --git a/src/gallium/auxiliary/indices/Makefile b/src/gallium/auxiliary/indices/Makefile index 8fa61d265e..f2ebc3f410 100644 --- a/src/gallium/auxiliary/indices/Makefile +++ b/src/gallium/auxiliary/indices/Makefile @@ -4,13 +4,13 @@ include $(TOP)/configs/current LIBNAME = indices C_SOURCES = \ - u_indices_gen.c + u_indices_gen.c \ + u_unfilled_gen.c include ../../Makefile.template u_indices_gen.c: u_indices_gen.py python $< > $@ - -symlinks: - +u_unfilled_gen.c: u_unfilled_gen.py + python $< > $@ diff --git a/src/gallium/auxiliary/indices/SConscript b/src/gallium/auxiliary/indices/SConscript index 65a43a9f64..e5f7ee9484 100644 --- a/src/gallium/auxiliary/indices/SConscript +++ b/src/gallium/auxiliary/indices/SConscript @@ -7,11 +7,20 @@ env.CodeGenerate( command = 'python $SCRIPT > $TARGET' ) +env.CodeGenerate( + target = 'u_unfilled_gen.c', + script = 'u_unfilled_gen.py', + source = [], + command = 'python $SCRIPT > $TARGET' +) + indices = env.ConvenienceLibrary( target = 'indices', source = [ # 'u_indices.c', +# 'u_unfilled_indices.c', 'u_indices_gen.c', + 'u_unfilled_gen.c', ]) auxiliaries.insert(0, indices) diff --git a/src/gallium/auxiliary/indices/u_indices.h b/src/gallium/auxiliary/indices/u_indices.h index abf5a3037d..be522c6725 100644 --- a/src/gallium/auxiliary/indices/u_indices.h +++ b/src/gallium/auxiliary/indices/u_indices.h @@ -80,4 +80,27 @@ int u_index_generator( unsigned hw_mask, u_generate_func *out_generate ); +void u_unfilled_init( void ); + +int u_unfilled_translator( unsigned prim, + unsigned in_index_size, + unsigned nr, + unsigned unfilled_mode, + unsigned *out_prim, + unsigned *out_index_size, + unsigned *out_nr, + u_translate_func *out_translate ); + +int u_unfilled_generator( unsigned prim, + unsigned start, + unsigned nr, + unsigned unfilled_mode, + unsigned *out_prim, + unsigned *out_index_size, + unsigned *out_nr, + u_generate_func *out_generate ); + + + + #endif diff --git a/src/gallium/auxiliary/indices/u_indices_gen.c b/src/gallium/auxiliary/indices/u_indices_gen.c index 4c05b3eedb..3c981e5d7f 100644 --- a/src/gallium/auxiliary/indices/u_indices_gen.c +++ b/src/gallium/auxiliary/indices/u_indices_gen.c @@ -34,7 +34,7 @@ #include "indices/u_indices.h" #include "indices/u_indices_priv.h" #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_defines.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/indices/u_indices_gen.py b/src/gallium/auxiliary/indices/u_indices_gen.py index 0dc58d0cd0..af63d09930 100644 --- a/src/gallium/auxiliary/indices/u_indices_gen.py +++ b/src/gallium/auxiliary/indices/u_indices_gen.py @@ -72,7 +72,7 @@ def prolog(): #include "indices/u_indices.h" #include "indices/u_indices_priv.h" #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_defines.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/indices/u_unfilled_gen.c b/src/gallium/auxiliary/indices/u_unfilled_gen.c new file mode 100644 index 0000000000..fd082ebbb3 --- /dev/null +++ b/src/gallium/auxiliary/indices/u_unfilled_gen.c @@ -0,0 +1,1152 @@ +/* File automatically generated by u_unfilled_gen.py */ + +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + + +/** + * @file + * Functions to translate and generate index lists + */ + +#include "indices/u_indices.h" +#include "indices/u_indices_priv.h" +#include "pipe/p_compiler.h" +#include "util/u_debug.h" +#include "pipe/p_defines.h" +#include "util/u_memory.h" + + +static unsigned out_size_idx( unsigned index_size ) +{ + switch (index_size) { + case 4: return OUT_UINT; + case 2: return OUT_USHORT; + default: assert(0); return OUT_USHORT; + } +} + +static unsigned in_size_idx( unsigned index_size ) +{ + switch (index_size) { + case 4: return IN_UINT; + case 2: return IN_USHORT; + case 1: return IN_UBYTE; + default: assert(0); return IN_UBYTE; + } +} + + +static u_generate_func generate_line[OUT_COUNT][PRIM_COUNT]; +static u_translate_func translate_line[IN_COUNT][OUT_COUNT][PRIM_COUNT]; + + +static void generate_tris_ushort( + unsigned nr, + void *_out ) +{ + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (ushort)(i); + (out+j)[1] = (ushort)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)(i+1); + (out+j+2)[1] = (ushort)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (ushort)(i+2); + (out+j+4)[1] = (ushort)(i); + } +} +static void generate_tristrip_ushort( + unsigned nr, + void *_out ) +{ + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (ushort)(i); + (out+j)[1] = (ushort)(i+1/*+(i&1)*/); + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (ushort)(i+1/*+(i&1)*/); + (out+j+2)[1] = (ushort)(i+2/*-(i&1)*/); + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (ushort)(i+2/*-(i&1)*/); + (out+j+4)[1] = (ushort)(i); + } +} +static void generate_trifan_ushort( + unsigned nr, + void *_out ) +{ + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)(0); + (out+j)[1] = (ushort)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)(i+1); + (out+j+2)[1] = (ushort)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)(i+2); + (out+j+4)[1] = (ushort)(0); + } +} +static void generate_quads_ushort( + unsigned nr, + void *_out ) +{ + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (ushort)(i+0); + (out+j)[1] = (ushort)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)(i+1); + (out+j+2)[1] = (ushort)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (ushort)(i+2); + (out+j+4)[1] = (ushort)(i+3); + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (ushort)(i+3); + (out+j+6)[1] = (ushort)(i+0); + } +} +static void generate_quadstrip_ushort( + unsigned nr, + void *_out ) +{ + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (ushort)(i+2); + (out+j)[1] = (ushort)(i+0); + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (ushort)(i+0); + (out+j+2)[1] = (ushort)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (ushort)(i+1); + (out+j+4)[1] = (ushort)(i+3); + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (ushort)(i+3); + (out+j+6)[1] = (ushort)(i+2); + } +} +static void generate_polygon_ushort( + unsigned nr, + void *_out ) +{ + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)(0); + (out+j)[1] = (ushort)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)(i+1); + (out+j+2)[1] = (ushort)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)(i+2); + (out+j+4)[1] = (ushort)(0); + } +} +static void generate_tris_uint( + unsigned nr, + void *_out ) +{ + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (uint)(i); + (out+j)[1] = (uint)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)(i+1); + (out+j+2)[1] = (uint)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (uint)(i+2); + (out+j+4)[1] = (uint)(i); + } +} +static void generate_tristrip_uint( + unsigned nr, + void *_out ) +{ + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (uint)(i); + (out+j)[1] = (uint)(i+1/*+(i&1)*/); + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (uint)(i+1/*+(i&1)*/); + (out+j+2)[1] = (uint)(i+2/*-(i&1)*/); + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (uint)(i+2/*-(i&1)*/); + (out+j+4)[1] = (uint)(i); + } +} +static void generate_trifan_uint( + unsigned nr, + void *_out ) +{ + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)(0); + (out+j)[1] = (uint)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)(i+1); + (out+j+2)[1] = (uint)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)(i+2); + (out+j+4)[1] = (uint)(0); + } +} +static void generate_quads_uint( + unsigned nr, + void *_out ) +{ + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (uint)(i+0); + (out+j)[1] = (uint)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)(i+1); + (out+j+2)[1] = (uint)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (uint)(i+2); + (out+j+4)[1] = (uint)(i+3); + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (uint)(i+3); + (out+j+6)[1] = (uint)(i+0); + } +} +static void generate_quadstrip_uint( + unsigned nr, + void *_out ) +{ + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (uint)(i+2); + (out+j)[1] = (uint)(i+0); + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (uint)(i+0); + (out+j+2)[1] = (uint)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (uint)(i+1); + (out+j+4)[1] = (uint)(i+3); + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (uint)(i+3); + (out+j+6)[1] = (uint)(i+2); + } +} +static void generate_polygon_uint( + unsigned nr, + void *_out ) +{ + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)(0); + (out+j)[1] = (uint)(i+1); + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)(i+1); + (out+j+2)[1] = (uint)(i+2); + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)(i+2); + (out+j+4)[1] = (uint)(0); + } +} +static void translate_tris_ubyte2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (ushort)in[i]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[i]; + } +} +static void translate_tristrip_ubyte2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (ushort)in[i]; + (out+j)[1] = (ushort)in[i+1/*+(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/]; + (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/]; + (out+j+4)[1] = (ushort)in[i]; + } +} +static void translate_trifan_ubyte2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)in[0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[0]; + } +} +static void translate_quads_ubyte2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (ushort)in[i+0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (ushort)in[i+3]; + (out+j+6)[1] = (ushort)in[i+0]; + } +} +static void translate_quadstrip_ubyte2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (ushort)in[i+2]; + (out+j)[1] = (ushort)in[i+0]; + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (ushort)in[i+0]; + (out+j+2)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (ushort)in[i+1]; + (out+j+4)[1] = (ushort)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (ushort)in[i+3]; + (out+j+6)[1] = (ushort)in[i+2]; + } +} +static void translate_polygon_ubyte2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)in[0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[0]; + } +} +static void translate_tris_ubyte2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (uint)in[i]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[i]; + } +} +static void translate_tristrip_ubyte2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (uint)in[i]; + (out+j)[1] = (uint)in[i+1/*+(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/]; + (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/]; + (out+j+4)[1] = (uint)in[i]; + } +} +static void translate_trifan_ubyte2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)in[0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[0]; + } +} +static void translate_quads_ubyte2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (uint)in[i+0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (uint)in[i+3]; + (out+j+6)[1] = (uint)in[i+0]; + } +} +static void translate_quadstrip_ubyte2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (uint)in[i+2]; + (out+j)[1] = (uint)in[i+0]; + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (uint)in[i+0]; + (out+j+2)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (uint)in[i+1]; + (out+j+4)[1] = (uint)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (uint)in[i+3]; + (out+j+6)[1] = (uint)in[i+2]; + } +} +static void translate_polygon_ubyte2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ubyte*in = (const ubyte*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)in[0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[0]; + } +} +static void translate_tris_ushort2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (ushort)in[i]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[i]; + } +} +static void translate_tristrip_ushort2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (ushort)in[i]; + (out+j)[1] = (ushort)in[i+1/*+(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/]; + (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/]; + (out+j+4)[1] = (ushort)in[i]; + } +} +static void translate_trifan_ushort2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)in[0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[0]; + } +} +static void translate_quads_ushort2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (ushort)in[i+0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (ushort)in[i+3]; + (out+j+6)[1] = (ushort)in[i+0]; + } +} +static void translate_quadstrip_ushort2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (ushort)in[i+2]; + (out+j)[1] = (ushort)in[i+0]; + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (ushort)in[i+0]; + (out+j+2)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (ushort)in[i+1]; + (out+j+4)[1] = (ushort)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (ushort)in[i+3]; + (out+j+6)[1] = (ushort)in[i+2]; + } +} +static void translate_polygon_ushort2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)in[0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[0]; + } +} +static void translate_tris_ushort2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (uint)in[i]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[i]; + } +} +static void translate_tristrip_ushort2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (uint)in[i]; + (out+j)[1] = (uint)in[i+1/*+(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/]; + (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/]; + (out+j+4)[1] = (uint)in[i]; + } +} +static void translate_trifan_ushort2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)in[0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[0]; + } +} +static void translate_quads_ushort2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (uint)in[i+0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (uint)in[i+3]; + (out+j+6)[1] = (uint)in[i+0]; + } +} +static void translate_quadstrip_ushort2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (uint)in[i+2]; + (out+j)[1] = (uint)in[i+0]; + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (uint)in[i+0]; + (out+j+2)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (uint)in[i+1]; + (out+j+4)[1] = (uint)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (uint)in[i+3]; + (out+j+6)[1] = (uint)in[i+2]; + } +} +static void translate_polygon_ushort2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const ushort*in = (const ushort*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)in[0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[0]; + } +} +static void translate_tris_uint2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (ushort)in[i]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[i]; + } +} +static void translate_tristrip_uint2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (ushort)in[i]; + (out+j)[1] = (ushort)in[i+1/*+(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/]; + (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/]; + (out+j+4)[1] = (ushort)in[i]; + } +} +static void translate_trifan_uint2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)in[0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[0]; + } +} +static void translate_quads_uint2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (ushort)in[i+0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (ushort)in[i+3]; + (out+j+6)[1] = (ushort)in[i+0]; + } +} +static void translate_quadstrip_uint2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (ushort)in[i+2]; + (out+j)[1] = (ushort)in[i+0]; + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (ushort)in[i+0]; + (out+j+2)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (ushort)in[i+1]; + (out+j+4)[1] = (ushort)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (ushort)in[i+3]; + (out+j+6)[1] = (ushort)in[i+2]; + } +} +static void translate_polygon_uint2ushort( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + ushort *out = (ushort*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (ushort)in[0]; + (out+j)[1] = (ushort)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (ushort)in[i+1]; + (out+j+2)[1] = (ushort)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (ushort)in[i+2]; + (out+j+4)[1] = (ushort)in[0]; + } +} +static void translate_tris_uint2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i+=3) { + debug_printf(" line %d %d\n", (int)i, (int)i+1); + (out+j)[0] = (uint)in[i]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[i]; + } +} +static void translate_tristrip_uint2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); + (out+j)[0] = (uint)in[i]; + (out+j)[1] = (uint)in[i+1/*+(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); + (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/]; + (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/]; + debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); + (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/]; + (out+j+4)[1] = (uint)in[i]; + } +} +static void translate_trifan_uint2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)in[0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[0]; + } +} +static void translate_quads_uint2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=4) { + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j)[0] = (uint)in[i+0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)i+3); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+0); + (out+j+6)[0] = (uint)in[i+3]; + (out+j+6)[1] = (uint)in[i+0]; + } +} +static void translate_quadstrip_uint2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=8, i+=2) { + debug_printf(" line %d %d\n", (int)i+2, (int)i+0); + (out+j)[0] = (uint)in[i+2]; + (out+j)[1] = (uint)in[i+0]; + debug_printf(" line %d %d\n", (int)i+0, (int)i+1); + (out+j+2)[0] = (uint)in[i+0]; + (out+j+2)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+3); + (out+j+4)[0] = (uint)in[i+1]; + (out+j+4)[1] = (uint)in[i+3]; + debug_printf(" line %d %d\n", (int)i+3, (int)i+2); + (out+j+6)[0] = (uint)in[i+3]; + (out+j+6)[1] = (uint)in[i+2]; + } +} +static void translate_polygon_uint2uint( + const void * _in, + unsigned nr, + void *_out ) +{ + const uint*in = (const uint*)_in; + uint *out = (uint*)_out; + unsigned i, j; + (void)j; + for (j = i = 0; j < nr; j+=6, i++) { + debug_printf(" line %d %d\n", (int)0, (int)i+1); + (out+j)[0] = (uint)in[0]; + (out+j)[1] = (uint)in[i+1]; + debug_printf(" line %d %d\n", (int)i+1, (int)i+2); + (out+j+2)[0] = (uint)in[i+1]; + (out+j+2)[1] = (uint)in[i+2]; + debug_printf(" line %d %d\n", (int)i+2, (int)0); + (out+j+4)[0] = (uint)in[i+2]; + (out+j+4)[1] = (uint)in[0]; + } +} +void u_unfilled_init( void ) +{ + static int firsttime = 1; + if (!firsttime) return; + firsttime = 0; +generate_line[OUT_USHORT][PIPE_PRIM_TRIANGLES] = generate_tris_ushort; +generate_line[OUT_USHORT][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_ushort; +generate_line[OUT_USHORT][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_ushort; +generate_line[OUT_USHORT][PIPE_PRIM_QUADS] = generate_quads_ushort; +generate_line[OUT_USHORT][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_ushort; +generate_line[OUT_USHORT][PIPE_PRIM_POLYGON] = generate_polygon_ushort; +generate_line[OUT_UINT][PIPE_PRIM_TRIANGLES] = generate_tris_uint; +generate_line[OUT_UINT][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_uint; +generate_line[OUT_UINT][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_uint; +generate_line[OUT_UINT][PIPE_PRIM_QUADS] = generate_quads_uint; +generate_line[OUT_UINT][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_uint; +generate_line[OUT_UINT][PIPE_PRIM_POLYGON] = generate_polygon_uint; +translate_line[IN_UBYTE][OUT_USHORT][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2ushort; +translate_line[IN_UBYTE][OUT_USHORT][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2ushort; +translate_line[IN_UBYTE][OUT_USHORT][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2ushort; +translate_line[IN_UBYTE][OUT_USHORT][PIPE_PRIM_QUADS] = translate_quads_ubyte2ushort; +translate_line[IN_UBYTE][OUT_USHORT][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2ushort; +translate_line[IN_UBYTE][OUT_USHORT][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2ushort; +translate_line[IN_UBYTE][OUT_UINT][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2uint; +translate_line[IN_UBYTE][OUT_UINT][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2uint; +translate_line[IN_UBYTE][OUT_UINT][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2uint; +translate_line[IN_UBYTE][OUT_UINT][PIPE_PRIM_QUADS] = translate_quads_ubyte2uint; +translate_line[IN_UBYTE][OUT_UINT][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2uint; +translate_line[IN_UBYTE][OUT_UINT][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2uint; +translate_line[IN_USHORT][OUT_USHORT][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2ushort; +translate_line[IN_USHORT][OUT_USHORT][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2ushort; +translate_line[IN_USHORT][OUT_USHORT][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2ushort; +translate_line[IN_USHORT][OUT_USHORT][PIPE_PRIM_QUADS] = translate_quads_ushort2ushort; +translate_line[IN_USHORT][OUT_USHORT][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2ushort; +translate_line[IN_USHORT][OUT_USHORT][PIPE_PRIM_POLYGON] = translate_polygon_ushort2ushort; +translate_line[IN_USHORT][OUT_UINT][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2uint; +translate_line[IN_USHORT][OUT_UINT][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2uint; +translate_line[IN_USHORT][OUT_UINT][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2uint; +translate_line[IN_USHORT][OUT_UINT][PIPE_PRIM_QUADS] = translate_quads_ushort2uint; +translate_line[IN_USHORT][OUT_UINT][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2uint; +translate_line[IN_USHORT][OUT_UINT][PIPE_PRIM_POLYGON] = translate_polygon_ushort2uint; +translate_line[IN_UINT][OUT_USHORT][PIPE_PRIM_TRIANGLES] = translate_tris_uint2ushort; +translate_line[IN_UINT][OUT_USHORT][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2ushort; +translate_line[IN_UINT][OUT_USHORT][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2ushort; +translate_line[IN_UINT][OUT_USHORT][PIPE_PRIM_QUADS] = translate_quads_uint2ushort; +translate_line[IN_UINT][OUT_USHORT][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2ushort; +translate_line[IN_UINT][OUT_USHORT][PIPE_PRIM_POLYGON] = translate_polygon_uint2ushort; +translate_line[IN_UINT][OUT_UINT][PIPE_PRIM_TRIANGLES] = translate_tris_uint2uint; +translate_line[IN_UINT][OUT_UINT][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2uint; +translate_line[IN_UINT][OUT_UINT][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2uint; +translate_line[IN_UINT][OUT_UINT][PIPE_PRIM_QUADS] = translate_quads_uint2uint; +translate_line[IN_UINT][OUT_UINT][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2uint; +translate_line[IN_UINT][OUT_UINT][PIPE_PRIM_POLYGON] = translate_polygon_uint2uint; +} +#include "indices/u_unfilled_indices.c" diff --git a/src/gallium/auxiliary/indices/u_unfilled_gen.py b/src/gallium/auxiliary/indices/u_unfilled_gen.py new file mode 100644 index 0000000000..d0344fe313 --- /dev/null +++ b/src/gallium/auxiliary/indices/u_unfilled_gen.py @@ -0,0 +1,247 @@ +#!/usr/bin/env python +copyright = ''' +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +''' + +GENERATE, UBYTE, USHORT, UINT = 'generate', 'ubyte', 'ushort', 'uint' +FIRST, LAST = 'first', 'last' + +INTYPES = (GENERATE, UBYTE, USHORT, UINT) +OUTTYPES = (USHORT, UINT) +PRIMS=('tris', + 'trifan', + 'tristrip', + 'quads', + 'quadstrip', + 'polygon') + +LONGPRIMS=('PIPE_PRIM_TRIANGLES', + 'PIPE_PRIM_TRIANGLE_FAN', + 'PIPE_PRIM_TRIANGLE_STRIP', + 'PIPE_PRIM_QUADS', + 'PIPE_PRIM_QUAD_STRIP', + 'PIPE_PRIM_POLYGON') + +longprim = dict(zip(PRIMS, LONGPRIMS)) +intype_idx = dict(ubyte='IN_UBYTE', ushort='IN_USHORT', uint='IN_UINT') +outtype_idx = dict(ushort='OUT_USHORT', uint='OUT_UINT') + + +def prolog(): + print '''/* File automatically generated by u_unfilled_gen.py */''' + print copyright + print r''' + +/** + * @file + * Functions to translate and generate index lists + */ + +#include "indices/u_indices.h" +#include "indices/u_indices_priv.h" +#include "pipe/p_compiler.h" +#include "util/u_debug.h" +#include "pipe/p_defines.h" +#include "util/u_memory.h" + + +static unsigned out_size_idx( unsigned index_size ) +{ + switch (index_size) { + case 4: return OUT_UINT; + case 2: return OUT_USHORT; + default: assert(0); return OUT_USHORT; + } +} + +static unsigned in_size_idx( unsigned index_size ) +{ + switch (index_size) { + case 4: return IN_UINT; + case 2: return IN_USHORT; + case 1: return IN_UBYTE; + default: assert(0); return IN_UBYTE; + } +} + + +static u_generate_func generate_line[OUT_COUNT][PRIM_COUNT]; +static u_translate_func translate_line[IN_COUNT][OUT_COUNT][PRIM_COUNT]; + +''' + +def vert( intype, outtype, v0 ): + if intype == GENERATE: + return '(' + outtype + ')(' + v0 + ')' + else: + return '(' + outtype + ')in[' + v0 + ']' + +def line( intype, outtype, ptr, v0, v1 ): + print ' debug_printf(" line %d %d\\n", (int)' + v0 + ', (int)' + v1 + ');' + print ' (' + ptr + ')[0] = ' + vert( intype, outtype, v0 ) + ';' + print ' (' + ptr + ')[1] = ' + vert( intype, outtype, v1 ) + ';' + +# XXX: have the opportunity here to avoid over-drawing shared lines in +# tristrips, fans, etc, by integrating this into the calling functions +# and only emitting each line at most once. +# +def do_tri( intype, outtype, ptr, v0, v1, v2 ): + line( intype, outtype, ptr, v0, v1 ) + line( intype, outtype, ptr + '+2', v1, v2 ) + line( intype, outtype, ptr + '+4', v2, v0 ) + +def do_quad( intype, outtype, ptr, v0, v1, v2, v3 ): + line( intype, outtype, ptr, v0, v1 ) + line( intype, outtype, ptr + '+2', v1, v2 ) + line( intype, outtype, ptr + '+4', v2, v3 ) + line( intype, outtype, ptr + '+6', v3, v0 ) + +def name(intype, outtype, prim): + if intype == GENERATE: + return 'generate_' + prim + '_' + outtype + else: + return 'translate_' + prim + '_' + intype + '2' + outtype + +def preamble(intype, outtype, prim): + print 'static void ' + name( intype, outtype, prim ) + '(' + if intype != GENERATE: + print ' const void * _in,' + print ' unsigned nr,' + print ' void *_out )' + print '{' + if intype != GENERATE: + print ' const ' + intype + '*in = (const ' + intype + '*)_in;' + print ' ' + outtype + ' *out = (' + outtype + '*)_out;' + print ' unsigned i, j;' + print ' (void)j;' + +def postamble(): + print '}' + + +def tris(intype, outtype): + preamble(intype, outtype, prim='tris') + print ' for (j = i = 0; j < nr; j+=6, i+=3) { ' + do_tri( intype, outtype, 'out+j', 'i', 'i+1', 'i+2' ); + print ' }' + postamble() + + +def tristrip(intype, outtype): + preamble(intype, outtype, prim='tristrip') + print ' for (j = i = 0; j < nr; j+=6, i++) { ' + do_tri( intype, outtype, 'out+j', 'i', 'i+1/*+(i&1)*/', 'i+2/*-(i&1)*/' ); + print ' }' + postamble() + + +def trifan(intype, outtype): + preamble(intype, outtype, prim='trifan') + print ' for (j = i = 0; j < nr; j+=6, i++) { ' + do_tri( intype, outtype, 'out+j', '0', 'i+1', 'i+2' ); + print ' }' + postamble() + + + +def polygon(intype, outtype): + preamble(intype, outtype, prim='polygon') + print ' for (j = i = 0; j < nr; j+=6, i++) { ' + do_tri( intype, outtype, 'out+j', '0', 'i+1', 'i+2' ); + print ' }' + postamble() + + +def quads(intype, outtype): + preamble(intype, outtype, prim='quads') + print ' for (j = i = 0; j < nr; j+=8, i+=4) { ' + do_quad( intype, outtype, 'out+j', 'i+0', 'i+1', 'i+2', 'i+3' ); + print ' }' + postamble() + + +def quadstrip(intype, outtype): + preamble(intype, outtype, prim='quadstrip') + print ' for (j = i = 0; j < nr; j+=8, i+=2) { ' + do_quad( intype, outtype, 'out+j', 'i+2', 'i+0', 'i+1', 'i+3' ); + print ' }' + postamble() + + +def emit_funcs(): + for intype in INTYPES: + for outtype in OUTTYPES: + tris(intype, outtype) + tristrip(intype, outtype) + trifan(intype, outtype) + quads(intype, outtype) + quadstrip(intype, outtype) + polygon(intype, outtype) + +def init(intype, outtype, prim): + if intype == GENERATE: + print ('generate_line[' + + outtype_idx[outtype] + + '][' + longprim[prim] + + '] = ' + name( intype, outtype, prim ) + ';') + else: + print ('translate_line[' + + intype_idx[intype] + + '][' + outtype_idx[outtype] + + '][' + longprim[prim] + + '] = ' + name( intype, outtype, prim ) + ';') + + +def emit_all_inits(): + for intype in INTYPES: + for outtype in OUTTYPES: + for prim in PRIMS: + init(intype, outtype, prim) + +def emit_init(): + print 'void u_unfilled_init( void )' + print '{' + print ' static int firsttime = 1;' + print ' if (!firsttime) return;' + print ' firsttime = 0;' + emit_all_inits() + print '}' + + + + +def epilog(): + print '#include "indices/u_unfilled_indices.c"' + + +def main(): + prolog() + emit_funcs() + emit_init() + epilog() + + +if __name__ == '__main__': + main() diff --git a/src/gallium/auxiliary/indices/u_unfilled_indices.c b/src/gallium/auxiliary/indices/u_unfilled_indices.c new file mode 100644 index 0000000000..26c5d4d4c7 --- /dev/null +++ b/src/gallium/auxiliary/indices/u_unfilled_indices.c @@ -0,0 +1,186 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "u_indices.h" +#include "u_indices_priv.h" + + +static void translate_ubyte_ushort( const void *in, + unsigned nr, + void *out ) +{ + const ubyte *in_ub = (const ubyte *)in; + ushort *out_us = (ushort *)out; + unsigned i; + for (i = 0; i < nr; i++) + out_us[i] = (ushort) in_ub[i]; +} + +static void translate_memcpy_ushort( const void *in, + unsigned nr, + void *out ) +{ + memcpy(out, in, nr*sizeof(short)); +} + +static void translate_memcpy_uint( const void *in, + unsigned nr, + void *out ) +{ + memcpy(out, in, nr*sizeof(int)); +} + + +static void generate_linear_ushort( unsigned nr, + void *out ) +{ + ushort *out_us = (ushort *)out; + unsigned i; + for (i = 0; i < nr; i++) + out_us[i] = (ushort) i; +} + +static void generate_linear_uint( unsigned nr, + void *out ) +{ + unsigned *out_ui = (unsigned *)out; + unsigned i; + for (i = 0; i < nr; i++) + out_ui[i] = i; +} + + +static unsigned nr_lines( unsigned prim, + unsigned nr ) +{ + switch (prim) { + case PIPE_PRIM_TRIANGLES: + return (nr / 3) * 6; + case PIPE_PRIM_TRIANGLE_STRIP: + return (nr - 2) * 6; + case PIPE_PRIM_TRIANGLE_FAN: + return (nr - 2) * 6; + case PIPE_PRIM_QUADS: + return (nr / 4) * 8; + case PIPE_PRIM_QUAD_STRIP: + return (nr - 2) / 2 * 8; + case PIPE_PRIM_POLYGON: + return (nr - 2) * 6; + default: + assert(0); + return 0; + } +} + + + +int u_unfilled_translator( unsigned prim, + unsigned in_index_size, + unsigned nr, + unsigned unfilled_mode, + unsigned *out_prim, + unsigned *out_index_size, + unsigned *out_nr, + u_translate_func *out_translate ) +{ + unsigned in_idx; + unsigned out_idx; + + u_unfilled_init(); + + in_idx = in_size_idx(in_index_size); + *out_index_size = (in_index_size == 4) ? 4 : 2; + out_idx = out_size_idx(*out_index_size); + + if (unfilled_mode == PIPE_POLYGON_MODE_POINT) + { + *out_prim = PIPE_PRIM_POINTS; + *out_nr = nr; + + switch (in_index_size) + { + case 1: + *out_translate = translate_ubyte_ushort; + return U_TRANSLATE_NORMAL; + case 2: + *out_translate = translate_memcpy_uint; + return U_TRANSLATE_MEMCPY; + case 4: + *out_translate = translate_memcpy_ushort; + return U_TRANSLATE_MEMCPY; + default: + *out_translate = translate_memcpy_uint; + *out_nr = 0; + assert(0); + return U_TRANSLATE_ERROR; + } + } + else { + assert(unfilled_mode == PIPE_POLYGON_MODE_LINE); + *out_prim = PIPE_PRIM_LINES; + *out_translate = translate_line[in_idx][out_idx][prim]; + *out_nr = nr_lines( prim, nr ); + return U_TRANSLATE_NORMAL; + } +} + + + +int u_unfilled_generator( unsigned prim, + unsigned start, + unsigned nr, + unsigned unfilled_mode, + unsigned *out_prim, + unsigned *out_index_size, + unsigned *out_nr, + u_generate_func *out_generate ) +{ + unsigned out_idx; + + u_unfilled_init(); + + *out_index_size = ((start + nr) > 0xfffe) ? 4 : 2; + out_idx = out_size_idx(*out_index_size); + + if (unfilled_mode == PIPE_POLYGON_MODE_POINT) { + + if (*out_index_size == 4) + *out_generate = generate_linear_uint; + else + *out_generate = generate_linear_ushort; + + *out_prim = PIPE_PRIM_POINTS; + *out_nr = nr; + return U_GENERATE_LINEAR; + } + else { + assert(unfilled_mode == PIPE_POLYGON_MODE_LINE); + *out_prim = PIPE_PRIM_LINES; + *out_generate = generate_line[out_idx][prim]; + *out_nr = nr_lines( prim, nr ); + + return U_GENERATE_REUSABLE; + } +} + diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index 3b501c51df..1c00ba8d98 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -17,6 +17,3 @@ C_SOURCES = \ pb_validate.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index d8f1f02d68..2a1315922a 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -45,10 +45,9 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_error.h" #include "pipe/p_state.h" -#include "pipe/p_inlines.h" #ifdef __cplusplus @@ -159,7 +158,7 @@ pb_map(struct pb_buffer *buf, assert(buf); if(!buf) return NULL; - assert(buf->base.refcount > 0); + assert(p_atomic_read(&buf->base.reference.count) > 0); return buf->vtbl->map(buf, flags); } @@ -170,7 +169,7 @@ pb_unmap(struct pb_buffer *buf) assert(buf); if(!buf) return; - assert(buf->base.refcount > 0); + assert(p_atomic_read(&buf->base.reference.count) > 0); buf->vtbl->unmap(buf); } @@ -186,7 +185,7 @@ pb_get_base_buffer( struct pb_buffer *buf, offset = 0; return; } - assert(buf->base.refcount > 0); + assert(p_atomic_read(&buf->base.reference.count) > 0); assert(buf->vtbl->get_base_buffer); buf->vtbl->get_base_buffer(buf, base_buf, offset); assert(*base_buf); @@ -222,29 +221,18 @@ pb_destroy(struct pb_buffer *buf) assert(buf); if(!buf) return; - assert(buf->base.refcount == 0); + assert(p_atomic_read(&buf->base.reference.count) == 0); buf->vtbl->destroy(buf); } - -/* XXX: thread safety issues! - */ static INLINE void pb_reference(struct pb_buffer **dst, struct pb_buffer *src) { - if (src) { - assert(src->base.refcount); - src->base.refcount++; - } - - if (*dst) { - assert((*dst)->base.refcount); - if(--(*dst)->base.refcount == 0) - pb_destroy( *dst ); - } + struct pb_buffer *old = *dst; - *dst = src; + if (pipe_reference((struct pipe_reference**)dst, &src->base.reference)) + pb_destroy( old ); } diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index f9e6226436..1bdf7a0b2d 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -36,14 +36,14 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #include <unistd.h> #include <sched.h> #endif #include "pipe/p_compiler.h" #include "pipe/p_error.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "util/u_memory.h" #include "util/u_double_list.h" @@ -66,8 +66,12 @@ struct fenced_buffer_list struct pb_fence_ops *ops; size_t numDelayed; - struct list_head delayed; + +#ifdef DEBUG + size_t numUnfenced; + struct list_head unfenced; +#endif }; @@ -111,12 +115,15 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf) { struct fenced_buffer_list *fenced_list = fenced_buf->list; - assert(fenced_buf->base.base.refcount); + assert(p_atomic_read(&fenced_buf->base.base.reference.count)); assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); assert(fenced_buf->fence); - assert(!fenced_buf->head.prev); - assert(!fenced_buf->head.next); +#ifdef DEBUG + LIST_DEL(&fenced_buf->head); + assert(fenced_list->numUnfenced); + --fenced_list->numUnfenced; +#endif LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); ++fenced_list->numDelayed; } @@ -128,8 +135,19 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf) static INLINE void _fenced_buffer_destroy(struct fenced_buffer *fenced_buf) { - assert(!fenced_buf->base.base.refcount); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + assert(p_atomic_read(&fenced_buf->base.base.reference.count) == 0); assert(!fenced_buf->fence); +#ifdef DEBUG + assert(fenced_buf->head.prev); + assert(fenced_buf->head.next); + LIST_DEL(&fenced_buf->head); + assert(fenced_list->numUnfenced); + --fenced_list->numUnfenced; +#else + (void)fenced_list; +#endif pb_reference(&fenced_buf->buffer, NULL); FREE(fenced_buf); } @@ -149,16 +167,21 @@ _fenced_buffer_remove(struct fenced_buffer_list *fenced_list, assert(fenced_buf->head.prev); assert(fenced_buf->head.next); - LIST_DEL(&fenced_buf->head); -#ifdef DEBUG - fenced_buf->head.prev = NULL; - fenced_buf->head.next = NULL; -#endif + LIST_DEL(&fenced_buf->head); assert(fenced_list->numDelayed); --fenced_list->numDelayed; - if(!fenced_buf->base.base.refcount) +#ifdef DEBUG + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->unfenced); + ++fenced_list->numUnfenced; +#endif + + /** + * FIXME!!! + */ + + if(!p_atomic_read(&fenced_buf->base.base.reference.count)) _fenced_buffer_destroy(fenced_buf); } @@ -234,7 +257,7 @@ fenced_buffer_destroy(struct pb_buffer *buf) struct fenced_buffer_list *fenced_list = fenced_buf->list; pipe_mutex_lock(fenced_list->mutex); - assert(fenced_buf->base.base.refcount == 0); + assert(p_atomic_read(&fenced_buf->base.base.reference.count) == 0); if (fenced_buf->fence) { struct pb_fence_ops *ops = fenced_list->ops; if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) { @@ -265,16 +288,26 @@ fenced_buffer_map(struct pb_buffer *buf, unsigned flags) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pb_fence_ops *ops = fenced_list->ops; void *map; - assert(flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE); - assert(!(flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE)); - flags &= PIPE_BUFFER_USAGE_CPU_READ_WRITE; + assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE)); - /* Check for GPU read/write access */ - if(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) { - /* Wait for the GPU to finish writing */ - _fenced_buffer_finish(fenced_buf); + /* Serialize writes */ + if((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) || + ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) && (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) { + if(flags & PIPE_BUFFER_USAGE_DONTBLOCK) { + /* Don't wait for the GPU to finish writing */ + if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) + _fenced_buffer_remove(fenced_list, fenced_buf); + else + return NULL; + } + else { + /* Wait for the GPU to finish writing */ + _fenced_buffer_finish(fenced_buf); + } } #if 0 @@ -288,7 +321,7 @@ fenced_buffer_map(struct pb_buffer *buf, map = pb_map(fenced_buf->buffer, flags); if(map) { ++fenced_buf->mapcount; - fenced_buf->flags |= flags; + fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE; } return map; @@ -432,7 +465,7 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, return NULL; } - buf->base.base.refcount = 1; + pipe_reference_init(&buf->base.base.reference, 1); buf->base.base.alignment = buffer->base.alignment; buf->base.base.usage = buffer->base.usage; buf->base.base.size = buffer->base.size; @@ -441,6 +474,13 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, buf->buffer = buffer; buf->list = fenced_list; +#ifdef DEBUG + pipe_mutex_lock(fenced_list->mutex); + LIST_ADDTAIL(&buf->head, &fenced_list->unfenced); + ++fenced_list->numUnfenced; + pipe_mutex_unlock(fenced_list->mutex); +#endif + return &buf->base; } @@ -457,9 +497,13 @@ fenced_buffer_list_create(struct pb_fence_ops *ops) fenced_list->ops = ops; LIST_INITHEAD(&fenced_list->delayed); - fenced_list->numDelayed = 0; +#ifdef DEBUG + LIST_INITHEAD(&fenced_list->unfenced); + fenced_list->numUnfenced = 0; +#endif + pipe_mutex_init(fenced_list->mutex); return fenced_list; @@ -476,6 +520,51 @@ fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, } +#ifdef DEBUG +void +fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) +{ + struct pb_fence_ops *ops = fenced_list->ops; + struct list_head *curr, *next; + struct fenced_buffer *fenced_buf; + + pipe_mutex_lock(fenced_list->mutex); + + debug_printf("%10s %7s %10s %s\n", + "buffer", "reference.count", "fence", "signalled"); + + curr = fenced_list->unfenced.next; + next = curr->next; + while(curr != &fenced_list->unfenced) { + fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + assert(!fenced_buf->fence); + debug_printf("%10p %7u\n", + fenced_buf, + fenced_buf->base.base.reference.count); + curr = next; + next = curr->next; + } + + curr = fenced_list->delayed.next; + next = curr->next; + while(curr != &fenced_list->delayed) { + int signaled; + fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + signaled = ops->fence_signalled(ops, fenced_buf->fence, 0); + debug_printf("%10p %7u %10p %s\n", + fenced_buf, + fenced_buf->base.base.reference.count, + fenced_buf->fence, + signaled == 0 ? "y" : "n"); + curr = next; + next = curr->next; + } + + pipe_mutex_unlock(fenced_list->mutex); +} +#endif + + void fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) { @@ -484,13 +573,17 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) /* Wait on outstanding fences */ while (fenced_list->numDelayed) { pipe_mutex_unlock(fenced_list->mutex); -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) sched_yield(); #endif _fenced_buffer_list_check_free(fenced_list, 1); pipe_mutex_lock(fenced_list->mutex); } +#ifdef DEBUG + //assert(!fenced_list->numUnfenced); +#endif + pipe_mutex_unlock(fenced_list->mutex); fenced_list->ops->destroy(fenced_list->ops); diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h index d1c9d4c17d..034ca1e024 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h @@ -51,7 +51,7 @@ #define PB_BUFFER_FENCED_H_ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #ifdef __cplusplus @@ -114,6 +114,13 @@ void fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, int wait); + +#ifdef DEBUG +void +fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list); +#endif + + void fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c index 53f497cfb0..689fd74771 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c @@ -34,7 +34,7 @@ */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "pb_buffer.h" #include "pb_bufmgr.h" @@ -132,7 +132,7 @@ pb_malloc_buffer_create(size_t size, if(!buf) return NULL; - buf->base.base.refcount = 1; + pipe_reference_init(&buf->base.base.reference, 1); buf->base.base.alignment = desc->alignment; buf->base.base.usage = desc->usage; buf->base.base.size = size; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index fec8db91c7..74077f8277 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -200,10 +200,11 @@ pb_ondemand_manager_create(struct pb_manager *provider); /** * Debug buffer manager to detect buffer under- and overflows. * - * Band size should be a multiple of the largest alignment + * Under/overflow sizes should be a multiple of the largest alignment */ struct pb_manager * -pb_debug_manager_create(struct pb_manager *provider, size_t band_size); +pb_debug_manager_create(struct pb_manager *provider, + size_t underflow_size, size_t overflow_size); #ifdef __cplusplus diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c index c956924cc7..db67d46c56 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c @@ -34,7 +34,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "pb_buffer.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index a168853713..010a2ecc1f 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -35,7 +35,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "util/u_memory.h" #include "util/u_double_list.h" @@ -112,7 +112,7 @@ _pb_cache_buffer_destroy(struct pb_cache_buffer *buf) LIST_DEL(&buf->head); assert(mgr->numDelayed); --mgr->numDelayed; - assert(!buf->base.base.refcount); + assert(p_atomic_read(&buf->base.base.reference.count) == 0); pb_reference(&buf->buffer, NULL); FREE(buf); } @@ -153,7 +153,7 @@ pb_cache_buffer_destroy(struct pb_buffer *_buf) struct pb_cache_manager *mgr = buf->mgr; pipe_mutex_lock(mgr->mutex); - assert(buf->base.base.refcount == 0); + assert(p_atomic_read(&buf->base.base.reference.count) == 0); _pb_cache_buffer_list_check_free(mgr); @@ -293,7 +293,8 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr, if(buf) { LIST_DEL(&buf->head); pipe_mutex_unlock(mgr->mutex); - ++buf->base.base.refcount; + /* Increase refcount */ + pb_reference((struct pb_buffer**)&buf, &buf->base); return &buf->base; } @@ -309,12 +310,12 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr, return NULL; } - assert(buf->buffer->base.refcount >= 1); + assert(p_atomic_read(&buf->buffer->base.reference.count) >= 1); assert(pb_check_alignment(desc->alignment, buf->buffer->base.alignment)); assert(pb_check_usage(desc->usage, buf->buffer->base.usage)); assert(buf->buffer->base.size >= size); - buf->base.base.refcount = 1; + pipe_reference_init(&buf->base.base.reference, 1); buf->base.base.alignment = buf->buffer->base.alignment; buf->base.base.usage = buf->buffer->base.usage; buf->base.base.size = buf->buffer->base.size; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index 26d9c24aec..478682dbee 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -34,7 +34,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -78,7 +78,8 @@ struct pb_debug_manager struct pb_manager *provider; - size_t band_size; + size_t underflow_size; + size_t overflow_size; }; @@ -205,9 +206,9 @@ pb_debug_buffer_check(struct pb_debug_buffer *buf) static void pb_debug_buffer_destroy(struct pb_buffer *_buf) { - struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + struct pb_debug_buffer *buf = pb_debug_buffer(_buf); - assert(!buf->base.base.refcount); + assert(p_atomic_read(&buf->base.base.reference.count) == 0); pb_debug_buffer_check(buf); @@ -301,7 +302,7 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr, if(!buf) return NULL; - real_size = size + 2*mgr->band_size; + real_size = mgr->underflow_size + size + mgr->overflow_size; real_desc = *desc; real_desc.usage |= PIPE_BUFFER_USAGE_CPU_WRITE; real_desc.usage |= PIPE_BUFFER_USAGE_CPU_READ; @@ -314,12 +315,12 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr, return NULL; } - assert(buf->buffer->base.refcount >= 1); + assert(p_atomic_read(&buf->buffer->base.reference.count) >= 1); assert(pb_check_alignment(real_desc.alignment, buf->buffer->base.alignment)); assert(pb_check_usage(real_desc.usage, buf->buffer->base.usage)); assert(buf->buffer->base.size >= real_size); - buf->base.base.refcount = 1; + pipe_reference_init(&buf->base.base.reference, 1); buf->base.base.alignment = desc->alignment; buf->base.base.usage = desc->usage; buf->base.base.size = size; @@ -327,7 +328,7 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr, buf->base.vtbl = &pb_debug_buffer_vtbl; buf->mgr = mgr; - buf->underflow_size = mgr->band_size; + buf->underflow_size = mgr->underflow_size; buf->overflow_size = buf->buffer->base.size - buf->underflow_size - size; pb_debug_buffer_fill(buf); @@ -356,7 +357,8 @@ pb_debug_manager_destroy(struct pb_manager *_mgr) struct pb_manager * -pb_debug_manager_create(struct pb_manager *provider, size_t band_size) +pb_debug_manager_create(struct pb_manager *provider, + size_t underflow_size, size_t overflow_size) { struct pb_debug_manager *mgr; @@ -371,7 +373,8 @@ pb_debug_manager_create(struct pb_manager *provider, size_t band_size) mgr->base.create_buffer = pb_debug_manager_create_buffer; mgr->base.flush = pb_debug_manager_flush; mgr->provider = provider; - mgr->band_size = band_size; + mgr->underflow_size = underflow_size; + mgr->overflow_size = overflow_size; return &mgr->base; } @@ -381,7 +384,8 @@ pb_debug_manager_create(struct pb_manager *provider, size_t band_size) struct pb_manager * -pb_debug_manager_create(struct pb_manager *provider, size_t band_size) +pb_debug_manager_create(struct pb_manager *provider, + size_t underflow_size, size_t overflow_size) { return provider; } diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c index 47e9fee533..144db5669b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -34,7 +34,7 @@ */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "pb_buffer.h" @@ -79,6 +79,10 @@ fenced_bufmgr_create_buffer(struct pb_manager *mgr, buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); if(!buf) { +#if 0 + fenced_buffer_list_dump(fenced_mgr->fenced_list); +#endif + /* give up */ return NULL; } diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 2f5a5d8ea0..fb18dcc5dc 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -34,7 +34,7 @@ #include "pipe/p_defines.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "util/u_memory.h" #include "util/u_double_list.h" @@ -97,11 +97,11 @@ mm_buffer_destroy(struct pb_buffer *buf) struct mm_buffer *mm_buf = mm_buffer(buf); struct mm_pb_manager *mm = mm_buf->mgr; - assert(buf->base.refcount == 0); + assert(p_atomic_read(&mm_buf->base.base.reference.count) == 0); pipe_mutex_lock(mm->mutex); u_mmFreeMem(mm_buf->block); - FREE(buf); + FREE(mm_buf); pipe_mutex_unlock(mm->mutex); } @@ -189,7 +189,7 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, return NULL; } - mm_buf->base.base.refcount = 1; + pipe_reference_init(&mm_buf->base.base.reference, 1); mm_buf->base.base.alignment = desc->alignment; mm_buf->base.base.usage = desc->usage; mm_buf->base.base.size = size; @@ -204,13 +204,9 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, #if 0 mmDumpMemInfo(mm->heap); #endif - - mm_buf->block = u_mmAllocMem(mm->heap, size, mm->align2, 0); - if(!mm_buf->block) { - FREE(mm_buf); - pipe_mutex_unlock(mm->mutex); - return NULL; - } + FREE(mm_buf); + pipe_mutex_unlock(mm->mutex); + return NULL; } /* Some sanity checks */ diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c index ba02a84e62..4f7e6b1c4d 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c @@ -34,7 +34,7 @@ */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "pb_buffer.h" #include "pb_bufmgr.h" @@ -242,7 +242,7 @@ pb_ondemand_manager_create_buffer(struct pb_manager *_mgr, if(!buf) return NULL; - buf->base.base.refcount = 1; + pipe_reference_init(&buf->base.base.reference, 1); buf->base.base.alignment = desc->alignment; buf->base.base.usage = desc->usage; buf->base.base.size = size; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c index a6ff37653e..75b95e132e 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -36,7 +36,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "pipe/p_defines.h" #include "util/u_memory.h" @@ -108,7 +108,7 @@ pool_buffer_destroy(struct pb_buffer *buf) struct pool_buffer *pool_buf = pool_buffer(buf); struct pool_pb_manager *pool = pool_buf->mgr; - assert(pool_buf->base.base.refcount == 0); + assert(p_atomic_read(&pool_buf->base.base.reference.count) == 0); pipe_mutex_lock(pool->mutex); LIST_ADD(&pool_buf->head, &pool->free); @@ -216,8 +216,8 @@ pool_bufmgr_create_buffer(struct pb_manager *mgr, pipe_mutex_unlock(pool->mutex); pool_buf = LIST_ENTRY(struct pool_buffer, item, head); - assert(pool_buf->base.base.refcount == 0); - pool_buf->base.base.refcount = 1; + assert(p_atomic_read(&pool_buf->base.base.reference.count) == 0); + pipe_reference_init(&pool_buf->base.base.reference, 1); pool_buf->base.base.alignment = desc->alignment; pool_buf->base.base.usage = desc->usage; @@ -295,7 +295,7 @@ pool_bufmgr_create(struct pb_manager *provider, pool_buf = pool->bufs; for (i = 0; i < numBufs; ++i) { - pool_buf->base.base.refcount = 0; + pipe_reference_init(&pool_buf->base.base.reference, 0); pool_buf->base.base.alignment = 0; pool_buf->base.base.usage = 0; pool_buf->base.base.size = bufSize; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index 9b9fedccb4..a431fd5211 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -38,7 +38,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_error.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "pipe/p_defines.h" #include "util/u_memory.h" @@ -202,7 +202,7 @@ pb_slab_buffer_destroy(struct pb_buffer *_buf) pipe_mutex_lock(mgr->mutex); - assert(buf->base.base.refcount == 0); + assert(p_atomic_read(&buf->base.base.reference.count) == 0); buf->mapCount = 0; @@ -340,7 +340,7 @@ pb_slab_create(struct pb_slab_manager *mgr) buf = slab->buffers; for (i=0; i < numBuffers; ++i) { - buf->base.base.refcount = 0; + pipe_reference_init(&buf->base.base.reference, 0); buf->base.base.size = mgr->bufSize; buf->base.base.alignment = 0; buf->base.base.usage = 0; @@ -419,7 +419,7 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr, pipe_mutex_unlock(mgr->mutex); buf = LIST_ENTRY(struct pb_slab_buffer, list, head); - ++buf->base.base.refcount; + pipe_reference_init(&buf->base.base.reference, 1); buf->base.base.alignment = desc->alignment; buf->base.base.usage = desc->usage; diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c index 94532bb4ce..150fd50618 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_validate.c +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c @@ -36,7 +36,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_error.h" #include "util/u_memory.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile index 252dc5274a..ab8ea464c6 100644 --- a/src/gallium/auxiliary/rtasm/Makefile +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -11,6 +11,3 @@ C_SOURCES = \ rtasm_ppc_spe.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c index 5499018b21..03bdd47238 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_cpu.c +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "rtasm_cpu.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c index be7433baf8..1f0923b683 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_execmem.c +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c @@ -31,14 +31,18 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "util/u_memory.h" #include "rtasm_execmem.h" +#if defined(PIPE_OS_BSD) +#define MAP_ANONYMOUS MAP_ANON +#endif -#if defined(PIPE_OS_LINUX) + +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) /* @@ -114,7 +118,7 @@ rtasm_exec_free(void *addr) } -#else /* PIPE_OS_LINUX */ +#else /* PIPE_OS_LINUX || PIPE_OS_BSD */ /* * Just use regular memory. @@ -134,4 +138,4 @@ rtasm_exec_free(void *addr) } -#endif /* PIPE_OS_LINUX */ +#endif /* PIPE_OS_LINUX || PIPE_OS_BSD */ diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc.c b/src/gallium/auxiliary/rtasm/rtasm_ppc.c index 1bb9026205..e3586482db 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc.c @@ -38,7 +38,7 @@ #include <stdio.h> #include "util/u_memory.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "rtasm_execmem.h" #include "rtasm_ppc.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 99ee74cf14..57fcf6de2a 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -26,7 +26,7 @@ #if defined(PIPE_ARCH_X86) #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_pointer.h" #include "rtasm_execmem.h" diff --git a/src/gallium/auxiliary/sct/Makefile b/src/gallium/auxiliary/sct/Makefile index 516d1756cf..a7d111b689 100644 --- a/src/gallium/auxiliary/sct/Makefile +++ b/src/gallium/auxiliary/sct/Makefile @@ -7,6 +7,3 @@ C_SOURCES = \ sct.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c index 49bb7ea92e..722d2b7e66 100644 --- a/src/gallium/auxiliary/sct/sct.c +++ b/src/gallium/auxiliary/sct/sct.c @@ -28,7 +28,6 @@ #include "util/u_memory.h" #include "pipe/p_state.h" -#include "pipe/p_inlines.h" #include "sct.h" @@ -372,7 +371,7 @@ sct_flush_textures(struct surface_context_tracker *sct, for (tl = ci->textures_used; tl; tl = next) { next = tl->next; - pipe_texture_release(&tl->texture); + pipe_texture_reference(&tl->texture, NULL); FREE(tl); } ci->textures_used = NULL; diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index d7df9490cf..b4900e8dba 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -19,6 +19,3 @@ C_SOURCES = \ tgsi_util.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt new file mode 100644 index 0000000000..b83abd4093 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt @@ -0,0 +1,1036 @@ +TGSI Instruction Specification +============================== +============================== + + +1 Instruction Set Operations +============================= + + +1.1 GL_NV_vertex_program +------------------------- + + +1.1.1 ARL - Address Register Load + + dst.x = floor(src.x) + dst.y = floor(src.y) + dst.z = floor(src.z) + dst.w = floor(src.w) + + +1.1.2 MOV - Move + + dst.x = src.x + dst.y = src.y + dst.z = src.z + dst.w = src.w + + +1.1.3 LIT - Light Coefficients + + dst.x = 1.0 + dst.y = max(src.x, 0.0) + dst.z = (src.x > 0.0) ? pow(max(src.y, 0.0), clamp(src.w, -128.0, 128.0)) : 0.0 + dst.w = 1.0 + + +1.1.4 RCP - Reciprocal + + dst.x = 1.0 / src.x + dst.y = 1.0 / src.x + dst.z = 1.0 / src.x + dst.w = 1.0 / src.x + + +1.1.5 RSQ - Reciprocal Square Root + + dst.x = 1.0 / sqrt(abs(src.x)) + dst.y = 1.0 / sqrt(abs(src.x)) + dst.z = 1.0 / sqrt(abs(src.x)) + dst.w = 1.0 / sqrt(abs(src.x)) + + +1.1.6 EXP - Approximate Exponential Base 2 + + dst.x = pow(2.0, floor(src.x)) + dst.y = src.x - floor(src.x) + dst.z = pow(2.0, src.x) + dst.w = 1.0 + + +1.1.7 LOG - Approximate Logarithm Base 2 + + dst.x = floor(lg2(abs(src.x))) + dst.y = abs(src.x) / pow(2.0, floor(lg2(abs(src.x)))) + dst.z = lg2(abs(src.x)) + dst.w = 1.0 + + +1.1.8 MUL - Multiply + + dst.x = src0.x * src1.x + dst.y = src0.y * src1.y + dst.z = src0.z * src1.z + dst.w = src0.w * src1.w + + +1.1.9 ADD - Add + + dst.x = src0.x + src1.x + dst.y = src0.y + src1.y + dst.z = src0.z + src1.z + dst.w = src0.w + src1.w + + +1.1.10 DP3 - 3-component Dot Product + + dst.x = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + dst.y = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + dst.z = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + + +1.1.11 DP4 - 4-component Dot Product + + dst.x = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src0.w * src1.w + dst.y = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src0.w * src1.w + dst.z = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src0.w * src1.w + dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src0.w * src1.w + + +1.1.12 DST - Distance Vector + + dst.x = 1.0 + dst.y = src0.y * src1.y + dst.z = src0.z + dst.w = src1.w + + +1.1.13 MIN - Minimum + + dst.x = min(src0.x, src1.x) + dst.y = min(src0.y, src1.y) + dst.z = min(src0.z, src1.z) + dst.w = min(src0.w, src1.w) + + +1.1.14 MAX - Maximum + + dst.x = max(src0.x, src1.x) + dst.y = max(src0.y, src1.y) + dst.z = max(src0.z, src1.z) + dst.w = max(src0.w, src1.w) + + +1.1.15 SLT - Set On Less Than + + dst.x = (src0.x < src1.x) ? 1.0 : 0.0 + dst.y = (src0.y < src1.y) ? 1.0 : 0.0 + dst.z = (src0.z < src1.z) ? 1.0 : 0.0 + dst.w = (src0.w < src1.w) ? 1.0 : 0.0 + + +1.1.16 SGE - Set On Greater Equal Than + + dst.x = (src0.x >= src1.x) ? 1.0 : 0.0 + dst.y = (src0.y >= src1.y) ? 1.0 : 0.0 + dst.z = (src0.z >= src1.z) ? 1.0 : 0.0 + dst.w = (src0.w >= src1.w) ? 1.0 : 0.0 + + +1.1.17 MAD - Multiply And Add + + dst.x = src0.x * src1.x + src2.x + dst.y = src0.y * src1.y + src2.y + dst.z = src0.z * src1.z + src2.z + dst.w = src0.w * src1.w + src2.w + + +1.2 GL_ATI_fragment_shader +--------------------------- + + +1.2.1 SUB - Subtract + + dst.x = src0.x - src1.x + dst.y = src0.y - src1.y + dst.z = src0.z - src1.z + dst.w = src0.w - src1.w + + +1.2.2 DOT3 - 3-component Dot Product + + Alias for DP3. + + +1.2.3 DOT4 - 4-component Dot Product + + Alias for DP4. + + +1.2.4 LERP - Linear Interpolate + + dst.x = src0.x * (src1.x - src2.x) + src2.x + dst.y = src0.y * (src1.y - src2.y) + src2.y + dst.z = src0.z * (src1.z - src2.z) + src2.z + dst.w = src0.w * (src1.w - src2.w) + src2.w + + +1.2.5 CND - Condition + + dst.x = (src2.x > 0.5) ? src0.x : src1.x + dst.y = (src2.y > 0.5) ? src0.y : src1.y + dst.z = (src2.z > 0.5) ? src0.z : src1.z + dst.w = (src2.w > 0.5) ? src0.w : src1.w + + +1.2.6 CND0 - Condition Zero + + dst.x = (src2.x >= 0.0) ? src0.x : src1.x + dst.y = (src2.y >= 0.0) ? src0.y : src1.y + dst.z = (src2.z >= 0.0) ? src0.z : src1.z + dst.w = (src2.w >= 0.0) ? src0.w : src1.w + + +1.2.7 DOT2ADD - 2-component Dot Product And Add + + dst.x = src0.x * src1.x + src0.y * src1.y + src2.x + dst.y = src0.x * src1.x + src0.y * src1.y + src2.x + dst.z = src0.x * src1.x + src0.y * src1.y + src2.x + dst.w = src0.x * src1.x + src0.y * src1.y + src2.x + + +1.3 GL_EXT_vertex_shader +------------------------- + + +1.3.1 INDEX - Array Lookup + + Considered for removal from language. + + +1.3.2 NEGATE - Negate + + Considered for removal from language. + + +1.3.3 MADD - Multiply And Add + + Alias for MAD. + + +1.3.4 FRAC - Fraction + + dst.x = src.x - floor(src.x) + dst.y = src.y - floor(src.y) + dst.z = src.z - floor(src.z) + dst.w = src.w - floor(src.w) + + +1.3.5 SETGE - Set On Greater Equal + + Alias for SGE. + + +1.3.6 SETLT - Set On Less Than + + Alias for SLT. + + +1.3.7 CLAMP - Clamp + + dst.x = clamp(src0.x, src1.x, src2.x) + dst.y = clamp(src0.y, src1.y, src2.y) + dst.z = clamp(src0.z, src1.z, src2.z) + dst.w = clamp(src0.w, src1.w, src2.w) + + +1.3.8 FLOOR - Floor + + dst.x = floor(src.x) + dst.y = floor(src.y) + dst.z = floor(src.z) + dst.w = floor(src.w) + + +1.3.9 ROUND - Round + + dst.x = round(src.x) + dst.y = round(src.y) + dst.z = round(src.z) + dst.w = round(src.w) + + +1.3.10 EXPBASE2 - Exponent Base 2 + + dst.x = pow(2.0, src.x) + dst.y = pow(2.0, src.x) + dst.z = pow(2.0, src.x) + dst.w = pow(2.0, src.x) + + +1.3.11 LOGBASE2 - Logarithm Base 2 + + dst.x = lg2(src.x) + dst.y = lg2(src.x) + dst.z = lg2(src.x) + dst.w = lg2(src.x) + + +1.3.12 POWER - Power + + dst.x = pow(src0.x, src1.x) + dst.y = pow(src0.x, src1.x) + dst.z = pow(src0.x, src1.x) + dst.w = pow(src0.x, src1.x) + + +1.3.13 RECIP - Reciprocal + + Alias for RCP. + + +1.3.14 RECIPSQRT - Reciprocal Square Root + + Alias for RSQ. + + +1.3.15 CROSSPRODUCT - Cross Product + + dst.x = src0.y * src1.z - src1.y * src0.z + dst.y = src0.z * src1.x - src1.z * src0.x + dst.z = src0.x * src1.y - src1.x * src0.y + dst.w = 1.0 + + +1.3.16 MULTIPLYMATRIX - Multiply Matrix + + Considered for removal from language. + + +1.4 GL_NV_vertex_program1_1 +---------------------------- + + +1.4.1 ABS - Absolute + + dst.x = abs(src.x) + dst.y = abs(src.y) + dst.z = abs(src.z) + dst.w = abs(src.w) + + +1.4.2 RCC - Reciprocal Clamped + + dst.x = (1.0 / src.x) > 0.0 ? clamp(1.0 / src.x, 5.42101e-020, 1.884467e+019) : clamp(1.0 / src.x, -1.884467e+019, -5.42101e-020) + dst.y = (1.0 / src.x) > 0.0 ? clamp(1.0 / src.x, 5.42101e-020, 1.884467e+019) : clamp(1.0 / src.x, -1.884467e+019, -5.42101e-020) + dst.z = (1.0 / src.x) > 0.0 ? clamp(1.0 / src.x, 5.42101e-020, 1.884467e+019) : clamp(1.0 / src.x, -1.884467e+019, -5.42101e-020) + dst.w = (1.0 / src.x) > 0.0 ? clamp(1.0 / src.x, 5.42101e-020, 1.884467e+019) : clamp(1.0 / src.x, -1.884467e+019, -5.42101e-020) + + +1.4.3 DPH - Homogeneous Dot Product + + dst.x = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w + dst.y = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w + dst.z = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w + dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w + + +1.5 GL_NV_fragment_program +--------------------------- + + +1.5.1 COS - Cosine + + dst.x = cos(src.x) + dst.y = cos(src.x) + dst.z = cos(src.x) + dst.w = cos(src.w) + + +1.5.2 DDX - Derivative Relative To X + + dst.x = partialx(src.x) + dst.y = partialx(src.y) + dst.z = partialx(src.z) + dst.w = partialx(src.w) + + +1.5.3 DDY - Derivative Relative To Y + + dst.x = partialy(src.x) + dst.y = partialy(src.y) + dst.z = partialy(src.z) + dst.w = partialy(src.w) + + +1.5.4 EX2 - Exponential Base 2 + + Alias for EXPBASE2. + + +1.5.5 FLR - Floor + + Alias for FLOOR. + + +1.5.6 FRC - Fraction + + Alias for FRAC. + + +1.5.7 KILP - Predicated Discard + + TBD + + +1.5.8 LG2 - Logarithm Base 2 + + Alias for LOGBASE2. + + +1.5.9 LRP - Linear Interpolate + + Alias for LERP. + + +1.5.10 PK2H - Pack Two 16-bit Floats + + TBD + + +1.5.11 PK2US - Pack Two Unsigned 16-bit Scalars + + TBD + + +1.5.12 PK4B - Pack Four Signed 8-bit Scalars + + TBD + + +1.5.13 PK4UB - Pack Four Unsigned 8-bit Scalars + + TBD + + +1.5.14 POW - Power + + Alias for POWER. + + +1.5.15 RFL - Reflection Vector + + dst.x = 2.0 * (src0.x * src1.x + src0.y * src1.y + src0.z * src1.z) / (src0.x * src0.x + src0.y * src0.y + src0.z * src0.z) * src0.x - src1.x + dst.y = 2.0 * (src0.x * src1.x + src0.y * src1.y + src0.z * src1.z) / (src0.x * src0.x + src0.y * src0.y + src0.z * src0.z) * src0.y - src1.y + dst.z = 2.0 * (src0.x * src1.x + src0.y * src1.y + src0.z * src1.z) / (src0.x * src0.x + src0.y * src0.y + src0.z * src0.z) * src0.z - src1.z + dst.w = 1.0 + + +1.5.16 SEQ - Set On Equal + + dst.x = (src0.x == src1.x) ? 1.0 : 0.0 + dst.y = (src0.y == src1.y) ? 1.0 : 0.0 + dst.z = (src0.z == src1.z) ? 1.0 : 0.0 + dst.w = (src0.w == src1.w) ? 1.0 : 0.0 + + +1.5.17 SFL - Set On False + + dst.x = 0.0 + dst.y = 0.0 + dst.z = 0.0 + dst.w = 0.0 + + +1.5.18 SGT - Set On Greater Than + + dst.x = (src0.x > src1.x) ? 1.0 : 0.0 + dst.y = (src0.y > src1.y) ? 1.0 : 0.0 + dst.z = (src0.z > src1.z) ? 1.0 : 0.0 + dst.w = (src0.w > src1.w) ? 1.0 : 0.0 + + +1.5.19 SIN - Sine + + dst.x = sin(src.x) + dst.y = sin(src.x) + dst.z = sin(src.x) + dst.w = sin(src.w) + + +1.5.20 SLE - Set On Less Equal Than + + dst.x = (src0.x <= src1.x) ? 1.0 : 0.0 + dst.y = (src0.y <= src1.y) ? 1.0 : 0.0 + dst.z = (src0.z <= src1.z) ? 1.0 : 0.0 + dst.w = (src0.w <= src1.w) ? 1.0 : 0.0 + + +1.5.21 SNE - Set On Not Equal + + dst.x = (src0.x != src1.x) ? 1.0 : 0.0 + dst.y = (src0.y != src1.y) ? 1.0 : 0.0 + dst.z = (src0.z != src1.z) ? 1.0 : 0.0 + dst.w = (src0.w != src1.w) ? 1.0 : 0.0 + + +1.5.22 STR - Set On True + + dst.x = 1.0 + dst.y = 1.0 + dst.z = 1.0 + dst.w = 1.0 + + +1.5.23 TEX - Texture Lookup + + TBD + + +1.5.24 TXD - Texture Lookup with Derivatives + + TBD + + +1.5.25 TXP - Projective Texture Lookup + + TBD + + +1.5.26 UP2H - Unpack Two 16-Bit Floats + + TBD + + +1.5.27 UP2US - Unpack Two Unsigned 16-Bit Scalars + + TBD + + +1.5.28 UP4B - Unpack Four Signed 8-Bit Values + + TBD + + +1.5.29 UP4UB - Unpack Four Unsigned 8-Bit Scalars + + TBD + + +1.5.30 X2D - 2D Coordinate Transformation + + dst.x = src0.x + src1.x * src2.x + src1.y * src2.y + dst.y = src0.y + src1.x * src2.z + src1.y * src2.w + dst.z = src0.x + src1.x * src2.x + src1.y * src2.y + dst.w = src0.y + src1.x * src2.z + src1.y * src2.w + + +1.6 GL_NV_vertex_program2 +-------------------------- + + +1.6.1 ARA - Address Register Add + + TBD + + +1.6.2 ARR - Address Register Load With Round + + dst.x = round(src.x) + dst.y = round(src.y) + dst.z = round(src.z) + dst.w = round(src.w) + + +1.6.3 BRA - Branch + + TBD + + +1.6.4 CAL - Subroutine Call + + push(pc) + pc = target + + +1.6.5 RET - Subroutine Call Return + + pc = pop() + + +1.6.6 SSG - Set Sign + + dst.x = (src.x > 0.0) ? 1.0 : (src.x < 0.0) ? -1.0 : 0.0 + dst.y = (src.y > 0.0) ? 1.0 : (src.y < 0.0) ? -1.0 : 0.0 + dst.z = (src.z > 0.0) ? 1.0 : (src.z < 0.0) ? -1.0 : 0.0 + dst.w = (src.w > 0.0) ? 1.0 : (src.w < 0.0) ? -1.0 : 0.0 + + +1.7 GL_ARB_vertex_program +-------------------------- + + +1.7.1 SWZ - Extended Swizzle + + dst.x = src.x + dst.y = src.y + dst.z = src.z + dst.w = src.w + + +1.7.2 XPD - Cross Product + + Alias for CROSSPRODUCT. + + +1.8 GL_ARB_fragment_program +---------------------------- + + +1.8.1 CMP - Compare + + dst.x = (src0.x < 0.0) ? src1.x : src2.x + dst.y = (src0.y < 0.0) ? src1.y : src2.y + dst.z = (src0.z < 0.0) ? src1.z : src2.z + dst.w = (src0.w < 0.0) ? src1.w : src2.w + + +1.8.2 KIL - Conditional Discard + + TBD + + +1.8.3 SCS - Sine Cosine + + dst.x = cos(src.x) + dst.y = sin(src.x) + dst.z = 0.0 + dst.y = 1.0 + + +1.8.4 TXB - Texture Lookup With Bias + + TBD + + +1.9 GL_NV_fragment_program2 +---------------------------- + + +1.9.1 NRM - 3-component Vector Normalise + + dst.x = src.x / (src.x * src.x + src.y * src.y + src.z * src.z) + dst.y = src.y / (src.x * src.x + src.y * src.y + src.z * src.z) + dst.z = src.z / (src.x * src.x + src.y * src.y + src.z * src.z) + dst.w = 1.0 + + +1.9.2 DIV - Divide + + dst.x = src0.x / src1.x + dst.y = src0.y / src1.y + dst.z = src0.z / src1.z + dst.w = src0.w / src1.w + + +1.9.3 DP2 - 2-component Dot Product + + dst.x = src0.x * src1.x + src0.y * src1.y + dst.y = src0.x * src1.x + src0.y * src1.y + dst.z = src0.x * src1.x + src0.y * src1.y + dst.w = src0.x * src1.x + src0.y * src1.y + + +1.9.4 DP2A - 2-component Dot Product And Add + + Alias for DOT2ADD. + + +1.9.5 TXL - Texture Lookup With LOD + + TBD + + +1.9.6 BRK - Break + + TBD + + +1.9.7 IF - If + + TBD + + +1.9.8 LOOP - Loop + + TBD + + +1.9.9 REP - Repeat + + TBD + + +1.9.10 ELSE - Else + + TBD + + +1.9.11 ENDIF - End If + + TBD + + +1.9.12 ENDLOOP - End Loop + + TBD + + +1.9.13 ENDREP - End Repeat + + TBD + + +1.10 GL_NV_vertex_program3 +--------------------------- + + +1.10.1 PUSHA - Push Address Register On Stack + + push(src.x) + push(src.y) + push(src.z) + push(src.w) + + +1.10.2 POPA - Pop Address Register From Stack + + dst.w = pop() + dst.z = pop() + dst.y = pop() + dst.x = pop() + + +1.11 GL_NV_gpu_program4 +------------------------ + + +1.11.1 CEIL - Ceiling + + dst.x = ceil(src.x) + dst.y = ceil(src.y) + dst.z = ceil(src.z) + dst.w = ceil(src.w) + + +1.11.2 I2F - Integer To Float + + dst.x = (float) src.x + dst.y = (float) src.y + dst.z = (float) src.z + dst.w = (float) src.w + + +1.11.3 NOT - Bitwise Not + + dst.x = ~src.x + dst.y = ~src.y + dst.z = ~src.z + dst.w = ~src.w + + +1.11.4 TRUNC - Truncate + + dst.x = trunc(src.x) + dst.y = trunc(src.y) + dst.z = trunc(src.z) + dst.w = trunc(src.w) + + +1.11.5 SHL - Shift Left + + dst.x = src0.x << src1.x + dst.y = src0.y << src1.x + dst.z = src0.z << src1.x + dst.w = src0.w << src1.x + + +1.11.6 SHR - Shift Right + + dst.x = src0.x >> src1.x + dst.y = src0.y >> src1.x + dst.z = src0.z >> src1.x + dst.w = src0.w >> src1.x + + +1.11.7 AND - Bitwise And + + dst.x = src0.x & src1.x + dst.y = src0.y & src1.y + dst.z = src0.z & src1.z + dst.w = src0.w & src1.w + + +1.11.8 OR - Bitwise Or + + dst.x = src0.x | src1.x + dst.y = src0.y | src1.y + dst.z = src0.z | src1.z + dst.w = src0.w | src1.w + + +1.11.9 MOD - Modulus + + dst.x = src0.x % src1.x + dst.y = src0.y % src1.y + dst.z = src0.z % src1.z + dst.w = src0.w % src1.w + + +1.11.10 XOR - Bitwise Xor + + dst.x = src0.x ^ src1.x + dst.y = src0.y ^ src1.y + dst.z = src0.z ^ src1.z + dst.w = src0.w ^ src1.w + + +1.11.11 SAD - Sum Of Absolute Differences + + dst.x = abs(src0.x - src1.x) + src2.x + dst.y = abs(src0.y - src1.y) + src2.y + dst.z = abs(src0.z - src1.z) + src2.z + dst.w = abs(src0.w - src1.w) + src2.w + + +1.11.12 TXF - Texel Fetch + + TBD + + +1.11.13 TXQ - Texture Size Query + + TBD + + +1.11.14 CONT - Continue + + TBD + + +1.12 GL_NV_geometry_program4 +----------------------------- + + +1.12.1 EMIT - Emit + + TBD + + +1.12.2 ENDPRIM - End Primitive + + TBD + + +1.13 GLSL +---------- + + +1.13.1 BGNLOOP2 - Begin Loop + + TBD + + +1.13.2 BGNSUB - Begin Subroutine + + TBD + + +1.13.3 ENDLOOP2 - End Loop + + TBD + + +1.13.4 ENDSUB - End Subroutine + + TBD + + +1.13.5 INT - Truncate + + Alias for TRUNC. + + +1.13.6 NOISE1 - 1D Noise + + TBD + + +1.13.7 NOISE2 - 2D Noise + + TBD + + +1.13.8 NOISE3 - 3D Noise + + TBD + + +1.13.9 NOISE4 - 4D Noise + + TBD + + +1.13.10 NOP - No Operation + + Do nothing. + + +1.14 ps_1_1 +------------ + + +1.14.1 TEXKILL - Conditional Discard + + Alias for KIL. + + +1.15 ps_1_4 +------------ + + +1.15.1 TEXLD - Texture Lookup + + Alias for TEX. + + +1.16 ps_2_0 +------------ + + +1.16.1 M4X4 - Multiply Matrix + + Alias for MULTIPLYMATRIX. + + +1.16.2 M4X3 - Multiply Matrix + + Considered for removal from language. + + +1.16.3 M3X4 - Multiply Matrix + + Considered for removal from language. + + +1.16.4 M3X3 - Multiply Matrix + + Considered for removal from language. + + +1.16.5 M3X2 - Multiply Matrix + + Considered for removal from language. + + +1.16.6 CRS - Cross Product + + Alias for XPD. + + +1.16.7 NRM4 - 4-component Vector Normalise + + dst.x = src.x / (src.x * src.x + src.y * src.y + src.z * src.z + src.w * src.w) + dst.y = src.y / (src.x * src.x + src.y * src.y + src.z * src.z + src.w * src.w) + dst.z = src.z / (src.x * src.x + src.y * src.y + src.z * src.z + src.w * src.w) + dst.w = src.w / (src.x * src.x + src.y * src.y + src.z * src.z + src.w * src.w) + + +1.16.8 SINCOS - Sine Cosine + + Alias for SCS. + + +1.16.9 TEXLDB - Texture Lookup With Bias + + Alias for TXB. + + +1.16.10 DP2ADD - 2-component Dot Product And Add + + Alias for DP2A. + + +1.17 ps_2_x +------------ + + +1.17.1 CALL - Subroutine Call + + Alias for CAL. + + +1.17.2 CALLNZ - Subroutine Call If Not Zero + + TBD + + +1.17.3 IFC - If + + TBD + + +1.17.4 BREAK - Break + + Alias for BRK. + + +1.17.5 BREAKC - Break Conditional + + TBD + + +1.17.6 DSX - Derivative Relative To X + + Alias for DDX. + + +1.17.7 DSY - Derivative Relative To Y + + Alias for DDY. + + +1.17.8 TEXLDD - Texture Lookup with Derivatives + + Alias for TXD. + + +1.18 vs_1_1 +------------ + + +1.18.1 EXPP - Approximate Exponential Base 2 + + Alias for EXP. + + +1.18.2 LOGP - Logarithm Base 2 + + Alias for LG2. + + +1.19 vs_2_0 +------------ + + +1.19.1 SGN - Set Sign + + Alias for SSG. + + +1.19.2 MOVA - Move Address Register + + Alias for ARR. + diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index 17886540cf..a1891a140a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_shader_tokens.h" #include "tgsi_build.h" #include "tgsi_parse.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index ab2b1f2c58..a784b7cc3c 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_string.h" #include "tgsi_dump.h" #include "tgsi_info.h" @@ -245,8 +245,12 @@ iter_declaration( } } - TXT( ", " ); - ENM( decl->Declaration.Interpolate, interpolate_names ); + if (iter->processor.Processor == TGSI_PROCESSOR_FRAGMENT && + decl->Declaration.File == TGSI_FILE_INPUT) + { + TXT( ", " ); + ENM( decl->Declaration.Interpolate, interpolate_names ); + } if (decl->Declaration.Centroid) { TXT( ", CENTROID" ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c index 2ecf1e2f14..3dc61c48ca 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_string.h" #include "tgsi_dump_c.h" #include "tgsi_build.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index ab641efb60..ba807e498f 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -486,6 +486,32 @@ micro_f2ut( #endif static void +micro_float_clamp(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + uint i; + + for (i = 0; i < 4; i++) { + if (src->f[i] > 0.0f) { + if (src->f[i] > 1.884467e+019f) + dst->f[i] = 1.884467e+019f; + else if (src->f[i] < 5.42101e-020f) + dst->f[i] = 5.42101e-020f; + else + dst->f[i] = src->f[i]; + } + else { + if (src->f[i] < -1.884467e+019f) + dst->f[i] = -1.884467e+019f; + else if (src->f[i] > -5.42101e-020f) + dst->f[i] = -5.42101e-020f; + else + dst->f[i] = src->f[i]; + } + } +} + +static void micro_flr( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) @@ -508,20 +534,6 @@ micro_frc( } static void -micro_ge( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void micro_i2f( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) @@ -1614,6 +1626,7 @@ exec_tex(struct tgsi_exec_machine *mach, switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: + case TGSI_TEXTURE_SHADOW1D: FETCH(&r[0], 0, CHAN_X); @@ -1636,6 +1649,8 @@ exec_tex(struct tgsi_exec_machine *mach, case TGSI_TEXTURE_2D: case TGSI_TEXTURE_RECT: + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 0, CHAN_Y); @@ -1826,12 +1841,14 @@ exec_instruction( int *pc ) { uint chan_index; - union tgsi_exec_channel r[8]; + union tgsi_exec_channel r[10]; (*pc)++; switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: + /* TGSI_OPCODE_FLOOR */ + /* TGSI_OPCODE_FLR */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); micro_flr( &r[0], &r[0] ); @@ -1849,31 +1866,31 @@ exec_instruction( case TGSI_OPCODE_LIT: if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Y ); - } + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - FETCH( &r[2], 0, CHAN_W ); - micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); - micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); - micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); - } + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } break; @@ -1882,17 +1899,18 @@ exec_instruction( FETCH( &r[0], 0, CHAN_X ); micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; case TGSI_OPCODE_RSQ: /* TGSI_OPCODE_RECIPSQRT */ FETCH( &r[0], 0, CHAN_X ); + micro_abs( &r[0], &r[0] ); micro_sqrt( &r[0], &r[0] ); micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; @@ -2005,30 +2023,30 @@ exec_instruction( micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; case TGSI_OPCODE_DST: if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - FETCH( &r[0], 0, CHAN_Y ); - FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, CHAN_Y ); + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); } break; @@ -2071,7 +2089,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + micro_le( &r[0], &r[1], &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); STORE( &r[0], 0, chan_index ); } break; @@ -2115,15 +2133,27 @@ exec_instruction( break; case TGSI_OPCODE_CND: - assert (0); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + micro_lt(&r[0], &mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C], &r[2], &r[0], &r[1]); + STORE(&r[0], 0, chan_index); + } break; case TGSI_OPCODE_CND0: - assert (0); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + micro_le(&r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[2], &r[0], &r[1]); + STORE(&r[0], 0, chan_index); + } break; case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ + /* TGSI_OPCODE_DP2A */ FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); micro_mul( &r[0], &r[0], &r[1] ); @@ -2142,10 +2172,12 @@ exec_instruction( break; case TGSI_OPCODE_INDEX: + /* XXX: considered for removal */ assert (0); break; case TGSI_OPCODE_NEGATE: + /* XXX: considered for removal */ assert (0); break; @@ -2159,15 +2191,13 @@ exec_instruction( break; case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + micro_max(&r[0], &r[0], &r[1]); + FETCH(&r[1], 2, chan_index); + micro_min(&r[0], &r[0], &r[1]); + STORE(&r[0], 0, chan_index); } break; @@ -2181,7 +2211,7 @@ exec_instruction( break; case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ + /* TGSI_OPCODE_EX2 */ FETCH(&r[0], 0, CHAN_X); #if FAST_MATH @@ -2191,7 +2221,7 @@ exec_instruction( #endif FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; @@ -2205,19 +2235,19 @@ exec_instruction( break; case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ + /* TGSI_OPCODE_POW */ FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); micro_pow( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ + /* TGSI_OPCODE_XPD */ FETCH(&r[0], 0, CHAN_Y); FETCH(&r[1], 1, CHAN_Z); @@ -2260,6 +2290,7 @@ exec_instruction( break; case TGSI_OPCODE_MULTIPLYMATRIX: + /* XXX: considered for removal */ assert (0); break; @@ -2274,7 +2305,12 @@ exec_instruction( break; case TGSI_OPCODE_RCC: - assert (0); + FETCH(&r[0], 0, CHAN_X); + micro_div(&r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0]); + micro_float_clamp(&r[0], &r[0]); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&r[0], 0, chan_index); + } break; case TGSI_OPCODE_DPH: @@ -2300,7 +2336,7 @@ exec_instruction( micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; @@ -2310,7 +2346,7 @@ exec_instruction( micro_cos( &r[0], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; @@ -2355,7 +2391,52 @@ exec_instruction( break; case TGSI_OPCODE_RFL: - assert (0); + if (IS_CHANNEL_ENABLED(*inst, CHAN_X) || + IS_CHANNEL_ENABLED(*inst, CHAN_Y) || + IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + /* r0 = dp3(src0, src0) */ + FETCH(&r[2], 0, CHAN_X); + micro_mul(&r[0], &r[2], &r[2]); + FETCH(&r[4], 0, CHAN_Y); + micro_mul(&r[8], &r[4], &r[4]); + micro_add(&r[0], &r[0], &r[8]); + FETCH(&r[6], 0, CHAN_Z); + micro_mul(&r[8], &r[6], &r[6]); + micro_add(&r[0], &r[0], &r[8]); + + /* r1 = dp3(src0, src1) */ + FETCH(&r[3], 1, CHAN_X); + micro_mul(&r[1], &r[2], &r[3]); + FETCH(&r[5], 1, CHAN_Y); + micro_mul(&r[8], &r[4], &r[5]); + micro_add(&r[1], &r[1], &r[8]); + FETCH(&r[7], 1, CHAN_Z); + micro_mul(&r[8], &r[6], &r[7]); + micro_add(&r[1], &r[1], &r[8]); + + /* r1 = 2 * r1 / r0 */ + micro_add(&r[1], &r[1], &r[1]); + micro_div(&r[1], &r[1], &r[0]); + + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + micro_mul(&r[2], &r[2], &r[1]); + micro_sub(&r[2], &r[2], &r[3]); + STORE(&r[2], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + micro_mul(&r[4], &r[4], &r[1]); + micro_sub(&r[4], &r[4], &r[5]); + STORE(&r[4], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + micro_mul(&r[6], &r[6], &r[1]); + micro_sub(&r[6], &r[6], &r[7]); + STORE(&r[6], 0, CHAN_Z); + } + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) { + STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W); + } break; case TGSI_OPCODE_SEQ: @@ -2370,7 +2451,9 @@ exec_instruction( break; case TGSI_OPCODE_SFL: - assert (0); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, chan_index); + } break; case TGSI_OPCODE_SGT: @@ -2409,7 +2492,9 @@ exec_instruction( break; case TGSI_OPCODE_STR: - assert (0); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, chan_index); + } break; case TGSI_OPCODE_TEX: @@ -2466,7 +2551,40 @@ exec_instruction( break; case TGSI_OPCODE_X2D: - assert (0); + FETCH(&r[0], 1, CHAN_X); + FETCH(&r[1], 1, CHAN_Y); + if (IS_CHANNEL_ENABLED(*inst, CHAN_X) || + IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + FETCH(&r[2], 2, CHAN_X); + micro_mul(&r[2], &r[2], &r[0]); + FETCH(&r[3], 2, CHAN_Y); + micro_mul(&r[3], &r[3], &r[1]); + micro_add(&r[2], &r[2], &r[3]); + FETCH(&r[3], 0, CHAN_X); + micro_add(&r[2], &r[2], &r[3]); + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + STORE(&r[2], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + STORE(&r[2], 0, CHAN_Z); + } + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y) || + IS_CHANNEL_ENABLED(*inst, CHAN_W)) { + FETCH(&r[2], 2, CHAN_Z); + micro_mul(&r[2], &r[2], &r[0]); + FETCH(&r[3], 2, CHAN_W); + micro_mul(&r[3], &r[3], &r[1]); + micro_add(&r[2], &r[2], &r[3]); + FETCH(&r[3], 0, CHAN_Y); + micro_add(&r[2], &r[2], &r[3]); + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + STORE(&r[2], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) { + STORE(&r[2], 0, CHAN_W); + } + } break; case TGSI_OPCODE_ARA: @@ -2551,14 +2669,14 @@ exec_instruction( case TGSI_OPCODE_SCS: if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { FETCH( &r[0], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - micro_cos( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - micro_sin( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_Y ); + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + micro_cos(&r[1], &r[0]); + STORE(&r[1], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + micro_sin(&r[1], &r[0]); + STORE(&r[1], 0, CHAN_Y); + } } if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); @@ -2570,32 +2688,36 @@ exec_instruction( case TGSI_OPCODE_NRM: /* 3-component vector normalize */ - { - union tgsi_exec_channel tmp, dot; - - /* tmp = dp3(src0, src0): */ - FETCH( &r[0], 0, CHAN_X ); - micro_mul( &tmp, &r[0], &r[0] ); - - FETCH( &r[1], 0, CHAN_Y ); - micro_mul( &dot, &r[1], &r[1] ); - micro_add( &tmp, &tmp, &dot ); - - FETCH( &r[2], 0, CHAN_Z ); - micro_mul( &dot, &r[2], &r[2] ); - micro_add( &tmp, &tmp, &dot ); - - /* tmp = 1 / sqrt(tmp) */ - micro_sqrt( &tmp, &tmp ); - micro_div( &tmp, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &tmp ); - - /* note: w channel is undefined */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - /* chan = chan * tmp */ - micro_mul( &r[chan_index], &tmp, &r[chan_index] ); - STORE( &r[chan_index], 0, chan_index ); + if(IS_CHANNEL_ENABLED(*inst, CHAN_X) || + IS_CHANNEL_ENABLED(*inst, CHAN_Y) || + IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + /* r3 = sqrt(dp3(src0, src0)) */ + FETCH(&r[0], 0, CHAN_X); + micro_mul(&r[3], &r[0], &r[0]); + FETCH(&r[1], 0, CHAN_Y); + micro_mul(&r[4], &r[1], &r[1]); + micro_add(&r[3], &r[3], &r[4]); + FETCH(&r[2], 0, CHAN_Z); + micro_mul(&r[4], &r[2], &r[2]); + micro_add(&r[3], &r[3], &r[4]); + micro_sqrt(&r[3], &r[3]); + + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + micro_div(&r[0], &r[0], &r[3]); + STORE(&r[0], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + micro_div(&r[1], &r[1], &r[3]); + STORE(&r[1], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + micro_div(&r[2], &r[2], &r[3]); + STORE(&r[2], 0, CHAN_Z); } } + if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) { + STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W); + } break; case TGSI_OPCODE_NRM4: @@ -2956,5 +3078,3 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; } - - diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c index 68c7a6b7f5..2b8a6f0fb1 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.c +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi_info.h" static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/tgsi_iterate.c index 5371a88b96..d88c2558d8 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_iterate.c +++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi_iterate.h" boolean diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c index d374b16f9a..0081f74ffc 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_shader_tokens.h" #include "tgsi_parse.h" #include "tgsi_build.h" @@ -154,10 +154,17 @@ tgsi_parse_token( switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: - imm->u.Pointer = MALLOC( - sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.NrTokens - 1) ); - for( i = 0; i < imm->Immediate.NrTokens - 1; i++ ) { - next_token( ctx, (struct tgsi_immediate_float32 *) &imm->u.ImmediateFloat32[i] ); + { + uint imm_count = imm->Immediate.NrTokens - 1; + struct tgsi_immediate_float32 *data; + + data = (struct tgsi_immediate_float32 *) MALLOC(sizeof(struct tgsi_immediate_float32) * imm_count); + if (data) { + for (i = 0; i < imm_count; i++) { + next_token(ctx, &data[i]); + } + imm->u.ImmediateFloat32 = data; + } } break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c index f365030e52..0c64ae5713 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c @@ -33,7 +33,7 @@ #if defined(PIPE_ARCH_PPC) -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_shader_tokens.h" #include "util/u_math.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index bc7b941b78..6f1f5c2b4b 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -25,23 +25,24 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi_sanity.h" #include "tgsi_info.h" #include "tgsi_iterate.h" -#define MAX_REGISTERS 256 - typedef uint reg_flag; #define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8) +#define MAX_REGISTERS 256 +#define MAX_REG_FLAGS ((MAX_REGISTERS + BITS_IN_REG_FLAG - 1) / BITS_IN_REG_FLAG) + struct sanity_check_ctx { struct tgsi_iterate_context iter; - reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; - reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; + reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REG_FLAGS]; + reg_flag regs_used[TGSI_FILE_COUNT][MAX_REG_FLAGS]; boolean regs_ind_used[TGSI_FILE_COUNT]; uint num_imms; uint num_instructions; @@ -89,7 +90,7 @@ check_file_name( uint file ) { if (file <= TGSI_FILE_NULL || file >= TGSI_FILE_COUNT) { - report_error( ctx, "Invalid register file name" ); + report_error( ctx, "(%u): Invalid register file name", file ); return FALSE; } return TRUE; @@ -113,7 +114,7 @@ is_any_register_declared( { uint i; - for (i = 0; i < MAX_REGISTERS / BITS_IN_REG_FLAG; i++) + for (i = 0; i < MAX_REG_FLAGS; i++) if (ctx->regs_decl[file][i]) return TRUE; return FALSE; @@ -162,9 +163,8 @@ check_register_usage( ctx->regs_ind_used[file] = TRUE; } else { - if (index < 0 || index > MAX_REGISTERS) { - report_error( ctx, "%s[%i]: Invalid index %s", - file_names[file], index, name ); + if (index < 0 || index >= MAX_REGISTERS) { + report_error( ctx, "%s[%d]: Invalid %s index", file_names[file], index, name ); return FALSE; } @@ -193,15 +193,15 @@ iter_instruction( info = tgsi_get_opcode_info( inst->Instruction.Opcode ); if (info == NULL) { - report_error( ctx, "Invalid instruction opcode" ); + report_error( ctx, "(%u): Invalid instruction opcode", inst->Instruction.Opcode ); return TRUE; } if (info->num_dst != inst->Instruction.NumDstRegs) { - report_error( ctx, "Invalid number of destination operands" ); + report_error( ctx, "Invalid number of destination operands, should be %u", info->num_dst ); } if (info->num_src != inst->Instruction.NumSrcRegs) { - report_error( ctx, "Invalid number of source operands" ); + report_error( ctx, "Invalid number of source operands, should be %u", info->num_src ); } /* Check destination and source registers' validity. @@ -266,7 +266,7 @@ iter_declaration( return TRUE; for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) { if (is_register_declared( ctx, file, i )) - report_error( ctx, "The same register declared twice" ); + report_error( ctx, "%s[%u]: The same register declared more than once", file_names[file], i ); ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG)); } @@ -295,7 +295,7 @@ iter_immediate( /* Check data type validity. */ if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) { - report_error( ctx, "Invalid immediate data type" ); + report_error( ctx, "(%u): Invalid immediate data type", imm->Immediate.DataType ); return TRUE; } @@ -322,7 +322,7 @@ epilog( for (i = 0; i < MAX_REGISTERS; i++) { if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) { - report_warning( ctx, "Register never used" ); + report_warning( ctx, "%s[%u]: Register never used", file_names[file], i ); } } } diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c index d02205a63e..c535788819 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c @@ -151,7 +151,14 @@ tgsi_scan_shader(const struct tgsi_token *tokens, break; case TGSI_TOKEN_TYPE_IMMEDIATE: - info->immediate_count++; + { + uint reg = info->immediate_count++; + uint file = TGSI_FILE_IMMEDIATE; + + info->file_mask[file] |= (1 << reg); + info->file_count[file]++; + info->file_max[file] = MAX2(info->file_max[file], (int)reg); + } break; default: diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 481ba89c5e..d70bcd03c5 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -29,7 +29,7 @@ #if defined(PIPE_ARCH_X86) -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_shader_tokens.h" #include "util/u_math.h" #if defined(PIPE_ARCH_SSE) @@ -1575,6 +1575,7 @@ emit_instruction( case TGSI_OPCODE_RSQ: /* TGSI_OPCODE_RECIPSQRT */ FETCH( func, *inst, 0, 0, CHAN_X ); + emit_abs( func, 0 ); emit_rsqrt( func, 1, 0 ); FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { STORE( func, *inst, 1, 0, chan_index ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c index 1e822fbbea..58fe07c11d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi_text.h" #include "tgsi_build.h" #include "tgsi_info.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c index ea87da31e5..062c1be938 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_transform.c +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c @@ -31,7 +31,7 @@ * Authors: Brian Paul */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi_transform.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c index 50101a9bb0..71f8a6ca40 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_util.c +++ b/src/gallium/auxiliary/tgsi/tgsi_util.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_shader_tokens.h" #include "tgsi_parse.h" #include "tgsi_build.h" diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile index ad2a5b705e..3c82f8ae03 100644 --- a/src/gallium/auxiliary/translate/Makefile +++ b/src/gallium/auxiliary/translate/Makefile @@ -10,6 +10,3 @@ C_SOURCES = \ translate_cache.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 44c2377721..160df8dfa7 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -4,7 +4,7 @@ include $(TOP)/configs/current LIBNAME = util C_SOURCES = \ - p_debug.c \ + u_debug.c \ u_blit.c \ u_cache.c \ u_draw_quad.c \ @@ -27,6 +27,3 @@ C_SOURCES = \ u_simple_screen.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 35f683fb8e..9d5dd006f0 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -3,11 +3,13 @@ Import('*') util = env.ConvenienceLibrary( target = 'util', source = [ - 'p_debug.c', - 'p_debug_mem.c', - 'p_debug_prof.c', + 'u_bitmask.c', 'u_blit.c', 'u_cache.c', + 'u_debug.c', + 'u_debug_memory.c', + 'u_debug_profile.c', + 'u_debug_stack.c', 'u_draw_quad.c', 'u_gen_mipmap.c', 'u_handle_table.c', diff --git a/src/gallium/auxiliary/util/u_bitmask.c b/src/gallium/auxiliary/util/u_bitmask.c new file mode 100644 index 0000000000..77587c07ec --- /dev/null +++ b/src/gallium/auxiliary/util/u_bitmask.c @@ -0,0 +1,320 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Generic bitmask implementation. + * + * @author Jose Fonseca <jfonseca@vmware.com> + */ + + +#include "pipe/p_compiler.h" +#include "util/u_debug.h" + +#include "util/u_memory.h" +#include "util/u_bitmask.h" + + +typedef uint32_t util_bitmask_word; + + +#define UTIL_BITMASK_INITIAL_WORDS 16 +#define UTIL_BITMASK_BITS_PER_BYTE 8 +#define UTIL_BITMASK_BITS_PER_WORD (sizeof(util_bitmask_word) * UTIL_BITMASK_BITS_PER_BYTE) + + +struct util_bitmask +{ + util_bitmask_word *words; + + /** Number of bits we can currently hold */ + unsigned size; + + /** Number of consecutive bits set at the start of the bitmask */ + unsigned filled; +}; + + +struct util_bitmask * +util_bitmask_create(void) +{ + struct util_bitmask *bm; + + bm = MALLOC_STRUCT(util_bitmask); + if(!bm) + return NULL; + + bm->words = (util_bitmask_word *)CALLOC(UTIL_BITMASK_INITIAL_WORDS, sizeof(util_bitmask_word)); + if(!bm->words) { + FREE(bm); + return NULL; + } + + bm->size = UTIL_BITMASK_INITIAL_WORDS * UTIL_BITMASK_BITS_PER_WORD; + bm->filled = 0; + + return bm; +} + + +/** + * Resize the bitmask if necessary + */ +static INLINE boolean +util_bitmask_resize(struct util_bitmask *bm, + unsigned minimum_index) +{ + unsigned minimum_size = minimum_index + 1; + unsigned new_size; + util_bitmask_word *new_words; + + /* Check integer overflow */ + if(!minimum_size) + return FALSE; + + if(bm->size > minimum_size) + return TRUE; + + assert(bm->size % UTIL_BITMASK_BITS_PER_WORD == 0); + new_size = bm->size; + while(!(new_size > minimum_size)) { + new_size *= 2; + /* Check integer overflow */ + if(new_size < bm->size) + return FALSE; + } + assert(new_size); + assert(new_size % UTIL_BITMASK_BITS_PER_WORD == 0); + + new_words = (util_bitmask_word *)REALLOC((void *)bm->words, + bm->size / UTIL_BITMASK_BITS_PER_BYTE, + new_size / UTIL_BITMASK_BITS_PER_BYTE); + if(!new_words) + return FALSE; + + memset(new_words + bm->size/UTIL_BITMASK_BITS_PER_WORD, + 0, + (new_size - bm->size)/UTIL_BITMASK_BITS_PER_BYTE); + + bm->size = new_size; + bm->words = new_words; + + return TRUE; +} + + +/** + * Lazily update the filled. + */ +static INLINE void +util_bitmask_filled_set(struct util_bitmask *bm, + unsigned index) +{ + assert(bm->filled <= bm->size); + assert(index <= bm->size); + + if(index == bm->filled) { + ++bm->filled; + assert(bm->filled <= bm->size); + } +} + +static INLINE void +util_bitmask_filled_unset(struct util_bitmask *bm, + unsigned index) +{ + assert(bm->filled <= bm->size); + assert(index <= bm->size); + + if(index < bm->filled) + bm->filled = index; +} + + +unsigned +util_bitmask_add(struct util_bitmask *bm) +{ + unsigned word; + unsigned bit; + util_bitmask_word mask; + + assert(bm); + + /* linear search for an empty index */ + word = bm->filled / UTIL_BITMASK_BITS_PER_WORD; + bit = bm->filled % UTIL_BITMASK_BITS_PER_WORD; + mask = 1 << bit; + while(word < bm->size / UTIL_BITMASK_BITS_PER_WORD) { + while(bit < UTIL_BITMASK_BITS_PER_WORD) { + if(!(bm->words[word] & mask)) + goto found; + ++bm->filled; + ++bit; + mask <<= 1; + } + ++word; + bit = 0; + mask = 1; + } +found: + + /* grow the bitmask if necessary */ + if(!util_bitmask_resize(bm, bm->filled)) + return UTIL_BITMASK_INVALID_INDEX; + + assert(!(bm->words[word] & mask)); + bm->words[word] |= mask; + + return bm->filled++; +} + + +unsigned +util_bitmask_set(struct util_bitmask *bm, + unsigned index) +{ + unsigned word = index / UTIL_BITMASK_BITS_PER_WORD; + unsigned bit = index % UTIL_BITMASK_BITS_PER_WORD; + util_bitmask_word mask = 1 << bit; + + assert(bm); + + /* grow the bitmask if necessary */ + if(!util_bitmask_resize(bm, index)) + return UTIL_BITMASK_INVALID_INDEX; + + bm->words[word] |= mask; + + util_bitmask_filled_set(bm, index); + + return index; +} + + +void +util_bitmask_clear(struct util_bitmask *bm, + unsigned index) +{ + unsigned word = index / UTIL_BITMASK_BITS_PER_WORD; + unsigned bit = index % UTIL_BITMASK_BITS_PER_WORD; + util_bitmask_word mask = 1 << bit; + + assert(bm); + + if(index >= bm->size) + return; + + bm->words[word] &= ~mask; + + util_bitmask_filled_unset(bm, index); +} + + +boolean +util_bitmask_get(struct util_bitmask *bm, + unsigned index) +{ + unsigned word = index / UTIL_BITMASK_BITS_PER_WORD; + unsigned bit = index % UTIL_BITMASK_BITS_PER_WORD; + util_bitmask_word mask = 1 << bit; + + assert(bm); + + if(index < bm->filled) { + assert(bm->words[word] & mask); + return TRUE; + } + + if(index > bm->size) + return FALSE; + + if(bm->words[word] & mask) { + util_bitmask_filled_set(bm, index); + return TRUE; + } + else + return FALSE; +} + + +unsigned +util_bitmask_get_next_index(struct util_bitmask *bm, + unsigned index) +{ + unsigned word = index / UTIL_BITMASK_BITS_PER_WORD; + unsigned bit = index % UTIL_BITMASK_BITS_PER_WORD; + util_bitmask_word mask = 1 << bit; + + if(index < bm->filled) { + assert(bm->words[word] & mask); + return index; + } + + if(index >= bm->size) { + return UTIL_BITMASK_INVALID_INDEX; + } + + /* Do a linear search */ + while(word < bm->size / UTIL_BITMASK_BITS_PER_WORD) { + while(bit < UTIL_BITMASK_BITS_PER_WORD) { + if(bm->words[word] & mask) { + if(index == bm->filled) { + ++bm->filled; + assert(bm->filled <= bm->size); + } + return index; + } + ++index; + ++bit; + mask <<= 1; + } + ++word; + bit = 0; + mask = 1; + } + + return UTIL_BITMASK_INVALID_INDEX; +} + + +unsigned +util_bitmask_get_first_index(struct util_bitmask *bm) +{ + return util_bitmask_get_next_index(bm, 0); +} + + +void +util_bitmask_destroy(struct util_bitmask *bm) +{ + assert(bm); + + FREE(bm->words); + FREE(bm); +} + diff --git a/src/gallium/auxiliary/util/u_bitmask.h b/src/gallium/auxiliary/util/u_bitmask.h new file mode 100644 index 0000000000..87f1110296 --- /dev/null +++ b/src/gallium/auxiliary/util/u_bitmask.h @@ -0,0 +1,114 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Generic bitmask. + * + * @author Jose Fonseca <jfonseca@vmware.com> + */ + +#ifndef U_HANDLE_BITMASK_H_ +#define U_HANDLE_BITMASK_H_ + + +#ifdef __cplusplus +extern "C" { +#endif + + +#define UTIL_BITMASK_INVALID_INDEX (~0U) + + +/** + * Abstract data type to represent arbitrary set of bits. + */ +struct util_bitmask; + + +struct util_bitmask * +util_bitmask_create(void); + + +/** + * Search a cleared bit and set it. + * + * It searches for the first cleared bit. + * + * Returns the bit index on success, or UTIL_BITMASK_INVALID_INDEX on out of + * memory growing the bitmask. + */ +unsigned +util_bitmask_add(struct util_bitmask *bm); + +/** + * Set a bit. + * + * Returns the input index on success, or UTIL_BITMASK_INVALID_INDEX on out of + * memory growing the bitmask. + */ +unsigned +util_bitmask_set(struct util_bitmask *bm, + unsigned index); + +void +util_bitmask_clear(struct util_bitmask *bm, + unsigned index); + +boolean +util_bitmask_get(struct util_bitmask *bm, + unsigned index); + + +void +util_bitmask_destroy(struct util_bitmask *bm); + + +/** + * Search for the first set bit. + * + * Returns UTIL_BITMASK_INVALID_INDEX if a set bit cannot be found. + */ +unsigned +util_bitmask_get_first_index(struct util_bitmask *bm); + + +/** + * Search for the first set bit, starting from the giving index. + * + * Returns UTIL_BITMASK_INVALID_INDEX if a set bit cannot be found. + */ +unsigned +util_bitmask_get_next_index(struct util_bitmask *bm, + unsigned index); + + +#ifdef __cplusplus +} +#endif + +#endif /* U_HANDLE_BITMASK_H_ */ diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 841e9c01e7..deb68c43a6 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -34,10 +34,11 @@ #include "pipe/p_context.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_state.h" #include "util/u_blit.h" #include "util/u_draw_quad.h" @@ -59,8 +60,6 @@ struct blit_state struct pipe_sampler_state sampler; struct pipe_viewport_state viewport; - struct pipe_shader_state vert_shader; - struct pipe_shader_state frag_shader; void *vs; void *fs; @@ -103,8 +102,7 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer)); ctx->rasterizer.front_winding = PIPE_WINDING_CW; ctx->rasterizer.cull_mode = PIPE_WINDING_NONE; - ctx->rasterizer.bypass_clipping = 1; - /*ctx->rasterizer.bypass_vs = 1;*/ + ctx->rasterizer.bypass_vs_clip_and_viewport = 1; ctx->rasterizer.gl_rasterization_rules = 1; /* samplers */ @@ -117,28 +115,20 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) ctx->sampler.mag_img_filter = 0; /* set later */ ctx->sampler.normalized_coords = 1; - /* viewport (identity, we setup vertices in wincoords) */ - ctx->viewport.scale[0] = 1.0; - ctx->viewport.scale[1] = 1.0; - ctx->viewport.scale[2] = 1.0; - ctx->viewport.scale[3] = 1.0; - ctx->viewport.translate[0] = 0.0; - ctx->viewport.translate[1] = 0.0; - ctx->viewport.translate[2] = 0.0; - ctx->viewport.translate[3] = 0.0; - - /* vertex shader */ + + /* vertex shader - still required to provide the linkage between + * fragment shader input semantics and vertex_element/buffers. + */ { const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, TGSI_SEMANTIC_GENERIC }; const uint semantic_indexes[] = { 0, 0 }; ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, - semantic_indexes, - &ctx->vert_shader); + semantic_indexes); } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe, &ctx->frag_shader); + ctx->fs = util_make_fragment_tex_shader(pipe); ctx->vbuf = NULL; /* init vertex data that doesn't change */ @@ -163,10 +153,7 @@ util_destroy_blit(struct blit_state *ctx) pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); - FREE((void*) ctx->vert_shader.tokens); - FREE((void*) ctx->frag_shader.tokens); - - pipe_buffer_reference(pipe->screen, &ctx->vbuf, NULL); + pipe_buffer_reference(&ctx->vbuf, NULL); FREE(ctx); } @@ -199,7 +186,6 @@ static unsigned setup_vertex_data(struct blit_state *ctx, float x0, float y0, float x1, float y1, float z) { - void *buf; unsigned offset; ctx->vertices[0][0][0] = x0; @@ -228,12 +214,8 @@ setup_vertex_data(struct blit_state *ctx, offset = get_next_slot( ctx ); - buf = pipe_buffer_map(ctx->pipe->screen, ctx->vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE); - - memcpy((char *)buf + offset, ctx->vertices, sizeof(ctx->vertices)); - - pipe_buffer_unmap(ctx->pipe->screen, ctx->vbuf); + pipe_buffer_write(ctx->pipe->screen, ctx->vbuf, + offset, sizeof(ctx->vertices), ctx->vertices); return offset; } @@ -249,7 +231,6 @@ setup_vertex_data_tex(struct blit_state *ctx, float s0, float t0, float s1, float t1, float z) { - void *buf; unsigned offset; ctx->vertices[0][0][0] = x0; @@ -278,12 +259,8 @@ setup_vertex_data_tex(struct blit_state *ctx, offset = get_next_slot( ctx ); - buf = pipe_buffer_map(ctx->pipe->screen, ctx->vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE); - - memcpy((char *)buf + offset, ctx->vertices, sizeof(ctx->vertices)); - - pipe_buffer_unmap(ctx->pipe->screen, ctx->vbuf); + pipe_buffer_write(ctx->pipe->screen, ctx->vbuf, + offset, sizeof(ctx->vertices), ctx->vertices); return offset; } @@ -337,7 +314,7 @@ util_blit_pixels(struct blit_state *ctx, if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) { /* FIXME: this will most surely fail for overlapping rectangles */ - pipe->surface_copy(pipe, FALSE, + pipe->surface_copy(pipe, dst, dstX0, dstY0, /* dest */ src, srcX0, srcY0, /* src */ srcW, srcH); /* size */ @@ -371,14 +348,14 @@ util_blit_pixels(struct blit_state *ctx, PIPE_BUFFER_USAGE_GPU_WRITE); /* load temp texture */ - pipe->surface_copy(pipe, FALSE, + pipe->surface_copy(pipe, texSurf, 0, 0, /* dest */ src, srcLeft, srcTop, /* src */ srcW, srcH); /* size */ /* free the surface, update the texture if necessary. */ - screen->tex_surface_release(screen, &texSurf); + pipe_surface_reference(&texSurf, NULL); /* save state (restored below) */ cso_save_blend(ctx->cso); @@ -389,13 +366,11 @@ util_blit_pixels(struct blit_state *ctx, cso_save_framebuffer(ctx->cso); cso_save_fragment_shader(ctx->cso); cso_save_vertex_shader(ctx->cso); - cso_save_viewport(ctx->cso); /* set misc state we care about */ cso_set_blend(ctx->cso, &ctx->blend); cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); - cso_set_viewport(ctx->cso, &ctx->viewport); /* sampler */ ctx->sampler.min_img_filter = filter; @@ -437,9 +412,8 @@ util_blit_pixels(struct blit_state *ctx, cso_restore_framebuffer(ctx->cso); cso_restore_fragment_shader(ctx->cso); cso_restore_vertex_shader(ctx->cso); - cso_restore_viewport(ctx->cso); - screen->texture_release(screen, &tex); + pipe_texture_reference(&tex, NULL); } @@ -448,7 +422,7 @@ util_blit_pixels(struct blit_state *ctx, */ void util_blit_flush( struct blit_state *ctx ) { - pipe_buffer_reference(ctx->pipe->screen, &ctx->vbuf, NULL); + pipe_buffer_reference(&ctx->vbuf, NULL); ctx->vbuf_slot = 0; } @@ -471,8 +445,6 @@ util_blit_pixels_tex(struct blit_state *ctx, int dstX1, int dstY1, float z, uint filter) { - struct pipe_context *pipe = ctx->pipe; - struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb; float s0, t0, s1, t1; unsigned offset; @@ -488,8 +460,10 @@ util_blit_pixels_tex(struct blit_state *ctx, t0 = srcY0 / (float)tex->height[0]; t1 = srcY1 / (float)tex->height[0]; - assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); + assert(ctx->pipe->screen->is_format_supported(ctx->pipe->screen, dst->format, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, + 0)); /* save state (restored below) */ cso_save_blend(ctx->cso); @@ -500,13 +474,11 @@ util_blit_pixels_tex(struct blit_state *ctx, cso_save_framebuffer(ctx->cso); cso_save_fragment_shader(ctx->cso); cso_save_vertex_shader(ctx->cso); - cso_save_viewport(ctx->cso); /* set misc state we care about */ cso_set_blend(ctx->cso, &ctx->blend); cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); - cso_set_viewport(ctx->cso, &ctx->viewport); /* sampler */ ctx->sampler.min_img_filter = filter; @@ -551,5 +523,4 @@ util_blit_pixels_tex(struct blit_state *ctx, cso_restore_framebuffer(ctx->cso); cso_restore_fragment_shader(ctx->cso); cso_restore_vertex_shader(ctx->cso); - cso_restore_viewport(ctx->cso); } diff --git a/src/gallium/auxiliary/util/u_cache.c b/src/gallium/auxiliary/util/u_cache.c index 0a1a64259f..41cd38171f 100644 --- a/src/gallium/auxiliary/util/u_cache.c +++ b/src/gallium/auxiliary/util/u_cache.c @@ -36,7 +36,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_math.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/u_debug.c index f373f941dd..f96e27e09f 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/u_debug.c @@ -50,6 +50,7 @@ #define WIN32_LEAN_AND_MEAN // Exclude rarely-used stuff from Windows headers #endif #include <windows.h> +#include <stdio.h> #else @@ -59,7 +60,7 @@ #endif #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_format.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" @@ -106,6 +107,12 @@ void _debug_vprintf(const char *format, va_list ap) OutputDebugStringA(buf); buf[0] = '\0'; } + + if(GetConsoleWindow() && !IsDebuggerPresent()) { + vfprintf(stderr, format, ap); + fflush(stderr); + } + #elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) wchar_t *wide_format; long wide_str_len; @@ -643,34 +650,37 @@ void debug_dump_image(const char *prefix, void debug_dump_surface(const char *prefix, struct pipe_surface *surface) { - unsigned surface_usage; + struct pipe_texture *texture; + struct pipe_screen *screen; + struct pipe_transfer *transfer; void *data; if (!surface) - goto error1; + return; + + texture = surface->texture; + screen = texture->screen; - /* XXX: force mappable surface */ - surface_usage = surface->usage; - surface->usage |= PIPE_BUFFER_USAGE_CPU_READ; + transfer = screen->get_tex_transfer(screen, texture, surface->face, + surface->level, surface->zslice, + PIPE_TRANSFER_READ, 0, 0, surface->width, + surface->height); - data = pipe_surface_map(surface, - PIPE_BUFFER_USAGE_CPU_READ); + data = screen->transfer_map(screen, transfer); if(!data) - goto error2; + goto error; debug_dump_image(prefix, - surface->format, - surface->block.size, - surface->nblocksx, - surface->nblocksy, - surface->stride, + transfer->format, + transfer->block.size, + transfer->nblocksx, + transfer->nblocksy, + transfer->stride, data); - pipe_surface_unmap(surface); -error2: - surface->usage = surface_usage; -error1: - ; + screen->transfer_unmap(screen, transfer); +error: + screen->tex_transfer_destroy(transfer); } @@ -710,8 +720,10 @@ debug_dump_surface_bmp(const char *filename, struct pipe_surface *surface) { #ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT + struct pipe_texture *texture; + struct pipe_screen *screen; struct util_stream *stream; - unsigned surface_usage; + struct pipe_transfer *transfer; struct bmp_file_header bmfh; struct bmp_info_header bmih; float *rgba; @@ -748,14 +760,18 @@ debug_dump_surface_bmp(const char *filename, util_stream_write(stream, &bmfh, 14); util_stream_write(stream, &bmih, 40); + + texture = surface->texture; + screen = texture->screen; - /* XXX: force mappable surface */ - surface_usage = surface->usage; - surface->usage |= PIPE_BUFFER_USAGE_CPU_READ; + transfer = screen->get_tex_transfer(screen, texture, surface->face, + surface->level, surface->zslice, + PIPE_TRANSFER_READ, 0, 0, surface->width, + surface->height); y = surface->height; while(y--) { - pipe_get_tile_rgba(surface, + pipe_get_tile_rgba(transfer, 0, y, surface->width, 1, rgba); for(x = 0; x < surface->width; ++x) @@ -768,9 +784,9 @@ debug_dump_surface_bmp(const char *filename, util_stream_write(stream, &pixel, 4); } } - - surface->usage = surface_usage; + screen->tex_transfer_destroy(transfer); + util_stream_close(stream); error2: FREE(rgba); diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/auxiliary/util/u_debug.h index e9c95982dd..7c829707b2 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/auxiliary/util/u_debug.h @@ -35,13 +35,13 @@ * @author Jose Fonseca <jrfonseca@tungstengraphics.com> */ -#ifndef P_DEBUG_H_ -#define P_DEBUG_H_ +#ifndef U_DEBUG_H_ +#define U_DEBUG_H_ #include <stdarg.h> -#include "p_compiler.h" +#include "pipe/p_compiler.h" #ifdef __cplusplus @@ -162,7 +162,7 @@ void _debug_assert_fail(const char *expr, #ifdef DEBUG #define debug_assert(expr) ((expr) ? (void)0 : _debug_assert_fail(#expr, __FILE__, __LINE__, __FUNCTION__)) #else -#define debug_assert(expr) ((void)0) +#define debug_assert(expr) ((void)(expr)) #endif @@ -358,4 +358,4 @@ void debug_dump_surface_bmp(const char *filename, } #endif -#endif /* P_DEBUG_H_ */ +#endif /* U_DEBUG_H_ */ diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/u_debug_memory.c index 250fd60f63..7623cb9398 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/u_debug_memory.c @@ -44,11 +44,13 @@ #include <stdlib.h> #endif -#include "pipe/p_debug.h" +#include "util/u_debug.h" +#include "util/u_debug_stack.h" #include "util/u_double_list.h" #define DEBUG_MEMORY_MAGIC 0x6e34090aU +#define DEBUG_MEMORY_STACK 0 /* XXX: disabled until we have symbol lookup */ #if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && !defined(WINCE) @@ -71,7 +73,11 @@ struct debug_memory_header const char *file; unsigned line; const char *function; +#if DEBUG_MEMORY_STACK + struct debug_stack_frame backtrace[DEBUG_MEMORY_STACK]; +#endif size_t size; + unsigned magic; }; @@ -136,6 +142,10 @@ debug_malloc(const char *file, unsigned line, const char *function, hdr->size = size; hdr->magic = DEBUG_MEMORY_MAGIC; +#if DEBUG_MEMORY_STACK + debug_backtrace_capture(hdr->backtrace, 0, DEBUG_MEMORY_STACK); +#endif + ftr = footer_from_header(hdr); ftr->magic = DEBUG_MEMORY_MAGIC; @@ -290,6 +300,9 @@ debug_memory_end(unsigned long start_no) debug_printf("%s:%u:%s: %u bytes at %p not freed\n", hdr->file, hdr->line, hdr->function, hdr->size, ptr); +#if DEBUG_MEMORY_STACK + debug_backtrace_dump(hdr->backtrace, DEBUG_MEMORY_STACK); +#endif total_size += hdr->size; } diff --git a/src/gallium/auxiliary/util/p_debug_prof.c b/src/gallium/auxiliary/util/u_debug_profile.c index 5f9772ef91..6d8b244c3a 100644 --- a/src/gallium/auxiliary/util/p_debug_prof.c +++ b/src/gallium/auxiliary/util/u_debug_profile.c @@ -42,7 +42,7 @@ #include <windows.h> #include <winddi.h> -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_string.h" diff --git a/src/gallium/auxiliary/util/u_debug_stack.c b/src/gallium/auxiliary/util/u_debug_stack.c new file mode 100644 index 0000000000..76068a6509 --- /dev/null +++ b/src/gallium/auxiliary/util/u_debug_stack.c @@ -0,0 +1,97 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Stack backtracing. + * + * @author Jose Fonseca <jfonseca@vmware.com> + */ + +#include "u_debug.h" +#include "u_debug_stack.h" + + +void +debug_backtrace_capture(struct debug_stack_frame *backtrace, + unsigned start_frame, + unsigned nr_frames) +{ + const void **frame_pointer = NULL; + unsigned i = 0; + + if(!nr_frames) + return; + +#if defined(PIPE_CC_GCC) + frame_pointer = ((const void **)__builtin_frame_address(1)); +#elif defined(PIPE_CC_MSVC) + __asm { + mov frame_pointer, ebp + } + frame_pointer = (const void **)frame_pointer[0]; +#else + frame_pointer = NULL; +#endif + + +#ifdef PIPE_ARCH_X86 + while(nr_frames) { + if(!frame_pointer) + break; + + if(start_frame) + --start_frame; + else { + backtrace[i++].function = frame_pointer[1]; + --nr_frames; + } + + frame_pointer = (const void **)frame_pointer[0]; + } +#endif + + while(nr_frames) { + backtrace[i++].function = NULL; + --nr_frames; + } +} + + +void +debug_backtrace_dump(const struct debug_stack_frame *backtrace, + unsigned nr_frames) +{ + unsigned i; + + for(i = 0; i < nr_frames; ++i) { + if(!backtrace[i].function) + break; + debug_printf("\t%p\n", backtrace[i].function); + } +} + diff --git a/src/gallium/state_trackers/wgl/stw.c b/src/gallium/auxiliary/util/u_debug_stack.h index 8bccdad221..f50f04e0f7 100644 --- a/src/gallium/state_trackers/wgl/stw.c +++ b/src/gallium/auxiliary/util/u_debug_stack.h @@ -1,8 +1,8 @@ /************************************************************************** - * - * Copyright 2009, VMware, Inc. + * + * Copyright 2009 VMware, Inc. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,48 +10,56 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ -#include "stw.h" -#include "shared/stw_winsys.h" +#ifndef U_DEBUG_STACK_H_ +#define U_DEBUG_STACK_H_ -boolean -st_init(const struct stw_winsys *stw_winsys) -{ - if (!stw_shared_init( stw_winsys )) - goto fail; - if (!stw_icd_init()) - goto fail; +/** + * @file + * Stack backtracing. + * + * @author Jose Fonseca <jfonseca@vmware.com> + */ - if (!stw_wgl_init()) - goto fail; - return TRUE; +#ifdef __cplusplus +extern "C" { +#endif -fail: - st_cleanup(); - return FALSE; -} +struct debug_stack_frame +{ + const void *function; +}; + void -st_cleanup(void) -{ - stw_icd_cleanup(); - stw_shared_cleanup(); - stw_wgl_cleanup(); +debug_backtrace_capture(struct debug_stack_frame *backtrace, + unsigned start_frame, + unsigned nr_frames); + +void +debug_backtrace_dump(const struct debug_stack_frame *backtrace, + unsigned nr_frames); + + +#ifdef __cplusplus } +#endif + +#endif /* U_DEBUG_STACK_H_ */ diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c index f282f3d289..4110485fb1 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.c +++ b/src/gallium/auxiliary/util/u_draw_quad.c @@ -51,9 +51,11 @@ util_draw_vertex_buffer(struct pipe_context *pipe, assert(num_attribs <= PIPE_MAX_ATTRIBS); /* tell pipe about the vertex buffer */ + memset(&vbuffer, 0, sizeof(vbuffer)); vbuffer.buffer = vbuf; vbuffer.stride = num_attribs * 4 * sizeof(float); /* vertex size */ vbuffer.buffer_offset = offset; + vbuffer.max_index = num_verts - 1; pipe->set_vertex_buffers(pipe, 1, &vbuffer); /* tell pipe about the vertex attributes */ @@ -127,6 +129,6 @@ util_draw_texquad(struct pipe_context *pipe, util_draw_vertex_buffer(pipe, vbuf, 0, PIPE_PRIM_TRIANGLE_FAN, 4, 2); } - pipe_buffer_reference(pipe->screen, &vbuf, NULL); + pipe_buffer_reference(&vbuf, NULL); } } diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 2b4cdab6cf..b32ad1cbe9 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -35,10 +35,11 @@ #include "pipe/p_context.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_state.h" #include "util/u_memory.h" #include "util/u_draw_quad.h" @@ -61,10 +62,7 @@ struct gen_mipmap_state struct pipe_depth_stencil_alpha_state depthstencil; struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; - struct pipe_viewport_state viewport; - struct pipe_shader_state vert_shader; - struct pipe_shader_state frag_shader; void *vs; void *fs; @@ -78,15 +76,15 @@ struct gen_mipmap_state enum dtype { - UBYTE, - UBYTE_3_3_2, - USHORT, - USHORT_4_4_4_4, - USHORT_5_6_5, - USHORT_1_5_5_5_REV, - UINT, - FLOAT, - HALF_FLOAT + DTYPE_UBYTE, + DTYPE_UBYTE_3_3_2, + DTYPE_USHORT, + DTYPE_USHORT_4_4_4_4, + DTYPE_USHORT_5_6_5, + DTYPE_USHORT_1_5_5_5_REV, + DTYPE_UINT, + DTYPE_FLOAT, + DTYPE_HALF_FLOAT }; @@ -194,7 +192,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth); */ - if (datatype == UBYTE && comps == 4) { + if (datatype == DTYPE_UBYTE && comps == 4) { uint i, j, k; const ubyte(*rowA)[4] = (const ubyte(*)[4]) srcRowA; const ubyte(*rowB)[4] = (const ubyte(*)[4]) srcRowB; @@ -207,7 +205,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; } } - else if (datatype == UBYTE && comps == 3) { + else if (datatype == DTYPE_UBYTE && comps == 3) { uint i, j, k; const ubyte(*rowA)[3] = (const ubyte(*)[3]) srcRowA; const ubyte(*rowB)[3] = (const ubyte(*)[3]) srcRowB; @@ -219,7 +217,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; } } - else if (datatype == UBYTE && comps == 2) { + else if (datatype == DTYPE_UBYTE && comps == 2) { uint i, j, k; const ubyte(*rowA)[2] = (const ubyte(*)[2]) srcRowA; const ubyte(*rowB)[2] = (const ubyte(*)[2]) srcRowB; @@ -230,7 +228,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2; } } - else if (datatype == UBYTE && comps == 1) { + else if (datatype == DTYPE_UBYTE && comps == 1) { uint i, j, k; const ubyte *rowA = (const ubyte *) srcRowA; const ubyte *rowB = (const ubyte *) srcRowB; @@ -241,7 +239,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } } - else if (datatype == USHORT && comps == 4) { + else if (datatype == DTYPE_USHORT && comps == 4) { uint i, j, k; const ushort(*rowA)[4] = (const ushort(*)[4]) srcRowA; const ushort(*rowB)[4] = (const ushort(*)[4]) srcRowB; @@ -254,7 +252,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; } } - else if (datatype == USHORT && comps == 3) { + else if (datatype == DTYPE_USHORT && comps == 3) { uint i, j, k; const ushort(*rowA)[3] = (const ushort(*)[3]) srcRowA; const ushort(*rowB)[3] = (const ushort(*)[3]) srcRowB; @@ -266,7 +264,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; } } - else if (datatype == USHORT && comps == 2) { + else if (datatype == DTYPE_USHORT && comps == 2) { uint i, j, k; const ushort(*rowA)[2] = (const ushort(*)[2]) srcRowA; const ushort(*rowB)[2] = (const ushort(*)[2]) srcRowB; @@ -277,7 +275,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; } } - else if (datatype == USHORT && comps == 1) { + else if (datatype == DTYPE_USHORT && comps == 1) { uint i, j, k; const ushort *rowA = (const ushort *) srcRowA; const ushort *rowB = (const ushort *) srcRowB; @@ -288,7 +286,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } } - else if (datatype == FLOAT && comps == 4) { + else if (datatype == DTYPE_FLOAT && comps == 4) { uint i, j, k; const float(*rowA)[4] = (const float(*)[4]) srcRowA; const float(*rowB)[4] = (const float(*)[4]) srcRowB; @@ -305,7 +303,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, rowB[j][3] + rowB[k][3]) * 0.25F; } } - else if (datatype == FLOAT && comps == 3) { + else if (datatype == DTYPE_FLOAT && comps == 3) { uint i, j, k; const float(*rowA)[3] = (const float(*)[3]) srcRowA; const float(*rowB)[3] = (const float(*)[3]) srcRowB; @@ -320,7 +318,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, rowB[j][2] + rowB[k][2]) * 0.25F; } } - else if (datatype == FLOAT && comps == 2) { + else if (datatype == DTYPE_FLOAT && comps == 2) { uint i, j, k; const float(*rowA)[2] = (const float(*)[2]) srcRowA; const float(*rowB)[2] = (const float(*)[2]) srcRowB; @@ -333,7 +331,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, rowB[j][1] + rowB[k][1]) * 0.25F; } } - else if (datatype == FLOAT && comps == 1) { + else if (datatype == DTYPE_FLOAT && comps == 1) { uint i, j, k; const float *rowA = (const float *) srcRowA; const float *rowB = (const float *) srcRowB; @@ -345,7 +343,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } #if 0 - else if (datatype == HALF_FLOAT && comps == 4) { + else if (datatype == HALF_DTYPE_FLOAT && comps == 4) { uint i, j, k, comp; const half_float(*rowA)[4] = (const half_float(*)[4]) srcRowA; const half_float(*rowB)[4] = (const half_float(*)[4]) srcRowB; @@ -362,7 +360,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } } } - else if (datatype == HALF_FLOAT && comps == 3) { + else if (datatype == DTYPE_HALF_FLOAT && comps == 3) { uint i, j, k, comp; const half_float(*rowA)[3] = (const half_float(*)[3]) srcRowA; const half_float(*rowB)[3] = (const half_float(*)[3]) srcRowB; @@ -379,7 +377,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } } } - else if (datatype == HALF_FLOAT && comps == 2) { + else if (datatype == DTYPE_HALF_FLOAT && comps == 2) { uint i, j, k, comp; const half_float(*rowA)[2] = (const half_float(*)[2]) srcRowA; const half_float(*rowB)[2] = (const half_float(*)[2]) srcRowB; @@ -396,7 +394,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } } } - else if (datatype == HALF_FLOAT && comps == 1) { + else if (datatype == DTYPE_HALF_FLOAT && comps == 1) { uint i, j, k; const half_float *rowA = (const half_float *) srcRowA; const half_float *rowB = (const half_float *) srcRowB; @@ -413,7 +411,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } #endif - else if (datatype == UINT && comps == 1) { + else if (datatype == DTYPE_UINT && comps == 1) { uint i, j, k; const uint *rowA = (const uint *) srcRowA; const uint *rowB = (const uint *) srcRowB; @@ -424,7 +422,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, } } - else if (datatype == USHORT_5_6_5 && comps == 3) { + else if (datatype == DTYPE_USHORT_5_6_5 && comps == 3) { uint i, j, k; const ushort *rowA = (const ushort *) srcRowA; const ushort *rowB = (const ushort *) srcRowB; @@ -449,7 +447,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i] = (blue << 11) | (green << 5) | red; } } - else if (datatype == USHORT_4_4_4_4 && comps == 4) { + else if (datatype == DTYPE_USHORT_4_4_4_4 && comps == 4) { uint i, j, k; const ushort *rowA = (const ushort *) srcRowA; const ushort *rowB = (const ushort *) srcRowB; @@ -479,7 +477,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red; } } - else if (datatype == USHORT_1_5_5_5_REV && comps == 4) { + else if (datatype == DTYPE_USHORT_1_5_5_5_REV && comps == 4) { uint i, j, k; const ushort *rowA = (const ushort *) srcRowA; const ushort *rowB = (const ushort *) srcRowB; @@ -509,7 +507,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red; } } - else if (datatype == UBYTE_3_3_2 && comps == 3) { + else if (datatype == DTYPE_UBYTE_3_3_2 && comps == 3) { uint i, j, k; const ubyte *rowA = (const ubyte *) srcRowA; const ubyte *rowB = (const ubyte *) srcRowB; @@ -570,7 +568,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, assert(comps >= 1); assert(comps <= 4); - if ((datatype == UBYTE) && (comps == 4)) { + if ((datatype == DTYPE_UBYTE) && (comps == 4)) { DECLARE_ROW_POINTERS(ubyte, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -581,7 +579,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(3); } } - else if ((datatype == UBYTE) && (comps == 3)) { + else if ((datatype == DTYPE_UBYTE) && (comps == 3)) { DECLARE_ROW_POINTERS(ubyte, 3); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -591,7 +589,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(2); } } - else if ((datatype == UBYTE) && (comps == 2)) { + else if ((datatype == DTYPE_UBYTE) && (comps == 2)) { DECLARE_ROW_POINTERS(ubyte, 2); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -600,7 +598,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(1); } } - else if ((datatype == UBYTE) && (comps == 1)) { + else if ((datatype == DTYPE_UBYTE) && (comps == 1)) { DECLARE_ROW_POINTERS(ubyte, 1); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -608,7 +606,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(0); } } - else if ((datatype == USHORT) && (comps == 4)) { + else if ((datatype == DTYPE_USHORT) && (comps == 4)) { DECLARE_ROW_POINTERS(ushort, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -619,7 +617,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(3); } } - else if ((datatype == USHORT) && (comps == 3)) { + else if ((datatype == DTYPE_USHORT) && (comps == 3)) { DECLARE_ROW_POINTERS(ushort, 3); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -629,7 +627,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(2); } } - else if ((datatype == USHORT) && (comps == 2)) { + else if ((datatype == DTYPE_USHORT) && (comps == 2)) { DECLARE_ROW_POINTERS(ushort, 2); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -638,7 +636,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(1); } } - else if ((datatype == USHORT) && (comps == 1)) { + else if ((datatype == DTYPE_USHORT) && (comps == 1)) { DECLARE_ROW_POINTERS(ushort, 1); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -646,7 +644,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_3D(0); } } - else if ((datatype == FLOAT) && (comps == 4)) { + else if ((datatype == DTYPE_FLOAT) && (comps == 4)) { DECLARE_ROW_POINTERS(float, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -657,7 +655,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_F_3D(3); } } - else if ((datatype == FLOAT) && (comps == 3)) { + else if ((datatype == DTYPE_FLOAT) && (comps == 3)) { DECLARE_ROW_POINTERS(float, 3); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -667,7 +665,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_F_3D(2); } } - else if ((datatype == FLOAT) && (comps == 2)) { + else if ((datatype == DTYPE_FLOAT) && (comps == 2)) { DECLARE_ROW_POINTERS(float, 2); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -676,7 +674,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_F_3D(1); } } - else if ((datatype == FLOAT) && (comps == 1)) { + else if ((datatype == DTYPE_FLOAT) && (comps == 1)) { DECLARE_ROW_POINTERS(float, 1); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -684,7 +682,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_F_3D(0); } } - else if ((datatype == HALF_FLOAT) && (comps == 4)) { + else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 4)) { DECLARE_ROW_POINTERS(half_float, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -695,7 +693,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_HF_3D(3); } } - else if ((datatype == HALF_FLOAT) && (comps == 3)) { + else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 3)) { DECLARE_ROW_POINTERS(half_float, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -705,7 +703,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_HF_3D(2); } } - else if ((datatype == HALF_FLOAT) && (comps == 2)) { + else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 2)) { DECLARE_ROW_POINTERS(half_float, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -714,7 +712,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_HF_3D(1); } } - else if ((datatype == HALF_FLOAT) && (comps == 1)) { + else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 1)) { DECLARE_ROW_POINTERS(half_float, 4); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -722,7 +720,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, FILTER_HF_3D(0); } } - else if ((datatype == UINT) && (comps == 1)) { + else if ((datatype == DTYPE_UINT) && (comps == 1)) { const uint *rowA = (const uint *) srcRowA; const uint *rowB = (const uint *) srcRowB; const uint *rowC = (const uint *) srcRowC; @@ -738,7 +736,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, dst[i] = (float)((double) tmp * 0.125); } } - else if ((datatype == USHORT_5_6_5) && (comps == 3)) { + else if ((datatype == DTYPE_USHORT_5_6_5) && (comps == 3)) { DECLARE_ROW_POINTERS0(ushort); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -776,7 +774,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, dst[i] = (b << 11) | (g << 5) | r; } } - else if ((datatype == USHORT_4_4_4_4) && (comps == 4)) { + else if ((datatype == DTYPE_USHORT_4_4_4_4) && (comps == 4)) { DECLARE_ROW_POINTERS0(ushort); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -825,7 +823,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, dst[i] = (a << 12) | (b << 8) | (g << 4) | r; } } - else if ((datatype == USHORT_1_5_5_5_REV) && (comps == 4)) { + else if ((datatype == DTYPE_USHORT_1_5_5_5_REV) && (comps == 4)) { DECLARE_ROW_POINTERS0(ushort); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -874,7 +872,7 @@ do_row_3D(enum dtype datatype, uint comps, int srcWidth, dst[i] = (a << 15) | (b << 10) | (g << 5) | r; } } - else if ((datatype == UBYTE_3_3_2) && (comps == 3)) { + else if ((datatype == DTYPE_UBYTE_3_3_2) && (comps == 3)) { DECLARE_ROW_POINTERS0(ushort); for (i = j = 0, k = k0; i < (uint) dstWidth; @@ -928,34 +926,34 @@ format_to_type_comps(enum pipe_format pformat, case PIPE_FORMAT_X8R8G8B8_UNORM: case PIPE_FORMAT_B8G8R8A8_UNORM: case PIPE_FORMAT_B8G8R8X8_UNORM: - *datatype = UBYTE; + *datatype = DTYPE_UBYTE; *comps = 4; return; case PIPE_FORMAT_A1R5G5B5_UNORM: - *datatype = USHORT_1_5_5_5_REV; + *datatype = DTYPE_USHORT_1_5_5_5_REV; *comps = 4; return; case PIPE_FORMAT_A4R4G4B4_UNORM: - *datatype = USHORT_4_4_4_4; + *datatype = DTYPE_USHORT_4_4_4_4; *comps = 4; return; case PIPE_FORMAT_R5G6B5_UNORM: - *datatype = USHORT_5_6_5; + *datatype = DTYPE_USHORT_5_6_5; *comps = 3; return; case PIPE_FORMAT_L8_UNORM: case PIPE_FORMAT_A8_UNORM: case PIPE_FORMAT_I8_UNORM: - *datatype = UBYTE; + *datatype = DTYPE_UBYTE; *comps = 1; return; case PIPE_FORMAT_A8L8_UNORM: - *datatype = UBYTE; + *datatype = DTYPE_UBYTE; *comps = 2; return; default: assert(0); - *datatype = UBYTE; + *datatype = DTYPE_UBYTE; *comps = 0; break; } @@ -1116,31 +1114,30 @@ make_1d_mipmap(struct gen_mipmap_state *ctx, for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; - struct pipe_surface *srcSurf, *dstSurf; + struct pipe_transfer *srcTrans, *dstTrans; void *srcMap, *dstMap; - srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice, - PIPE_BUFFER_USAGE_CPU_READ); - - dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice, - PIPE_BUFFER_USAGE_CPU_WRITE); - - srcMap = ((ubyte *) pipe_surface_map(srcSurf, - PIPE_BUFFER_USAGE_CPU_READ) - + srcSurf->offset); - dstMap = ((ubyte *) pipe_surface_map(dstSurf, - PIPE_BUFFER_USAGE_CPU_WRITE) - + dstSurf->offset); + srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, + PIPE_TRANSFER_READ, 0, 0, + pt->width[srcLevel], + pt->height[srcLevel]); + dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, + PIPE_TRANSFER_WRITE, 0, 0, + pt->width[dstLevel], + pt->height[dstLevel]); + + srcMap = (ubyte *) screen->transfer_map(screen, srcTrans); + dstMap = (ubyte *) screen->transfer_map(screen, dstTrans); reduce_1d(pt->format, - srcSurf->width, srcMap, - dstSurf->width, dstMap); + srcTrans->width, srcMap, + dstTrans->width, dstMap); - pipe_surface_unmap(srcSurf); - pipe_surface_unmap(dstSurf); + screen->transfer_unmap(screen, srcTrans); + screen->transfer_unmap(screen, dstTrans); - pipe_surface_reference(&srcSurf, NULL); - pipe_surface_reference(&dstSurf, NULL); + screen->tex_transfer_destroy(srcTrans); + screen->tex_transfer_destroy(dstTrans); } } @@ -1160,32 +1157,32 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; - struct pipe_surface *srcSurf, *dstSurf; + struct pipe_transfer *srcTrans, *dstTrans; ubyte *srcMap, *dstMap; - srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice, - PIPE_BUFFER_USAGE_CPU_READ); - dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice, - PIPE_BUFFER_USAGE_CPU_WRITE); - - srcMap = ((ubyte *) pipe_surface_map(srcSurf, - PIPE_BUFFER_USAGE_CPU_READ) - + srcSurf->offset); - dstMap = ((ubyte *) pipe_surface_map(dstSurf, - PIPE_BUFFER_USAGE_CPU_WRITE) - + dstSurf->offset); + srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, + PIPE_TRANSFER_READ, 0, 0, + pt->width[srcLevel], + pt->height[srcLevel]); + dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, + PIPE_TRANSFER_WRITE, 0, 0, + pt->width[dstLevel], + pt->height[dstLevel]); + + srcMap = (ubyte *) screen->transfer_map(screen, srcTrans); + dstMap = (ubyte *) screen->transfer_map(screen, dstTrans); reduce_2d(pt->format, - srcSurf->width, srcSurf->height, - srcSurf->stride, srcMap, - dstSurf->width, dstSurf->height, - dstSurf->stride, dstMap); + srcTrans->width, srcTrans->height, + srcTrans->stride, srcMap, + dstTrans->width, dstTrans->height, + dstTrans->stride, dstMap); - pipe_surface_unmap(srcSurf); - pipe_surface_unmap(dstSurf); + screen->transfer_unmap(screen, srcTrans); + screen->transfer_unmap(screen, dstTrans); - pipe_surface_reference(&srcSurf, NULL); - pipe_surface_reference(&dstSurf, NULL); + screen->tex_transfer_destroy(srcTrans); + screen->tex_transfer_destroy(dstTrans); } } @@ -1195,6 +1192,7 @@ make_3d_mipmap(struct gen_mipmap_state *ctx, struct pipe_texture *pt, uint face, uint baseLevel, uint lastLevel) { +#if 0 struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; uint dstLevel, zslice = 0; @@ -1204,37 +1202,36 @@ make_3d_mipmap(struct gen_mipmap_state *ctx, for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; - struct pipe_surface *srcSurf, *dstSurf; + struct pipe_transfer *srcTrans, *dstTrans; ubyte *srcMap, *dstMap; - srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice, - PIPE_BUFFER_USAGE_CPU_READ); - dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice, - PIPE_BUFFER_USAGE_CPU_WRITE); - - srcMap = ((ubyte *) pipe_surface_map(srcSurf, - PIPE_BUFFER_USAGE_CPU_READ) - + srcSurf->offset); - dstMap = ((ubyte *) pipe_surface_map(dstSurf, - PIPE_BUFFER_USAGE_CPU_WRITE) - + dstSurf->offset); + srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, + PIPE_TRANSFER_READ, 0, 0, + pt->width[srcLevel], + pt->height[srcLevel]); + dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, + PIPE_TRANSFER_WRITE, 0, 0, + pt->width[dstLevel], + pt->height[dstLevel]); + + srcMap = (ubyte *) screen->transfer_map(screen, srcTrans); + dstMap = (ubyte *) screen->transfer_map(screen, dstTrans); -#if 0 reduce_3d(pt->format, - srcSurf->width, srcSurf->height, - srcSurf->stride, srcMap, - dstSurf->width, dstSurf->height, - dstSurf->stride, dstMap); -#else - (void) reduce_3d; -#endif + srcTrans->width, srcTrans->height, + srcTrans->stride, srcMap, + dstTrans->width, dstTrans->height, + dstTrans->stride, dstMap); - pipe_surface_unmap(srcSurf); - pipe_surface_unmap(dstSurf); + screen->transfer_unmap(screen, srcTrans); + screen->transfer_unmap(screen, dstTrans); - pipe_surface_reference(&srcSurf, NULL); - pipe_surface_reference(&dstSurf, NULL); + screen->tex_transfer_destroy(srcTrans); + screen->tex_transfer_destroy(dstTrans); } +#else + (void) reduce_3d; +#endif } @@ -1294,8 +1291,7 @@ util_create_gen_mipmap(struct pipe_context *pipe, memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer)); ctx->rasterizer.front_winding = PIPE_WINDING_CW; ctx->rasterizer.cull_mode = PIPE_WINDING_NONE; - ctx->rasterizer.bypass_clipping = 1; - /*ctx->rasterizer.bypass_vs = 1;*/ + ctx->rasterizer.bypass_vs_clip_and_viewport = 1; ctx->rasterizer.gl_rasterization_rules = 1; /* sampler state */ @@ -1306,28 +1302,19 @@ util_create_gen_mipmap(struct pipe_context *pipe, ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; ctx->sampler.normalized_coords = 1; - /* viewport state (identity, verts are in wincoords) */ - ctx->viewport.scale[0] = 1.0; - ctx->viewport.scale[1] = 1.0; - ctx->viewport.scale[2] = 1.0; - ctx->viewport.scale[3] = 1.0; - ctx->viewport.translate[0] = 0.0; - ctx->viewport.translate[1] = 0.0; - ctx->viewport.translate[2] = 0.0; - ctx->viewport.translate[3] = 0.0; - - /* vertex shader */ + /* vertex shader - still needed to specify mapping from fragment + * shader input semantics to vertex elements + */ { const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, TGSI_SEMANTIC_GENERIC }; const uint semantic_indexes[] = { 0, 0 }; ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, - semantic_indexes, - &ctx->vert_shader); + semantic_indexes); } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe, &ctx->frag_shader); + ctx->fs = util_make_fragment_tex_shader(pipe); /* vertex data that doesn't change */ for (i = 0; i < 4; i++) { @@ -1369,7 +1356,6 @@ get_next_slot(struct gen_mipmap_state *ctx) static unsigned set_vertex_data(struct gen_mipmap_state *ctx, float width, float height) { - void *buf; unsigned offset; ctx->vertices[0][0][0] = 0.0f; /*x*/ @@ -1394,12 +1380,8 @@ set_vertex_data(struct gen_mipmap_state *ctx, float width, float height) offset = get_next_slot( ctx ); - buf = pipe_buffer_map(ctx->pipe->screen, ctx->vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE); - - memcpy((char *)buf + offset, ctx->vertices, sizeof(ctx->vertices)); - - pipe_buffer_unmap(ctx->pipe->screen, ctx->vbuf); + pipe_buffer_write(ctx->pipe->screen, ctx->vbuf, + offset, sizeof(ctx->vertices), ctx->vertices); return offset; } @@ -1417,10 +1399,7 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); - FREE((void*) ctx->vert_shader.tokens); - FREE((void*) ctx->frag_shader.tokens); - - pipe_buffer_reference(pipe->screen, &ctx->vbuf, NULL); + pipe_buffer_reference(&ctx->vbuf, NULL); FREE(ctx); } @@ -1432,7 +1411,7 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) */ void util_gen_mipmap_flush( struct gen_mipmap_state *ctx ) { - pipe_buffer_reference(ctx->pipe->screen, &ctx->vbuf, NULL); + pipe_buffer_reference(&ctx->vbuf, NULL); ctx->vbuf_slot = 0; } @@ -1476,13 +1455,11 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_save_framebuffer(ctx->cso); cso_save_fragment_shader(ctx->cso); cso_save_vertex_shader(ctx->cso); - cso_save_viewport(ctx->cso); /* bind our state */ cso_set_blend(ctx->cso, &ctx->blend); cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); - cso_set_viewport(ctx->cso, &ctx->viewport); cso_set_fragment_shader_handle(ctx->cso, ctx->fs); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); @@ -1528,7 +1505,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_set_sampler_textures(ctx->cso, 1, &pt); - /* quad coords in window coords (bypassing clipping, viewport mapping) */ + /* quad coords in window coords (bypassing vs, clip and viewport) */ offset = set_vertex_data(ctx, (float) pt->width[dstLevel], (float) pt->height[dstLevel]); @@ -1555,5 +1532,4 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_restore_framebuffer(ctx->cso); cso_restore_fragment_shader(ctx->cso); cso_restore_vertex_shader(ctx->cso); - cso_restore_viewport(ctx->cso); } diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 2d15932ce3..6da7353e25 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -34,7 +34,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "util/u_handle_table.h" diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c index 0bc8de9632..2f83e318e4 100644 --- a/src/gallium/auxiliary/util/u_hash_table.c +++ b/src/gallium/auxiliary/util/u_hash_table.c @@ -39,7 +39,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "cso_cache/cso_hash.h" diff --git a/src/gallium/auxiliary/util/u_keymap.c b/src/gallium/auxiliary/util/u_keymap.c index 01b17ddb1b..3f70809efd 100644 --- a/src/gallium/auxiliary/util/u_keymap.c +++ b/src/gallium/auxiliary/util/u_keymap.c @@ -35,7 +35,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_error.h" #include "cso_cache/cso_hash.h" diff --git a/src/gallium/auxiliary/util/u_linear.c b/src/gallium/auxiliary/util/u_linear.c index e999cefe74..6be365e53b 100644 --- a/src/gallium/auxiliary/util/u_linear.c +++ b/src/gallium/auxiliary/util/u_linear.c @@ -1,5 +1,5 @@ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "u_linear.h" void diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index ab6f39ac31..1ecde7a912 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -40,7 +40,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #ifdef __cplusplus diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h index 1a6b596421..0b18d043ad 100644 --- a/src/gallium/auxiliary/util/u_memory.h +++ b/src/gallium/auxiliary/util/u_memory.h @@ -36,7 +36,7 @@ #include "util/u_pointer.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #ifdef __cplusplus @@ -52,11 +52,11 @@ extern "C" { #endif -#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) +#if defined(PIPE_OS_WINDOWS) && defined(DEBUG) /* memory debugging */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #define MALLOC( _size ) \ debug_malloc( __FILE__, __LINE__, __FUNCTION__, _size ) @@ -191,9 +191,11 @@ align_free(void *ptr) #if defined(HAVE_POSIX_MEMALIGN) FREE(ptr); #else - void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); - void *realAddr = *cubbyHole; - FREE(realAddr); + if (ptr) { + void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); + void *realAddr = *cubbyHole; + FREE(realAddr); + } #endif /* defined(HAVE_POSIX_MEMALIGN) */ } diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c index 45ce257b5e..151a480d34 100644 --- a/src/gallium/auxiliary/util/u_mm.c +++ b/src/gallium/auxiliary/util/u_mm.c @@ -24,7 +24,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "util/u_mm.h" diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index fe81a685be..74259d453b 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -169,46 +169,35 @@ util_surface_copy(struct pipe_context *pipe, unsigned w, unsigned h) { struct pipe_screen *screen = pipe->screen; - struct pipe_surface *new_src = NULL, *new_dst = NULL; + struct pipe_transfer *src_trans, *dst_trans; void *dst_map; const void *src_map; - assert(dst->block.size == src->block.size); - assert(dst->block.width == src->block.width); - assert(dst->block.height == src->block.height); - - if ((src->usage & PIPE_BUFFER_USAGE_CPU_READ) == 0) { - /* Need to create new src surface which is CPU readable */ - assert(src->texture); - if (!src->texture) - return; - new_src = screen->get_tex_surface(screen, + assert(src->texture && dst->texture); + if (!src->texture || !dst->texture) + return; + src_trans = screen->get_tex_transfer(screen, src->texture, src->face, src->level, src->zslice, - PIPE_BUFFER_USAGE_CPU_READ); - src = new_src; - } + PIPE_TRANSFER_READ, + src_x, src_y, w, h); - if ((dst->usage & PIPE_BUFFER_USAGE_CPU_WRITE) == 0) { - /* Need to create new dst surface which is CPU writable */ - assert(dst->texture); - if (!dst->texture) - return; - new_dst = screen->get_tex_surface(screen, + dst_trans = screen->get_tex_transfer(screen, dst->texture, dst->face, dst->level, dst->zslice, - PIPE_BUFFER_USAGE_CPU_WRITE); - dst = new_dst; - } + PIPE_TRANSFER_WRITE, + dst_x, dst_y, w, h); - src_map = pipe->screen->surface_map(screen, - src, PIPE_BUFFER_USAGE_CPU_READ); - dst_map = pipe->screen->surface_map(screen, - dst, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(dst_trans->block.size == src_trans->block.size); + assert(dst_trans->block.width == src_trans->block.width); + assert(dst_trans->block.height == src_trans->block.height); + + src_map = pipe->screen->transfer_map(screen, src_trans); + dst_map = pipe->screen->transfer_map(screen, dst_trans); assert(src_map); assert(dst_map); @@ -216,36 +205,25 @@ util_surface_copy(struct pipe_context *pipe, if (src_map && dst_map) { /* If do_flip, invert src_y position and pass negative src stride */ pipe_copy_rect(dst_map, - &dst->block, - dst->stride, - dst_x, dst_y, + &dst_trans->block, + dst_trans->stride, + 0, 0, w, h, src_map, - do_flip ? -(int) src->stride : src->stride, - src_x, - do_flip ? src_y + h - 1 : src_y); + do_flip ? -(int) src_trans->stride : src_trans->stride, + 0, + do_flip ? h - 1 : 0); } - pipe->screen->surface_unmap(pipe->screen, src); - pipe->screen->surface_unmap(pipe->screen, dst); + pipe->screen->transfer_unmap(pipe->screen, src_trans); + pipe->screen->transfer_unmap(pipe->screen, dst_trans); - if (new_src) - screen->tex_surface_release(screen, &new_src); - if (new_dst) - screen->tex_surface_release(screen, &new_dst); + screen->tex_transfer_destroy(src_trans); + screen->tex_transfer_destroy(dst_trans); } -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map - + y / dst->block.height * dst->stride - + x / dst->block.width * dst->block.size; -} - - #define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) @@ -260,42 +238,38 @@ util_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { struct pipe_screen *screen = pipe->screen; - struct pipe_surface *new_dst = NULL; + struct pipe_transfer *dst_trans; void *dst_map; - if ((dst->usage & PIPE_BUFFER_USAGE_CPU_WRITE) == 0) { - /* Need to create new dst surface which is CPU writable */ - assert(dst->texture); - if (!dst->texture) - return; - new_dst = screen->get_tex_surface(screen, + assert(dst->texture); + if (!dst->texture) + return; + dst_trans = screen->get_tex_transfer(screen, dst->texture, dst->face, dst->level, dst->zslice, - PIPE_BUFFER_USAGE_CPU_WRITE); - dst = new_dst; - } + PIPE_TRANSFER_WRITE, + dstx, dsty, width, height); - dst_map = pipe->screen->surface_map(screen, - dst, PIPE_BUFFER_USAGE_CPU_WRITE); + dst_map = pipe->screen->transfer_map(screen, dst_trans); assert(dst_map); if (dst_map) { - assert(dst->stride > 0); + assert(dst_trans->stride > 0); - switch (dst->block.size) { + switch (dst_trans->block.size) { case 1: case 2: case 4: - pipe_fill_rect(dst_map, &dst->block, dst->stride, - dstx, dsty, width, height, value); + pipe_fill_rect(dst_map, &dst_trans->block, dst_trans->stride, + 0, 0, width, height, value); break; case 8: { /* expand the 4-byte clear value to an 8-byte value */ - ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); + ushort *row = (ushort *) dst_map; ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); @@ -312,7 +286,7 @@ util_surface_fill(struct pipe_context *pipe, row[j*4+2] = val2; row[j*4+3] = val3; } - row += dst->stride/2; + row += dst_trans->stride/2; } } break; @@ -322,8 +296,6 @@ util_surface_fill(struct pipe_context *pipe, } } - pipe->screen->surface_unmap(pipe->screen, dst); - - if (new_dst) - screen->tex_surface_release(screen, &new_dst); + pipe->screen->transfer_unmap(pipe->screen, dst_trans); + screen->tex_transfer_destroy(dst_trans); } diff --git a/src/gallium/auxiliary/util/u_simple_screen.c b/src/gallium/auxiliary/util/u_simple_screen.c index 089bbbc48a..8114b53cd0 100644 --- a/src/gallium/auxiliary/util/u_simple_screen.c +++ b/src/gallium/auxiliary/util/u_simple_screen.c @@ -28,6 +28,7 @@ #include "u_simple_screen.h" #include "pipe/p_screen.h" +#include "pipe/p_state.h" #include "pipe/internal/p_winsys_screen.h" @@ -37,8 +38,12 @@ pass_buffer_create(struct pipe_screen *screen, unsigned usage, unsigned size) { - return screen->winsys->buffer_create(screen->winsys, - alignment, usage, size); + struct pipe_buffer *buffer = + screen->winsys->buffer_create(screen->winsys, alignment, usage, size); + + buffer->screen = screen; + + return buffer; } static struct pipe_buffer * @@ -46,8 +51,13 @@ pass_user_buffer_create(struct pipe_screen *screen, void *ptr, unsigned bytes) { - return screen->winsys->user_buffer_create(screen->winsys, + struct pipe_buffer *buffer = + screen->winsys->user_buffer_create(screen->winsys, ptr, bytes); + + buffer->screen = screen; + + return buffer; } static struct pipe_buffer * @@ -57,9 +67,14 @@ pass_surface_buffer_create(struct pipe_screen *screen, unsigned usage, unsigned *stride) { - return screen->winsys->surface_buffer_create(screen->winsys, + struct pipe_buffer *buffer = + screen->winsys->surface_buffer_create(screen->winsys, width, height, format, usage, stride); + + buffer->screen = screen; + + return buffer; } static void * @@ -79,10 +94,9 @@ pass_buffer_unmap(struct pipe_screen *screen, } static void -pass_buffer_destroy(struct pipe_screen *screen, - struct pipe_buffer *buf) +pass_buffer_destroy(struct pipe_buffer *buf) { - screen->winsys->buffer_destroy(screen->winsys, buf); + buf->screen->winsys->buffer_destroy(buf); } diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 706155e99a..e519c354d2 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -34,9 +34,8 @@ #include "pipe/p_context.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_defines.h" -#include "pipe/p_inlines.h" #include "pipe/p_screen.h" #include "pipe/p_shader_tokens.h" @@ -56,12 +55,11 @@ void * util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint num_attribs, const uint *semantic_names, - const uint *semantic_indexes, - struct pipe_shader_state *shader) + const uint *semantic_indexes) { - uint maxTokens = 100; - struct tgsi_token *tokens; + struct pipe_shader_state shader; + struct tgsi_token tokens[100]; struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_declaration decl; @@ -69,8 +67,6 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, const uint procType = TGSI_PROCESSOR_VERTEX; uint ti, i; - tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); - /* shader header */ *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -97,7 +93,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); } /* declare outputs */ @@ -112,7 +108,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); } /* emit MOV instructions */ @@ -129,7 +125,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, - maxTokens - ti ); + Elements(tokens) - ti ); } /* END instruction */ @@ -140,16 +136,15 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, - maxTokens - ti ); + Elements(tokens) - ti ); #if 0 /*debug*/ tgsi_dump(tokens, 0); #endif - shader->tokens = tokens; - /*shader->num_tokens = ti;*/ + shader.tokens = tokens; - return pipe->create_vs_state(pipe, shader); + return pipe->create_vs_state(pipe, &shader); } @@ -161,11 +156,10 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, * END; */ void * -util_make_fragment_tex_shader(struct pipe_context *pipe, - struct pipe_shader_state *shader) +util_make_fragment_tex_shader(struct pipe_context *pipe) { - uint maxTokens = 100; - struct tgsi_token *tokens; + struct pipe_shader_state shader; + struct tgsi_token tokens[100]; struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_declaration decl; @@ -173,8 +167,6 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, const uint procType = TGSI_PROCESSOR_FRAGMENT; uint ti; - tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); - /* shader header */ *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -200,7 +192,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); /* declare color[0] output */ decl = tgsi_default_full_declaration(); @@ -213,7 +205,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); /* declare sampler */ decl = tgsi_default_full_declaration(); @@ -223,7 +215,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); /* TEX instruction */ inst = tgsi_default_full_instruction(); @@ -240,7 +232,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, - maxTokens - ti ); + Elements(tokens) - ti ); /* END instruction */ inst = tgsi_default_full_instruction(); @@ -250,16 +242,15 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, - maxTokens - ti ); + Elements(tokens) - ti ); #if 0 /*debug*/ tgsi_dump(tokens, 0); #endif - shader->tokens = tokens; - /*shader->num_tokens = ti;*/ + shader.tokens = tokens; - return pipe->create_fs_state(pipe, shader); + return pipe->create_fs_state(pipe, &shader); } @@ -270,11 +261,10 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, * Make simple fragment color pass-through shader. */ void * -util_make_fragment_passthrough_shader(struct pipe_context *pipe, - struct pipe_shader_state *shader) +util_make_fragment_passthrough_shader(struct pipe_context *pipe) { - uint maxTokens = 40; - struct tgsi_token *tokens; + struct pipe_shader_state shader; + struct tgsi_token tokens[40]; struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_declaration decl; @@ -282,8 +272,6 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, const uint procType = TGSI_PROCESSOR_FRAGMENT; uint ti; - tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); - /* shader header */ *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -307,7 +295,7 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); /* declare output */ decl = tgsi_default_full_declaration(); @@ -320,7 +308,7 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, - maxTokens - ti); + Elements(tokens) - ti); /* MOVE out[0], in[0]; */ @@ -335,7 +323,7 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, - maxTokens - ti ); + Elements(tokens) - ti ); /* END instruction */ inst = tgsi_default_full_instruction(); @@ -345,17 +333,17 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, - maxTokens - ti ); + Elements(tokens) - ti ); - assert(ti < maxTokens); + assert(ti < Elements(tokens)); #if 0 /*debug*/ tgsi_dump(tokens, 0); #endif - shader->tokens = tokens; - /*shader->num_tokens = ti;*/ + shader.tokens = tokens; - return pipe->create_fs_state(pipe, shader); + return pipe->create_fs_state(pipe, &shader); } + diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h index 8ca4977d71..6f8d96af9b 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.h +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -46,18 +46,15 @@ extern void * util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint num_attribs, const uint *semantic_names, - const uint *semantic_indexes, - struct pipe_shader_state *shader); + const uint *semantic_indexes); extern void * -util_make_fragment_tex_shader(struct pipe_context *pipe, - struct pipe_shader_state *shader); +util_make_fragment_tex_shader(struct pipe_context *pipe); extern void * -util_make_fragment_passthrough_shader(struct pipe_context *pipe, - struct pipe_shader_state *shader); +util_make_fragment_passthrough_shader(struct pipe_context *pipe); #ifdef __cplusplus diff --git a/src/gallium/auxiliary/util/u_stream_stdc.c b/src/gallium/auxiliary/util/u_stream_stdc.c index ca80bef0f3..0ead45a749 100644 --- a/src/gallium/auxiliary/util/u_stream_stdc.c +++ b/src/gallium/auxiliary/util/u_stream_stdc.c @@ -32,7 +32,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) #include <stdio.h> diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index 32f6b072a0..d31ca9c029 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -28,7 +28,6 @@ /** * RGBA/float tile get/put functions. * Usable both by drivers and state trackers. - * Surfaces should already be in a mapped state. */ @@ -42,58 +41,58 @@ /** - * Move raw block of pixels from surface to user memory. - * This should be usable by any hw driver that has mappable surfaces. + * Move raw block of pixels from transfer object to user memory. */ void -pipe_get_tile_raw(struct pipe_surface *ps, +pipe_get_tile_raw(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, void *dst, int dst_stride) { + struct pipe_screen *screen = pt->texture->screen; const void *src; if (dst_stride == 0) - dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; + dst_stride = pf_get_nblocksx(&pt->block, w) * pt->block.size; - if (pipe_clip_tile(x, y, &w, &h, ps)) + if (pipe_clip_tile(x, y, &w, &h, pt)) return; - src = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); + src = screen->transfer_map(screen, pt); assert(src); if(!src) return; - pipe_copy_rect(dst, &ps->block, dst_stride, 0, 0, w, h, src, ps->stride, x, y); + pipe_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y); - pipe_surface_unmap(ps); + screen->transfer_unmap(screen, pt); } /** - * Move raw block of pixels from user memory to surface. - * This should be usable by any hw driver that has mappable surfaces. + * Move raw block of pixels from user memory to transfer object. */ void -pipe_put_tile_raw(struct pipe_surface *ps, +pipe_put_tile_raw(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, const void *src, int src_stride) { + struct pipe_screen *screen = pt->texture->screen; void *dst; if (src_stride == 0) - src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; + src_stride = pf_get_nblocksx(&pt->block, w) * pt->block.size; - if (pipe_clip_tile(x, y, &w, &h, ps)) + if (pipe_clip_tile(x, y, &w, &h, pt)) return; - dst = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); + dst = screen->transfer_map(screen, pt); assert(dst); if(!dst) return; - pipe_copy_rect(dst, &ps->block, ps->stride, x, y, w, h, src, src_stride, 0, 0); + pipe_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0); - pipe_surface_unmap(ps); + screen->transfer_unmap(screen, pt); } @@ -883,6 +882,27 @@ ycbcr_get_tile_rgba(const ushort *src, } +static void +fake_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (i ^ j) & 1 ? 1.0f : 0.0f; + } + p += dst_stride; + } +} + + void pipe_tile_raw_to_rgba(enum pipe_format format, void *src, @@ -949,55 +969,56 @@ pipe_tile_raw_to_rgba(enum pipe_format format, ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE); break; default: - assert(0); + debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(format)); + fake_get_tile_rgba(src, w, h, dst, dst_stride); } } void -pipe_get_tile_rgba(struct pipe_surface *ps, +pipe_get_tile_rgba(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, float *p) { unsigned dst_stride = w * 4; void *packed; - if (pipe_clip_tile(x, y, &w, &h, ps)) + if (pipe_clip_tile(x, y, &w, &h, pt)) return; - packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); + packed = MALLOC(pf_get_nblocks(&pt->block, w, h) * pt->block.size); if (!packed) return; - if(ps->format == PIPE_FORMAT_YCBCR || ps->format == PIPE_FORMAT_YCBCR_REV) + if(pt->format == PIPE_FORMAT_YCBCR || pt->format == PIPE_FORMAT_YCBCR_REV) assert((x & 1) == 0); - pipe_get_tile_raw(ps, x, y, w, h, packed, 0); + pipe_get_tile_raw(pt, x, y, w, h, packed, 0); - pipe_tile_raw_to_rgba(ps->format, packed, w, h, p, dst_stride); + pipe_tile_raw_to_rgba(pt->format, packed, w, h, p, dst_stride); FREE(packed); } void -pipe_put_tile_rgba(struct pipe_surface *ps, +pipe_put_tile_rgba(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, const float *p) { unsigned src_stride = w * 4; void *packed; - if (pipe_clip_tile(x, y, &w, &h, ps)) + if (pipe_clip_tile(x, y, &w, &h, pt)) return; - packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); + packed = MALLOC(pf_get_nblocks(&pt->block, w, h) * pt->block.size); if (!packed) return; - switch (ps->format) { + switch (pt->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); break; @@ -1051,10 +1072,10 @@ pipe_put_tile_rgba(struct pipe_surface *ps, /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; default: - assert(0); + debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(pt->format)); } - pipe_put_tile_raw(ps, x, y, w, h, packed, 0); + pipe_put_tile_raw(pt, x, y, w, h, packed, 0); FREE(packed); } @@ -1064,62 +1085,63 @@ pipe_put_tile_rgba(struct pipe_surface *ps, * Get a block of Z values, converted to 32-bit range. */ void -pipe_get_tile_z(struct pipe_surface *ps, +pipe_get_tile_z(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, uint *z) { + struct pipe_screen *screen = pt->texture->screen; const uint dstStride = w; ubyte *map; uint *pDest = z; uint i, j; - if (pipe_clip_tile(x, y, &w, &h, ps)) + if (pipe_clip_tile(x, y, &w, &h, pt)) return; - map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); + map = (ubyte *)screen->transfer_map(screen, pt); if (!map) { assert(0); return; } - switch (ps->format) { + switch (pt->format) { case PIPE_FORMAT_Z32_UNORM: { - const uint *pSrc - = (const uint *)(map + y * ps->stride + x*4); + const uint *ptrc + = (const uint *)(map + y * pt->stride + x*4); for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, 4 * w); + memcpy(pDest, ptrc, 4 * w); pDest += dstStride; - pSrc += ps->stride/4; + ptrc += pt->stride/4; } } break; case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: { - const uint *pSrc - = (const uint *)(map + y * ps->stride + x*4); + const uint *ptrc + = (const uint *)(map + y * pt->stride + x*4); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 24-bit Z to 32-bit Z */ - pDest[j] = (pSrc[j] << 8) | (pSrc[j] & 0xff); + pDest[j] = (ptrc[j] << 8) | (ptrc[j] & 0xff); } pDest += dstStride; - pSrc += ps->stride/4; + ptrc += pt->stride/4; } } break; case PIPE_FORMAT_Z16_UNORM: { - const ushort *pSrc - = (const ushort *)(map + y * ps->stride + x*2); + const ushort *ptrc + = (const ushort *)(map + y * pt->stride + x*2); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 16-bit Z to 32-bit Z */ - pDest[j] = (pSrc[j] << 16) | pSrc[j]; + pDest[j] = (ptrc[j] << 16) | ptrc[j]; } pDest += dstStride; - pSrc += ps->stride/2; + ptrc += pt->stride/2; } } break; @@ -1127,64 +1149,65 @@ pipe_get_tile_z(struct pipe_surface *ps, assert(0); } - pipe_surface_unmap(ps); + screen->transfer_unmap(screen, pt); } void -pipe_put_tile_z(struct pipe_surface *ps, +pipe_put_tile_z(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, const uint *zSrc) { + struct pipe_screen *screen = pt->texture->screen; const uint srcStride = w; - const uint *pSrc = zSrc; + const uint *ptrc = zSrc; ubyte *map; uint i, j; - if (pipe_clip_tile(x, y, &w, &h, ps)) + if (pipe_clip_tile(x, y, &w, &h, pt)) return; - map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); + map = (ubyte *)screen->transfer_map(screen, pt); if (!map) { assert(0); return; } - switch (ps->format) { + switch (pt->format) { case PIPE_FORMAT_Z32_UNORM: { - uint *pDest = (uint *) (map + y * ps->stride + x*4); + uint *pDest = (uint *) (map + y * pt->stride + x*4); for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, 4 * w); - pDest += ps->stride/4; - pSrc += srcStride; + memcpy(pDest, ptrc, 4 * w); + pDest += pt->stride/4; + ptrc += srcStride; } } break; case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: { - uint *pDest = (uint *) (map + y * ps->stride + x*4); + uint *pDest = (uint *) (map + y * pt->stride + x*4); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 24-bit Z (0 stencil) */ - pDest[j] = pSrc[j] >> 8; + pDest[j] = ptrc[j] >> 8; } - pDest += ps->stride/4; - pSrc += srcStride; + pDest += pt->stride/4; + ptrc += srcStride; } } break; case PIPE_FORMAT_Z16_UNORM: { - ushort *pDest = (ushort *) (map + y * ps->stride + x*2); + ushort *pDest = (ushort *) (map + y * pt->stride + x*2); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 16-bit Z */ - pDest[j] = pSrc[j] >> 16; + pDest[j] = ptrc[j] >> 16; } - pDest += ps->stride/2; - pSrc += srcStride; + pDest += pt->stride/2; + ptrc += srcStride; } } break; @@ -1192,7 +1215,7 @@ pipe_put_tile_z(struct pipe_surface *ps, assert(0); } - pipe_surface_unmap(ps); + screen->transfer_unmap(screen, pt); } diff --git a/src/gallium/auxiliary/util/u_tile.h b/src/gallium/auxiliary/util/u_tile.h index a8ac805308..1453af38b8 100644 --- a/src/gallium/auxiliary/util/u_tile.h +++ b/src/gallium/auxiliary/util/u_tile.h @@ -30,24 +30,24 @@ #include "pipe/p_compiler.h" -struct pipe_surface; +struct pipe_transfer; /** - * Clip tile against surface dims. + * Clip tile against transfer dims. * \return TRUE if tile is totally clipped, FALSE otherwise */ static INLINE boolean -pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) +pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_transfer *pt) { - if (x >= ps->width) + if (x >= pt->width) return TRUE; - if (y >= ps->height) + if (y >= pt->height) return TRUE; - if (x + *w > ps->width) - *w = ps->width - x; - if (y + *h > ps->height) - *h = ps->height - y; + if (x + *w > pt->width) + *w = pt->width - x; + if (y + *h > pt->height) + *h = pt->height - y; return FALSE; } @@ -56,34 +56,34 @@ extern "C" { #endif void -pipe_get_tile_raw(struct pipe_surface *ps, +pipe_get_tile_raw(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, void *p, int dst_stride); void -pipe_put_tile_raw(struct pipe_surface *ps, +pipe_put_tile_raw(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, const void *p, int src_stride); void -pipe_get_tile_rgba(struct pipe_surface *ps, +pipe_get_tile_rgba(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, float *p); void -pipe_put_tile_rgba(struct pipe_surface *ps, +pipe_put_tile_rgba(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, const float *p); void -pipe_get_tile_z(struct pipe_surface *ps, +pipe_get_tile_z(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, uint *z); void -pipe_put_tile_z(struct pipe_surface *ps, +pipe_put_tile_z(struct pipe_transfer *pt, uint x, uint y, uint w, uint h, const uint *z); diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index dde2c74fa8..357d9360c9 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -35,7 +35,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #include <sys/time.h> #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include <windows.h> @@ -77,7 +77,7 @@ util_time_get_frequency(void) void util_time_get(struct util_time *t) { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) gettimeofday(&t->tv, NULL); #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) LONGLONG temp; @@ -102,7 +102,7 @@ util_time_add(const struct util_time *t1, int64_t usecs, struct util_time *t2) { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) @@ -124,7 +124,7 @@ int64_t util_time_diff(const struct util_time *t1, const struct util_time *t2) { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) return (t2->tv.tv_usec - t1->tv.tv_usec) + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) @@ -144,7 +144,7 @@ util_time_micros( void ) util_time_get(&t1); -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) return t1.tv.tv_usec + t1.tv.tv_sec*1000000LL; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) util_time_get_frequency(); @@ -166,7 +166,7 @@ static INLINE int util_time_compare(const struct util_time *t1, const struct util_time *t2) { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) if (t1->tv.tv_sec < t2->tv.tv_sec) return -1; else if(t1->tv.tv_sec > t2->tv.tv_sec) diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index 35d97d16c7..4346ce1fa4 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -38,7 +38,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #include <time.h> /* timeval */ #include <unistd.h> /* usleep */ #endif @@ -58,7 +58,7 @@ extern "C" { */ struct util_time { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) struct timeval tv; #else int64_t counter; @@ -89,7 +89,7 @@ util_time_timeout(const struct util_time *start, const struct util_time *end, const struct util_time *curr); -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #define util_time_sleep usleep #else void diff --git a/src/gallium/auxiliary/util/u_timed_winsys.c b/src/gallium/auxiliary/util/u_timed_winsys.c index f237e12d73..77b2a3a1c8 100644 --- a/src/gallium/auxiliary/util/u_timed_winsys.c +++ b/src/gallium/auxiliary/util/u_timed_winsys.c @@ -29,6 +29,7 @@ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com> */ +#include "pipe/p_state.h" #include "pipe/internal/p_winsys_screen.h" #include "u_timed_winsys.h" #include "util/u_memory.h" @@ -178,13 +179,13 @@ timed_buffer_unmap(struct pipe_winsys *winsys, static void -timed_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) +timed_buffer_destroy(struct pipe_buffer *buf) { + struct pipe_winsys *winsys = buf->screen->winsys; struct pipe_winsys *backend = timed_winsys(winsys)->backend; uint64_t start = time_start(); - backend->buffer_destroy( backend, buf ); + backend->buffer_destroy( buf ); time_finish(winsys, start, 4, __FUNCTION__); } diff --git a/src/gallium/drivers/Makefile b/src/gallium/drivers/Makefile index 6161cb6ff8..9fe9b2c11d 100644 --- a/src/gallium/drivers/Makefile +++ b/src/gallium/drivers/Makefile @@ -1,20 +1,12 @@ +# src/gallium/drivers/Makefile TOP = ../../.. include $(TOP)/configs/current +SUBDIRS = $(GALLIUM_DRIVERS_DIRS) -SUBDIRS = $(GALLIUM_DRIVER_DIRS) - - -default: subdirs - - -subdirs: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - - -clean: - rm -f `find . -name \*.[oa]` diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c index c2e276988c..edc06747ac 100644 --- a/src/gallium/drivers/cell/ppu/cell_clear.c +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -70,18 +70,12 @@ void cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { - struct pipe_screen *screen = pipe->screen; struct cell_context *cell = cell_context(pipe); uint surfIndex; if (cell->dirty) cell_update_derived(cell); - - if (!cell->cbuf_map[0]) - cell->cbuf_map[0] = screen->surface_map(screen, ps, - PIPE_BUFFER_USAGE_GPU_WRITE); - if (ps == cell->framebuffer.zsbuf) { /* clear z/stencil buffer */ surfIndex = 1; diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index eb1397bb3f..5c3188e7f9 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -130,9 +130,6 @@ struct cell_context ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; ubyte *zsbuf_map; - struct pipe_surface *tex_surf; - uint *tex_map; - uint dirty; uint dirty_textures; /* bitmask of texture units */ uint dirty_samplers; /* bitmask of sampler units */ diff --git a/src/gallium/drivers/cell/ppu/cell_fence.c b/src/gallium/drivers/cell/ppu/cell_fence.c index 867b5dcaa0..13125a9fa3 100644 --- a/src/gallium/drivers/cell/ppu/cell_fence.c +++ b/src/gallium/drivers/cell/ppu/cell_fence.c @@ -96,7 +96,7 @@ cell_add_buffer_to_list(struct cell_context *cell, struct cell_buffer_node *node = CALLOC_STRUCT(cell_buffer_node); /* create new list node which references the buffer, insert at head */ if (node) { - pipe_buffer_reference(ps, &node->buffer, buffer); + pipe_buffer_reference(&node->buffer, buffer); node->next = list->head; list->head = node; } @@ -127,10 +127,10 @@ cell_free_fenced_buffers(struct cell_context *cell, pipe_buffer_unmap(ps, node->buffer); #if 0 printf("Unref buffer %p\n", node->buffer); - if (node->buffer->refcount == 1) + if (node->buffer->reference.count == 1) printf(" Delete!\n"); #endif - pipe_buffer_reference(ps, &node->buffer, NULL); + pipe_buffer_reference(&node->buffer, NULL); FREE(node); node = next; } @@ -153,16 +153,12 @@ cell_add_fenced_textures(struct cell_context *cell) for (i = 0; i < cell->num_textures; i++) { struct cell_texture *ct = cell->texture[i]; if (ct) { - uint level; - for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) { - if (ct->tiled_buffer[level]) { #if 0 - printf("Adding texture %p buffer %p to list\n", - ct, ct->tiled_buffer[level]); + printf("Adding texture %p buffer %p to list\n", + ct, ct->tiled_buffer[level]); #endif - cell_add_buffer_to_list(cell, list, ct->tiled_buffer[level]); - } - } + if (ct->buffer) + cell_add_buffer_to_list(cell, list, ct->buffer); } } } diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index ca358ed031..ccd0fef6e8 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -253,10 +253,13 @@ cell_set_sampler_textures(struct pipe_context *pipe, assert(num <= CELL_MAX_SAMPLERS); for (i = 0; i < CELL_MAX_SAMPLERS; i++) { - struct pipe_texture *new_tex = i < num ? texture[i] : NULL; - if ((struct pipe_texture *) cell->texture[i] != new_tex) { + struct cell_texture *new_tex = cell_texture(i < num ? texture[i] : NULL); + struct cell_texture *old_tex = cell->texture[i]; + if (old_tex != new_tex) { + pipe_texture_reference((struct pipe_texture **) &cell->texture[i], - new_tex); + (struct pipe_texture *) new_tex); + changed |= (1 << i); } } @@ -270,6 +273,70 @@ cell_set_sampler_textures(struct pipe_context *pipe, } +/** + * Map color and z/stencil framebuffer surfaces. + */ +static void +cell_map_surfaces(struct cell_context *cell) +{ + struct pipe_screen *screen = cell->pipe.screen; + uint i; + + for (i = 0; i < 1; i++) { + struct pipe_surface *ps = cell->framebuffer.cbufs[i]; + if (ps) { + struct cell_texture *ct = cell_texture(ps->texture); + cell->cbuf_map[i] = screen->buffer_map(screen, + ct->buffer, + (PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE)); + } + } + + { + struct pipe_surface *ps = cell->framebuffer.zsbuf; + if (ps) { + struct cell_texture *ct = cell_texture(ps->texture); + cell->zsbuf_map = screen->buffer_map(screen, + ct->buffer, + (PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE)); + } + } +} + + +/** + * Unmap color and z/stencil framebuffer surfaces. + */ +static void +cell_unmap_surfaces(struct cell_context *cell) +{ + struct pipe_screen *screen = cell->pipe.screen; + uint i; + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + struct pipe_surface *ps = cell->framebuffer.cbufs[i]; + if (ps && cell->cbuf_map[i]) { + struct cell_texture *ct = cell_texture(ps->texture); + assert(ps->texture); + assert(ct->buffer); + + screen->buffer_unmap(screen, ct->buffer); + cell->cbuf_map[i] = NULL; + } + } + + { + struct pipe_surface *ps = cell->framebuffer.zsbuf; + if (ps && cell->zsbuf_map) { + struct cell_texture *ct = cell_texture(ps->texture); + screen->buffer_unmap(screen, ct->buffer); + cell->zsbuf_map = NULL; + } + } +} + static void cell_set_framebuffer_state(struct pipe_context *pipe, @@ -278,24 +345,10 @@ cell_set_framebuffer_state(struct pipe_context *pipe, struct cell_context *cell = cell_context(pipe); if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) { - struct pipe_surface *csurf = fb->cbufs[0]; - struct pipe_surface *zsurf = fb->zsbuf; uint i; - uint flags = (PIPE_BUFFER_USAGE_GPU_WRITE | - PIPE_BUFFER_USAGE_GPU_READ); /* unmap old surfaces */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { - if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { - pipe_surface_unmap(cell->framebuffer.cbufs[i]); - cell->cbuf_map[i] = NULL; - } - } - - if (cell->framebuffer.zsbuf && cell->zsbuf_map) { - pipe_surface_unmap(cell->framebuffer.zsbuf); - cell->zsbuf_map = NULL; - } + cell_unmap_surfaces(cell); /* Finish any pending rendering to the current surface before * installing a new surface! @@ -314,11 +367,7 @@ cell_set_framebuffer_state(struct pipe_context *pipe, pipe_surface_reference(&cell->framebuffer.zsbuf, fb->zsbuf); /* map new surfaces */ - if (csurf) - cell->cbuf_map[0] = pipe_surface_map(csurf, flags); - - if (zsurf) - cell->zsbuf_map = pipe_surface_map(zsurf, flags); + cell_map_surfaces(cell); cell->dirty |= CELL_NEW_FRAMEBUFFER; } diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index ff529fe22c..9479c0898f 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -287,19 +287,23 @@ cell_emit_state(struct cell_context *cell) for (i = 0;i < CELL_MAX_SAMPLERS; i++) { if (cell->dirty_textures & (1 << i)) { STATIC_ASSERT(sizeof(struct cell_command_texture) % 16 == 0); - struct cell_command_texture *texture - = (struct cell_command_texture *)cell_batch_alloc16(cell, sizeof(*texture)); + struct cell_command_texture *texture = + (struct cell_command_texture *) + cell_batch_alloc16(cell, sizeof(*texture)); + texture->opcode[0] = CELL_CMD_STATE_TEXTURE; texture->unit = i; if (cell->texture[i]) { + struct cell_texture *ct = cell->texture[i]; uint level; for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) { - texture->start[level] = cell->texture[i]->tiled_mapped[level]; - texture->width[level] = cell->texture[i]->base.width[level]; - texture->height[level] = cell->texture[i]->base.height[level]; - texture->depth[level] = cell->texture[i]->base.depth[level]; + texture->start[level] = (ct->mapped + + ct->level_offset[level]); + texture->width[level] = ct->base.width[level]; + texture->height[level] = ct->base.height[level]; + texture->depth[level] = ct->base.depth[level]; } - texture->target = cell->texture[i]->base.target; + texture->target = ct->base.target; } else { uint level; diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index bf517ea563..6568c784fe 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -193,9 +193,7 @@ cell_set_constant_buffer(struct pipe_context *pipe, draw_flush(cell->draw); /* note: reference counting */ - pipe_buffer_reference(pipe->screen, - &cell->constants[shader].buffer, - buf->buffer); + pipe_buffer_reference(&cell->constants[shader].buffer, buf->buffer); if (shader == PIPE_SHADER_VERTEX) cell->dirty |= CELL_NEW_VS_CONSTANTS; diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index c9203fee08..ffb8595d82 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -30,9 +30,21 @@ #include "cell_surface.h" +static void +cell_surface_copy(struct pipe_context *pipe, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + util_surface_copy(pipe, FALSE, + dest, destx, desty, + src, srcx, srcy, + width, height); +} + void cell_init_surface_functions(struct cell_context *cell) { - cell->pipe.surface_copy = util_surface_copy; + cell->pipe.surface_copy = cell_surface_copy; cell->pipe.surface_fill = util_surface_fill; } diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 9ba995ab7d..e26594448f 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -62,7 +62,7 @@ cell_texture_layout(struct cell_texture *ct) ct->buffer_size = 0; - for ( level = 0 ; level <= pt->last_level ; level++ ) { + for (level = 0; level <= pt->last_level; level++) { unsigned size; unsigned w_tile, h_tile; @@ -90,7 +90,7 @@ cell_texture_layout(struct cell_texture *ct) ct->buffer_size += size; - width = minify(width); + width = minify(width); height = minify(height); depth = minify(depth); } @@ -101,18 +101,17 @@ static struct pipe_texture * cell_texture_create(struct pipe_screen *screen, const struct pipe_texture *templat) { - struct pipe_winsys *ws = screen->winsys; struct cell_texture *ct = CALLOC_STRUCT(cell_texture); if (!ct) return NULL; ct->base = *templat; - ct->base.refcount = 1; + pipe_reference_init(&ct->base.reference, 1); ct->base.screen = screen; cell_texture_layout(ct); - ct->buffer = ws->buffer_create(ws, 32, PIPE_BUFFER_USAGE_PIXEL, + ct->buffer = screen->buffer_create(screen, 32, PIPE_BUFFER_USAGE_PIXEL, ct->buffer_size); if (!ct->buffer) { @@ -125,42 +124,18 @@ cell_texture_create(struct pipe_screen *screen, static void -cell_texture_release(struct pipe_screen *screen, - struct pipe_texture **pt) +cell_texture_destroy(struct pipe_texture *pt) { - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - /* Delete this texture now. - * But note that the underlying pipe_buffer may linger... - */ - struct cell_texture *ct = cell_texture(*pt); - uint i; + struct cell_texture *ct = cell_texture(pt); - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) ct); - */ + if (ct->mapped) { + pipe_buffer_unmap(ct->buffer->screen, ct->buffer); + ct->mapped = NULL; + } - pipe_buffer_reference(screen, &ct->buffer, NULL); + pipe_buffer_reference(&ct->buffer, NULL); - for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) { - /* Unreference the tiled image buffer. - * It may not actually be deleted until a fence is hit. - */ - if (ct->tiled_buffer[i]) { - ct->tiled_mapped[i] = NULL; - pipe_buffer_reference(screen, &ct->tiled_buffer[i], NULL); - } - } - - FREE(ct); - } - *pt = NULL; + FREE(ct); } @@ -294,103 +269,6 @@ untwiddle_image_uint(uint w, uint h, uint tile_size, uint *dst, } -/** - * Convert linear texture image data to tiled format for SPU usage. - */ -static void -cell_twiddle_texture(struct pipe_screen *screen, - struct pipe_surface *surface) -{ - struct cell_texture *ct = cell_texture(surface->texture); - const uint level = surface->level; - const uint texWidth = ct->base.width[level]; - const uint texHeight = ct->base.height[level]; - const uint bufWidth = align(texWidth, TILE_SIZE); - const uint bufHeight = align(texHeight, TILE_SIZE); - const void *map = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_READ); - const uint *src = (const uint *) map; - - switch (ct->base.format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - { - int numFaces = ct->base.target == PIPE_TEXTURE_CUBE ? 6 : 1; - int offset = bufWidth * bufHeight * 4 * surface->face; - uint *dst; - - if (!ct->tiled_buffer[level]) { - /* allocate buffer for tiled data now */ - struct pipe_winsys *ws = screen->winsys; - uint bytes = bufWidth * bufHeight * 4 * numFaces; - ct->tiled_buffer[level] = - ws->buffer_create(ws, 16, PIPE_BUFFER_USAGE_PIXEL, bytes); - /* and map it */ - ct->tiled_mapped[level] = - ws->buffer_map(ws, ct->tiled_buffer[level], - PIPE_BUFFER_USAGE_GPU_READ); - } - dst = (uint *) ((ubyte *) ct->tiled_mapped[level] + offset); - - twiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst, - surface->stride, src); - } - break; - default: - printf("Cell: twiddle unsupported texture format %s\n", - pf_name(ct->base.format)); - } - - screen->surface_unmap(screen, surface); -} - - -/** - * Convert SPU tiled texture image data to linear format for app usage. - */ -static void -cell_untwiddle_texture(struct pipe_screen *screen, - struct pipe_surface *surface) -{ - struct cell_texture *ct = cell_texture(surface->texture); - const uint level = surface->level; - const uint texWidth = ct->base.width[level]; - const uint texHeight = ct->base.height[level]; - const void *map = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_READ); - const uint *src = (const uint *) ((const ubyte *) map + surface->offset); - - switch (ct->base.format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - { - int numFaces = ct->base.target == PIPE_TEXTURE_CUBE ? 6 : 1; - int offset = surface->stride * texHeight * 4 * surface->face; - uint *dst; - - if (!ct->untiled_data[level]) { - ct->untiled_data[level] = - align_malloc(surface->stride * texHeight * 4 * numFaces, 16); - } - - dst = (uint *) ((ubyte *) ct->untiled_data[level] + offset); - - untwiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst, - surface->stride, src); - } - break; - default: - { - ct->untiled_data[level] = NULL; - printf("Cell: untwiddle unsupported texture format %s\n", - pf_name(ct->base.format)); - } - } - - screen->surface_unmap(screen, surface); -} - - static struct pipe_surface * cell_get_tex_surface(struct pipe_screen *screen, struct pipe_texture *pt, @@ -402,130 +280,223 @@ cell_get_tex_surface(struct pipe_screen *screen, ps = CALLOC_STRUCT(pipe_surface); if (ps) { - ps->refcount = 1; + pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->block = pt->block; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = ct->stride[level]; ps->offset = ct->level_offset[level]; - ps->usage = usage; - /* XXX may need to override usage flags (see sp_texture.c) */ - - pipe_texture_reference(&ps->texture, pt); + ps->usage = usage; ps->face = face; ps->level = level; ps->zslice = zslice; - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - ps->nblocksy * - ps->stride; + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset += face * pt->nblocksy[level] * ct->stride[level]; + } + else if (pt->target == PIPE_TEXTURE_3D) { + ps->offset += zslice * pt->nblocksy[level] * ct->stride[level]; } else { assert(face == 0); assert(zslice == 0); } - - if (ps->usage & PIPE_BUFFER_USAGE_CPU_READ) { - /* convert from tiled to linear layout */ - cell_untwiddle_texture(screen, ps); - } } return ps; } static void -cell_tex_surface_release(struct pipe_screen *screen, - struct pipe_surface **s) +cell_tex_surface_destroy(struct pipe_surface *surf) { - struct cell_texture *ct = cell_texture((*s)->texture); - const uint level = (*s)->level; - struct pipe_surface *surf = *s; - - if ((surf->usage & PIPE_BUFFER_USAGE_CPU_READ) && (ct->untiled_data[level])) - { - align_free(ct->untiled_data[level]); - ct->untiled_data[level] = NULL; - } + pipe_texture_reference(&surf->texture, NULL); + FREE(surf); +} - if ((ct->base.tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) && - (surf->usage & PIPE_BUFFER_USAGE_CPU_WRITE)) { - /* convert from linear to tiled layout */ - cell_twiddle_texture(screen, surf); + +/** + * Create new pipe_transfer object. + * This is used by the user to put tex data into a texture (and get it + * back out for glGetTexImage). + */ +static struct pipe_transfer * +cell_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct cell_texture *ct = cell_texture(texture); + struct cell_transfer *ctrans; + + assert(texture); + assert(level <= texture->last_level); + + ctrans = CALLOC_STRUCT(cell_transfer); + if (ctrans) { + struct pipe_transfer *pt = &ctrans->base; + pipe_texture_reference(&pt->texture, texture); + pt->format = texture->format; + pt->block = texture->block; + pt->x = x; + pt->y = y; + pt->width = w; + pt->height = h; + pt->nblocksx = texture->nblocksx[level]; + pt->nblocksy = texture->nblocksy[level]; + pt->stride = ct->stride[level]; + pt->usage = usage; + pt->face = face; + pt->level = level; + pt->zslice = zslice; + + ctrans->offset = ct->level_offset[level]; + + if (texture->target == PIPE_TEXTURE_CUBE) { + ctrans->offset += face * pt->nblocksy * pt->stride; + } + else if (texture->target == PIPE_TEXTURE_3D) { + ctrans->offset += zslice * pt->nblocksy * pt->stride; + } + else { + assert(face == 0); + assert(zslice == 0); + } + return pt; } + return NULL; +} - /* XXX if done rendering to teximage, re-tile */ - if (--surf->refcount == 0) { - pipe_texture_reference(&surf->texture, NULL); - FREE(surf); - } - *s = NULL; +static void +cell_tex_transfer_destroy(struct pipe_transfer *t) +{ + struct cell_transfer *transfer = cell_transfer(t); + /* Effectively do the texture_update work here - if texture images + * needed post-processing to put them into hardware layout, this is + * where it would happen. For cell, nothing to do. + */ + assert (transfer->base.texture); + pipe_texture_reference(&transfer->base.texture, NULL); + FREE(transfer); } +/** + * Return pointer to texture image data in linear layout. + */ static void * -cell_surface_map(struct pipe_screen *screen, - struct pipe_surface *surface, - unsigned flags) +cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) { - ubyte *map; - struct cell_texture *ct = cell_texture(surface->texture); - const uint level = surface->level; + struct cell_transfer *ctrans = cell_transfer(transfer); + struct pipe_texture *pt = transfer->texture; + struct cell_texture *ct = cell_texture(pt); + const uint level = ctrans->base.level; + const uint texWidth = pt->width[level]; + const uint texHeight = pt->height[level]; + const uint stride = ct->stride[level]; + unsigned flags = 0x0; + unsigned size; - assert(ct); + assert(transfer->texture); -#if 0 - if (flags & ~surface->usage) { - assert(0); - return NULL; + if (transfer->usage != PIPE_TRANSFER_READ) { + flags |= PIPE_BUFFER_USAGE_CPU_WRITE; } -#endif - map = pipe_buffer_map( screen, ct->buffer, flags ); - if (map == NULL) { - return NULL; + if (transfer->usage != PIPE_TRANSFER_WRITE) { + flags |= PIPE_BUFFER_USAGE_CPU_READ; + } + + if (!ct->mapped) { + /* map now */ + ct->mapped = pipe_buffer_map(screen, ct->buffer, flags); } - else { - if ((surface->usage & PIPE_BUFFER_USAGE_CPU_READ) && - (ct->untiled_data[level])) { - return (void *) ((ubyte *) ct->untiled_data[level] + surface->offset); + + /* + * Create a buffer of ordinary memory for the linear texture. + * This is the memory that the user will read/write. + */ + size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size; + + ctrans->map = align_malloc(size, 16); + if (!ctrans->map) + return NULL; /* out of memory */ + + if (transfer->usage & PIPE_TRANSFER_READ) { + /* need to untwiddle the texture to make a linear version */ + const uint bpp = pf_get_size(ct->base.format); + if (bpp == 4) { + const uint *src = (uint *) (ct->mapped + ctrans->offset); + uint *dst = ctrans->map; + untwiddle_image_uint(texWidth, texHeight, TILE_SIZE, + dst, stride, src); } else { - return (void *) (map + surface->offset); + // xxx fix } } + + return ctrans->map; } +/** + * Called when user is done reading/writing texture data. + * If new data was written, this is where we convert the linear data + * to tiled data. + */ static void -cell_surface_unmap(struct pipe_screen *screen, - struct pipe_surface *surface) +cell_transfer_unmap(struct pipe_screen *screen, + struct pipe_transfer *transfer) { - struct cell_texture *ct = cell_texture(surface->texture); + struct cell_transfer *ctrans = cell_transfer(transfer); + struct pipe_texture *pt = transfer->texture; + struct cell_texture *ct = cell_texture(pt); + const uint level = ctrans->base.level; + const uint texWidth = pt->width[level]; + const uint texHeight = pt->height[level]; + const uint stride = ct->stride[level]; + + if (!ct->mapped) { + /* map now */ + ct->mapped = pipe_buffer_map(screen, ct->buffer, + PIPE_BUFFER_USAGE_CPU_READ); + } - assert(ct); + if (transfer->usage & PIPE_TRANSFER_WRITE) { + /* The user wrote new texture data into the mapped buffer. + * We need to convert the new linear data into the twiddled/tiled format. + */ + const uint bpp = pf_get_size(ct->base.format); + if (bpp == 4) { + const uint *src = ctrans->map; + uint *dst = (uint *) (ct->mapped + ctrans->offset); + twiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst, stride, src); + } + else { + // xxx fix + } + } - pipe_buffer_unmap( screen, ct->buffer ); + align_free(ctrans->map); + ctrans->map = NULL; } - void cell_init_screen_texture_funcs(struct pipe_screen *screen) { screen->texture_create = cell_texture_create; - screen->texture_release = cell_texture_release; + screen->texture_destroy = cell_texture_destroy; screen->get_tex_surface = cell_get_tex_surface; - screen->tex_surface_release = cell_tex_surface_release; + screen->tex_surface_destroy = cell_tex_surface_destroy; + + screen->get_tex_transfer = cell_get_tex_transfer; + screen->tex_transfer_destroy = cell_tex_transfer_destroy; - screen->surface_map = cell_surface_map; - screen->surface_unmap = cell_surface_unmap; + screen->transfer_map = cell_transfer_map; + screen->transfer_unmap = cell_transfer_unmap; } diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h index 7018b0c9bf..3ffc0bfdb5 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.h +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -43,16 +43,23 @@ struct cell_texture unsigned long level_offset[CELL_MAX_TEXTURE_LEVELS]; unsigned long stride[CELL_MAX_TEXTURE_LEVELS]; - /* The data is held here: - */ + /** The tiled texture data is held in this buffer */ struct pipe_buffer *buffer; unsigned long buffer_size; - /** Texture data in tiled layout is held here */ - struct pipe_buffer *tiled_buffer[CELL_MAX_TEXTURE_LEVELS]; - /** Mapped, tiled texture data */ - void *tiled_mapped[CELL_MAX_TEXTURE_LEVELS]; - void *untiled_data[CELL_MAX_TEXTURE_LEVELS]; + /** The buffer above, mapped. This is the memory from which the + * SPUs will fetch texels. This texture data is in the tiled layout. + */ + ubyte *mapped; +}; + + +struct cell_transfer +{ + struct pipe_transfer base; + + unsigned long offset; + void *map; }; @@ -64,6 +71,13 @@ cell_texture(struct pipe_texture *pt) } +/** cast wrapper */ +static INLINE struct cell_transfer * +cell_transfer(struct pipe_transfer *pt) +{ + return (struct cell_transfer *) pt; +} + extern void cell_init_screen_texture_funcs(struct pipe_screen *screen); diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c index ab54e79689..cfaffb52a8 100644 --- a/src/gallium/drivers/cell/ppu/cell_vbuf.c +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c @@ -62,6 +62,7 @@ struct cell_vbuf_render uint vertex_size; /**< in bytes */ void *vertex_buffer; /**< just for debug, really */ uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ + uint vertex_buffer_size; /**< size in bytes */ }; @@ -82,24 +83,26 @@ cell_vbuf_get_vertex_info(struct vbuf_render *vbr) } -static void * +static boolean cell_vbuf_allocate_vertices(struct vbuf_render *vbr, ushort vertex_size, ushort nr_vertices) { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + unsigned size = vertex_size * nr_vertices; /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ assert(cvbr->vertex_buf == ~0); cvbr->vertex_buf = cell_get_empty_buffer(cvbr->cell); cvbr->vertex_buffer = cvbr->cell->buffer[cvbr->vertex_buf]; + cvbr->vertex_buffer_size = size; cvbr->vertex_size = vertex_size; - return cvbr->vertex_buffer; + + return cvbr->vertex_buffer != NULL; } static void -cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, - unsigned vertex_size, unsigned vertices_used) +cell_vbuf_release_vertices(struct vbuf_render *vbr) { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); struct cell_context *cell = cvbr->cell; @@ -127,11 +130,29 @@ cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, cvbr->vertex_buf = ~0; cell_flush_int(cell, 0x0); - assert(vertices == cvbr->vertex_buffer); cvbr->vertex_buffer = NULL; } +static void * +cell_vbuf_map_vertices(struct vbuf_render *vbr) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + return cvbr->vertex_buffer; +} + + +static void +cell_vbuf_unmap_vertices(struct vbuf_render *vbr, + ushort min_index, + ushort max_index ) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + assert( cvbr->vertex_buffer_size >= (max_index+1) * cvbr->vertex_size ); + /* do nothing */ +} + + static boolean cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) @@ -295,6 +316,8 @@ cell_init_vbuf(struct cell_context *cell) cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; + cell->vbuf_render->base.map_vertices = cell_vbuf_map_vertices; + cell->vbuf_render->base.unmap_vertices = cell_vbuf_unmap_vertices; cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive; cell->vbuf_render->base.draw = cell_vbuf_draw; cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices; diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index 116453b79c..3cc52301da 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -33,9 +33,10 @@ OLD_SOURCES = \ spu_vertex_shader.c -SPU_OBJECTS = $(SOURCES:.c=.o) \ +SPU_OBJECTS = $(SOURCES:.c=.o) + +SPU_ASM_OUT = $(SOURCES:.c=.s) -SPU_ASM_OUT = $(SOURCES:.c=.s) \ INCLUDE_DIRS = \ -I$(TOP)/src/mesa \ diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 0d9fcb9997..d727268475 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -29,7 +29,6 @@ * Triangle rendering within a tile. */ -#include <transpose_matrix4x4.h> #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "util/u_math.h" @@ -71,6 +70,12 @@ struct vertex_header { #define MASK_ALL 0xf +#define CHAN0 0 +#define CHAN1 1 +#define CHAN2 2 +#define CHAN3 3 + + #define DEBUG_VERTS 0 /** @@ -144,99 +149,97 @@ struct setup_stage { static struct setup_stage setup; -/** - * Evaluate attribute coefficients (plane equations) to compute - * attribute values for the four fragments in a quad. - * Eg: four colors will be computed (in AoS format). - */ -static INLINE void -eval_coeff(uint slot, float x, float y, vector float w, vector float result[4]) +static INLINE vector float +splatx(vector float v) { - switch (spu.vertex_info.attrib[slot].interp_mode) { - case INTERP_CONSTANT: - result[QUAD_TOP_LEFT] = - result[QUAD_TOP_RIGHT] = - result[QUAD_BOTTOM_LEFT] = - result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0; - break; - case INTERP_LINEAR: - { - vector float dadx = setup.coef[slot].dadx; - vector float dady = setup.coef[slot].dady; - vector float topLeft = - spu_add(setup.coef[slot].a0, - spu_add(spu_mul(spu_splats(x), dadx), - spu_mul(spu_splats(y), dady))); - - result[QUAD_TOP_LEFT] = topLeft; - result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx); - result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady); - result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady); - } - break; - case INTERP_PERSPECTIVE: - { - vector float dadx = setup.coef[slot].dadx; - vector float dady = setup.coef[slot].dady; - vector float topLeft = - spu_add(setup.coef[slot].a0, - spu_add(spu_mul(spu_splats(x), dadx), - spu_mul(spu_splats(y), dady))); - - vector float wInv = spu_re(w); /* 1.0 / w */ - - result[QUAD_TOP_LEFT] = spu_mul(topLeft, wInv); - result[QUAD_TOP_RIGHT] = spu_mul(spu_add(topLeft, dadx), wInv); - result[QUAD_BOTTOM_LEFT] = spu_mul(spu_add(topLeft, dady), wInv); - result[QUAD_BOTTOM_RIGHT] = spu_mul(spu_add(spu_add(topLeft, dadx), dady), wInv); - } - break; - case INTERP_POS: - case INTERP_NONE: - break; - default: - ASSERT(0); - } + return spu_splats(spu_extract(v, CHAN0)); } - -/** - * As above, but return 4 vectors in SOA format. - * XXX this will all be re-written someday. - */ -static INLINE void -eval_coeff_soa(uint slot, float x, float y, vector float w, vector float result[4]) +static INLINE vector float +splaty(vector float v) { - eval_coeff(slot, x, y, w, result); - _transpose_matrix4x4(result, result); + return spu_splats(spu_extract(v, CHAN1)); } +static INLINE vector float +splatz(vector float v) +{ + return spu_splats(spu_extract(v, CHAN2)); +} -/** Evalute coefficients to get Z for four pixels in a quad */ static INLINE vector float -eval_z(float x, float y) +splatw(vector float v) { - const uint slot = 0; - const float dzdx = spu_extract(setup.coef[slot].dadx, 2); - const float dzdy = spu_extract(setup.coef[slot].dady, 2); - const float topLeft = spu_extract(setup.coef[slot].a0, 2) + x * dzdx + y * dzdy; - const vector float topLeftv = spu_splats(topLeft); - const vector float derivs = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; - return spu_add(topLeftv, derivs); + return spu_splats(spu_extract(v, CHAN3)); } -/** Evalute coefficients to get W for four pixels in a quad */ -static INLINE vector float -eval_w(float x, float y) +/** + * Setup fragment shader inputs by evaluating triangle's vertex + * attribute coefficient info. + * \param x quad x pos + * \param y quad y pos + * \param fragZ returns quad Z values + * \param fragInputs returns fragment program inputs + * Note: this code could be incorporated into the fragment program + * itself to avoid the loop and switch. + */ +static void +eval_inputs(float x, float y, vector float *fragZ, vector float fragInputs[]) { - const uint slot = 0; - const float dwdx = spu_extract(setup.coef[slot].dadx, 3); - const float dwdy = spu_extract(setup.coef[slot].dady, 3); - const float topLeft = spu_extract(setup.coef[slot].a0, 3) + x * dwdx + y * dwdy; - const vector float topLeftv = spu_splats(topLeft); - const vector float derivs = (vector float) { 0.0, dwdx, dwdy, dwdx + dwdy }; - return spu_add(topLeftv, derivs); + static const vector float deltaX = (const vector float) {0, 1, 0, 1}; + static const vector float deltaY = (const vector float) {0, 0, 1, 1}; + + const uint posSlot = 0; + const vector float pos = setup.coef[posSlot].a0; + const vector float dposdx = setup.coef[posSlot].dadx; + const vector float dposdy = setup.coef[posSlot].dady; + const vector float fragX = spu_splats(x) + deltaX; + const vector float fragY = spu_splats(y) + deltaY; + vector float fragW, wInv; + uint i; + + *fragZ = splatz(pos) + fragX * splatz(dposdx) + fragY * splatz(dposdy); + fragW = splatw(pos) + fragX * splatw(dposdx) + fragY * splatw(dposdy); + wInv = spu_re(fragW); /* 1 / w */ + + /* loop over fragment program inputs */ + for (i = 0; i < spu.vertex_info.num_attribs; i++) { + uint attr = i + 1; + enum interp_mode interp = spu.vertex_info.attrib[attr].interp_mode; + + /* constant term */ + vector float a0 = setup.coef[attr].a0; + vector float r0 = splatx(a0); + vector float r1 = splaty(a0); + vector float r2 = splatz(a0); + vector float r3 = splatw(a0); + + if (interp == INTERP_LINEAR || interp == INTERP_PERSPECTIVE) { + /* linear term */ + vector float dadx = setup.coef[attr].dadx; + vector float dady = setup.coef[attr].dady; + /* Use SPU intrinsics here to get slightly better code. + * originally: r0 += fragX * splatx(dadx) + fragY * splatx(dady); + */ + r0 = spu_madd(fragX, splatx(dadx), spu_madd(fragY, splatx(dady), r0)); + r1 = spu_madd(fragX, splaty(dadx), spu_madd(fragY, splaty(dady), r1)); + r2 = spu_madd(fragX, splatz(dadx), spu_madd(fragY, splatz(dady), r2)); + r3 = spu_madd(fragX, splatw(dadx), spu_madd(fragY, splatw(dady), r3)); + if (interp == INTERP_PERSPECTIVE) { + /* perspective term */ + r0 *= wInv; + r1 *= wInv; + r2 *= wInv; + r3 *= wInv; + } + } + fragInputs[CHAN0] = r0; + fragInputs[CHAN1] = r1; + fragInputs[CHAN2] = r2; + fragInputs[CHAN3] = r3; + fragInputs += 4; + } } @@ -262,19 +265,11 @@ emit_quad( int x, int y, mask_t mask) * Run fragment shader, execute per-fragment ops, update fb/tile. */ vector float inputs[4*4], outputs[2*4]; - vector float fragZ = eval_z((float) x, (float) y); - vector float fragW = eval_w((float) x, (float) y); vector unsigned int kill_mask; + vector float fragZ; + + eval_inputs((float) x, (float) y, &fragZ, inputs); - /* setup inputs */ -#if 0 - eval_coeff_soa(1, (float) x, (float) y, fragW, inputs); -#else - uint i; - for (i = 0; i < spu.vertex_info.num_attribs; i++) { - eval_coeff_soa(i+1, (float) x, (float) y, fragW, inputs + i * 4); - } -#endif ASSERT(spu.fragment_program); ASSERT(spu.fragment_ops); diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c index b8a0d4a265..af25dd3718 100644 --- a/src/gallium/drivers/cell/spu/spu_util.c +++ b/src/gallium/drivers/cell/spu/spu_util.c @@ -1,7 +1,7 @@ #include "cell/common.h" #include "pipe/p_shader_tokens.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "tgsi/tgsi_parse.h" //#include "tgsi_build.h" #include "tgsi/tgsi_util.h" diff --git a/src/gallium/drivers/failover/Makefile b/src/gallium/drivers/failover/Makefile index f08b8df07a..dfb7f5dcf6 100644 --- a/src/gallium/drivers/failover/Makefile +++ b/src/gallium/drivers/failover/Makefile @@ -9,6 +9,3 @@ C_SOURCES = \ fo_context.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index 0742b27b8f..fcad717cf8 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -145,7 +145,7 @@ struct pipe_context *failover_create( struct pipe_context *hw, #if 0 failover->pipe.texture_create = hw->texture_create; - failover->pipe.texture_release = hw->texture_release; + failover->pipe.texture_destroy = hw->texture_destroy; failover->pipe.get_tex_surface = hw->get_tex_surface; failover->pipe.texture_update = hw->texture_update; #endif diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c index 6a79706632..c8eb926299 100644 --- a/src/gallium/drivers/failover/fo_state.c +++ b/src/gallium/drivers/failover/fo_state.c @@ -28,8 +28,6 @@ /* Authors: Keith Whitwell <keith@tungstengraphics.com> */ -#include "pipe/p_inlines.h" - #include "fo_context.h" diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index 41a61a0020..12821c5a76 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -26,6 +26,3 @@ C_SOURCES = \ i915_surface.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index a300b61c3b..e08582efab 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -31,7 +31,7 @@ #include "i915_debug.h" #include "i915_batch.h" #include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" static void diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index d92bdc1bc6..961c1bf213 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -321,16 +321,27 @@ static uint translate_tex_src_target(struct i915_fp_compile *p, uint tex) { switch (tex) { + case TGSI_TEXTURE_SHADOW1D: + /* fall-through */ case TGSI_TEXTURE_1D: return D0_SAMPLE_TYPE_2D; + + case TGSI_TEXTURE_SHADOW2D: + /* fall-through */ case TGSI_TEXTURE_2D: return D0_SAMPLE_TYPE_2D; + + case TGSI_TEXTURE_SHADOWRECT: + /* fall-through */ case TGSI_TEXTURE_RECT: return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_3D: return D0_SAMPLE_TYPE_VOLUME; + case TGSI_TEXTURE_CUBE: return D0_SAMPLE_TYPE_CUBE; + default: i915_program_error(p, "TexSrc type"); return 0; diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index f49f6d6ed1..9bdd91f288 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -40,7 +40,7 @@ #include "draw/draw_context.h" #include "draw/draw_vbuf.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" #include "util/u_math.h" @@ -62,7 +62,7 @@ struct i915_vbuf_render { struct i915_context *i915; /** Vertex size in bytes */ - unsigned vertex_size; + size_t vertex_size; /** Software primitive */ unsigned prim; @@ -79,6 +79,7 @@ struct i915_vbuf_render { size_t vbo_offset; void *vbo_ptr; size_t vbo_alloc_size; + size_t vbo_max_used; }; @@ -108,7 +109,7 @@ i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) } -static void * +static boolean i915_vbuf_render_allocate_vertices( struct vbuf_render *render, ushort vertex_size, ushort nr_vertices ) @@ -124,7 +125,8 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, if (i915_render->vbo_size > size + i915_render->vbo_offset && !i915->vbo_flushed) { } else { i915->vbo_flushed = 0; - pipe_buffer_reference(screen, &i915_render->vbo, NULL); + if (i915_render->vbo) + pipe_buffer_reference(&i915_render->vbo, NULL); } if (!i915_render->vbo) { @@ -134,19 +136,49 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, 64, I915_BUFFER_USAGE_LIT_VERTEX, i915_render->vbo_size); - i915_render->vbo_ptr = pipe_buffer_map(screen, - i915_render->vbo, - PIPE_BUFFER_USAGE_CPU_WRITE); - pipe_buffer_unmap(screen, i915_render->vbo); + } + i915_render->vertex_size = vertex_size; i915->vbo = i915_render->vbo; i915->vbo_offset = i915_render->vbo_offset; i915->dirty |= I915_NEW_VBO; + if (!i915_render->vbo) + return FALSE; + return TRUE; +} + + +static void * +i915_vbuf_render_map_vertices( struct vbuf_render *render ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + struct pipe_screen *screen = i915->pipe.screen; + + if (i915->vbo_flushed) + debug_printf("%s bad vbo flush occured stalling on hw\n"); + + i915_render->vbo_ptr = pipe_buffer_map(screen, + i915_render->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); + return (unsigned char *)i915_render->vbo_ptr + i915->vbo_offset; } +static void +i915_vbuf_render_unmap_vertices( struct vbuf_render *render, + ushort min_index, + ushort max_index ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + struct pipe_screen *screen = i915->pipe.screen; + + i915_render->vbo_max_used = MAX2(i915_render->vbo_max_used, i915_render->vertex_size * (max_index + 1)); + pipe_buffer_unmap(screen, i915_render->vbo); +} static boolean i915_vbuf_render_set_primitive( struct vbuf_render *render, @@ -454,18 +486,15 @@ out: static void -i915_vbuf_render_release_vertices( struct vbuf_render *render, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ) +i915_vbuf_render_release_vertices( struct vbuf_render *render ) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - size_t size = (size_t)vertex_size * (size_t)vertices_used; assert(i915->vbo); - i915_render->vbo_offset += size; + i915_render->vbo_offset += i915_render->vbo_max_used; + i915_render->vbo_max_used = 0; i915->vbo = NULL; i915->dirty |= I915_NEW_VBO; } @@ -499,6 +528,8 @@ i915_vbuf_render_create( struct i915_context *i915 ) i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info; i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; + i915_render->base.map_vertices = i915_vbuf_render_map_vertices; + i915_render->base.unmap_vertices = i915_vbuf_render_unmap_vertices; i915_render->base.set_primitive = i915_vbuf_render_set_primitive; i915_render->base.draw = i915_vbuf_render_draw; i915_render->base.draw_arrays = i915_vbuf_render_draw_arrays; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 39e48105b3..f4aa8e60d8 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -36,6 +36,7 @@ #include "i915_context.h" #include "i915_screen.h" #include "i915_texture.h" +#include "i915_winsys.h" static const char * @@ -204,17 +205,71 @@ i915_destroy_screen( struct pipe_screen *screen ) } +static struct pipe_transfer* +i915_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, unsigned x, unsigned y, + unsigned w, unsigned h) +{ + struct i915_texture *tex = (struct i915_texture *)texture; + struct i915_transfer *trans; + unsigned offset; /* in bytes */ + + if (texture->target == PIPE_TEXTURE_CUBE) { + offset = tex->image_offset[level][face]; + } + else if (texture->target == PIPE_TEXTURE_3D) { + offset = tex->image_offset[level][zslice]; + } + else { + offset = tex->image_offset[level][0]; + assert(face == 0); + assert(zslice == 0); + } + + trans = CALLOC_STRUCT(i915_transfer); + if (trans) { + pipe_texture_reference(&trans->base.texture, texture); + trans->base.format = trans->base.format; + trans->base.width = w; + trans->base.height = h; + trans->base.block = texture->block; + trans->base.nblocksx = texture->nblocksx[level]; + trans->base.nblocksy = texture->nblocksy[level]; + trans->base.stride = tex->stride; + trans->offset = offset; + trans->base.usage = usage; + } + return &trans->base; +} + +static void +i915_tex_transfer_destroy(struct pipe_transfer *trans) +{ + pipe_texture_reference(&trans->texture, NULL); + FREE(trans); +} + static void * -i915_surface_map( struct pipe_screen *screen, - struct pipe_surface *surface, - unsigned flags ) +i915_transfer_map( struct pipe_screen *screen, + struct pipe_transfer *transfer ) { - struct i915_texture *tex = (struct i915_texture *)surface->texture; - char *map = pipe_buffer_map( screen, tex->buffer, flags ); + struct i915_texture *tex = (struct i915_texture *)transfer->texture; + char *map; + unsigned flags = 0; + + if (transfer->usage != PIPE_TRANSFER_WRITE) + flags |= PIPE_BUFFER_USAGE_CPU_READ; + + if (transfer->usage != PIPE_TRANSFER_READ) + flags |= PIPE_BUFFER_USAGE_CPU_WRITE; + + map = pipe_buffer_map( screen, tex->buffer, flags ); if (map == NULL) return NULL; - if (surface->texture && + if (transfer->texture && (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) { /* Do something to notify contexts of a texture change. @@ -222,14 +277,16 @@ i915_surface_map( struct pipe_screen *screen, /* i915_screen(screen)->timestamp++; */ } - return map + surface->offset; + return map + i915_transfer(transfer)->offset + + transfer->y / transfer->block.height * transfer->stride + + transfer->x / transfer->block.width * transfer->block.size; } static void -i915_surface_unmap(struct pipe_screen *screen, - struct pipe_surface *surface) +i915_transfer_unmap(struct pipe_screen *screen, + struct pipe_transfer *transfer) { - struct i915_texture *tex = (struct i915_texture *)surface->texture; + struct i915_texture *tex = (struct i915_texture *)transfer->texture; pipe_buffer_unmap( screen, tex->buffer ); } @@ -278,8 +335,10 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) i915screen->screen.get_param = i915_get_param; i915screen->screen.get_paramf = i915_get_paramf; i915screen->screen.is_format_supported = i915_is_format_supported; - i915screen->screen.surface_map = i915_surface_map; - i915screen->screen.surface_unmap = i915_surface_unmap; + i915screen->screen.get_tex_transfer = i915_get_tex_transfer; + i915screen->screen.tex_transfer_destroy = i915_tex_transfer_destroy; + i915screen->screen.transfer_map = i915_transfer_map; + i915screen->screen.transfer_unmap = i915_transfer_unmap; i915_init_screen_texture_functions(&i915screen->screen); u_simple_screen_init(&i915screen->screen); diff --git a/src/gallium/drivers/i915simple/i915_screen.h b/src/gallium/drivers/i915simple/i915_screen.h index 73b0ff05ce..5284c32595 100644 --- a/src/gallium/drivers/i915simple/i915_screen.h +++ b/src/gallium/drivers/i915simple/i915_screen.h @@ -50,16 +50,29 @@ struct i915_screen }; -/** cast wrapper */ +/** + * Subclass of pipe_transfer + */ +struct i915_transfer +{ + struct pipe_transfer base; + + unsigned offset; +}; + + +/** cast wrappers */ static INLINE struct i915_screen * i915_screen(struct pipe_screen *pscreen) { return (struct i915_screen *) pscreen; } - -extern struct pipe_screen * -i915_create_screen(struct pipe_winsys *winsys, uint pci_id); +static INLINE struct i915_transfer * +i915_transfer( struct pipe_transfer *transfer ) +{ + return (struct i915_transfer *)transfer; +} #ifdef __cplusplus diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 6558cf1c3e..1e1fb968b4 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -211,11 +211,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; if (cbuf_surface) { - unsigned cpitch = cbuf_surface->stride; unsigned ctile = BUF_3D_USE_FENCE; struct i915_texture *tex = (struct i915_texture *) cbuf_surface->texture; - struct pipe_buffer *buffer = tex->buffer; assert(tex); if (tex && tex->tiled) { @@ -225,7 +223,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(_3DSTATE_BUF_INFO_CMD); OUT_BATCH(BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(cpitch) | /* pitch in bytes */ + BUF_3D_PITCH(tex->stride) | /* pitch in bytes */ ctile); OUT_RELOC(tex->buffer, @@ -236,11 +234,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* What happens if no zbuf?? */ if (depth_surface) { - unsigned zpitch = depth_surface->stride; unsigned ztile = BUF_3D_USE_FENCE; struct i915_texture *tex = (struct i915_texture *) depth_surface->texture; - struct pipe_buffer *buffer = tex->buffer; assert(tex); if (tex && tex->tiled) { @@ -250,7 +246,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(_3DSTATE_BUF_INFO_CMD); OUT_BATCH(BUF_3D_ID_DEPTH | - BUF_3D_PITCH(zpitch) | /* pitch in bytes */ + BUF_3D_PITCH(tex->stride) | /* pitch in bytes */ ztile); OUT_RELOC(tex->buffer, diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index c09c10601b..3667ed1afa 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -116,7 +116,7 @@ static void update_sampler(struct i915_context *i915, ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER || wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) { - if (i915->strict_conformance) { + if (i915->conformance_mode > 0) { assert(0); /* sampler->fallback = true; */ /* TODO */ diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 94e2deaf61..09b2c499b8 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -41,50 +41,27 @@ */ static void i915_surface_copy(struct pipe_context *pipe, - boolean do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - assert( dst != src ); - assert( dst->block.size == src->block.size ); - assert( dst->block.width == src->block.height ); - assert( dst->block.height == src->block.height ); + struct i915_texture *dst_tex = (struct i915_texture *)dst->texture; + struct i915_texture *src_tex = (struct i915_texture *)src->texture; - if (0) { - void *dst_map = pipe->screen->surface_map( pipe->screen, - dst, - PIPE_BUFFER_USAGE_CPU_WRITE ); - - const void *src_map = pipe->screen->surface_map( pipe->screen, - src, - PIPE_BUFFER_USAGE_CPU_READ ); - - pipe_copy_rect(dst_map, - &dst->block, - dst->stride, - dstx, dsty, - width, height, - src_map, - do_flip ? -(int) src->stride : src->stride, - srcx, do_flip ? height - 1 - srcy : srcy); + assert( dst != src ); + assert( dst_tex->base.block.size == src_tex->base.block.size ); + assert( dst_tex->base.block.width == src_tex->base.block.height ); + assert( dst_tex->base.block.height == src_tex->base.block.height ); + assert( dst_tex->base.block.width == 1 ); + assert( dst_tex->base.block.height == 1 ); - pipe->screen->surface_unmap(pipe->screen, src); - pipe->screen->surface_unmap(pipe->screen, dst); - } - else { - struct i915_texture *dst_tex = (struct i915_texture *)dst->texture; - struct i915_texture *src_tex = (struct i915_texture *)src->texture; - assert(dst->block.width == 1); - assert(dst->block.height == 1); - i915_copy_blit( i915_context(pipe), - do_flip, - dst->block.size, - (unsigned short) src->stride, src_tex->buffer, src->offset, - (unsigned short) dst->stride, dst_tex->buffer, dst->offset, - (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); - } + i915_copy_blit( i915_context(pipe), + FALSE, + dst_tex->base.block.size, + (unsigned short) src_tex->stride, src_tex->buffer, src->offset, + (unsigned short) dst_tex->stride, dst_tex->buffer, dst->offset, + (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); } @@ -94,27 +71,18 @@ i915_surface_fill(struct pipe_context *pipe, unsigned dstx, unsigned dsty, unsigned width, unsigned height, unsigned value) { - if (0) { - void *dst_map = pipe->screen->surface_map( pipe->screen, - dst, - PIPE_BUFFER_USAGE_CPU_WRITE ); + struct i915_texture *tex = (struct i915_texture *)dst->texture; - pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); + assert(tex->base.block.width == 1); + assert(tex->base.block.height == 1); - pipe->screen->surface_unmap(pipe->screen, dst); - } - else { - struct i915_texture *tex = (struct i915_texture *)dst->texture; - assert(dst->block.width == 1); - assert(dst->block.height == 1); - i915_fill_blit( i915_context(pipe), - dst->block.size, - (unsigned short) dst->stride, - tex->buffer, dst->offset, - (short) dstx, (short) dsty, - (short) width, (short) height, - value ); - } + i915_fill_blit( i915_context(pipe), + tex->base.block.size, + (unsigned short) tex->stride, + tex->buffer, dst->offset, + (short) dstx, (short) dsty, + (short) width, (short) height, + value ); } diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index b2ca3a2286..39aca9f817 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -42,6 +42,7 @@ #include "i915_texture.h" #include "i915_debug.h" #include "i915_screen.h" +#include "i915_winsys.h" /* * Helper function and arrays @@ -581,7 +582,6 @@ i915_texture_create(struct pipe_screen *screen, const struct pipe_texture *templat) { struct i915_screen *i915screen = i915_screen(screen); - struct pipe_winsys *ws = screen->winsys; struct i915_texture *tex = CALLOC_STRUCT(i915_texture); size_t tex_size; @@ -589,7 +589,7 @@ i915_texture_create(struct pipe_screen *screen, return NULL; tex->base = *templat; - tex->base.refcount = 1; + pipe_reference_init(&tex->base.reference, 1); tex->base.screen = screen; tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]); @@ -605,7 +605,7 @@ i915_texture_create(struct pipe_screen *screen, tex_size = tex->stride * tex->total_nblocksy; - tex->buffer = ws->buffer_create(ws, 64, + tex->buffer = screen->buffer_create(screen, 64, PIPE_BUFFER_USAGE_PIXEL, tex_size); @@ -628,33 +628,22 @@ fail: static void -i915_texture_release(struct pipe_screen *screen, - struct pipe_texture **pt) +i915_texture_destroy(struct pipe_texture *pt) { - if (!*pt) - return; + struct i915_texture *tex = (struct i915_texture *)pt; + uint i; /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - if (--(*pt)->refcount <= 0) { - struct i915_texture *tex = (struct i915_texture *)*pt; - uint i; - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); - */ - pipe_buffer_reference(screen, &tex->buffer, NULL); + pipe_buffer_reference(&tex->buffer, NULL); - for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) - if (tex->image_offset[i]) - FREE(tex->image_offset[i]); + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (tex->image_offset[i]) + FREE(tex->image_offset[i]); - FREE(tex); - } - *pt = NULL; + FREE(tex); } static struct pipe_surface * @@ -681,15 +670,11 @@ i915_get_tex_surface(struct pipe_screen *screen, ps = CALLOC_STRUCT(pipe_surface); if (ps) { - ps->refcount = 1; + pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = tex->stride; ps->offset = offset; ps->usage = flags; ps->status = PIPE_SURFACE_STATUS_DEFINED; @@ -718,7 +703,7 @@ i915_texture_blanket(struct pipe_screen * screen, return NULL; tex->base = *base; - tex->base.refcount = 1; + pipe_reference_init(&tex->base.reference, 1); tex->base.screen = screen; tex->stride = stride[0]; @@ -726,7 +711,7 @@ i915_texture_blanket(struct pipe_screen * screen, i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - pipe_buffer_reference(screen, &tex->buffer, buffer); + pipe_buffer_reference(&tex->buffer, buffer); return &tex->base; } @@ -738,34 +723,43 @@ i915_init_texture_functions(struct i915_context *i915) } static void -i915_tex_surface_release(struct pipe_screen *screen, - struct pipe_surface **surface) +i915_tex_surface_destroy(struct pipe_surface *surf) { - struct pipe_surface *surf = *surface; - - if (--surf->refcount == 0) { - - /* This really should not be possible, but it's actually - * happening quite a bit... Will fix. - */ - if (surf->status == PIPE_SURFACE_STATUS_CLEAR) { - debug_printf("XXX destroying a surface with pending clears...\n"); - assert(0); - } - - pipe_texture_reference(&surf->texture, NULL); - FREE(surf); + /* This really should not be possible, but it's actually + * happening quite a bit... Will fix. + */ + if (surf->status == PIPE_SURFACE_STATUS_CLEAR) { + debug_printf("XXX destroying a surface with pending clears...\n"); + assert(0); } - *surface = NULL; + pipe_texture_reference(&surf->texture, NULL); + FREE(surf); } void i915_init_screen_texture_functions(struct pipe_screen *screen) { screen->texture_create = i915_texture_create; - screen->texture_release = i915_texture_release; + screen->texture_destroy = i915_texture_destroy; screen->get_tex_surface = i915_get_tex_surface; screen->texture_blanket = i915_texture_blanket; - screen->tex_surface_release = i915_tex_surface_release; + screen->tex_surface_destroy = i915_tex_surface_destroy; +} + +boolean i915_get_texture_buffer( struct pipe_texture *texture, + struct pipe_buffer **buf, + unsigned *stride ) +{ + struct i915_texture *tex = (struct i915_texture *)texture; + + if (!tex) + return FALSE; + + pipe_buffer_reference(buf, tex->buffer); + + if (stride) + *stride = tex->stride; + + return TRUE; } diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index 81904c2a74..ff5b34f193 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -30,7 +30,8 @@ * This is the interface that i915simple requires any window system * hosting it to implement. This is the only include file in i915simple * which is public. - * + * + * This isn't currently true as the winsys needs i915_batchbuffer.h */ #ifndef I915_WINSYS_H @@ -45,10 +46,9 @@ extern "C" { #endif -/* Pipe drivers are (meant to be!) independent of both GL and the - * window system. The window system provides a buffer manager and a - * set of additional hooks for things like command buffer submission, - * etc. +/* Pipe drivers are independent of both GL and the window system. + * The window system provides a buffer manager and a set of additional + * hooks for things like command buffer submission, etc. * * There clearly has to be some agreement between the window system * driver and the hardware driver about the format of command buffers, @@ -56,6 +56,7 @@ extern "C" { */ struct i915_batchbuffer; +struct pipe_texture; struct pipe_buffer; struct pipe_fence_handle; struct pipe_winsys; @@ -64,7 +65,7 @@ struct pipe_screen; /** * Additional winsys interface for i915simple. - * + * * It is an over-simple batchbuffer mechanism. Will want to improve the * performance of this, perhaps based on the cmdstream stuff. It * would be pretty impossible to implement swz on top of this @@ -110,12 +111,33 @@ struct i915_winsys { #define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) -struct pipe_context *i915_create_context( struct pipe_screen *, - struct pipe_winsys *, - struct i915_winsys * ); +/** + * Create i915 pipe_screen. + */ +struct pipe_screen *i915_create_screen( struct pipe_winsys *winsys, + uint pci_id ); + +/** + * Create a i915 pipe_context. + */ +struct pipe_context *i915_create_context( struct pipe_screen *screen, + struct pipe_winsys *winsys, + struct i915_winsys *i915 ); + +/** + * Used for the winsys to get the buffer used for a texture + * and also the stride used for the texture. + * + * Buffer is referenced for you so you need to unref after use. + * + * This is needed for example kms. + */ +boolean i915_get_texture_buffer( struct pipe_texture *texture, + struct pipe_buffer **buf, + unsigned *stride ); #ifdef __cplusplus } #endif -#endif +#endif diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile index e97146e57c..19182afa75 100644 --- a/src/gallium/drivers/i965simple/Makefile +++ b/src/gallium/drivers/i965simple/Makefile @@ -50,5 +50,3 @@ C_SOURCES = \ brw_wm_surface_state.c include ../../Makefile.template - -symlinks: diff --git a/src/gallium/drivers/i965simple/brw_eu_debug.c b/src/gallium/drivers/i965simple/brw_eu_debug.c index 4a94ddefa6..4adfb0c02f 100644 --- a/src/gallium/drivers/i965simple/brw_eu_debug.c +++ b/src/gallium/drivers/i965simple/brw_eu_debug.c @@ -30,7 +30,7 @@ */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "brw_eu.h" diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 0a95dce194..511779dbfa 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -41,7 +41,6 @@ */ static void brw_surface_copy(struct pipe_context *pipe, - boolean do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, @@ -64,11 +63,11 @@ brw_surface_copy(struct pipe_context *pipe, pipe_copy_rect(dst_map, &dst->block, dst->stride, - dstx, dsty, - width, height, - src_map, - do_flip ? -(int) src->stride : src->stride, - srcx, do_flip ? height - 1 - srcy : srcy); + dstx, dsty, + width, height, + src_map, + src->stride, + srcx, srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); @@ -79,7 +78,7 @@ brw_surface_copy(struct pipe_context *pipe, assert(dst->block.width == 1); assert(dst->block.height == 1); brw_copy_blit(brw_context(pipe), - do_flip, + FALSE, dst->block.size, (short) src->stride/src->block.size, src_tex->buffer, src->offset, FALSE, (short) dst->stride/dst->block.size, dst_tex->buffer, dst->offset, FALSE, diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 448229ed4e..c921c0d38b 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -284,18 +284,17 @@ static struct pipe_texture * brw_texture_create_screen(struct pipe_screen *screen, const struct pipe_texture *templat) { - struct pipe_winsys *ws = screen->winsys; struct brw_texture *tex = CALLOC_STRUCT(brw_texture); if (tex) { tex->base = *templat; - tex->base.refcount = 1; + pipe_reference_init(&tex->base.reference, 1); tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]); tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]); if (brw_miptree_layout(tex)) - tex->buffer = ws->buffer_create(ws, 64, + tex->buffer = screen->buffer_create(screen, 64, PIPE_BUFFER_USAGE_PIXEL, tex->stride * tex->total_nblocksy); @@ -311,33 +310,22 @@ brw_texture_create_screen(struct pipe_screen *screen, static void -brw_texture_release_screen(struct pipe_screen *screen, - struct pipe_texture **pt) +brw_texture_destroy_screen(struct pipe_texture *pt) { - if (!*pt) - return; + struct brw_texture *tex = (struct brw_texture *)pt; + uint i; /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - if (--(*pt)->refcount <= 0) { - struct brw_texture *tex = (struct brw_texture *)*pt; - uint i; - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); - */ - pipe_buffer_reference(screen, &tex->buffer, NULL); + pipe_buffer_reference(&tex->buffer, NULL); - for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) - if (tex->image_offset[i]) - free(tex->image_offset[i]); + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (tex->image_offset[i]) + free(tex->image_offset[i]); - free(tex); - } - *pt = NULL; + free(tex); } @@ -365,7 +353,7 @@ brw_get_tex_surface_screen(struct pipe_screen *screen, ps = CALLOC_STRUCT(pipe_surface); if (ps) { - ps->refcount = 1; + pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; ps->width = pt->width[level]; @@ -392,7 +380,7 @@ void brw_init_screen_texture_funcs(struct pipe_screen *screen) { screen->texture_create = brw_texture_create_screen; - screen->texture_release = brw_texture_release_screen; + screen->texture_destroy = brw_texture_destroy_screen; screen->get_tex_surface = brw_get_tex_surface_screen; } diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c index 1bab5bfdb3..b5b9e0e702 100644 --- a/src/gallium/drivers/i965simple/brw_wm_surface_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c @@ -242,7 +242,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) const struct brw_texture *texUnit = brw->attribs.Texture[i]; if (texUnit && - texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { + texUnit->base.reference.count/*(texUnit->reference.count > 0) == really used */) { brw_update_texture_surface(brw, i); diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h index 4ae4ff4940..97859110b5 100644 --- a/src/gallium/drivers/nouveau/nouveau_stateobj.h +++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h @@ -1,7 +1,7 @@ #ifndef __NOUVEAU_STATEOBJ_H__ #define __NOUVEAU_STATEOBJ_H__ -#include "pipe/p_debug.h" +#include "util/u_debug.h" struct nouveau_stateobj_reloc { struct pipe_buffer *bo; @@ -16,7 +16,7 @@ struct nouveau_stateobj_reloc { }; struct nouveau_stateobj { - int refcount; + struct pipe_reference reference; unsigned *push; struct nouveau_stateobj_reloc *reloc; @@ -32,7 +32,7 @@ so_new(unsigned push, unsigned reloc) struct nouveau_stateobj *so; so = MALLOC(sizeof(struct nouveau_stateobj)); - so->refcount = 0; + pipe_reference_init(&so->reference, 1); so->push = MALLOC(sizeof(unsigned) * push); so->reloc = MALLOC(sizeof(struct nouveau_stateobj_reloc) * reloc); @@ -47,17 +47,11 @@ so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso) { struct nouveau_stateobj *so = *pso; - if (ref) { - ref->refcount++; - } - - if (so && --so->refcount <= 0) { + if (pipe_reference((struct pipe_reference**)pso, &ref->reference)) { free(so->push); free(so->reloc); free(so); } - - *pso = ref; } static INLINE void diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h index b86c4b9338..ff7dd1c51c 100644 --- a/src/gallium/drivers/nouveau/nouveau_winsys.h +++ b/src/gallium/drivers/nouveau/nouveau_winsys.h @@ -19,11 +19,12 @@ #define NOUVEAU_TEXTURE_USAGE_LINEAR (1 << 16) -#define NOUVEAU_BUFFER_USAGE_TEXTURE (1 << 16) -#define NOUVEAU_BUFFER_USAGE_ZETA (1 << 17) +#define NOUVEAU_BUFFER_USAGE_TEXTURE (1 << 16) +#define NOUVEAU_BUFFER_USAGE_ZETA (1 << 17) +#define NOUVEAU_BUFFER_USAGE_TRANSFER (1 << 18) struct nouveau_winsys { - struct nouveau_context *nv; + struct pipe_winsys *ws; struct nouveau_channel *channel; diff --git a/src/gallium/drivers/nv04/Makefile b/src/gallium/drivers/nv04/Makefile index 4ed62dae95..7c14bacb1d 100644 --- a/src/gallium/drivers/nv04/Makefile +++ b/src/gallium/drivers/nv04/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = nv04 -DRIVER_SOURCES = \ +C_SOURCES = \ nv04_surface_2d.c \ nv04_clear.c \ nv04_context.c \ @@ -15,15 +15,7 @@ DRIVER_SOURCES = \ nv04_state.c \ nv04_state_emit.c \ nv04_surface.c \ + nv04_transfer.c \ nv04_vbo.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c index 993c5ef5dd..85dc017fbc 100644 --- a/src/gallium/drivers/nv04/nv04_miptree.c +++ b/src/gallium/drivers/nv04/nv04_miptree.c @@ -41,23 +41,20 @@ nv04_miptree_layout(struct nv04_miptree *nv04mt) static struct pipe_texture * nv04_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) { - struct pipe_winsys *ws = pscreen->winsys; struct nv04_miptree *mt; mt = MALLOC(sizeof(struct nv04_miptree)); if (!mt) return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; - mt->shadow_tex = NULL; - mt->shadow_surface = NULL; //mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; nv04_miptree_layout(mt); - mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL | + mt->buffer = pscreen->buffer_create(pscreen, 256, PIPE_BUFFER_USAGE_PIXEL | NOUVEAU_BUFFER_USAGE_TEXTURE, mt->total_size); if (!mt->buffer) { @@ -85,38 +82,27 @@ nv04_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; mt->level[0].pitch = stride[0]; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(pscreen, &mt->buffer, pb); + pipe_buffer_reference(&mt->buffer, pb); return &mt->base; } static void -nv04_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) +nv04_miptree_destroy(struct pipe_texture *pt) { - struct pipe_texture *pt = *ppt; struct nv04_miptree *mt = (struct nv04_miptree *)pt; int l; - *ppt = NULL; - if (--pt->refcount) - return; - - pipe_buffer_reference(pscreen, &mt->buffer, NULL); + pipe_buffer_reference(&mt->buffer, NULL); for (l = 0; l <= pt->last_level; l++) { if (mt->level[l].image_offset) FREE(mt->level[l].image_offset); } - if (mt->shadow_tex) { - assert(mt->shadow_surface); - pscreen->tex_surface_release(pscreen, &mt->shadow_surface); - nv04_miptree_release(pscreen, &mt->shadow_tex); - } - FREE(mt); } @@ -126,41 +112,31 @@ nv04_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, unsigned flags) { struct nv04_miptree *nv04mt = (struct nv04_miptree *)pt; - struct pipe_surface *ps; + struct nv04_surface *ns; - ps = CALLOC_STRUCT(pipe_surface); - if (!ps) + ns = CALLOC_STRUCT(nv04_surface); + if (!ns) return NULL; - pipe_texture_reference(&ps->texture, pt); - ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = nv04mt->level[level].pitch; - ps->usage = flags; - ps->status = PIPE_SURFACE_STATUS_DEFINED; - ps->refcount = 1; - ps->face = face; - ps->level = level; - ps->zslice = zslice; - - ps->offset = nv04mt->level[level].image_offset; - - return ps; + pipe_texture_reference(&ns->base.texture, pt); + ns->base.format = pt->format; + ns->base.width = pt->width[level]; + ns->base.height = pt->height[level]; + ns->base.usage = flags; + ns->base.status = PIPE_SURFACE_STATUS_DEFINED; + pipe_reference_init(&ns->base.reference, 1); + ns->base.face = face; + ns->base.level = level; + ns->base.zslice = zslice; + ns->pitch = nv04mt->level[level].pitch; + + ns->base.offset = nv04mt->level[level].image_offset; + + return &ns->base; } static void -nv04_miptree_surface_del(struct pipe_screen *pscreen, - struct pipe_surface **psurface) +nv04_miptree_surface_del(struct pipe_surface *ps) { - struct pipe_surface *ps = *psurface; - - *psurface = NULL; - if (--ps->refcount > 0) - return; - pipe_texture_reference(&ps->texture, NULL); FREE(ps); } @@ -170,8 +146,8 @@ nv04_screen_init_miptree_functions(struct pipe_screen *pscreen) { pscreen->texture_create = nv04_miptree_create; pscreen->texture_blanket = nv04_miptree_blanket; - pscreen->texture_release = nv04_miptree_release; + pscreen->texture_destroy = nv04_miptree_destroy; pscreen->get_tex_surface = nv04_miptree_surface_new; - pscreen->tex_surface_release = nv04_miptree_surface_del; + pscreen->tex_surface_destroy = nv04_miptree_surface_del; } diff --git a/src/gallium/drivers/nv04/nv04_prim_vbuf.c b/src/gallium/drivers/nv04/nv04_prim_vbuf.c index 18a8872ae3..f6458232ae 100644 --- a/src/gallium/drivers/nv04/nv04_prim_vbuf.c +++ b/src/gallium/drivers/nv04/nv04_prim_vbuf.c @@ -1,5 +1,5 @@ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" #include "pipe/p_compiler.h" @@ -51,7 +51,7 @@ nv04_vbuf_render_get_vertex_info( struct vbuf_render *render ) } -static void * +static boolean nv04_vbuf_render_allocate_vertices( struct vbuf_render *render, ushort vertex_size, ushort nr_vertices ) @@ -61,9 +61,22 @@ nv04_vbuf_render_allocate_vertices( struct vbuf_render *render, nv04_render->buffer = (unsigned char*) MALLOC(VERTEX_BUFFER_SIZE); assert(!nv04_render->buffer); + return nv04_render->buffer ? TRUE : FALSE; +} + +static void * +nv04_vbuf_render_map_vertices( struct vbuf_render *render ) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); return nv04_render->buffer; } +static void +nv04_vbuf_render_unmap_vertices( struct vbuf_render *render, + ushort min_index, + ushort max_index ) +{ +} static boolean nv04_vbuf_render_set_primitive( struct vbuf_render *render, @@ -244,10 +257,7 @@ nv04_vbuf_render_draw( struct vbuf_render *render, static void -nv04_vbuf_render_release_vertices( struct vbuf_render *render, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ) +nv04_vbuf_render_release_vertices( struct vbuf_render *render ) { struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); @@ -278,6 +288,8 @@ nv04_vbuf_render_create( struct nv04_context *nv04 ) nv04_render->base.max_indices = 65536; nv04_render->base.get_vertex_info = nv04_vbuf_render_get_vertex_info; nv04_render->base.allocate_vertices = nv04_vbuf_render_allocate_vertices; + nv04_render->base.map_vertices = nv04_vbuf_render_map_vertices; + nv04_render->base.unmap_vertices = nv04_vbuf_render_unmap_vertices; nv04_render->base.set_primitive = nv04_vbuf_render_set_primitive; nv04_render->base.draw = nv04_vbuf_render_draw; nv04_render->base.release_vertices = nv04_vbuf_render_release_vertices; diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c index 9ef38bc244..f9f6d97426 100644 --- a/src/gallium/drivers/nv04/nv04_screen.c +++ b/src/gallium/drivers/nv04/nv04_screen.c @@ -119,28 +119,6 @@ nv04_screen_is_format_supported(struct pipe_screen *screen, return FALSE; } -static void * -nv04_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, - unsigned flags ) -{ - void *map; - struct nv04_miptree *nv04mt = (struct nv04_miptree *)surface->texture; - - map = pipe_buffer_map(screen, nv04mt->buffer, flags); - if (!map) - return NULL; - - return map + surface->offset; -} - -static void -nv04_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) -{ - struct nv04_miptree *nv04mt = (struct nv04_miptree *)surface->texture; - - pipe_buffer_unmap(screen, nv04mt->buffer); -} - static void nv04_screen_destroy(struct pipe_screen *pscreen) { @@ -226,10 +204,8 @@ nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) screen->pipe.is_format_supported = nv04_screen_is_format_supported; - screen->pipe.surface_map = nv04_surface_map; - screen->pipe.surface_unmap = nv04_surface_unmap; - nv04_screen_init_miptree_functions(&screen->pipe); + nv04_screen_init_transfer_functions(&screen->pipe); u_simple_screen_init(&screen->pipe); return &screen->pipe; diff --git a/src/gallium/drivers/nv04/nv04_screen.h b/src/gallium/drivers/nv04/nv04_screen.h index 540aec907b..ee6fb6db44 100644 --- a/src/gallium/drivers/nv04/nv04_screen.h +++ b/src/gallium/drivers/nv04/nv04_screen.h @@ -24,4 +24,7 @@ nv04_screen(struct pipe_screen *screen) return (struct nv04_screen *)screen; } +void +nv04_screen_init_transfer_functions(struct pipe_screen *pscreen); + #endif diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h index 15d4685ec1..0d51439e3f 100644 --- a/src/gallium/drivers/nv04/nv04_state.h +++ b/src/gallium/drivers/nv04/nv04_state.h @@ -35,9 +35,6 @@ struct nv04_miptree { struct pipe_buffer *buffer; uint total_size; - struct pipe_texture *shadow_tex; - struct pipe_surface *shadow_surface; - struct { uint pitch; uint image_offset; diff --git a/src/gallium/drivers/nv04/nv04_state_emit.c b/src/gallium/drivers/nv04/nv04_state_emit.c index bd8ef1adbf..eb2c1c57c6 100644 --- a/src/gallium/drivers/nv04/nv04_state_emit.c +++ b/src/gallium/drivers/nv04/nv04_state_emit.c @@ -93,7 +93,7 @@ static void nv04_emit_sampler(struct nv04_context *nv04, int unit) static void nv04_state_emit_framebuffer(struct nv04_context* nv04) { struct pipe_framebuffer_state* fb = nv04->framebuffer; - struct pipe_surface *rt, *zeta; + struct nv04_surface *rt, *zeta; uint32_t rt_format, w, h; int colour_format = 0, zeta_format = 0; struct nv04_miptree *nv04mt = 0; @@ -101,7 +101,7 @@ static void nv04_state_emit_framebuffer(struct nv04_context* nv04) w = fb->cbufs[0]->width; h = fb->cbufs[0]->height; colour_format = fb->cbufs[0]->format; - rt = fb->cbufs[0]; + rt = (struct nv04_surface *)fb->cbufs[0]; if (fb->zsbuf) { if (colour_format) { @@ -113,7 +113,7 @@ static void nv04_state_emit_framebuffer(struct nv04_context* nv04) } zeta_format = fb->zsbuf->format; - zeta = fb->zsbuf; + zeta = (struct nv04_surface *)fb->zsbuf; } switch (colour_format) { @@ -131,13 +131,13 @@ static void nv04_state_emit_framebuffer(struct nv04_context* nv04) BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_FORMAT, 1); OUT_RING(rt_format); - nv04mt = (struct nv04_miptree *)rt->texture; + nv04mt = (struct nv04_miptree *)rt->base.texture; /* FIXME pitches have to be aligned ! */ BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_PITCH, 2); - OUT_RING(rt->stride|(zeta->stride<<16)); + OUT_RING(rt->pitch|(zeta->pitch<<16)); OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); if (fb->zsbuf) { - nv04mt = (struct nv04_miptree *)zeta->texture; + nv04mt = (struct nv04_miptree *)zeta->base.texture; BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1); OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); } @@ -202,8 +202,11 @@ nv04_emit_hw_state(struct nv04_context *nv04) */ /* Render target */ + unsigned rt_pitch = ((struct nv04_surface *)nv04->rt)->pitch; + unsigned zeta_pitch = ((struct nv04_surface *)nv04->zeta)->pitch; + BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_PITCH, 2); - OUT_RING(nv04->rt->stride|(nv04->zeta->stride<<16)); + OUT_RING(rt_pitch|(zeta_pitch<<16)); OUT_RELOCl(nv04->rt, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); if (nv04->zeta) { BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1); diff --git a/src/gallium/drivers/nv04/nv04_surface.c b/src/gallium/drivers/nv04/nv04_surface.c index 14abf16679..0387ff4e78 100644 --- a/src/gallium/drivers/nv04/nv04_surface.c +++ b/src/gallium/drivers/nv04/nv04_surface.c @@ -33,7 +33,7 @@ #include "util/u_tile.h" static void -nv04_surface_copy(struct pipe_context *pipe, boolean do_flip, +nv04_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) @@ -41,15 +41,6 @@ nv04_surface_copy(struct pipe_context *pipe, boolean do_flip, struct nv04_context *nv04 = nv04_context(pipe); struct nv04_surface_2d *eng2d = nv04->screen->eng2d; - if (do_flip) { - desty += height; - while (height--) { - eng2d->copy(eng2d, dest, destx, desty--, src, - srcx, srcy++, width, 1); - } - return; - } - eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height); } diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c index 230cfd17dd..f3a8d7efee 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.c +++ b/src/gallium/drivers/nv04/nv04_surface_2d.c @@ -101,6 +101,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, struct nouveau_grobj *sifm = ctx->sifm; struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src)); struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + const unsigned src_pitch = ((struct nv04_surface *)src)->pitch; const unsigned max_w = 1024; const unsigned max_h = 1024; const unsigned sub_w = w > max_w ? max_w : w; @@ -110,6 +111,8 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, /* POT or GTFO */ assert(!(w & (w - 1)) && !(h & (h - 1))); + /* That's the way she likes it */ + assert(src_pitch == ((struct nv04_surface *)dst)->pitch); BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_DMA_IMAGE, 1); OUT_RELOCo(chan, dst_bo, @@ -130,7 +133,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, for (cx = 0; cx < w; cx += sub_w) { BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1); OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(cx, cy) * - dst->block.size, NOUVEAU_BO_GART | + dst->texture->block.size, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION, 9); @@ -146,11 +149,11 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_SIZE, 4); OUT_RING (chan, sub_h << 16 | sub_w); - OUT_RING (chan, src->stride | + OUT_RING (chan, src_pitch | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE); - OUT_RELOCl(chan, src_bo, src->offset + cy * src->stride + - cx * src->block.size, NOUVEAU_BO_GART | + OUT_RELOCl(chan, src_bo, src->offset + cy * src_pitch + + cx * src->texture->block.size, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); OUT_RING (chan, 0); } @@ -168,10 +171,12 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, struct nouveau_grobj *m2mf = ctx->m2mf; struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src)); struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); - unsigned dst_offset, src_offset; - - dst_offset = dst->offset + (dy * dst->stride) + (dx * dst->block.size); - src_offset = src->offset + (sy * src->stride) + (sx * src->block.size); + unsigned src_pitch = ((struct nv04_surface *)src)->pitch; + unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; + unsigned dst_offset = dst->offset + dy * dst_pitch + + dx * dst->texture->block.size; + unsigned src_offset = src->offset + sy * src_pitch + + sx * src->texture->block.size; WAIT_RING (chan, 3 + ((h / 2047) + 1) * 9); BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2); @@ -188,16 +193,16 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); OUT_RELOCl(chan, dst_bo, dst_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR); - OUT_RING (chan, src->stride); - OUT_RING (chan, dst->stride); - OUT_RING (chan, w * src->block.size); + OUT_RING (chan, src_pitch); + OUT_RING (chan, dst_pitch); + OUT_RING (chan, w * src->texture->block.size); OUT_RING (chan, count); OUT_RING (chan, 0x0101); OUT_RING (chan, 0); h -= count; - src_offset += src->stride * count; - dst_offset += dst->stride * count; + src_offset += src_pitch * count; + dst_offset += dst_pitch * count; } return 0; @@ -213,6 +218,8 @@ nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst, struct nouveau_grobj *blit = ctx->blit; struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src)); struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + unsigned src_pitch = ((struct nv04_surface *)src)->pitch; + unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; int format; format = nv04_surface_format(dst->format); @@ -225,7 +232,7 @@ nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst, OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_FORMAT, 4); OUT_RING (chan, format); - OUT_RING (chan, (dst->stride << 16) | src->stride); + OUT_RING (chan, (dst_pitch << 16) | src_pitch); OUT_RELOCl(chan, src_bo, src->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); OUT_RELOCl(chan, dst_bo, dst->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); @@ -242,6 +249,8 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst, int dx, int dy, struct pipe_surface *src, int sx, int sy, int w, int h) { + unsigned src_pitch = ((struct nv04_surface *)src)->pitch; + unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; int src_linear = src->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR; int dst_linear = dst->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR; @@ -257,7 +266,8 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst, * to NV_MEMORY_TO_MEMORY_FORMAT in this case. */ if ((src->offset & 63) || (dst->offset & 63) || - (src->stride & 63) || (dst->stride & 63)) { + (src_pitch & 63) || (dst_pitch & 63) || + debug_get_bool_option("NOUVEAU_NO_COPYBLIT", FALSE)) { nv04_surface_copy_m2mf(ctx, dst, dx, dy, src, sx, sy, w, h); return; } @@ -273,6 +283,7 @@ nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst, struct nouveau_grobj *surf2d = ctx->surf2d; struct nouveau_grobj *rect = ctx->rect; struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; int cs2d_format, gdirect_format; cs2d_format = nv04_surface_format(dst->format); @@ -287,7 +298,7 @@ nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst, OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_FORMAT, 4); OUT_RING (chan, cs2d_format); - OUT_RING (chan, (dst->stride << 16) | dst->stride); + OUT_RING (chan, (dst_pitch << 16) | dst_pitch); OUT_RELOCl(chan, dst_bo, dst->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RELOCl(chan, dst_bo, dst->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); @@ -426,13 +437,21 @@ nv04_surface_2d_init(struct nouveau_winsys *nvws) return NULL; } - if (chan->device->chipset < 0x10) { - class = NV04_SCALED_IMAGE_FROM_MEMORY; - } else - if (chan->device->chipset < 0x40) { + switch (chan->device->chipset & 0xf0) { + case 0x10: + case 0x20: class = NV10_SCALED_IMAGE_FROM_MEMORY; - } else { + break; + case 0x30: + class = NV30_SCALED_IMAGE_FROM_MEMORY; + break; + case 0x40: + case 0x60: class = NV40_SCALED_IMAGE_FROM_MEMORY; + break; + default: + class = NV04_SCALED_IMAGE_FROM_MEMORY; + break; } ret = nouveau_grobj_alloc(chan, handle++, class, &ctx->sifm); diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.h b/src/gallium/drivers/nv04/nv04_surface_2d.h index 21b8f86960..82ce7189c8 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.h +++ b/src/gallium/drivers/nv04/nv04_surface_2d.h @@ -1,6 +1,11 @@ #ifndef __NV04_SURFACE_2D_H__ #define __NV04_SURFACE_2D_H__ +struct nv04_surface { + struct pipe_surface base; + unsigned pitch; +}; + struct nv04_surface_2d { struct nouveau_winsys *nvws; struct nouveau_notifier *ntfy; diff --git a/src/gallium/drivers/nv04/nv04_transfer.c b/src/gallium/drivers/nv04/nv04_transfer.c new file mode 100644 index 0000000000..e925a44e29 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_transfer.c @@ -0,0 +1,196 @@ +#include <pipe/p_state.h> +#include <pipe/p_defines.h> +#include <pipe/p_inlines.h> +#include <util/u_memory.h> +#include <nouveau/nouveau_winsys.h> +#include "nv04_context.h" +#include "nv04_screen.h" +#include "nv04_state.h" + +struct nv04_transfer { + struct pipe_transfer base; + struct pipe_surface *surface; + bool direct; +}; + +static unsigned nv04_usage_tx_to_buf(unsigned tx_usage) +{ + switch (tx_usage) { + case PIPE_TRANSFER_READ: + return PIPE_BUFFER_USAGE_CPU_READ; + case PIPE_TRANSFER_WRITE: + return PIPE_BUFFER_USAGE_CPU_WRITE; + case PIPE_TRANSFER_READ_WRITE: + return PIPE_BUFFER_USAGE_CPU_READ_WRITE; + default: + assert(0); + } + + return -1; +} + +static void +nv04_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, + struct pipe_texture *template) +{ + memset(template, 0, sizeof(struct pipe_texture)); + template->target = pt->target; + template->format = pt->format; + template->width[0] = pt->width[level]; + template->height[0] = pt->height[level]; + template->depth[0] = 1; + template->block = pt->block; + template->nblocksx[0] = pt->nblocksx[level]; + template->nblocksy[0] = pt->nblocksx[level]; + template->last_level = 0; + template->compressed = pt->compressed; + template->nr_samples = pt->nr_samples; + + template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | + NOUVEAU_TEXTURE_USAGE_LINEAR; +} + +static struct pipe_transfer * +nv04_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct nv04_miptree *mt = (struct nv04_miptree *)pt; + struct nv04_transfer *tx; + struct pipe_texture tx_tex_template, *tx_tex; + + tx = CALLOC_STRUCT(nv04_transfer); + if (!tx) + return NULL; + + pipe_texture_reference(&tx->base.texture, pt); + tx->base.format = pt->format; + tx->base.x = x; + tx->base.y = y; + tx->base.width = w; + tx->base.height = h; + tx->base.block = pt->block; + tx->base.nblocksx = pt->nblocksx[level]; + tx->base.nblocksy = pt->nblocksy[level]; + tx->base.stride = mt->level[level].pitch; + tx->base.usage = usage; + tx->base.face = face; + tx->base.level = level; + tx->base.zslice = zslice; + + /* Direct access to texture */ + if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC || + debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) && + pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) + { + tx->direct = true; + tx->surface = pscreen->get_tex_surface(pscreen, pt, + 0, 0, 0, + nv04_usage_tx_to_buf(usage)); + return &tx->base; + } + + tx->direct = false; + + nv04_compatible_transfer_tex(pt, level, &tx_tex_template); + + tx_tex = pscreen->texture_create(pscreen, &tx_tex_template); + if (!tx_tex) + { + FREE(tx); + return NULL; + } + + tx->surface = pscreen->get_tex_surface(pscreen, tx_tex, + face, level, zslice, + nv04_usage_tx_to_buf(usage)); + + pipe_texture_reference(&tx_tex, NULL); + + if (!tx->surface) + { + pipe_surface_reference(&tx->surface, NULL); + FREE(tx); + return NULL; + } + + if (usage != PIPE_TRANSFER_WRITE) { + struct nv04_screen *nvscreen = nv04_screen(pscreen); + struct pipe_surface *src; + + src = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + PIPE_BUFFER_USAGE_GPU_READ); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + /* TODO: Check if SIFM can un-swizzle */ + nvscreen->eng2d->copy(nvscreen->eng2d, + tx->surface, 0, 0, + src, 0, 0, + src->width, src->height); + + pipe_surface_reference(&src, NULL); + } + + return &tx->base; +} + +static void +nv04_transfer_del(struct pipe_transfer *ptx) +{ + struct nv04_transfer *tx = (struct nv04_transfer *)ptx; + + if (!tx->direct && ptx->usage != PIPE_TRANSFER_READ) { + struct pipe_screen *pscreen = ptx->texture->screen; + struct nv04_screen *nvscreen = nv04_screen(pscreen); + struct pipe_surface *dst; + + dst = pscreen->get_tex_surface(pscreen, ptx->texture, + ptx->face, ptx->level, ptx->zslice, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + nvscreen->eng2d->copy(nvscreen->eng2d, + dst, 0, 0, + tx->surface, 0, 0, + dst->width, dst->height); + + pipe_surface_reference(&dst, NULL); + } + + pipe_surface_reference(&tx->surface, NULL); + pipe_texture_reference(&ptx->texture, NULL); + FREE(ptx); +} + +static void * +nv04_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv04_transfer *tx = (struct nv04_transfer *)ptx; + struct nv04_surface *ns = (struct nv04_surface *)tx->surface; + struct nv04_miptree *mt = (struct nv04_miptree *)tx->surface->texture; + void *map = pipe_buffer_map(pscreen, mt->buffer, + nv04_usage_tx_to_buf(ptx->usage)); + + return map + ns->base.offset + + ptx->y * ns->pitch + ptx->x * ptx->block.size; +} + +static void +nv04_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv04_transfer *tx = (struct nv04_transfer *)ptx; + struct nv04_miptree *mt = (struct nv04_miptree *)tx->surface->texture; + + pipe_buffer_unmap(pscreen, mt->buffer); +} + +void +nv04_screen_init_transfer_functions(struct pipe_screen *pscreen) +{ + pscreen->get_tex_transfer = nv04_transfer_new; + pscreen->tex_transfer_destroy = nv04_transfer_del; + pscreen->transfer_map = nv04_transfer_map; + pscreen->transfer_unmap = nv04_transfer_unmap; +} diff --git a/src/gallium/drivers/nv10/Makefile b/src/gallium/drivers/nv10/Makefile index 4ba7ce586d..62677f5194 100644 --- a/src/gallium/drivers/nv10/Makefile +++ b/src/gallium/drivers/nv10/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = nv10 -DRIVER_SOURCES = \ +C_SOURCES = \ nv10_clear.c \ nv10_context.c \ nv10_fragprog.c \ @@ -14,15 +14,7 @@ DRIVER_SOURCES = \ nv10_state.c \ nv10_state_emit.c \ nv10_surface.c \ + nv10_transfer.c \ nv10_vbo.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c index 9616135461..bb3a1c0f19 100644 --- a/src/gallium/drivers/nv10/nv10_miptree.c +++ b/src/gallium/drivers/nv10/nv10_miptree.c @@ -66,31 +66,30 @@ nv10_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; mt->level[0].pitch = stride[0]; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(pscreen, &mt->buffer, pb); + pipe_buffer_reference(&mt->buffer, pb); return &mt->base; } static struct pipe_texture * nv10_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) { - struct pipe_winsys *ws = screen->winsys; struct nv10_miptree *mt; mt = MALLOC(sizeof(struct nv10_miptree)); if (!mt) return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = screen; nv10_miptree_layout(mt); - mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL, + mt->buffer = screen->buffer_create(screen, 256, PIPE_BUFFER_USAGE_PIXEL, mt->total_size); if (!mt->buffer) { FREE(mt); @@ -101,22 +100,17 @@ nv10_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) } static void -nv10_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt) +nv10_miptree_destroy(struct pipe_texture *pt) { - struct pipe_texture *mt = *pt; - - *pt = NULL; - if (--mt->refcount <= 0) { - struct nv10_miptree *nv10mt = (struct nv10_miptree *)mt; - int l; - - pipe_buffer_reference(screen, &nv10mt->buffer, NULL); - for (l = 0; l <= mt->last_level; l++) { - if (nv10mt->level[l].image_offset) - FREE(nv10mt->level[l].image_offset); - } - FREE(nv10mt); - } + struct nv10_miptree *nv10mt = (struct nv10_miptree *)pt; + int l; + + pipe_buffer_reference(&nv10mt->buffer, NULL); + for (l = 0; l <= pt->last_level; l++) { + if (nv10mt->level[l].image_offset) + FREE(nv10mt->level[l].image_offset); + } + FREE(nv10mt); } static void @@ -131,35 +125,35 @@ nv10_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice, unsigned flags) { - struct pipe_winsys *ws = screen->winsys; struct nv10_miptree *nv10mt = (struct nv10_miptree *)pt; - struct pipe_surface *ps; + struct nv04_surface *ns; - ps = CALLOC_STRUCT(pipe_surface); - if (!ps) + ns = CALLOC_STRUCT(nv04_surface); + if (!ns) return NULL; - pipe_texture_reference(&ps->texture, pt); - ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = nv10mt->level[level].pitch; - ps->refcount = 1; + pipe_texture_reference(&ns->base.texture, pt); + ns->base.format = pt->format; + ns->base.width = pt->width[level]; + ns->base.height = pt->height[level]; + ns->base.usage = flags; + ns->base.status = PIPE_SURFACE_STATUS_DEFINED; + pipe_reference_init(&ns->base.reference, 1); + ns->base.face = face; + ns->base.level = level; + ns->base.zslice = zslice; + ns->pitch = nv10mt->level[level].pitch; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset = nv10mt->level[level].image_offset[face]; + ns->base.offset = nv10mt->level[level].image_offset[face]; } else { - ps->offset = nv10mt->level[level].image_offset[0]; + ns->base.offset = nv10mt->level[level].image_offset[0]; } - return ps; + return &ns->base; } static void -nv10_miptree_surface_release(struct pipe_screen *screen, - struct pipe_surface **surface) +nv10_miptree_surface_destroy(struct pipe_surface *surface) { } @@ -167,8 +161,8 @@ void nv10_screen_init_miptree_functions(struct pipe_screen *pscreen) { pscreen->texture_create = nv10_miptree_create; pscreen->texture_blanket = nv10_miptree_blanket; - pscreen->texture_release = nv10_miptree_release; + pscreen->texture_destroy = nv10_miptree_destroy; pscreen->get_tex_surface = nv10_miptree_surface_get; - pscreen->tex_surface_release = nv10_miptree_surface_release; + pscreen->tex_surface_destroy = nv10_miptree_surface_destroy; } diff --git a/src/gallium/drivers/nv10/nv10_prim_vbuf.c b/src/gallium/drivers/nv10/nv10_prim_vbuf.c index 7435d87315..089c236302 100644 --- a/src/gallium/drivers/nv10/nv10_prim_vbuf.c +++ b/src/gallium/drivers/nv10/nv10_prim_vbuf.c @@ -38,7 +38,7 @@ */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" @@ -99,27 +99,50 @@ nv10_vbuf_render_get_vertex_info( struct vbuf_render *render ) return &nv10->vertex_info; } - -static void * +static boolean nv10_vbuf_render_allocate_vertices( struct vbuf_render *render, ushort vertex_size, ushort nr_vertices ) { struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); struct nv10_context *nv10 = nv10_render->nv10; - struct pipe_winsys *winsys = nv10->pipe.winsys; + struct pipe_screen *screen = nv10->pipe.screen; size_t size = (size_t)vertex_size * (size_t)nr_vertices; assert(!nv10_render->buffer); - nv10_render->buffer = winsys->buffer_create(winsys, 64, PIPE_BUFFER_USAGE_VERTEX, size); + nv10_render->buffer = screen->buffer_create(screen, 64, PIPE_BUFFER_USAGE_VERTEX, size); nv10->dirty |= NV10_NEW_VTXARRAYS; + if (nv10_render->buffer) + return FALSE; + return TRUE; +} + +static void * +nv10_vbuf_render_map_vertices( struct vbuf_render *render ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + struct nv10_context *nv10 = nv10_render->nv10; + struct pipe_winsys *winsys = nv10->pipe.winsys; + return winsys->buffer_map(winsys, nv10_render->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); } +static void +nv10_vbuf_render_unmap_vertices( struct vbuf_render *render, + ushort min_index, + ushort max_index ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + struct nv10_context *nv10 = nv10_render->nv10; + struct pipe_winsys *winsys = nv10->pipe.winsys; + + assert(!nv10_render->buffer); + winsys->buffer_unmap(winsys, nv10_render->buffer); +} static boolean nv10_vbuf_render_set_primitive( struct vbuf_render *render, @@ -176,19 +199,14 @@ nv10_vbuf_render_draw( struct vbuf_render *render, static void -nv10_vbuf_render_release_vertices( struct vbuf_render *render, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ) +nv10_vbuf_render_release_vertices( struct vbuf_render *render ) { struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); struct nv10_context *nv10 = nv10_render->nv10; - struct pipe_winsys *winsys = nv10->pipe.winsys; struct pipe_screen *pscreen = &nv10->screen->pipe; assert(nv10_render->buffer); - winsys->buffer_unmap(winsys, nv10_render->buffer); - pipe_buffer_reference(pscreen, &nv10_render->buffer, NULL); + pipe_buffer_reference(&nv10_render->buffer, NULL); } @@ -214,6 +232,8 @@ nv10_vbuf_render_create( struct nv10_context *nv10 ) nv10_render->base.max_indices = 1024; nv10_render->base.get_vertex_info = nv10_vbuf_render_get_vertex_info; nv10_render->base.allocate_vertices = nv10_vbuf_render_allocate_vertices; + nv10_render->base.map_vertices = nv10_vbuf_render_map_vertices; + nv10_render->base.unmap_vertices = nv10_vbuf_render_unmap_vertices; nv10_render->base.set_primitive = nv10_vbuf_render_set_primitive; nv10_render->base.draw = nv10_vbuf_render_draw; nv10_render->base.release_vertices = nv10_vbuf_render_release_vertices; diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c index f417b06c94..6532a93c7b 100644 --- a/src/gallium/drivers/nv10/nv10_screen.c +++ b/src/gallium/drivers/nv10/nv10_screen.c @@ -116,30 +116,6 @@ nv10_screen_is_format_supported(struct pipe_screen *screen, return FALSE; } -static void * -nv10_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, - unsigned flags ) -{ - struct pipe_winsys *ws = screen->winsys; - void *map; - struct nv10_miptree *nv10mt = (struct nv10_miptree *)surface->texture; - - map = ws->buffer_map(ws, nv10mt->buffer, flags); - if (!map) - return NULL; - - return map + surface->offset; -} - -static void -nv10_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) -{ - struct pipe_winsys *ws = screen->winsys; - struct nv10_miptree *nv10mt = (struct nv10_miptree *)surface->texture; - - ws->buffer_unmap(ws, nv10mt->buffer); -} - static void nv10_screen_destroy(struct pipe_screen *pscreen) { @@ -215,10 +191,8 @@ nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) screen->pipe.is_format_supported = nv10_screen_is_format_supported; - screen->pipe.surface_map = nv10_surface_map; - screen->pipe.surface_unmap = nv10_surface_unmap; - nv10_screen_init_miptree_functions(&screen->pipe); + nv10_screen_init_transfer_functions(&screen->pipe); u_simple_screen_init(&screen->pipe); return &screen->pipe; diff --git a/src/gallium/drivers/nv10/nv10_screen.h b/src/gallium/drivers/nv10/nv10_screen.h index 60102a369a..ad829ee3fd 100644 --- a/src/gallium/drivers/nv10/nv10_screen.h +++ b/src/gallium/drivers/nv10/nv10_screen.h @@ -21,4 +21,8 @@ nv10_screen(struct pipe_screen *screen) return (struct nv10_screen *)screen; } + +void +nv10_screen_init_transfer_functions(struct pipe_screen *pscreen); + #endif diff --git a/src/gallium/drivers/nv10/nv10_state_emit.c b/src/gallium/drivers/nv10/nv10_state_emit.c index 5dec618b93..d8691ef9c6 100644 --- a/src/gallium/drivers/nv10/nv10_state_emit.c +++ b/src/gallium/drivers/nv10/nv10_state_emit.c @@ -103,7 +103,7 @@ static void nv10_state_emit_scissor(struct nv10_context* nv10) static void nv10_state_emit_framebuffer(struct nv10_context* nv10) { struct pipe_framebuffer_state* fb = nv10->framebuffer; - struct pipe_surface *rt, *zeta = NULL; + struct nv04_surface *rt, *zeta = NULL; uint32_t rt_format, w, h; int colour_format = 0, zeta_format = 0; struct nv10_miptree *nv10mt = 0; @@ -111,7 +111,7 @@ static void nv10_state_emit_framebuffer(struct nv10_context* nv10) w = fb->cbufs[0]->width; h = fb->cbufs[0]->height; colour_format = fb->cbufs[0]->format; - rt = fb->cbufs[0]; + rt = (struct nv04_surface *)fb->cbufs[0]; if (fb->zsbuf) { if (colour_format) { @@ -123,7 +123,7 @@ static void nv10_state_emit_framebuffer(struct nv10_context* nv10) } zeta_format = fb->zsbuf->format; - zeta = fb->zsbuf; + zeta = (struct nv04_surface *)fb->zsbuf; } rt_format = NV10TCL_RT_FORMAT_TYPE_LINEAR; @@ -142,18 +142,18 @@ static void nv10_state_emit_framebuffer(struct nv10_context* nv10) if (zeta) { BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1); - OUT_RING (rt->stride | (zeta->stride << 16)); + OUT_RING (rt->pitch | (zeta->pitch << 16)); } else { BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1); - OUT_RING (rt->stride | (rt->stride << 16)); + OUT_RING (rt->pitch | (rt->pitch << 16)); } - nv10mt = (struct nv10_miptree *)rt->texture; + nv10mt = (struct nv10_miptree *)rt->base.texture; nv10->rt[0] = nv10mt->buffer; if (zeta_format) { - nv10mt = (struct nv10_miptree *)zeta->texture; + nv10mt = (struct nv10_miptree *)zeta->base.texture; nv10->zeta = nv10mt->buffer; } diff --git a/src/gallium/drivers/nv10/nv10_surface.c b/src/gallium/drivers/nv10/nv10_surface.c index 2538151063..5b52246a9c 100644 --- a/src/gallium/drivers/nv10/nv10_surface.c +++ b/src/gallium/drivers/nv10/nv10_surface.c @@ -33,7 +33,7 @@ #include "util/u_tile.h" static void -nv10_surface_copy(struct pipe_context *pipe, boolean do_flip, +nv10_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) @@ -41,15 +41,6 @@ nv10_surface_copy(struct pipe_context *pipe, boolean do_flip, struct nv10_context *nv10 = nv10_context(pipe); struct nv04_surface_2d *eng2d = nv10->screen->eng2d; - if (do_flip) { - desty += height; - while (height--) { - eng2d->copy(eng2d, dest, destx, desty--, src, - srcx, srcy++, width, 1); - } - return; - } - eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height); } diff --git a/src/gallium/drivers/nv10/nv10_transfer.c b/src/gallium/drivers/nv10/nv10_transfer.c new file mode 100644 index 0000000000..5a99225409 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_transfer.c @@ -0,0 +1,196 @@ +#include <pipe/p_state.h> +#include <pipe/p_defines.h> +#include <pipe/p_inlines.h> +#include <util/u_memory.h> +#include <nouveau/nouveau_winsys.h> +#include "nv10_context.h" +#include "nv10_screen.h" +#include "nv10_state.h" + +struct nv10_transfer { + struct pipe_transfer base; + struct pipe_surface *surface; + bool direct; +}; + +static unsigned nv10_usage_tx_to_buf(unsigned tx_usage) +{ + switch (tx_usage) { + case PIPE_TRANSFER_READ: + return PIPE_BUFFER_USAGE_CPU_READ; + case PIPE_TRANSFER_WRITE: + return PIPE_BUFFER_USAGE_CPU_WRITE; + case PIPE_TRANSFER_READ_WRITE: + return PIPE_BUFFER_USAGE_CPU_READ_WRITE; + default: + assert(0); + } + + return -1; +} + +static void +nv10_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, + struct pipe_texture *template) +{ + memset(template, 0, sizeof(struct pipe_texture)); + template->target = pt->target; + template->format = pt->format; + template->width[0] = pt->width[level]; + template->height[0] = pt->height[level]; + template->depth[0] = 1; + template->block = pt->block; + template->nblocksx[0] = pt->nblocksx[level]; + template->nblocksy[0] = pt->nblocksx[level]; + template->last_level = 0; + template->compressed = pt->compressed; + template->nr_samples = pt->nr_samples; + + template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | + NOUVEAU_TEXTURE_USAGE_LINEAR; +} + +static struct pipe_transfer * +nv10_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct nv10_miptree *mt = (struct nv10_miptree *)pt; + struct nv10_transfer *tx; + struct pipe_texture tx_tex_template, *tx_tex; + + tx = CALLOC_STRUCT(nv10_transfer); + if (!tx) + return NULL; + + pipe_texture_reference(&tx->base.texture, pt); + tx->base.format = pt->format; + tx->base.x = x; + tx->base.y = y; + tx->base.width = w; + tx->base.height = h; + tx->base.block = pt->block; + tx->base.nblocksx = pt->nblocksx[level]; + tx->base.nblocksy = pt->nblocksy[level]; + tx->base.stride = mt->level[level].pitch; + tx->base.usage = usage; + tx->base.face = face; + tx->base.level = level; + tx->base.zslice = zslice; + + /* Direct access to texture */ + if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC || + debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) && + pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) + { + tx->direct = true; + tx->surface = pscreen->get_tex_surface(pscreen, pt, + 0, 0, 0, + nv10_usage_tx_to_buf(usage)); + return &tx->base; + } + + tx->direct = false; + + nv10_compatible_transfer_tex(pt, level, &tx_tex_template); + + tx_tex = pscreen->texture_create(pscreen, &tx_tex_template); + if (!tx_tex) + { + FREE(tx); + return NULL; + } + + tx->surface = pscreen->get_tex_surface(pscreen, tx_tex, + face, level, zslice, + nv10_usage_tx_to_buf(usage)); + + pipe_texture_reference(&tx_tex, NULL); + + if (!tx->surface) + { + pipe_surface_reference(&tx->surface, NULL); + FREE(tx); + return NULL; + } + + if (usage != PIPE_TRANSFER_WRITE) { + struct nv10_screen *nvscreen = nv10_screen(pscreen); + struct pipe_surface *src; + + src = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + PIPE_BUFFER_USAGE_GPU_READ); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + /* TODO: Check if SIFM can un-swizzle */ + nvscreen->eng2d->copy(nvscreen->eng2d, + tx->surface, 0, 0, + src, 0, 0, + src->width, src->height); + + pipe_surface_reference(&src, NULL); + } + + return &tx->base; +} + +static void +nv10_transfer_del(struct pipe_transfer *ptx) +{ + struct nv10_transfer *tx = (struct nv10_transfer *)ptx; + + if (!tx->direct && ptx->usage != PIPE_TRANSFER_READ) { + struct pipe_screen *pscreen = ptx->texture->screen; + struct nv10_screen *nvscreen = nv10_screen(pscreen); + struct pipe_surface *dst; + + dst = pscreen->get_tex_surface(pscreen, ptx->texture, + ptx->face, ptx->level, ptx->zslice, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + nvscreen->eng2d->copy(nvscreen->eng2d, + dst, 0, 0, + tx->surface, 0, 0, + dst->width, dst->height); + + pipe_surface_reference(&dst, NULL); + } + + pipe_surface_reference(&tx->surface, NULL); + pipe_texture_reference(&ptx->texture, NULL); + FREE(ptx); +} + +static void * +nv10_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv10_transfer *tx = (struct nv10_transfer *)ptx; + struct nv04_surface *ns = (struct nv04_surface *)tx->surface; + struct nv10_miptree *mt = (struct nv10_miptree *)tx->surface->texture; + void *map = pipe_buffer_map(pscreen, mt->buffer, + nv10_usage_tx_to_buf(ptx->usage)); + + return map + ns->base.offset + + ptx->y * ns->pitch + ptx->x * ptx->block.size; +} + +static void +nv10_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv10_transfer *tx = (struct nv10_transfer *)ptx; + struct nv10_miptree *mt = (struct nv10_miptree *)tx->surface->texture; + + pipe_buffer_unmap(pscreen, mt->buffer); +} + +void +nv10_screen_init_transfer_functions(struct pipe_screen *pscreen) +{ + pscreen->get_tex_transfer = nv10_transfer_new; + pscreen->tex_transfer_destroy = nv10_transfer_del; + pscreen->transfer_map = nv10_transfer_map; + pscreen->transfer_unmap = nv10_transfer_unmap; +} diff --git a/src/gallium/drivers/nv20/Makefile b/src/gallium/drivers/nv20/Makefile index d777fd3d8b..1305f26c59 100644 --- a/src/gallium/drivers/nv20/Makefile +++ b/src/gallium/drivers/nv20/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = nv20 -DRIVER_SOURCES = \ +C_SOURCES = \ nv20_clear.c \ nv20_context.c \ nv20_fragprog.c \ @@ -14,16 +14,8 @@ DRIVER_SOURCES = \ nv20_state.c \ nv20_state_emit.c \ nv20_surface.c \ + nv20_transfer.c \ nv20_vbo.c # nv20_vertprog.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c index ef7e9c5428..b2f29aff8d 100644 --- a/src/gallium/drivers/nv20/nv20_miptree.c +++ b/src/gallium/drivers/nv20/nv20_miptree.c @@ -9,10 +9,14 @@ static void nv20_miptree_layout(struct nv20_miptree *nv20mt) { struct pipe_texture *pt = &nv20mt->base; - boolean swizzled = FALSE; uint width = pt->width[0], height = pt->height[0]; uint offset = 0; int nr_faces, l, f; + uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER | + PIPE_TEXTURE_USAGE_DEPTH_STENCIL | + PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_PRIMARY); if (pt->target == PIPE_TEXTURE_CUBE) { nr_faces = 6; @@ -26,25 +30,31 @@ nv20_miptree_layout(struct nv20_miptree *nv20mt) pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); - if (swizzled) - nv20mt->level[l].pitch = pt->nblocksx[l] * pt->block.size; + if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) + nv20mt->level[l].pitch = align(pt->width[0] * pt->block.size, 64); else - nv20mt->level[l].pitch = pt->nblocksx[0] * pt->block.size; - nv20mt->level[l].pitch = (nv20mt->level[l].pitch + 63) & ~63; + nv20mt->level[l].pitch = pt->width[l] * pt->block.size; nv20mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); width = MAX2(1, width >> 1); height = MAX2(1, height >> 1); - } for (f = 0; f < nr_faces; f++) { - for (l = 0; l <= pt->last_level; l++) { + for (l = 0; l < pt->last_level; l++) { nv20mt->level[l].image_offset[f] = offset; - offset += nv20mt->level[l].pitch * pt->height[l]; + + if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) && + pt->width[l + 1] > 1 && pt->height[l + 1] > 1) + offset += align(nv20mt->level[l].pitch * pt->height[l], 64); + else + offset += nv20mt->level[l].pitch * pt->height[l]; } + + nv20mt->level[l].image_offset[f] = offset; + offset += nv20mt->level[l].pitch * pt->height[l]; } nv20mt->total_size = offset; @@ -66,19 +76,18 @@ nv20_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; mt->level[0].pitch = stride[0]; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(pscreen, &mt->buffer, pb); + pipe_buffer_reference(&mt->buffer, pb); return &mt->base; } static struct pipe_texture * nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) { - struct pipe_winsys *ws = screen->winsys; struct nv20_miptree *mt; unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL | NOUVEAU_BUFFER_USAGE_TEXTURE; @@ -87,7 +96,7 @@ nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) if (!mt) return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = screen; /* Swizzled textures must be POT */ @@ -96,7 +105,8 @@ nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | - PIPE_TEXTURE_USAGE_DISPLAY_TARGET)) + PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_DEPTH_STENCIL)) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC) @@ -107,7 +117,11 @@ nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_X8R8G8B8_UNORM: case PIPE_FORMAT_R16_SNORM: + { + if (debug_get_bool_option("NOUVEAU_NO_SWIZZLE", FALSE)) + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; break; + } default: mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; } @@ -118,7 +132,7 @@ nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) nv20_miptree_layout(mt); - mt->buffer = ws->buffer_create(ws, 256, buf_usage, mt->total_size); + mt->buffer = screen->buffer_create(screen, 256, buf_usage, mt->total_size); if (!mt->buffer) { FREE(mt); return NULL; @@ -128,22 +142,16 @@ nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) } static void -nv20_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt) +nv20_miptree_destroy(struct pipe_texture *pt) { - struct pipe_texture *mt = *pt; - - *pt = NULL; - if (--mt->refcount <= 0) { - struct nv20_miptree *nv20mt = (struct nv20_miptree *)mt; - int l; + struct nv20_miptree *nv20mt = (struct nv20_miptree *)pt; + int l; - pipe_buffer_reference(screen, &nv20mt->buffer, NULL); - for (l = 0; l <= mt->last_level; l++) { - if (nv20mt->level[l].image_offset) - FREE(nv20mt->level[l].image_offset); - } - FREE(nv20mt); - } + pipe_buffer_reference(&nv20mt->buffer, NULL); + for (l = 0; l <= pt->last_level; l++) { + if (nv20mt->level[l].image_offset) + FREE(nv20mt->level[l].image_offset); + } } static struct pipe_surface * @@ -152,45 +160,38 @@ nv20_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt, unsigned flags) { struct nv20_miptree *nv20mt = (struct nv20_miptree *)pt; - struct pipe_surface *ps; + struct nv04_surface *ns; - ps = CALLOC_STRUCT(pipe_surface); - if (!ps) + ns = CALLOC_STRUCT(nv04_surface); + if (!ns) return NULL; - pipe_texture_reference(&ps->texture, pt); - ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = nv20mt->level[level].pitch; - ps->usage = flags; - ps->status = PIPE_SURFACE_STATUS_DEFINED; - ps->refcount = 1; + pipe_texture_reference(&ns->base.texture, pt); + ns->base.format = pt->format; + ns->base.width = pt->width[level]; + ns->base.height = pt->height[level]; + ns->base.usage = flags; + ns->base.status = PIPE_SURFACE_STATUS_DEFINED; + pipe_reference_init(&ns->base.reference, 1); + ns->base.face = face; + ns->base.level = level; + ns->base.zslice = zslice; + ns->pitch = nv20mt->level[level].pitch; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset = nv20mt->level[level].image_offset[face]; + ns->base.offset = nv20mt->level[level].image_offset[face]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset = nv20mt->level[level].image_offset[zslice]; + ns->base.offset = nv20mt->level[level].image_offset[zslice]; } else { - ps->offset = nv20mt->level[level].image_offset[0]; + ns->base.offset = nv20mt->level[level].image_offset[0]; } - return ps; + return &ns->base; } static void -nv20_miptree_surface_release(struct pipe_screen *pscreen, - struct pipe_surface **psurface) +nv20_miptree_surface_destroy(struct pipe_surface *ps) { - struct pipe_surface *ps = *psurface; - - *psurface = NULL; - if (--ps->refcount > 0) - return; - pipe_texture_reference(&ps->texture, NULL); FREE(ps); } @@ -199,8 +200,8 @@ void nv20_screen_init_miptree_functions(struct pipe_screen *pscreen) { pscreen->texture_create = nv20_miptree_create; pscreen->texture_blanket = nv20_miptree_blanket; - pscreen->texture_release = nv20_miptree_release; + pscreen->texture_destroy = nv20_miptree_destroy; pscreen->get_tex_surface = nv20_miptree_surface_get; - pscreen->tex_surface_release = nv20_miptree_surface_release; + pscreen->tex_surface_destroy = nv20_miptree_surface_destroy; } diff --git a/src/gallium/drivers/nv20/nv20_prim_vbuf.c b/src/gallium/drivers/nv20/nv20_prim_vbuf.c index 4dd7052814..8aa342cd2d 100644 --- a/src/gallium/drivers/nv20/nv20_prim_vbuf.c +++ b/src/gallium/drivers/nv20/nv20_prim_vbuf.c @@ -38,7 +38,7 @@ */ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" @@ -109,18 +109,15 @@ nv20__allocate_mbuffer(struct nv20_vbuf_render *nv20_render, size_t size) return nv20_render->mbuffer; } -static void * +static void nv20__allocate_pbuffer(struct nv20_vbuf_render *nv20_render, size_t size) { - struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys; - nv20_render->pbuffer = winsys->buffer_create(winsys, 64, + struct pipe_screen *screen = nv20_render->nv20->pipe.screen; + nv20_render->pbuffer = screen->buffer_create(screen, 64, PIPE_BUFFER_USAGE_VERTEX, size); - return winsys->buffer_map(winsys, - nv20_render->pbuffer, - PIPE_BUFFER_USAGE_CPU_WRITE); } -static void * +static boolean nv20_vbuf_render_allocate_vertices( struct vbuf_render *render, ushort vertex_size, ushort nr_vertices ) @@ -137,15 +134,49 @@ nv20_vbuf_render_allocate_vertices( struct vbuf_render *render, * buffer, the data will be passed directly via the fifo. */ /* XXX: Pipe vertex buffers don't work. */ - if (0 && size > 16 * 1024) - buf = nv20__allocate_pbuffer(nv20_render, size); - else + if (0 && size > 16 * 1024) { + nv20__allocate_pbuffer(nv20_render, size); + /* umm yeah so this is ugly */ + buf = nv20_render->pbuffer; + } else { buf = nv20__allocate_mbuffer(nv20_render, size); + } if (buf) nv20_render->nv20->dirty |= NV20_NEW_VTXARRAYS; - return buf; + return buf ? TRUE : FALSE; +} + +static void * +nv20_vbuf_render_map_vertices( struct vbuf_render *render ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys; + + if (nv20_render->pbuffer) { + return winsys->buffer_map(winsys, + nv20_render->pbuffer, + PIPE_BUFFER_USAGE_CPU_WRITE); + } else if (nv20_render->mbuffer) { + return nv20_render->mbuffer; + } else + assert(0); + + /* warnings be gone */ + return NULL; +} + +static void +nv20_vbuf_render_unmap_vertices( struct vbuf_render *render, + ushort min_index, + ushort max_index ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys; + + if (nv20_render->pbuffer) + winsys->buffer_unmap(winsys, nv20_render->pbuffer); } static boolean @@ -323,19 +354,14 @@ nv20_vbuf_render_draw( struct vbuf_render *render, static void -nv20_vbuf_render_release_vertices( struct vbuf_render *render, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ) +nv20_vbuf_render_release_vertices( struct vbuf_render *render ) { struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); struct nv20_context *nv20 = nv20_render->nv20; - struct pipe_winsys *winsys = nv20->pipe.winsys; struct pipe_screen *pscreen = &nv20->screen->pipe; if (nv20_render->pbuffer) { - winsys->buffer_unmap(winsys, nv20_render->pbuffer); - pipe_buffer_reference(pscreen, &nv20_render->pbuffer, NULL); + pipe_buffer_reference(&nv20_render->pbuffer, NULL); } else if (nv20_render->mbuffer) { FREE(nv20_render->mbuffer); nv20_render->mbuffer = NULL; @@ -371,6 +397,8 @@ nv20_vbuf_render_create( struct nv20_context *nv20 ) nv20_render->base.get_vertex_info = nv20_vbuf_render_get_vertex_info; nv20_render->base.allocate_vertices = nv20_vbuf_render_allocate_vertices; + nv20_render->base.map_vertices = nv20_vbuf_render_map_vertices; + nv20_render->base.unmap_vertices = nv20_vbuf_render_unmap_vertices; nv20_render->base.set_primitive = nv20_vbuf_render_set_primitive; nv20_render->base.draw = nv20_vbuf_render_draw; nv20_render->base.release_vertices = nv20_vbuf_render_release_vertices; diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c index 5f2b7b4f71..7760ae27c0 100644 --- a/src/gallium/drivers/nv20/nv20_screen.c +++ b/src/gallium/drivers/nv20/nv20_screen.c @@ -116,30 +116,6 @@ nv20_screen_is_format_supported(struct pipe_screen *screen, return FALSE; } -static void * -nv20_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, - unsigned flags ) -{ - struct pipe_winsys *ws = screen->winsys; - void *map; - struct nv20_miptree *nv20mt = (struct nv20_miptree *)surface->texture; - - map = ws->buffer_map(ws, nv20mt->buffer, flags); - if (!map) - return NULL; - - return map + surface->offset; -} - -static void -nv20_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) -{ - struct pipe_winsys *ws = screen->winsys; - struct nv20_miptree *nv20mt = (struct nv20_miptree *)surface->texture; - - ws->buffer_unmap(ws, nv20mt->buffer); -} - static void nv20_screen_destroy(struct pipe_screen *pscreen) { @@ -211,10 +187,8 @@ nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) screen->pipe.is_format_supported = nv20_screen_is_format_supported; - screen->pipe.surface_map = nv20_surface_map; - screen->pipe.surface_unmap = nv20_surface_unmap; - nv20_screen_init_miptree_functions(&screen->pipe); + nv20_screen_init_transfer_functions(&screen->pipe); u_simple_screen_init(&screen->pipe); return &screen->pipe; diff --git a/src/gallium/drivers/nv20/nv20_screen.h b/src/gallium/drivers/nv20/nv20_screen.h index bf2f2c0d9f..d9fce2bced 100644 --- a/src/gallium/drivers/nv20/nv20_screen.h +++ b/src/gallium/drivers/nv20/nv20_screen.h @@ -21,4 +21,8 @@ nv20_screen(struct pipe_screen *screen) return (struct nv20_screen *)screen; } + +void +nv20_screen_init_transfer_functions(struct pipe_screen *pscreen); + #endif diff --git a/src/gallium/drivers/nv20/nv20_state_emit.c b/src/gallium/drivers/nv20/nv20_state_emit.c index 0f4df9ca31..4042f46d05 100644 --- a/src/gallium/drivers/nv20/nv20_state_emit.c +++ b/src/gallium/drivers/nv20/nv20_state_emit.c @@ -109,7 +109,7 @@ static void nv20_state_emit_scissor(struct nv20_context* nv20) static void nv20_state_emit_framebuffer(struct nv20_context* nv20) { struct pipe_framebuffer_state* fb = nv20->framebuffer; - struct pipe_surface *rt, *zeta = NULL; + struct nv04_surface *rt, *zeta = NULL; uint32_t rt_format, w, h; int colour_format = 0, zeta_format = 0; struct nv20_miptree *nv20mt = 0; @@ -117,7 +117,7 @@ static void nv20_state_emit_framebuffer(struct nv20_context* nv20) w = fb->cbufs[0]->width; h = fb->cbufs[0]->height; colour_format = fb->cbufs[0]->format; - rt = fb->cbufs[0]; + rt = (struct nv04_surface *)fb->cbufs[0]; if (fb->zsbuf) { if (colour_format) { @@ -129,7 +129,7 @@ static void nv20_state_emit_framebuffer(struct nv20_context* nv20) } zeta_format = fb->zsbuf->format; - zeta = fb->zsbuf; + zeta = (struct nv04_surface *)fb->zsbuf; } rt_format = NV20TCL_RT_FORMAT_TYPE_LINEAR | 0x20; @@ -148,18 +148,18 @@ static void nv20_state_emit_framebuffer(struct nv20_context* nv20) if (zeta) { BEGIN_RING(kelvin, NV20TCL_RT_PITCH, 1); - OUT_RING (rt->stride | (zeta->stride << 16)); + OUT_RING (rt->pitch | (zeta->pitch << 16)); } else { BEGIN_RING(kelvin, NV20TCL_RT_PITCH, 1); - OUT_RING (rt->stride | (rt->stride << 16)); + OUT_RING (rt->pitch | (rt->pitch << 16)); } - nv20mt = (struct nv20_miptree *)rt->texture; + nv20mt = (struct nv20_miptree *)rt->base.texture; nv20->rt[0] = nv20mt->buffer; if (zeta_format) { - nv20mt = (struct nv20_miptree *)zeta->texture; + nv20mt = (struct nv20_miptree *)zeta->base.texture; nv20->zeta = nv20mt->buffer; } diff --git a/src/gallium/drivers/nv20/nv20_surface.c b/src/gallium/drivers/nv20/nv20_surface.c index 6cd607583c..4224bdd6af 100644 --- a/src/gallium/drivers/nv20/nv20_surface.c +++ b/src/gallium/drivers/nv20/nv20_surface.c @@ -33,7 +33,7 @@ #include "util/u_tile.h" static void -nv20_surface_copy(struct pipe_context *pipe, boolean do_flip, +nv20_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) @@ -41,15 +41,6 @@ nv20_surface_copy(struct pipe_context *pipe, boolean do_flip, struct nv20_context *nv20 = nv20_context(pipe); struct nv04_surface_2d *eng2d = nv20->screen->eng2d; - if (do_flip) { - desty += height; - while (height--) { - eng2d->copy(eng2d, dest, destx, desty--, src, - srcx, srcy++, width, 1); - } - return; - } - eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height); } diff --git a/src/gallium/drivers/nv20/nv20_transfer.c b/src/gallium/drivers/nv20/nv20_transfer.c new file mode 100644 index 0000000000..e5255296aa --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_transfer.c @@ -0,0 +1,196 @@ +#include <pipe/p_state.h> +#include <pipe/p_defines.h> +#include <pipe/p_inlines.h> +#include <util/u_memory.h> +#include <nouveau/nouveau_winsys.h> +#include "nv20_context.h" +#include "nv20_screen.h" +#include "nv20_state.h" + +struct nv20_transfer { + struct pipe_transfer base; + struct pipe_surface *surface; + bool direct; +}; + +static unsigned nv20_usage_tx_to_buf(unsigned tx_usage) +{ + switch (tx_usage) { + case PIPE_TRANSFER_READ: + return PIPE_BUFFER_USAGE_CPU_READ; + case PIPE_TRANSFER_WRITE: + return PIPE_BUFFER_USAGE_CPU_WRITE; + case PIPE_TRANSFER_READ_WRITE: + return PIPE_BUFFER_USAGE_CPU_READ_WRITE; + default: + assert(0); + } + + return -1; +} + +static void +nv20_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, + struct pipe_texture *template) +{ + memset(template, 0, sizeof(struct pipe_texture)); + template->target = pt->target; + template->format = pt->format; + template->width[0] = pt->width[level]; + template->height[0] = pt->height[level]; + template->depth[0] = 1; + template->block = pt->block; + template->nblocksx[0] = pt->nblocksx[level]; + template->nblocksy[0] = pt->nblocksx[level]; + template->last_level = 0; + template->compressed = pt->compressed; + template->nr_samples = pt->nr_samples; + + template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | + NOUVEAU_TEXTURE_USAGE_LINEAR; +} + +static struct pipe_transfer * +nv20_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct nv20_miptree *mt = (struct nv20_miptree *)pt; + struct nv20_transfer *tx; + struct pipe_texture tx_tex_template, *tx_tex; + + tx = CALLOC_STRUCT(nv20_transfer); + if (!tx) + return NULL; + + pipe_texture_reference(&tx->base.texture, pt); + tx->base.format = pt->format; + tx->base.x = x; + tx->base.y = y; + tx->base.width = w; + tx->base.height = h; + tx->base.block = pt->block; + tx->base.nblocksx = pt->nblocksx[level]; + tx->base.nblocksy = pt->nblocksy[level]; + tx->base.stride = mt->level[level].pitch; + tx->base.usage = usage; + tx->base.face = face; + tx->base.level = level; + tx->base.zslice = zslice; + + /* Direct access to texture */ + if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC || + debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) && + pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) + { + tx->direct = true; + tx->surface = pscreen->get_tex_surface(pscreen, pt, + 0, 0, 0, + nv20_usage_tx_to_buf(usage)); + return &tx->base; + } + + tx->direct = false; + + nv20_compatible_transfer_tex(pt, level, &tx_tex_template); + + tx_tex = pscreen->texture_create(pscreen, &tx_tex_template); + if (!tx_tex) + { + FREE(tx); + return NULL; + } + + tx->surface = pscreen->get_tex_surface(pscreen, tx_tex, + face, level, zslice, + nv20_usage_tx_to_buf(usage)); + + pipe_texture_reference(&tx_tex, NULL); + + if (!tx->surface) + { + pipe_surface_reference(&tx->surface, NULL); + FREE(tx); + return NULL; + } + + if (usage != PIPE_TRANSFER_WRITE) { + struct nv20_screen *nvscreen = nv20_screen(pscreen); + struct pipe_surface *src; + + src = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + PIPE_BUFFER_USAGE_GPU_READ); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + /* TODO: Check if SIFM can un-swizzle */ + nvscreen->eng2d->copy(nvscreen->eng2d, + tx->surface, 0, 0, + src, 0, 0, + src->width, src->height); + + pipe_surface_reference(&src, NULL); + } + + return &tx->base; +} + +static void +nv20_transfer_del(struct pipe_transfer *ptx) +{ + struct nv20_transfer *tx = (struct nv20_transfer *)ptx; + + if (!tx->direct && ptx->usage != PIPE_TRANSFER_READ) { + struct pipe_screen *pscreen = ptx->texture->screen; + struct nv20_screen *nvscreen = nv20_screen(pscreen); + struct pipe_surface *dst; + + dst = pscreen->get_tex_surface(pscreen, ptx->texture, + ptx->face, ptx->level, ptx->zslice, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + nvscreen->eng2d->copy(nvscreen->eng2d, + dst, 0, 0, + tx->surface, 0, 0, + dst->width, dst->height); + + pipe_surface_reference(&dst, NULL); + } + + pipe_surface_reference(&tx->surface, NULL); + pipe_texture_reference(&ptx->texture, NULL); + FREE(ptx); +} + +static void * +nv20_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv20_transfer *tx = (struct nv20_transfer *)ptx; + struct nv04_surface *ns = (struct nv04_surface *)tx->surface; + struct nv20_miptree *mt = (struct nv20_miptree *)tx->surface->texture; + void *map = pipe_buffer_map(pscreen, mt->buffer, + nv20_usage_tx_to_buf(ptx->usage)); + + return map + ns->base.offset + + ptx->y * ns->pitch + ptx->x * ptx->block.size; +} + +static void +nv20_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv20_transfer *tx = (struct nv20_transfer *)ptx; + struct nv20_miptree *mt = (struct nv20_miptree *)tx->surface->texture; + + pipe_buffer_unmap(pscreen, mt->buffer); +} + +void +nv20_screen_init_transfer_functions(struct pipe_screen *pscreen) +{ + pscreen->get_tex_transfer = nv20_transfer_new; + pscreen->tex_transfer_destroy = nv20_transfer_del; + pscreen->transfer_map = nv20_transfer_map; + pscreen->transfer_unmap = nv20_transfer_unmap; +} diff --git a/src/gallium/drivers/nv30/Makefile b/src/gallium/drivers/nv30/Makefile index 69f2790dfe..364c80d8f3 100644 --- a/src/gallium/drivers/nv30/Makefile +++ b/src/gallium/drivers/nv30/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = nv30 -DRIVER_SOURCES = \ +C_SOURCES = \ nv30_clear.c \ nv30_context.c \ nv30_draw.c \ @@ -22,16 +22,8 @@ DRIVER_SOURCES = \ nv30_state_viewport.c \ nv30_state_zsa.c \ nv30_surface.c \ + nv30_transfer.c \ nv30_vbo.c \ nv30_vertprog.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c index 320ba3f4bf..bdfe1425d2 100644 --- a/src/gallium/drivers/nv30/nv30_fragprog.c +++ b/src/gallium/drivers/nv30/nv30_fragprog.c @@ -834,6 +834,7 @@ nv30_fragprog_validate(struct nv30_context *nv30) struct nv30_fragment_program *fp = nv30->fragprog; struct pipe_buffer *constbuf = nv30->constbuf[PIPE_SHADER_FRAGMENT]; + struct pipe_screen *screen = nv30->pipe.screen; struct pipe_winsys *ws = nv30->pipe.winsys; struct nouveau_stateobj *so; boolean new_consts = FALSE; @@ -849,7 +850,7 @@ nv30_fragprog_validate(struct nv30_context *nv30) return FALSE; } - fp->buffer = ws->buffer_create(ws, 0x100, 0, fp->insn_len * 4); + fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4); nv30_fragprog_upload(nv30, fp); so = so_new(8, 1); @@ -864,6 +865,7 @@ nv30_fragprog_validate(struct nv30_context *nv30) so_method(so, nv30->screen->rankine, NV34TCL_TX_UNITS_ENABLE, 1); so_data (so, fp->samplers); so_ref(so, &fp->so); + so_ref(NULL, &so); update_constants: if (fp->nr_consts) { diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c index b1d2663af3..8b6ab992d1 100644 --- a/src/gallium/drivers/nv30/nv30_fragtex.c +++ b/src/gallium/drivers/nv30/nv30_fragtex.c @@ -137,6 +137,7 @@ nv30_fragtex_validate(struct nv30_context *nv30) so_method(so, nv30->screen->rankine, NV34TCL_TX_ENABLE(unit), 1); so_data (so, 0); so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]); + so_ref(NULL, &so); state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit)); } @@ -147,6 +148,7 @@ nv30_fragtex_validate(struct nv30_context *nv30) so = nv30_fragtex_build(nv30, unit); so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]); + so_ref(NULL, &so); state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit)); } diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c index 510c94d4e6..d6dc621c9e 100644 --- a/src/gallium/drivers/nv30/nv30_miptree.c +++ b/src/gallium/drivers/nv30/nv30_miptree.c @@ -67,17 +67,16 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt) static struct pipe_texture * nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) { - struct pipe_winsys *ws = pscreen->winsys; struct nv30_miptree *mt; + unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL | + NOUVEAU_BUFFER_USAGE_TEXTURE; mt = MALLOC(sizeof(struct nv30_miptree)); if (!mt) return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; - mt->shadow_tex = NULL; - mt->shadow_surface = NULL; /* Swizzled textures must be POT */ if (pt->width[0] & (pt->width[0] - 1) || @@ -85,8 +84,8 @@ nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | -+ PIPE_TEXTURE_USAGE_DISPLAY_TARGET | -+ PIPE_TEXTURE_USAGE_DEPTH_STENCIL)) + PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_DEPTH_STENCIL)) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC) @@ -107,11 +106,12 @@ nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) } } + if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC) + buf_usage |= PIPE_BUFFER_USAGE_CPU_READ_WRITE; + nv30_miptree_layout(mt); - mt->buffer = ws->buffer_create(ws, 256, - PIPE_BUFFER_USAGE_PIXEL | - NOUVEAU_BUFFER_USAGE_TEXTURE, + mt->buffer = pscreen->buffer_create(pscreen, 256, buf_usage, mt->total_size); if (!mt->buffer) { FREE(mt); @@ -137,38 +137,27 @@ nv30_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; mt->level[0].pitch = stride[0]; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(pscreen, &mt->buffer, pb); + pipe_buffer_reference(&mt->buffer, pb); return &mt->base; } static void -nv30_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) +nv30_miptree_destroy(struct pipe_texture *pt) { - struct pipe_texture *pt = *ppt; struct nv30_miptree *mt = (struct nv30_miptree *)pt; int l; - *ppt = NULL; - if (--pt->refcount) - return; - - pipe_buffer_reference(pscreen, &mt->buffer, NULL); + pipe_buffer_reference(&mt->buffer, NULL); for (l = 0; l <= pt->last_level; l++) { if (mt->level[l].image_offset) FREE(mt->level[l].image_offset); } - if (mt->shadow_tex) { - if (mt->shadow_surface) - pscreen->tex_surface_release(pscreen, &mt->shadow_surface); - nv30_miptree_release(pscreen, &mt->shadow_tex); - } - FREE(mt); } @@ -178,48 +167,38 @@ nv30_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, unsigned flags) { struct nv30_miptree *nv30mt = (struct nv30_miptree *)pt; - struct pipe_surface *ps; + struct nv04_surface *ns; - ps = CALLOC_STRUCT(pipe_surface); - if (!ps) + ns = CALLOC_STRUCT(nv04_surface); + if (!ns) return NULL; - pipe_texture_reference(&ps->texture, pt); - ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = nv30mt->level[level].pitch; - ps->usage = flags; - ps->status = PIPE_SURFACE_STATUS_DEFINED; - ps->refcount = 1; - ps->face = face; - ps->level = level; - ps->zslice = zslice; + pipe_texture_reference(&ns->base.texture, pt); + ns->base.format = pt->format; + ns->base.width = pt->width[level]; + ns->base.height = pt->height[level]; + ns->base.usage = flags; + ns->base.status = PIPE_SURFACE_STATUS_DEFINED; + pipe_reference_init(&ns->base.reference, 1); + ns->base.face = face; + ns->base.level = level; + ns->base.zslice = zslice; + ns->pitch = nv30mt->level[level].pitch; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset = nv30mt->level[level].image_offset[face]; + ns->base.offset = nv30mt->level[level].image_offset[face]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset = nv30mt->level[level].image_offset[zslice]; + ns->base.offset = nv30mt->level[level].image_offset[zslice]; } else { - ps->offset = nv30mt->level[level].image_offset[0]; + ns->base.offset = nv30mt->level[level].image_offset[0]; } - return ps; + return &ns->base; } static void -nv30_miptree_surface_del(struct pipe_screen *pscreen, - struct pipe_surface **psurface) +nv30_miptree_surface_del(struct pipe_surface *ps) { - struct pipe_surface *ps = *psurface; - - *psurface = NULL; - if (--ps->refcount > 0) - return; - pipe_texture_reference(&ps->texture, NULL); FREE(ps); } @@ -229,7 +208,7 @@ nv30_screen_init_miptree_functions(struct pipe_screen *pscreen) { pscreen->texture_create = nv30_miptree_create; pscreen->texture_blanket = nv30_miptree_blanket; - pscreen->texture_release = nv30_miptree_release; + pscreen->texture_destroy = nv30_miptree_destroy; pscreen->get_tex_surface = nv30_miptree_surface_new; - pscreen->tex_surface_release = nv30_miptree_surface_del; + pscreen->tex_surface_destroy = nv30_miptree_surface_del; } diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c index c97a73f0b1..d395c5e1b7 100644 --- a/src/gallium/drivers/nv30/nv30_screen.c +++ b/src/gallium/drivers/nv30/nv30_screen.c @@ -135,82 +135,6 @@ nv30_surface_buffer(struct pipe_surface *surf) return mt->buffer; } -static void * -nv30_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, - unsigned flags ) -{ - struct pipe_winsys *ws = screen->winsys; - struct pipe_surface *surface_to_map; - void *map; - - if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { - struct nv30_miptree *mt = (struct nv30_miptree *)surface->texture; - - if (!mt->shadow_tex) { - unsigned old_tex_usage = surface->texture->tex_usage; - surface->texture->tex_usage = NOUVEAU_TEXTURE_USAGE_LINEAR | - PIPE_TEXTURE_USAGE_DYNAMIC; - mt->shadow_tex = screen->texture_create(screen, surface->texture); - surface->texture->tex_usage = old_tex_usage; - - assert(mt->shadow_tex->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR); - } - - mt->shadow_surface = screen->get_tex_surface - ( - screen, mt->shadow_tex, - surface->face, surface->level, surface->zslice, - surface->usage - ); - - surface_to_map = mt->shadow_surface; - } - else - surface_to_map = surface; - - assert(surface_to_map); - - map = ws->buffer_map(ws, nv30_surface_buffer(surface_to_map), flags); - if (!map) - return NULL; - - return map + surface_to_map->offset; -} - -static void -nv30_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) -{ - struct pipe_winsys *ws = screen->winsys; - struct pipe_surface *surface_to_unmap; - - /* TODO: Copy from shadow just before push buffer is flushed instead. - There are probably some programs that map/unmap excessively - before rendering. */ - if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { - struct nv30_miptree *mt = (struct nv30_miptree *)surface->texture; - - assert(mt->shadow_tex); - - surface_to_unmap = mt->shadow_surface; - } - else - surface_to_unmap = surface; - - assert(surface_to_unmap); - - ws->buffer_unmap(ws, nv30_surface_buffer(surface_to_unmap)); - - if (surface_to_unmap != surface) { - struct nv30_screen *nvscreen = nv30_screen(screen); - - nvscreen->eng2d->copy(nvscreen->eng2d, surface, 0, 0, - surface_to_unmap, 0, 0, - surface->width, surface->height); - - screen->tex_surface_release(screen, &surface_to_unmap); - } -} - static void nv30_screen_destroy(struct pipe_screen *pscreen) { @@ -391,10 +315,8 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) screen->pipe.is_format_supported = nv30_screen_surface_format_supported; - screen->pipe.surface_map = nv30_surface_map; - screen->pipe.surface_unmap = nv30_surface_unmap; - nv30_screen_init_miptree_functions(&screen->pipe); + nv30_screen_init_transfer_functions(&screen->pipe); u_simple_screen_init(&screen->pipe); return &screen->pipe; diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nv30/nv30_screen.h index b11e470f94..8e36883975 100644 --- a/src/gallium/drivers/nv30/nv30_screen.h +++ b/src/gallium/drivers/nv30/nv30_screen.h @@ -34,4 +34,7 @@ nv30_screen(struct pipe_screen *screen) return (struct nv30_screen *)screen; } +void +nv30_screen_init_transfer_functions(struct pipe_screen *pscreen); + #endif diff --git a/src/gallium/drivers/nv30/nv30_state.c b/src/gallium/drivers/nv30/nv30_state.c index 26147565a5..b91e972c12 100644 --- a/src/gallium/drivers/nv30/nv30_state.c +++ b/src/gallium/drivers/nv30/nv30_state.c @@ -51,6 +51,7 @@ nv30_blend_state_create(struct pipe_context *pipe, so_data (so, cso->dither ? 1 : 0); so_ref(so, &bso->so); + so_ref(NULL, &so); bso->pipe = *cso; return (void *)bso; } @@ -404,6 +405,7 @@ nv30_rasterizer_state_create(struct pipe_context *pipe, } so_ref(so, &rsso->so); + so_ref(NULL, &so); rsso->pipe = *cso; return (void *)rsso; } @@ -477,6 +479,7 @@ nv30_depth_stencil_alpha_state_create(struct pipe_context *pipe, } so_ref(so, &zsaso->so); + so_ref(NULL, &so); zsaso->pipe = *cso; return (void *)zsaso; } diff --git a/src/gallium/drivers/nv30/nv30_state.h b/src/gallium/drivers/nv30/nv30_state.h index 2023278e37..e6f23bf166 100644 --- a/src/gallium/drivers/nv30/nv30_state.h +++ b/src/gallium/drivers/nv30/nv30_state.h @@ -76,9 +76,6 @@ struct nv30_miptree { struct pipe_buffer *buffer; uint total_size; - struct pipe_texture *shadow_tex; - struct pipe_surface *shadow_surface; - struct { uint pitch; uint *image_offset; diff --git a/src/gallium/drivers/nv30/nv30_state_blend.c b/src/gallium/drivers/nv30/nv30_state_blend.c index 44d43e132a..64cf9ae93a 100644 --- a/src/gallium/drivers/nv30/nv30_state_blend.c +++ b/src/gallium/drivers/nv30/nv30_state_blend.c @@ -28,6 +28,7 @@ nv30_state_blend_colour_validate(struct nv30_context *nv30) (float_to_ubyte(bcol->color[2]) << 0))); so_ref(so, &nv30->state.hw[NV30_STATE_BCOL]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c index 77368cb205..fdc1cade90 100644 --- a/src/gallium/drivers/nv30/nv30_state_fb.c +++ b/src/gallium/drivers/nv30/nv30_state_fb.c @@ -5,7 +5,7 @@ static boolean nv30_state_framebuffer_validate(struct nv30_context *nv30) { struct pipe_framebuffer_state *fb = &nv30->framebuffer; - struct pipe_surface *rt[2], *zeta = NULL; + struct nv04_surface *rt[2], *zeta = NULL; uint32_t rt_enable, rt_format; int i, colour_format = 0, zeta_format = 0; struct nouveau_stateobj *so = so_new(64, 10); @@ -21,7 +21,7 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) } else { colour_format = fb->cbufs[i]->format; rt_enable |= (NV34TCL_RT_ENABLE_COLOR0 << i); - rt[i] = fb->cbufs[i]; + rt[i] = (struct nv04_surface *)fb->cbufs[i]; } } @@ -30,13 +30,13 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) if (fb->zsbuf) { zeta_format = fb->zsbuf->format; - zeta = fb->zsbuf; + zeta = (struct nv04_surface *)fb->zsbuf; } - if (!(rt[0]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1))); for (i = 1; i < fb->nr_cbufs; i++) - assert(!(rt[i]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)); + assert(!(rt[i]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)); /* FIXME: NV34TCL_RT_FORMAT_LOG2_[WIDTH/HEIGHT] */ rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED | @@ -71,44 +71,44 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) } if (rt_enable & NV34TCL_RT_ENABLE_COLOR0) { - uint32_t pitch = rt[0]->stride; + uint32_t pitch = rt[0]->pitch; if (zeta) { - pitch |= (zeta->stride << 16); + pitch |= (zeta->pitch << 16); } else { pitch |= (pitch << 16); } - nv30mt = (struct nv30_miptree *)rt[0]->texture; + nv30mt = (struct nv30_miptree *)rt[0]->base.texture; so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR0, 1); so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR, nv30->nvws->channel->vram->handle, nv30->nvws->channel->gart->handle); so_method(so, nv30->screen->rankine, NV34TCL_COLOR0_PITCH, 2); so_data (so, pitch); - so_reloc (so, nv30mt->buffer, rt[0]->offset, rt_flags | + so_reloc (so, nv30mt->buffer, rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); } if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) { - nv30mt = (struct nv30_miptree *)rt[1]->texture; + nv30mt = (struct nv30_miptree *)rt[1]->base.texture; so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR1, 1); so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR, nv30->nvws->channel->vram->handle, nv30->nvws->channel->gart->handle); so_method(so, nv30->screen->rankine, NV34TCL_COLOR1_OFFSET, 2); - so_reloc (so, nv30mt->buffer, rt[1]->offset, rt_flags | + so_reloc (so, nv30mt->buffer, rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); - so_data (so, rt[1]->stride); + so_data (so, rt[1]->pitch); } if (zeta_format) { - nv30mt = (struct nv30_miptree *)zeta->texture; + nv30mt = (struct nv30_miptree *)zeta->base.texture; so_method(so, nv30->screen->rankine, NV34TCL_DMA_ZETA, 1); so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR, nv30->nvws->channel->vram->handle, nv30->nvws->channel->gart->handle); so_method(so, nv30->screen->rankine, NV34TCL_ZETA_OFFSET, 1); - so_reloc (so, nv30mt->buffer, zeta->offset, rt_flags | + so_reloc (so, nv30mt->buffer, zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); /* TODO: allocate LMA depth buffer */ } @@ -132,6 +132,7 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) so_data (so, 0); so_ref(so, &nv30->state.hw[NV30_STATE_FB]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv30/nv30_state_scissor.c b/src/gallium/drivers/nv30/nv30_state_scissor.c index 1db9bc1795..3ac7a8471e 100644 --- a/src/gallium/drivers/nv30/nv30_state_scissor.c +++ b/src/gallium/drivers/nv30/nv30_state_scissor.c @@ -23,6 +23,7 @@ nv30_state_scissor_validate(struct nv30_context *nv30) } so_ref(so, &nv30->state.hw[NV30_STATE_SCISSOR]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv30/nv30_state_stipple.c b/src/gallium/drivers/nv30/nv30_state_stipple.c index 41b42813b4..d0c791ac08 100644 --- a/src/gallium/drivers/nv30/nv30_state_stipple.c +++ b/src/gallium/drivers/nv30/nv30_state_stipple.c @@ -27,6 +27,7 @@ nv30_state_stipple_validate(struct nv30_context *nv30) } so_ref(so, &nv30->state.hw[NV30_STATE_STIPPLE]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv30/nv30_state_viewport.c b/src/gallium/drivers/nv30/nv30_state_viewport.c index 951d40ebfd..c3eb413dac 100644 --- a/src/gallium/drivers/nv30/nv30_state_viewport.c +++ b/src/gallium/drivers/nv30/nv30_state_viewport.c @@ -7,7 +7,8 @@ nv30_state_viewport_validate(struct nv30_context *nv30) struct nouveau_stateobj *so; unsigned bypass; - if (/*nv30->render_mode == HW &&*/ !nv30->rasterizer->pipe.bypass_clipping) + if (/*nv30->render_mode == HW &&*/ + !nv30->rasterizer->pipe.bypass_vs_clip_and_viewport) bypass = 0; else bypass = 1; @@ -58,6 +59,7 @@ nv30_state_viewport_validate(struct nv30_context *nv30) so_data (so, 1); so_ref(so, &nv30->state.hw[NV30_STATE_VIEWPORT]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv30/nv30_surface.c b/src/gallium/drivers/nv30/nv30_surface.c index 0f8dc12045..5e237e13eb 100644 --- a/src/gallium/drivers/nv30/nv30_surface.c +++ b/src/gallium/drivers/nv30/nv30_surface.c @@ -33,7 +33,7 @@ #include "util/u_tile.h" static void -nv30_surface_copy(struct pipe_context *pipe, boolean do_flip, +nv30_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) @@ -41,15 +41,6 @@ nv30_surface_copy(struct pipe_context *pipe, boolean do_flip, struct nv30_context *nv30 = nv30_context(pipe); struct nv04_surface_2d *eng2d = nv30->screen->eng2d; - if (do_flip) { - desty += height; - while (height--) { - eng2d->copy(eng2d, dest, destx, desty--, src, - srcx, srcy++, width, 1); - } - return; - } - eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height); } diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nv30/nv30_transfer.c new file mode 100644 index 0000000000..8b915b35bd --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_transfer.c @@ -0,0 +1,196 @@ +#include <pipe/p_state.h> +#include <pipe/p_defines.h> +#include <pipe/p_inlines.h> +#include <util/u_memory.h> +#include <nouveau/nouveau_winsys.h> +#include "nv30_context.h" +#include "nv30_screen.h" +#include "nv30_state.h" + +struct nv30_transfer { + struct pipe_transfer base; + struct pipe_surface *surface; + bool direct; +}; + +static unsigned nv30_usage_tx_to_buf(unsigned tx_usage) +{ + switch (tx_usage) { + case PIPE_TRANSFER_READ: + return PIPE_BUFFER_USAGE_CPU_READ; + case PIPE_TRANSFER_WRITE: + return PIPE_BUFFER_USAGE_CPU_WRITE; + case PIPE_TRANSFER_READ_WRITE: + return PIPE_BUFFER_USAGE_CPU_READ_WRITE; + default: + assert(0); + } + + return -1; +} + +static void +nv30_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, + struct pipe_texture *template) +{ + memset(template, 0, sizeof(struct pipe_texture)); + template->target = pt->target; + template->format = pt->format; + template->width[0] = pt->width[level]; + template->height[0] = pt->height[level]; + template->depth[0] = 1; + template->block = pt->block; + template->nblocksx[0] = pt->nblocksx[level]; + template->nblocksy[0] = pt->nblocksx[level]; + template->last_level = 0; + template->compressed = pt->compressed; + template->nr_samples = pt->nr_samples; + + template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | + NOUVEAU_TEXTURE_USAGE_LINEAR; +} + +static struct pipe_transfer * +nv30_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct nv30_miptree *mt = (struct nv30_miptree *)pt; + struct nv30_transfer *tx; + struct pipe_texture tx_tex_template, *tx_tex; + + tx = CALLOC_STRUCT(nv30_transfer); + if (!tx) + return NULL; + + pipe_texture_reference(&tx->base.texture, pt); + tx->base.format = pt->format; + tx->base.x = x; + tx->base.y = y; + tx->base.width = w; + tx->base.height = h; + tx->base.block = pt->block; + tx->base.nblocksx = pt->nblocksx[level]; + tx->base.nblocksy = pt->nblocksy[level]; + tx->base.stride = mt->level[level].pitch; + tx->base.usage = usage; + tx->base.face = face; + tx->base.level = level; + tx->base.zslice = zslice; + + /* Direct access to texture */ + if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC || + debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) && + pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) + { + tx->direct = true; + tx->surface = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + nv30_usage_tx_to_buf(usage)); + return &tx->base; + } + + tx->direct = false; + + nv30_compatible_transfer_tex(pt, level, &tx_tex_template); + + tx_tex = pscreen->texture_create(pscreen, &tx_tex_template); + if (!tx_tex) + { + FREE(tx); + return NULL; + } + + tx->surface = pscreen->get_tex_surface(pscreen, tx_tex, + 0, 0, 0, + nv30_usage_tx_to_buf(usage)); + + pipe_texture_reference(&tx_tex, NULL); + + if (!tx->surface) + { + pipe_surface_reference(&tx->surface, NULL); + FREE(tx); + return NULL; + } + + if (usage != PIPE_TRANSFER_WRITE) { + struct nv30_screen *nvscreen = nv30_screen(pscreen); + struct pipe_surface *src; + + src = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + PIPE_BUFFER_USAGE_GPU_READ); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + /* TODO: Check if SIFM can un-swizzle */ + nvscreen->eng2d->copy(nvscreen->eng2d, + tx->surface, 0, 0, + src, 0, 0, + src->width, src->height); + + pipe_surface_reference(&src, NULL); + } + + return &tx->base; +} + +static void +nv30_transfer_del(struct pipe_transfer *ptx) +{ + struct nv30_transfer *tx = (struct nv30_transfer *)ptx; + + if (!tx->direct && ptx->usage != PIPE_TRANSFER_READ) { + struct pipe_screen *pscreen = ptx->texture->screen; + struct nv30_screen *nvscreen = nv30_screen(pscreen); + struct pipe_surface *dst; + + dst = pscreen->get_tex_surface(pscreen, ptx->texture, + ptx->face, ptx->level, ptx->zslice, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + nvscreen->eng2d->copy(nvscreen->eng2d, + dst, 0, 0, + tx->surface, 0, 0, + dst->width, dst->height); + + pipe_surface_reference(&dst, NULL); + } + + pipe_surface_reference(&tx->surface, NULL); + pipe_texture_reference(&ptx->texture, NULL); + FREE(ptx); +} + +static void * +nv30_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv30_transfer *tx = (struct nv30_transfer *)ptx; + struct nv04_surface *ns = (struct nv04_surface *)tx->surface; + struct nv30_miptree *mt = (struct nv30_miptree *)tx->surface->texture; + void *map = pipe_buffer_map(pscreen, mt->buffer, + nv30_usage_tx_to_buf(ptx->usage)); + + return map + ns->base.offset + + ptx->y * ns->pitch + ptx->x * ptx->block.size; +} + +static void +nv30_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv30_transfer *tx = (struct nv30_transfer *)ptx; + struct nv30_miptree *mt = (struct nv30_miptree *)tx->surface->texture; + + pipe_buffer_unmap(pscreen, mt->buffer); +} + +void +nv30_screen_init_transfer_functions(struct pipe_screen *pscreen) +{ + pscreen->get_tex_transfer = nv30_transfer_new; + pscreen->tex_transfer_destroy = nv30_transfer_del; + pscreen->transfer_map = nv30_transfer_map; + pscreen->transfer_unmap = nv30_transfer_unmap; +} diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c index 2d6d48ac16..990a876382 100644 --- a/src/gallium/drivers/nv30/nv30_vbo.c +++ b/src/gallium/drivers/nv30/nv30_vbo.c @@ -539,10 +539,13 @@ nv30_vbo_validate(struct nv30_context *nv30) so_data (vtxbuf, 0); so_ref(vtxbuf, &nv30->state.hw[NV30_STATE_VTXBUF]); + so_ref(NULL, &vtxbuf); nv30->state.dirty |= (1ULL << NV30_STATE_VTXBUF); so_ref(vtxfmt, &nv30->state.hw[NV30_STATE_VTXFMT]); + so_ref(NULL, &vtxfmt); nv30->state.dirty |= (1ULL << NV30_STATE_VTXFMT); so_ref(sattr, &nv30->state.hw[NV30_STATE_VTXATTR]); + so_ref(NULL, &sattr); nv30->state.dirty |= (1ULL << NV30_STATE_VTXATTR); return FALSE; } diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c index d262725057..eaf543b8f7 100644 --- a/src/gallium/drivers/nv30/nv30_vertprog.c +++ b/src/gallium/drivers/nv30/nv30_vertprog.c @@ -685,6 +685,7 @@ nv30_vertprog_validate(struct nv30_context *nv30) so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1); so_data (so, vp->exec->start); so_ref(so, &vp->so); + so_ref(NULL, &so); upload_code = TRUE; } diff --git a/src/gallium/drivers/nv40/Makefile b/src/gallium/drivers/nv40/Makefile index 9c8eadf7e4..0ecae2b491 100644 --- a/src/gallium/drivers/nv40/Makefile +++ b/src/gallium/drivers/nv40/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = nv40 -DRIVER_SOURCES = \ +C_SOURCES = \ nv40_clear.c \ nv40_context.c \ nv40_draw.c \ @@ -22,16 +22,8 @@ DRIVER_SOURCES = \ nv40_state_viewport.c \ nv40_state_zsa.c \ nv40_surface.c \ + nv40_transfer.c \ nv40_vbo.c \ nv40_vertprog.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c index 91dcbebda0..16e40889ec 100644 --- a/src/gallium/drivers/nv40/nv40_fragprog.c +++ b/src/gallium/drivers/nv40/nv40_fragprog.c @@ -917,6 +917,7 @@ nv40_fragprog_validate(struct nv40_context *nv40) struct nv40_fragment_program *fp = nv40->fragprog; struct pipe_buffer *constbuf = nv40->constbuf[PIPE_SHADER_FRAGMENT]; + struct pipe_screen *screen = nv40->pipe.screen; struct pipe_winsys *ws = nv40->pipe.winsys; struct nouveau_stateobj *so; boolean new_consts = FALSE; @@ -932,7 +933,7 @@ nv40_fragprog_validate(struct nv40_context *nv40) return FALSE; } - fp->buffer = ws->buffer_create(ws, 0x100, 0, fp->insn_len * 4); + fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4); nv40_fragprog_upload(nv40, fp); so = so_new(4, 1); @@ -943,6 +944,7 @@ nv40_fragprog_validate(struct nv40_context *nv40) so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1); so_data (so, fp->fp_control); so_ref(so, &fp->so); + so_ref(NULL, &so); update_constants: if (fp->nr_consts) { diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c index 0227d22620..eb3002dc05 100644 --- a/src/gallium/drivers/nv40/nv40_fragtex.c +++ b/src/gallium/drivers/nv40/nv40_fragtex.c @@ -151,6 +151,7 @@ nv40_fragtex_validate(struct nv40_context *nv40) so = nv40_fragtex_build(nv40, unit); so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]); + so_ref(NULL, &so); state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit)); } diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c index e38b1e7f5c..abadca8c93 100644 --- a/src/gallium/drivers/nv40/nv40_miptree.c +++ b/src/gallium/drivers/nv40/nv40_miptree.c @@ -67,7 +67,6 @@ nv40_miptree_layout(struct nv40_miptree *mt) static struct pipe_texture * nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) { - struct pipe_winsys *ws = pscreen->winsys; struct nv40_miptree *mt; unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL | NOUVEAU_BUFFER_USAGE_TEXTURE; @@ -76,10 +75,8 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) if (!mt) return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; - mt->shadow_tex = NULL; - mt->shadow_surface = NULL; /* Swizzled textures must be POT */ if (pt->width[0] & (pt->width[0] - 1) || @@ -114,7 +111,7 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) nv40_miptree_layout(mt); - mt->buffer = ws->buffer_create(ws, 256, buf_usage, mt->total_size); + mt->buffer = pscreen->buffer_create(pscreen, 256, buf_usage, mt->total_size); if (!mt->buffer) { FREE(mt); return NULL; @@ -139,38 +136,27 @@ nv40_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; mt->level[0].pitch = stride[0]; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(pscreen, &mt->buffer, pb); + pipe_buffer_reference(&mt->buffer, pb); return &mt->base; } static void -nv40_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) +nv40_miptree_destroy(struct pipe_texture *pt) { - struct pipe_texture *pt = *ppt; struct nv40_miptree *mt = (struct nv40_miptree *)pt; int l; - *ppt = NULL; - if (--pt->refcount) - return; - - pipe_buffer_reference(pscreen, &mt->buffer, NULL); + pipe_buffer_reference(&mt->buffer, NULL); for (l = 0; l <= pt->last_level; l++) { if (mt->level[l].image_offset) FREE(mt->level[l].image_offset); } - if (mt->shadow_tex) { - if (mt->shadow_surface) - pscreen->tex_surface_release(pscreen, &mt->shadow_surface); - nv40_miptree_release(pscreen, &mt->shadow_tex); - } - FREE(mt); } @@ -180,48 +166,38 @@ nv40_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, unsigned flags) { struct nv40_miptree *mt = (struct nv40_miptree *)pt; - struct pipe_surface *ps; + struct nv04_surface *ns; - ps = CALLOC_STRUCT(pipe_surface); - if (!ps) + ns = CALLOC_STRUCT(nv04_surface); + if (!ns) return NULL; - pipe_texture_reference(&ps->texture, pt); - ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = mt->level[level].pitch; - ps->usage = flags; - ps->status = PIPE_SURFACE_STATUS_DEFINED; - ps->refcount = 1; - ps->face = face; - ps->level = level; - ps->zslice = zslice; + pipe_texture_reference(&ns->base.texture, pt); + ns->base.format = pt->format; + ns->base.width = pt->width[level]; + ns->base.height = pt->height[level]; + ns->base.usage = flags; + ns->base.status = PIPE_SURFACE_STATUS_DEFINED; + pipe_reference_init(&ns->base.reference, 1); + ns->base.face = face; + ns->base.level = level; + ns->base.zslice = zslice; + ns->pitch = mt->level[level].pitch; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset = mt->level[level].image_offset[face]; + ns->base.offset = mt->level[level].image_offset[face]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset = mt->level[level].image_offset[zslice]; + ns->base.offset = mt->level[level].image_offset[zslice]; } else { - ps->offset = mt->level[level].image_offset[0]; + ns->base.offset = mt->level[level].image_offset[0]; } - return ps; + return &ns->base; } static void -nv40_miptree_surface_del(struct pipe_screen *pscreen, - struct pipe_surface **psurface) +nv40_miptree_surface_del(struct pipe_surface *ps) { - struct pipe_surface *ps = *psurface; - - *psurface = NULL; - if (--ps->refcount > 0) - return; - pipe_texture_reference(&ps->texture, NULL); FREE(ps); } @@ -231,8 +207,8 @@ nv40_screen_init_miptree_functions(struct pipe_screen *pscreen) { pscreen->texture_create = nv40_miptree_create; pscreen->texture_blanket = nv40_miptree_blanket; - pscreen->texture_release = nv40_miptree_release; + pscreen->texture_destroy = nv40_miptree_destroy; pscreen->get_tex_surface = nv40_miptree_surface_new; - pscreen->tex_surface_release = nv40_miptree_surface_del; + pscreen->tex_surface_destroy = nv40_miptree_surface_del; } diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c index 2372bc8441..0d4baefaea 100644 --- a/src/gallium/drivers/nv40/nv40_screen.c +++ b/src/gallium/drivers/nv40/nv40_screen.c @@ -144,81 +144,6 @@ nv40_surface_buffer(struct pipe_surface *surf) return mt->buffer; } -static void * -nv40_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, - unsigned flags ) -{ - struct pipe_winsys *ws = screen->winsys; - struct pipe_surface *surface_to_map; - void *map; - - if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { - struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture; - - if (!mt->shadow_tex) { - unsigned old_tex_usage = surface->texture->tex_usage; - surface->texture->tex_usage = NOUVEAU_TEXTURE_USAGE_LINEAR | - PIPE_TEXTURE_USAGE_DYNAMIC; - mt->shadow_tex = screen->texture_create(screen, surface->texture); - surface->texture->tex_usage = old_tex_usage; - - assert(mt->shadow_tex->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR); - } - - mt->shadow_surface = screen->get_tex_surface - ( - screen, mt->shadow_tex, - surface->face, surface->level, surface->zslice, - surface->usage - ); - - surface_to_map = mt->shadow_surface; - } - else - surface_to_map = surface; - - assert(surface_to_map); - map = ws->buffer_map(ws, nv40_surface_buffer(surface_to_map), flags); - if (!map) - return NULL; - - return map + surface_to_map->offset; -} - -static void -nv40_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) -{ - struct pipe_winsys *ws = screen->winsys; - struct pipe_surface *surface_to_unmap; - - /* TODO: Copy from shadow just before push buffer is flushed instead. - There are probably some programs that map/unmap excessively - before rendering. */ - if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { - struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture; - - assert(mt->shadow_tex); - - surface_to_unmap = mt->shadow_surface; - } - else - surface_to_unmap = surface; - - assert(surface_to_unmap); - - ws->buffer_unmap(ws, nv40_surface_buffer(surface_to_unmap)); - - if (surface_to_unmap != surface) { - struct nv40_screen *nvscreen = nv40_screen(screen); - - nvscreen->eng2d->copy(nvscreen->eng2d, surface, 0, 0, - surface_to_unmap, 0, 0, - surface->width, surface->height); - - screen->tex_surface_release(screen, &surface_to_unmap); - } -} - static void nv40_screen_destroy(struct pipe_screen *pscreen) { @@ -240,7 +165,7 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) { struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen); struct nouveau_stateobj *so; - unsigned curie_class; + unsigned curie_class = 0; unsigned chipset = nvws->channel->device->chipset; int ret; @@ -265,8 +190,6 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) if (NV6X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f))) curie_class = NV44TCL; break; - default: - break; } if (!curie_class) { @@ -372,10 +295,8 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) screen->pipe.is_format_supported = nv40_screen_surface_format_supported; - screen->pipe.surface_map = nv40_surface_map; - screen->pipe.surface_unmap = nv40_surface_unmap; - nv40_screen_init_miptree_functions(&screen->pipe); + nv40_screen_init_transfer_functions(&screen->pipe); u_simple_screen_init(&screen->pipe); return &screen->pipe; diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h index 4500aa0e5c..7b503bd207 100644 --- a/src/gallium/drivers/nv40/nv40_screen.h +++ b/src/gallium/drivers/nv40/nv40_screen.h @@ -34,4 +34,7 @@ nv40_screen(struct pipe_screen *screen) return (struct nv40_screen *)screen; } +void +nv40_screen_init_transfer_functions(struct pipe_screen *pscreen); + #endif diff --git a/src/gallium/drivers/nv40/nv40_state.c b/src/gallium/drivers/nv40/nv40_state.c index 2eff25aa83..c3ee4d2345 100644 --- a/src/gallium/drivers/nv40/nv40_state.c +++ b/src/gallium/drivers/nv40/nv40_state.c @@ -52,6 +52,7 @@ nv40_blend_state_create(struct pipe_context *pipe, so_data (so, cso->dither ? 1 : 0); so_ref(so, &bso->so); + so_ref(NULL, &so); bso->pipe = *cso; return (void *)bso; } @@ -414,6 +415,7 @@ nv40_rasterizer_state_create(struct pipe_context *pipe, } so_ref(so, &rsso->so); + so_ref(NULL, &so); rsso->pipe = *cso; return (void *)rsso; } @@ -487,6 +489,7 @@ nv40_depth_stencil_alpha_state_create(struct pipe_context *pipe, } so_ref(so, &zsaso->so); + so_ref(NULL, &so); zsaso->pipe = *cso; return (void *)zsaso; } diff --git a/src/gallium/drivers/nv40/nv40_state.h b/src/gallium/drivers/nv40/nv40_state.h index 9c55903ae3..8a9d8c8fdf 100644 --- a/src/gallium/drivers/nv40/nv40_state.h +++ b/src/gallium/drivers/nv40/nv40_state.h @@ -79,9 +79,6 @@ struct nv40_miptree { struct pipe_buffer *buffer; uint total_size; - struct pipe_texture *shadow_tex; - struct pipe_surface *shadow_surface; - struct { uint pitch; uint *image_offset; diff --git a/src/gallium/drivers/nv40/nv40_state_blend.c b/src/gallium/drivers/nv40/nv40_state_blend.c index 95e6d7394f..8cd05ce66e 100644 --- a/src/gallium/drivers/nv40/nv40_state_blend.c +++ b/src/gallium/drivers/nv40/nv40_state_blend.c @@ -28,6 +28,7 @@ nv40_state_blend_colour_validate(struct nv40_context *nv40) (float_to_ubyte(bcol->color[2]) << 0))); so_ref(so, &nv40->state.hw[NV40_STATE_BCOL]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c index 454abad31f..be618a306b 100644 --- a/src/gallium/drivers/nv40/nv40_state_fb.c +++ b/src/gallium/drivers/nv40/nv40_state_fb.c @@ -12,7 +12,7 @@ static boolean nv40_state_framebuffer_validate(struct nv40_context *nv40) { struct pipe_framebuffer_state *fb = &nv40->framebuffer; - struct pipe_surface *rt[4], *zeta; + struct nv04_surface *rt[4], *zeta; uint32_t rt_enable, rt_format; int i, colour_format = 0, zeta_format = 0; struct nouveau_stateobj *so = so_new(64, 10); @@ -27,7 +27,7 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) } else { colour_format = fb->cbufs[i]->format; rt_enable |= (NV40TCL_RT_ENABLE_COLOR0 << i); - rt[i] = fb->cbufs[i]; + rt[i] = (struct nv04_surface *)fb->cbufs[i]; } } @@ -37,13 +37,13 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) if (fb->zsbuf) { zeta_format = fb->zsbuf->format; - zeta = fb->zsbuf; + zeta = (struct nv04_surface *)fb->zsbuf; } - if (!(rt[0]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1))); for (i = 1; i < fb->nr_cbufs; i++) - assert(!(rt[i]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)); + assert(!(rt[i]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)); rt_format = NV40TCL_RT_FORMAT_TYPE_SWIZZLED | log2i(fb->width) << NV40TCL_RT_FORMAT_LOG2_WIDTH_SHIFT | @@ -78,60 +78,60 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) { so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR0, 1); - so_reloc (so, nv40_surface_buffer(rt[0]), 0, rt_flags | NOUVEAU_BO_OR, + so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0, rt_flags | NOUVEAU_BO_OR, nv40->nvws->channel->vram->handle, nv40->nvws->channel->gart->handle); so_method(so, nv40->screen->curie, NV40TCL_COLOR0_PITCH, 2); - so_data (so, rt[0]->stride); - so_reloc (so, nv40_surface_buffer(rt[0]), rt[0]->offset, rt_flags | + so_data (so, rt[0]->pitch); + so_reloc (so, nv40_surface_buffer(&rt[0]->base), rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); } if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) { so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR1, 1); - so_reloc (so, nv40_surface_buffer(rt[1]), 0, rt_flags | NOUVEAU_BO_OR, + so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0, rt_flags | NOUVEAU_BO_OR, nv40->nvws->channel->vram->handle, nv40->nvws->channel->gart->handle); so_method(so, nv40->screen->curie, NV40TCL_COLOR1_OFFSET, 2); - so_reloc (so, nv40_surface_buffer(rt[1]), rt[1]->offset, rt_flags | + so_reloc (so, nv40_surface_buffer(&rt[1]->base), rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); - so_data (so, rt[1]->stride); + so_data (so, rt[1]->pitch); } if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) { so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR2, 1); - so_reloc (so, nv40_surface_buffer(rt[2]), 0, rt_flags | NOUVEAU_BO_OR, + so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0, rt_flags | NOUVEAU_BO_OR, nv40->nvws->channel->vram->handle, nv40->nvws->channel->gart->handle); so_method(so, nv40->screen->curie, NV40TCL_COLOR2_OFFSET, 1); - so_reloc (so, nv40_surface_buffer(rt[2]), rt[2]->offset, rt_flags | + so_reloc (so, nv40_surface_buffer(&rt[2]->base), rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); so_method(so, nv40->screen->curie, NV40TCL_COLOR2_PITCH, 1); - so_data (so, rt[2]->stride); + so_data (so, rt[2]->pitch); } if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) { so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR3, 1); - so_reloc (so, nv40_surface_buffer(rt[3]), 0, rt_flags | NOUVEAU_BO_OR, + so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0, rt_flags | NOUVEAU_BO_OR, nv40->nvws->channel->vram->handle, nv40->nvws->channel->gart->handle); so_method(so, nv40->screen->curie, NV40TCL_COLOR3_OFFSET, 1); - so_reloc (so, nv40_surface_buffer(rt[3]), rt[3]->offset, rt_flags | + so_reloc (so, nv40_surface_buffer(&rt[3]->base), rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); so_method(so, nv40->screen->curie, NV40TCL_COLOR3_PITCH, 1); - so_data (so, rt[3]->stride); + so_data (so, rt[3]->pitch); } if (zeta_format) { so_method(so, nv40->screen->curie, NV40TCL_DMA_ZETA, 1); - so_reloc (so, nv40_surface_buffer(zeta), 0, rt_flags | NOUVEAU_BO_OR, + so_reloc (so, nv40_surface_buffer(&zeta->base), 0, rt_flags | NOUVEAU_BO_OR, nv40->nvws->channel->vram->handle, nv40->nvws->channel->gart->handle); so_method(so, nv40->screen->curie, NV40TCL_ZETA_OFFSET, 1); - so_reloc (so, nv40_surface_buffer(zeta), zeta->offset, rt_flags | + so_reloc (so, nv40_surface_buffer(&zeta->base), zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); so_method(so, nv40->screen->curie, NV40TCL_ZETA_PITCH, 1); - so_data (so, zeta->stride); + so_data (so, zeta->pitch); } so_method(so, nv40->screen->curie, NV40TCL_RT_ENABLE, 1); @@ -150,6 +150,7 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) so_data (so, (1 << 12) | h); so_ref(so, &nv40->state.hw[NV40_STATE_FB]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv40/nv40_state_scissor.c b/src/gallium/drivers/nv40/nv40_state_scissor.c index 285239ef41..cf58d33906 100644 --- a/src/gallium/drivers/nv40/nv40_state_scissor.c +++ b/src/gallium/drivers/nv40/nv40_state_scissor.c @@ -23,6 +23,7 @@ nv40_state_scissor_validate(struct nv40_context *nv40) } so_ref(so, &nv40->state.hw[NV40_STATE_SCISSOR]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv40/nv40_state_viewport.c b/src/gallium/drivers/nv40/nv40_state_viewport.c index 869a55b405..665d2d5fca 100644 --- a/src/gallium/drivers/nv40/nv40_state_viewport.c +++ b/src/gallium/drivers/nv40/nv40_state_viewport.c @@ -7,7 +7,8 @@ nv40_state_viewport_validate(struct nv40_context *nv40) struct nouveau_stateobj *so; unsigned bypass; - if (nv40->render_mode == HW && !nv40->rasterizer->pipe.bypass_clipping) + if (nv40->render_mode == HW && + !nv40->rasterizer->pipe.bypass_vs_clip_and_viewport) bypass = 0; else bypass = 1; @@ -55,6 +56,7 @@ nv40_state_viewport_validate(struct nv40_context *nv40) } so_ref(so, &nv40->state.hw[NV40_STATE_VIEWPORT]); + so_ref(NULL, &so); return TRUE; } diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c index c4a5fb20d9..1a849da32d 100644 --- a/src/gallium/drivers/nv40/nv40_surface.c +++ b/src/gallium/drivers/nv40/nv40_surface.c @@ -33,7 +33,7 @@ #include "util/u_tile.h" static void -nv40_surface_copy(struct pipe_context *pipe, boolean do_flip, +nv40_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) @@ -41,15 +41,6 @@ nv40_surface_copy(struct pipe_context *pipe, boolean do_flip, struct nv40_context *nv40 = nv40_context(pipe); struct nv04_surface_2d *eng2d = nv40->screen->eng2d; - if (do_flip) { - desty += height; - while (height--) { - eng2d->copy(eng2d, dest, destx, desty--, src, - srcx, srcy++, width, 1); - } - return; - } - eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height); } diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c new file mode 100644 index 0000000000..728e8b5674 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_transfer.c @@ -0,0 +1,196 @@ +#include <pipe/p_state.h> +#include <pipe/p_defines.h> +#include <pipe/p_inlines.h> +#include <util/u_memory.h> +#include <nouveau/nouveau_winsys.h> +#include "nv40_context.h" +#include "nv40_screen.h" +#include "nv40_state.h" + +struct nv40_transfer { + struct pipe_transfer base; + struct pipe_surface *surface; + bool direct; +}; + +static unsigned nv40_usage_tx_to_buf(unsigned tx_usage) +{ + switch (tx_usage) { + case PIPE_TRANSFER_READ: + return PIPE_BUFFER_USAGE_CPU_READ; + case PIPE_TRANSFER_WRITE: + return PIPE_BUFFER_USAGE_CPU_WRITE; + case PIPE_TRANSFER_READ_WRITE: + return PIPE_BUFFER_USAGE_CPU_READ_WRITE; + default: + assert(0); + } + + return -1; +} + +static void +nv40_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, + struct pipe_texture *template) +{ + memset(template, 0, sizeof(struct pipe_texture)); + template->target = pt->target; + template->format = pt->format; + template->width[0] = pt->width[level]; + template->height[0] = pt->height[level]; + template->depth[0] = 1; + template->block = pt->block; + template->nblocksx[0] = pt->nblocksx[level]; + template->nblocksy[0] = pt->nblocksx[level]; + template->last_level = 0; + template->compressed = pt->compressed; + template->nr_samples = pt->nr_samples; + + template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | + NOUVEAU_TEXTURE_USAGE_LINEAR; +} + +static struct pipe_transfer * +nv40_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct nv40_miptree *mt = (struct nv40_miptree *)pt; + struct nv40_transfer *tx; + struct pipe_texture tx_tex_template, *tx_tex; + + tx = CALLOC_STRUCT(nv40_transfer); + if (!tx) + return NULL; + + pipe_texture_reference(&tx->base.texture, pt); + tx->base.format = pt->format; + tx->base.x = x; + tx->base.y = y; + tx->base.width = w; + tx->base.height = h; + tx->base.block = pt->block; + tx->base.nblocksx = pt->nblocksx[level]; + tx->base.nblocksy = pt->nblocksy[level]; + tx->base.stride = mt->level[level].pitch; + tx->base.usage = usage; + tx->base.face = face; + tx->base.level = level; + tx->base.zslice = zslice; + + /* Direct access to texture */ + if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC || + debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) && + pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) + { + tx->direct = true; + tx->surface = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + nv40_usage_tx_to_buf(usage)); + return &tx->base; + } + + tx->direct = false; + + nv40_compatible_transfer_tex(pt, level, &tx_tex_template); + + tx_tex = pscreen->texture_create(pscreen, &tx_tex_template); + if (!tx_tex) + { + FREE(tx); + return NULL; + } + + tx->surface = pscreen->get_tex_surface(pscreen, tx_tex, + 0, 0, 0, + nv40_usage_tx_to_buf(usage)); + + pipe_texture_reference(&tx_tex, NULL); + + if (!tx->surface) + { + pipe_surface_reference(&tx->surface, NULL); + FREE(tx); + return NULL; + } + + if (usage != PIPE_TRANSFER_WRITE) { + struct nv40_screen *nvscreen = nv40_screen(pscreen); + struct pipe_surface *src; + + src = pscreen->get_tex_surface(pscreen, pt, + face, level, zslice, + PIPE_BUFFER_USAGE_GPU_READ); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + /* TODO: Check if SIFM can un-swizzle */ + nvscreen->eng2d->copy(nvscreen->eng2d, + tx->surface, 0, 0, + src, 0, 0, + src->width, src->height); + + pipe_surface_reference(&src, NULL); + } + + return &tx->base; +} + +static void +nv40_transfer_del(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv40_transfer *tx = (struct nv40_transfer *)ptx; + + if (!tx->direct && ptx->usage != PIPE_TRANSFER_READ) { + struct pipe_screen *pscreen = ptx->texture->screen; + struct nv40_screen *nvscreen = nv40_screen(pscreen); + struct pipe_surface *dst; + + dst = pscreen->get_tex_surface(pscreen, ptx->texture, + ptx->face, ptx->level, ptx->zslice, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* TODO: Check if SIFM can deal with x,y,w,h when swizzling */ + nvscreen->eng2d->copy(nvscreen->eng2d, + dst, 0, 0, + tx->surface, 0, 0, + dst->width, dst->height); + + pipe_surface_reference(&dst, NULL); + } + + pipe_surface_reference(&tx->surface, NULL); + pipe_texture_reference(&ptx->texture, NULL); + FREE(ptx); +} + +static void * +nv40_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv40_transfer *tx = (struct nv40_transfer *)ptx; + struct nv04_surface *ns = (struct nv04_surface *)tx->surface; + struct nv40_miptree *mt = (struct nv40_miptree *)tx->surface->texture; + void *map = pipe_buffer_map(pscreen, mt->buffer, + nv40_usage_tx_to_buf(ptx->usage)); + + return map + ns->base.offset + + ptx->y * ns->pitch + ptx->x * ptx->block.size; +} + +static void +nv40_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv40_transfer *tx = (struct nv40_transfer *)ptx; + struct nv40_miptree *mt = (struct nv40_miptree *)tx->surface->texture; + + pipe_buffer_unmap(pscreen, mt->buffer); +} + +void +nv40_screen_init_transfer_functions(struct pipe_screen *pscreen) +{ + pscreen->get_tex_transfer = nv40_transfer_new; + pscreen->tex_transfer_destroy = nv40_transfer_del; + pscreen->transfer_map = nv40_transfer_map; + pscreen->transfer_unmap = nv40_transfer_unmap; +} diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c index 8f1834628f..f3518b2e4f 100644 --- a/src/gallium/drivers/nv40/nv40_vbo.c +++ b/src/gallium/drivers/nv40/nv40_vbo.c @@ -537,10 +537,13 @@ nv40_vbo_validate(struct nv40_context *nv40) so_data (vtxbuf, 0); so_ref(vtxbuf, &nv40->state.hw[NV40_STATE_VTXBUF]); + so_ref(NULL, &vtxbuf); nv40->state.dirty |= (1ULL << NV40_STATE_VTXBUF); so_ref(vtxfmt, &nv40->state.hw[NV40_STATE_VTXFMT]); + so_ref(NULL, &vtxfmt); nv40->state.dirty |= (1ULL << NV40_STATE_VTXFMT); so_ref(sattr, &nv40->state.hw[NV40_STATE_VTXATTR]); + so_ref(NULL, &sattr); nv40->state.dirty |= (1ULL << NV40_STATE_VTXATTR); return FALSE; } diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c index 0862386638..7df9a4d326 100644 --- a/src/gallium/drivers/nv40/nv40_vertprog.c +++ b/src/gallium/drivers/nv40/nv40_vertprog.c @@ -916,6 +916,7 @@ check_gpu_resources: so_method(so, curie, NV40TCL_CLIP_PLANE_ENABLE, 1); so_data (so, vp->clip_ctrl); so_ref(so, &vp->so); + so_ref(NULL, &so); upload_code = TRUE; } diff --git a/src/gallium/drivers/nv50/Makefile b/src/gallium/drivers/nv50/Makefile index be30400c03..612aea28a3 100644 --- a/src/gallium/drivers/nv50/Makefile +++ b/src/gallium/drivers/nv50/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = nv50 -DRIVER_SOURCES = \ +C_SOURCES = \ nv50_clear.c \ nv50_context.c \ nv50_draw.c \ @@ -15,15 +15,7 @@ DRIVER_SOURCES = \ nv50_state_validate.c \ nv50_surface.c \ nv50_tex.c \ + nv50_transfer.c \ nv50_vbo.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 1e9d45cb34..313e435e7a 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -64,10 +64,8 @@ struct nv50_rasterizer_stateobj { }; struct nv50_miptree_level { - struct pipe_buffer **image; int *image_offset; - unsigned image_dirty_cpu[512/32]; - unsigned image_dirty_gpu[512/32]; + unsigned pitch; }; struct nv50_miptree { @@ -200,8 +198,4 @@ extern boolean nv50_state_validate(struct nv50_context *nv50); /* nv50_tex.c */ extern void nv50_tex_validate(struct nv50_context *); -/* nv50_miptree.c */ -extern void nv50_miptree_sync(struct pipe_screen *, struct nv50_miptree *, - unsigned level, unsigned image); - #endif diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index 91091d53f5..dc4688ccdc 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -29,7 +29,6 @@ static struct pipe_texture * nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) { - struct pipe_winsys *ws = pscreen->winsys; struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree); struct pipe_texture *pt = &mt->base; unsigned usage, width = tmp->width[0], height = tmp->height[0]; @@ -37,7 +36,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) int i, l; mt->base = *tmp; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; usage = PIPE_BUFFER_USAGE_PIXEL; @@ -72,7 +71,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); lvl->image_offset = CALLOC(mt->image_nr, sizeof(int)); - lvl->image = CALLOC(mt->image_nr, sizeof(struct pipe_buffer *)); + lvl->pitch = align(pt->width[l] * pt->block.size, 64); width = MAX2(1, width >> 1); height = MAX2(1, height >> 1); @@ -88,14 +87,13 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) size = align(size, 64); size *= align(pt->height[l], 8) * pt->block.size; - lvl->image[i] = ws->buffer_create(ws, 256, 0, size); lvl->image_offset[i] = mt->total_size; mt->total_size += size; } } - mt->buffer = ws->buffer_create(ws, 256, usage, mt->total_size); + mt->buffer = pscreen->buffer_create(pscreen, 256, usage, mt->total_size); if (!mt->buffer) { FREE(mt); return NULL; @@ -120,119 +118,23 @@ nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, return NULL; mt->base = *pt; - mt->base.refcount = 1; + pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; mt->image_nr = 1; + mt->level[0].pitch = *stride; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(pscreen, &mt->buffer, pb); + pipe_buffer_reference(&mt->buffer, pb); return &mt->base; } -static INLINE void -mark_dirty(uint32_t *flags, unsigned image) -{ - flags[image / 32] |= (1 << (image % 32)); -} - -static INLINE void -mark_clean(uint32_t *flags, unsigned image) -{ - flags[image / 32] &= ~(1 << (image % 32)); -} - -static INLINE int -is_dirty(uint32_t *flags, unsigned image) -{ - return !!(flags[image / 32] & (1 << (image % 32))); -} - static void -nv50_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) -{ - struct pipe_texture *pt = *ppt; - - *ppt = NULL; - - if (--pt->refcount <= 0) { - struct nv50_miptree *mt = nv50_miptree(pt); - - pipe_buffer_reference(pscreen, &mt->buffer, NULL); - FREE(mt); - } -} - -void -nv50_miptree_sync(struct pipe_screen *pscreen, struct nv50_miptree *mt, - unsigned level, unsigned image) +nv50_miptree_destroy(struct pipe_texture *pt) { - struct nv50_screen *nvscreen = nv50_screen(pscreen); - struct nv50_miptree_level *lvl = &mt->level[level]; - struct pipe_surface *dst, *src; - unsigned face = 0, zslice = 0; - - if (!is_dirty(lvl->image_dirty_cpu, image)) - return; - - if (mt->base.target == PIPE_TEXTURE_CUBE) - face = image; - else - if (mt->base.target == PIPE_TEXTURE_3D) - zslice = image; - - /* Mark as clean already - so we don't continually call this function - * trying to get a GPU_WRITE pipe_surface! - */ - mark_clean(lvl->image_dirty_cpu, image); - - /* Pretend we're doing CPU access so we get the backing pipe_surface - * and not a view into the larger miptree. - */ - src = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice, - PIPE_BUFFER_USAGE_CPU_READ); - - /* Pretend we're only reading with the GPU so surface doesn't get marked - * as dirtied by the GPU. - */ - dst = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice, - PIPE_BUFFER_USAGE_GPU_READ); - - nv50_surface_do_copy(nvscreen, dst, 0, 0, src, 0, 0, dst->width, dst->height); - - pscreen->tex_surface_release(pscreen, &dst); - pscreen->tex_surface_release(pscreen, &src); -} - -/* The reverse of the above */ -static void -nv50_miptree_sync_cpu(struct pipe_screen *pscreen, struct nv50_miptree *mt, - unsigned level, unsigned image) -{ - struct nv50_screen *nvscreen = nv50_screen(pscreen); - struct nv50_miptree_level *lvl = &mt->level[level]; - struct pipe_surface *dst, *src; - unsigned face = 0, zslice = 0; - - if (!is_dirty(lvl->image_dirty_gpu, image)) - return; - - if (mt->base.target == PIPE_TEXTURE_CUBE) - face = image; - else - if (mt->base.target == PIPE_TEXTURE_3D) - zslice = image; - - mark_clean(lvl->image_dirty_gpu, image); - - src = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice, - PIPE_BUFFER_USAGE_GPU_READ); - dst = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice, - PIPE_BUFFER_USAGE_CPU_READ); - - nv50_surface_do_copy(nvscreen, dst, 0, 0, src, 0, 0, dst->width, dst->height); + struct nv50_miptree *mt = nv50_miptree(pt); - pscreen->tex_surface_release(pscreen, &dst); - pscreen->tex_surface_release(pscreen, &src); + pipe_buffer_reference(&mt->buffer, NULL); + FREE(mt); } static struct pipe_surface * @@ -260,52 +162,24 @@ nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, ps->format = pt->format; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->block = pt->block; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = ps->width * ps->block.size; ps->usage = flags; ps->status = PIPE_SURFACE_STATUS_DEFINED; - ps->refcount = 1; + pipe_reference_init(&ps->reference, 1); ps->face = face; ps->level = level; ps->zslice = zslice; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { - assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE)); - nv50_miptree_sync_cpu(pscreen, mt, level, img); - - ps->offset = 0; - pipe_texture_reference(&ps->texture, pt); - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - mark_dirty(lvl->image_dirty_cpu, img); - } else { - nv50_miptree_sync(pscreen, mt, level, img); - - ps->offset = lvl->image_offset[img]; - pipe_texture_reference(&ps->texture, pt); - - if (flags & PIPE_BUFFER_USAGE_GPU_WRITE) - mark_dirty(lvl->image_dirty_gpu, img); - } + ps->offset = lvl->image_offset[img]; return ps; } static void -nv50_miptree_surface_del(struct pipe_screen *pscreen, - struct pipe_surface **psurface) +nv50_miptree_surface_del(struct pipe_surface *ps) { - struct pipe_surface *ps = *psurface; struct nv50_surface *s = nv50_surface(ps); - *psurface = NULL; - - if (--ps->refcount <= 0) { - pipe_texture_reference(&ps->texture, NULL); - FREE(s); - } + pipe_texture_reference(&ps->texture, NULL); + FREE(s); } void @@ -313,8 +187,8 @@ nv50_screen_init_miptree_functions(struct pipe_screen *pscreen) { pscreen->texture_create = nv50_miptree_create; pscreen->texture_blanket = nv50_miptree_blanket; - pscreen->texture_release = nv50_miptree_release; + pscreen->texture_destroy = nv50_miptree_destroy; pscreen->get_tex_surface = nv50_miptree_surface_new; - pscreen->tex_surface_release = nv50_miptree_surface_del; + pscreen->tex_surface_destroy = nv50_miptree_surface_del; } diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index 14c5d47e79..2d15868ae8 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -1603,7 +1603,7 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) { struct nouveau_channel *chan = nv50->screen->nvws->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; - struct pipe_winsys *ws = nv50->pipe.winsys; + struct pipe_screen *screen = nv50->pipe.screen; struct nv50_program_exec *e; struct nouveau_stateobj *so; const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR; @@ -1611,7 +1611,7 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) boolean upload = FALSE; if (!p->buffer) { - p->buffer = ws->buffer_create(ws, 0x100, 0, p->exec_size * 4); + p->buffer = screen->buffer_create(screen, 0x100, 0, p->exec_size * 4); upload = TRUE; } @@ -1719,6 +1719,7 @@ nv50_vertprog_validate(struct nv50_context *nv50) so_method(so, tesla, 0x140c, 1); so_data (so, 0); /* program start offset */ so_ref(so, &nv50->state.vertprog); + so_ref(NULL, &so); } void @@ -1758,6 +1759,7 @@ nv50_fragprog_validate(struct nv50_context *nv50) so_method(so, tesla, 0x1414, 1); so_data (so, 0); /* program start offset */ so_ref(so, &nv50->state.fragprog); + so_ref(NULL, &so); } void @@ -1775,7 +1777,7 @@ nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p) p->exec_size = 0; if (p->buffer) - pipe_buffer_reference(pscreen, &p->buffer, NULL); + pipe_buffer_reference(&p->buffer, NULL); nv50->screen->nvws->res_free(&p->data); diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c index 20745ceab8..a2c56f99a8 100644 --- a/src/gallium/drivers/nv50/nv50_query.c +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -41,13 +41,13 @@ nv50_query(struct pipe_query *pipe) static struct pipe_query * nv50_query_create(struct pipe_context *pipe, unsigned type) { - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *screen = pipe->winsys; struct nv50_query *q = CALLOC_STRUCT(nv50_query); assert (q->type == PIPE_QUERY_OCCLUSION_COUNTER); q->type = type; - q->buffer = ws->buffer_create(ws, 256, 0, 16); + q->buffer = screen->buffer_create(screen, 256, 0, 16); if (!q->buffer) { FREE(q); return NULL; @@ -62,7 +62,7 @@ nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *pq) struct nv50_query *q = nv50_query(pq); if (q) { - pipe_buffer_reference(pipe->screen, &q->buffer, NULL); + pipe_buffer_reference(&q->buffer, NULL); FREE(q); } } @@ -90,11 +90,12 @@ nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq) struct nouveau_channel *chan = nv50->screen->nvws->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; struct nv50_query *q = nv50_query(pq); + struct nouveau_bo *bo = nv50->screen->nvws->get_bo(q->buffer); WAIT_RING (chan, 5); BEGIN_RING(chan, tesla, 0x1b00, 4); - OUT_RELOCh(chan, q->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RELOCl(chan, q->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + OUT_RELOCh(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + OUT_RELOCl(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RING (chan, 0x00000000); OUT_RING (chan, 0x0100f002); FIRE_RING (chan); diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index 58d7a621a8..2980564594 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -29,10 +29,6 @@ #include "nouveau/nouveau_stateobj.h" -#define NV5X_GRCLASS5097_CHIPSETS 0x00000001 -#define NV8X_GRCLASS8297_CHIPSETS 0x00000050 -#define NV9X_GRCLASS8297_CHIPSETS 0x00000014 - static boolean nv50_screen_is_format_supported(struct pipe_screen *pscreen, enum pipe_format format, @@ -124,9 +120,9 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param) return 1; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: return 0; - case NOUVEAU_CAP_HW_VTXBUF: + case NOUVEAU_CAP_HW_VTXBUF: return 1; - case NOUVEAU_CAP_HW_IDXBUF: + case NOUVEAU_CAP_HW_IDXBUF: return 0; default: NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); @@ -173,36 +169,38 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) return NULL; screen->nvws = nvws; - /* 2D object */ - ret = nvws->grobj_alloc(nvws, NV50_2D, &screen->eng2d); + /* DMA engine object */ + ret = nvws->grobj_alloc(nvws, 0x5039, &screen->m2mf); if (ret) { - NOUVEAU_ERR("Error creating 2D object: %d\n", ret); + NOUVEAU_ERR("Error creating M2MF object: %d\n", ret); nv50_screen_destroy(&screen->pipe); return NULL; } - /* 3D object */ - if ((chipset & 0xf0) != 0x50 && (chipset & 0xf0) != 0x80) { - NOUVEAU_ERR("Not a G8x chipset\n"); + /* 2D object */ + ret = nvws->grobj_alloc(nvws, NV50_2D, &screen->eng2d); + if (ret) { + NOUVEAU_ERR("Error creating 2D object: %d\n", ret); nv50_screen_destroy(&screen->pipe); return NULL; } + /* 3D object */ switch (chipset & 0xf0) { case 0x50: - if (NV5X_GRCLASS5097_CHIPSETS & (1 << (chipset & 0x0f))) - tesla_class = 0x5097; + tesla_class = 0x5097; break; case 0x80: - if (NV8X_GRCLASS8297_CHIPSETS & (1 << (chipset & 0x0f))) - tesla_class = 0x8297; - break; case 0x90: - if (NV9X_GRCLASS8297_CHIPSETS & (1 << (chipset & 0x0f))) - tesla_class = 0x8297; + tesla_class = 0x8297; break; - default: + case 0xa0: + tesla_class = 0x8397; break; + default: + NOUVEAU_ERR("Not a known NV50 chipset: NV%02x\n", chipset); + nv50_screen_destroy(&screen->pipe); + return NULL; } if (tesla_class == 0) { @@ -226,6 +224,31 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) return NULL; } + /* Setup the pipe */ + screen->pipe.winsys = ws; + + screen->pipe.destroy = nv50_screen_destroy; + + screen->pipe.get_name = nv50_screen_get_name; + screen->pipe.get_vendor = nv50_screen_get_vendor; + screen->pipe.get_param = nv50_screen_get_param; + screen->pipe.get_paramf = nv50_screen_get_paramf; + + screen->pipe.is_format_supported = nv50_screen_is_format_supported; + + nv50_screen_init_miptree_functions(&screen->pipe); + nv50_transfer_init_screen_functions(&screen->pipe); + u_simple_screen_init(&screen->pipe); + + /* Static M2MF init */ + so = so_new(32, 0); + so_method(so, screen->m2mf, 0x0180, 3); + so_data (so, screen->sync->handle); + so_data (so, screen->nvws->channel->vram->handle); + so_data (so, screen->nvws->channel->vram->handle); + so_emit(nvws, so); + so_ref (NULL, &so); + /* Static 2D init */ so = so_new(64, 0); so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4); @@ -268,7 +291,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_data (so, 8); /* Shared constant buffer */ - screen->constbuf = ws->buffer_create(ws, 0, 0, 128 * 4 * 4); + screen->constbuf = screen->pipe.buffer_create(&screen->pipe, 0, 0, 128 * 4 * 4); if (nvws->res_init(&screen->vp_data_heap, 0, 128)) { NOUVEAU_ERR("Error initialising constant buffer\n"); nv50_screen_destroy(&screen->pipe); @@ -287,7 +310,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) * blocks. At some point we *may* want to go the NVIDIA way of doing * things? */ - screen->tic = ws->buffer_create(ws, 0, 0, 32 * 8 * 4); + screen->tic = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4); so_method(so, screen->tesla, 0x1280, 3); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); @@ -301,7 +324,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, 0x00000800); - screen->tsc = ws->buffer_create(ws, 0, 0, 32 * 8 * 4); + screen->tsc = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4); so_method(so, screen->tesla, 0x1280, 3); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); @@ -333,24 +356,10 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_data (so, 1); so_emit(nvws, so); - so_ref(so, &screen->static_init); + so_ref (so, &screen->static_init); + so_ref (NULL, &so); nvws->push_flush(nvws, 0, NULL); - screen->pipe.winsys = ws; - - screen->pipe.destroy = nv50_screen_destroy; - - screen->pipe.get_name = nv50_screen_get_name; - screen->pipe.get_vendor = nv50_screen_get_vendor; - screen->pipe.get_param = nv50_screen_get_param; - screen->pipe.get_paramf = nv50_screen_get_paramf; - - screen->pipe.is_format_supported = nv50_screen_is_format_supported; - - nv50_screen_init_miptree_functions(&screen->pipe); - nv50_surface_init_screen_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); - return &screen->pipe; } diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h index c888ca071c..db567aaac8 100644 --- a/src/gallium/drivers/nv50/nv50_screen.h +++ b/src/gallium/drivers/nv50/nv50_screen.h @@ -12,6 +12,7 @@ struct nv50_screen { struct nouveau_grobj *tesla; struct nouveau_grobj *eng2d; + struct nouveau_grobj *m2mf; struct nouveau_notifier *sync; struct pipe_buffer *constbuf; @@ -29,6 +30,6 @@ nv50_screen(struct pipe_screen *screen) return (struct nv50_screen *)screen; } -void nv50_surface_init_screen_functions(struct pipe_screen *); +void nv50_transfer_init_screen_functions(struct pipe_screen *); #endif diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index 787ff958ec..ba852194cd 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -85,6 +85,7 @@ nv50_blend_state_create(struct pipe_context *pipe, bso->pipe = *cso; so_ref(so, &bso->so); + so_ref(NULL, &so); return (void *)bso; } @@ -352,6 +353,7 @@ nv50_rasterizer_state_create(struct pipe_context *pipe, rso->pipe = *cso; so_ref(so, &rso->so); + so_ref(NULL, &so); return (void *)rso; } @@ -438,6 +440,7 @@ nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe, zsa->pipe = *cso; so_ref(so, &zsa->so); + so_ref(NULL, &so); return (void *)zsa; } diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 948112ffa9..85098a78a2 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -124,6 +124,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_data (so, h); so_ref(so, &nv50->state.fb); + so_ref(NULL, &so); } static void @@ -214,6 +215,7 @@ nv50_state_validate(struct nv50_context *nv50) so_data (so, fui(nv50->blend_colour.color[2])); so_data (so, fui(nv50->blend_colour.color[3])); so_ref(so, &nv50->state.blend_colour); + so_ref(NULL, &so); } if (nv50->dirty & NV50_NEW_STIPPLE) { @@ -222,6 +224,7 @@ nv50_state_validate(struct nv50_context *nv50) for (i = 0; i < 32; i++) so_data(so, nv50->stipple.stipple[i]); so_ref(so, &nv50->state.stipple); + so_ref(NULL, &so); } if (nv50->dirty & (NV50_NEW_SCISSOR | NV50_NEW_RASTERIZER)) { @@ -243,6 +246,7 @@ nv50_state_validate(struct nv50_context *nv50) so_data(so, (8192 << 16)); } so_ref(so, &nv50->state.scissor); + so_ref(NULL, &so); nv50->state.dirty |= NV50_NEW_SCISSOR; } scissor_uptodate: @@ -250,7 +254,7 @@ scissor_uptodate: if (nv50->dirty & NV50_NEW_VIEWPORT) { unsigned bypass; - if (!nv50->rasterizer->pipe.bypass_clipping) + if (!nv50->rasterizer->pipe.bypass_vs_clip_and_viewport) bypass = 0; else bypass = 1; @@ -283,6 +287,7 @@ scissor_uptodate: } so_ref(so, &nv50->state.viewport); + so_ref(NULL, &so); } viewport_uptodate: @@ -296,6 +301,7 @@ viewport_uptodate: for (i = 0; i < nv50->sampler_nr; i++) so_datap (so, nv50->sampler[i], 8); so_ref(so, &nv50->state.tsc_upload); + so_ref(NULL, &so); } if (nv50->dirty & NV50_NEW_TEXTURE) diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index f2dd2eb30b..0cc5168144 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -51,6 +51,7 @@ nv50_format(enum pipe_format format) static int nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) { + struct nv50_miptree *mt = nv50_miptree(ps->texture); struct nouveau_channel *chan = screen->nvws->channel; struct nouveau_grobj *eng2d = screen->eng2d; struct nouveau_bo *bo; @@ -70,7 +71,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) OUT_RING (chan, format); OUT_RING (chan, 1); BEGIN_RING(chan, eng2d, mthd + 0x14, 5); - OUT_RING (chan, ps->stride); + OUT_RING (chan, mt->level[0].pitch); OUT_RING (chan, ps->width); OUT_RING (chan, ps->height); OUT_RELOCh(chan, bo, ps->offset, flags); @@ -143,7 +144,7 @@ nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst, } static void -nv50_surface_copy(struct pipe_context *pipe, boolean flip, +nv50_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) @@ -153,16 +154,8 @@ nv50_surface_copy(struct pipe_context *pipe, boolean flip, assert(src->format == dest->format); - if (flip) { - desty += height; - while (height--) { - nv50_surface_do_copy(screen, dest, destx, desty--, src, - srcx, srcy++, width, 1); - } - } else { - nv50_surface_do_copy(screen, dest, destx, desty, src, srcx, + nv50_surface_do_copy(screen, dest, destx, desty, src, srcx, srcy, width, height); - } } static void @@ -197,23 +190,6 @@ nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, OUT_RING (chan, height); } -static void * -nv50_surface_map(struct pipe_screen *screen, struct pipe_surface *ps, - unsigned flags ) -{ - struct pipe_winsys *ws = screen->winsys; - - return ws->buffer_map(ws, nv50_surface_buffer(ps), flags); -} - -static void -nv50_surface_unmap(struct pipe_screen *pscreen, struct pipe_surface *ps) -{ - struct pipe_winsys *ws = pscreen->winsys; - - ws->buffer_unmap(ws, nv50_surface_buffer(ps)); -} - void nv50_init_surface_functions(struct nv50_context *nv50) { @@ -221,10 +197,4 @@ nv50_init_surface_functions(struct nv50_context *nv50) nv50->pipe.surface_fill = nv50_surface_fill; } -void -nv50_surface_init_screen_functions(struct pipe_screen *pscreen) -{ - pscreen->surface_map = nv50_surface_map; - pscreen->surface_unmap = nv50_surface_unmap; -} diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 675f9b20cb..223c8a3a45 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -135,7 +135,7 @@ nv50_tex_validate(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; struct nouveau_stateobj *so; - int unit, level, image; + int unit; so = so_new(nv50->miptree_nr * 8 + 3, nv50->miptree_nr * 2); so_method(so, tesla, 0x0f00, 1); @@ -144,13 +144,6 @@ nv50_tex_validate(struct nv50_context *nv50) for (unit = 0; unit < nv50->miptree_nr; unit++) { struct nv50_miptree *mt = nv50->miptree[unit]; - for (level = 0; level <= mt->base.last_level; level++) { - for (image = 0; image < mt->image_nr; image++) { - nv50_miptree_sync(&nv50->screen->pipe, mt, - level, image); - } - } - if (nv50_tex_construct(so, mt)) { NOUVEAU_ERR("failed tex validate\n"); so_ref(NULL, &so); @@ -159,5 +152,6 @@ nv50_tex_validate(struct nv50_context *nv50) } so_ref(so, &nv50->state.tic_upload); + so_ref(NULL, &so); } diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c new file mode 100644 index 0000000000..747195b4f6 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -0,0 +1,211 @@ + +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" + +#include "nv50_context.h" + +struct nv50_transfer { + struct pipe_transfer base; + struct pipe_buffer *buffer; + struct nv50_miptree_level *level; + int level_pitch; + int level_width; + int level_height; + int level_x; + int level_y; +}; + +static void +nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src, + int src_pitch, int sx, int sy, int sw, int sh, + struct pipe_buffer *dst, int dst_pitch, int dx, int dy, + int dw, int dh, int cpp, int width, int height, + unsigned src_reloc, unsigned dst_reloc) +{ + struct nv50_screen *screen = nv50_screen(pscreen); + struct nouveau_winsys *nvws = screen->nvws; + struct nouveau_channel *chan = nvws->channel; + struct nouveau_grobj *m2mf = screen->m2mf; + struct nouveau_bo *src_bo = nvws->get_bo(src); + struct nouveau_bo *dst_bo = nvws->get_bo(dst); + unsigned src_offset = 0, dst_offset = 0; + + src_reloc |= NOUVEAU_BO_RD; + dst_reloc |= NOUVEAU_BO_WR; + + WAIT_RING (chan, 14); + + if (!src_bo->tiled) { + BEGIN_RING(chan, m2mf, 0x0200, 1); + OUT_RING (chan, 1); + BEGIN_RING(chan, m2mf, 0x0314, 1); + OUT_RING (chan, src_pitch); + src_offset = (sy * src_pitch) + (sx * cpp); + } else { + BEGIN_RING(chan, m2mf, 0x0200, 6); + OUT_RING (chan, 0); + OUT_RING (chan, 0); + OUT_RING (chan, sw * cpp); + OUT_RING (chan, sh); + OUT_RING (chan, 1); + OUT_RING (chan, 0); + } + + if (!dst_bo->tiled) { + BEGIN_RING(chan, m2mf, 0x021c, 1); + OUT_RING (chan, 1); + BEGIN_RING(chan, m2mf, 0x0318, 1); + OUT_RING (chan, dst_pitch); + dst_offset = (dy * dst_pitch) + (dx * cpp); + } else { + BEGIN_RING(chan, m2mf, 0x021c, 6); + OUT_RING (chan, 0); + OUT_RING (chan, 0); + OUT_RING (chan, dw * cpp); + OUT_RING (chan, dh); + OUT_RING (chan, 1); + OUT_RING (chan, 0); + } + + while (height) { + int line_count = height > 2047 ? 2047 : height; + + WAIT_RING (chan, 15); + BEGIN_RING(chan, m2mf, 0x0238, 2); + OUT_RELOCh(chan, src_bo, src_offset, src_reloc); + OUT_RELOCh(chan, dst_bo, dst_offset, dst_reloc); + BEGIN_RING(chan, m2mf, 0x030c, 2); + OUT_RELOCl(chan, src_bo, src_offset, src_reloc); + OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc); + if (src_bo->tiled) { + BEGIN_RING(chan, m2mf, 0x0218, 1); + OUT_RING (chan, (dy << 16) | sx); + } else { + src_offset += (line_count * src_pitch); + } + if (dst_bo->tiled) { + BEGIN_RING(chan, m2mf, 0x0234, 1); + OUT_RING (chan, (sy << 16) | dx); + } else { + dst_offset += (line_count * dst_pitch); + } + BEGIN_RING(chan, m2mf, 0x031c, 4); + OUT_RING (chan, width * cpp); + OUT_RING (chan, line_count); + OUT_RING (chan, 0x00000101); + OUT_RING (chan, 0); + FIRE_RING (chan); + + height -= line_count; + sy += line_count; + dy += line_count; + } +} + +static struct pipe_transfer * +nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct nv50_miptree *mt = nv50_miptree(pt); + struct nv50_miptree_level *lvl = &mt->level[level]; + struct nv50_transfer *tx; + unsigned image = 0; + + if (pt->target == PIPE_TEXTURE_CUBE) + image = face; + else + if (pt->target == PIPE_TEXTURE_3D) + image = zslice; + + tx = CALLOC_STRUCT(nv50_transfer); + if (!tx) + return NULL; + + pipe_texture_reference(&tx->base.texture, pt); + tx->base.format = pt->format; + tx->base.width = w; + tx->base.height = h; + tx->base.block = pt->block; + tx->base.nblocksx = pt->nblocksx[level]; + tx->base.nblocksy = pt->nblocksy[level]; + tx->base.stride = (w * pt->block.size); + tx->base.usage = usage; + + tx->level = lvl; + tx->level_pitch = lvl->pitch; + tx->level_width = mt->base.width[level]; + tx->level_height = mt->base.height[level]; + tx->level_x = x; + tx->level_y = y; + tx->buffer = + pipe_buffer_create(pscreen, 0, NOUVEAU_BUFFER_USAGE_TRANSFER, + w * tx->base.block.size * h); + + if (usage != PIPE_TRANSFER_WRITE) { + nv50_transfer_rect_m2mf(pscreen, mt->buffer, tx->level_pitch, + x, y, tx->level_width, tx->level_height, + tx->buffer, tx->base.stride, 0, 0, + tx->base.width, tx->base.height, + tx->base.block.size, w, h, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART, + NOUVEAU_BO_GART); + } + + return &tx->base; +} + +static void +nv50_transfer_del(struct pipe_transfer *ptx) +{ + struct nv50_transfer *tx = (struct nv50_transfer *)ptx; + struct nv50_miptree *mt = nv50_miptree(ptx->texture); + + if (ptx->usage != PIPE_TRANSFER_READ) { + struct pipe_screen *pscreen = ptx->texture->screen; + nv50_transfer_rect_m2mf(pscreen, tx->buffer, tx->base.stride, + 0, 0, tx->base.width, tx->base.height, + mt->buffer, tx->level_pitch, + tx->level_x, tx->level_y, + tx->level_width, tx->level_height, + tx->base.block.size, tx->base.width, + tx->base.height, NOUVEAU_BO_GART, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART); + } + + pipe_buffer_reference(&tx->buffer, NULL); + pipe_texture_reference(&ptx->texture, NULL); + FREE(ptx); +} + +static void * +nv50_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv50_transfer *tx = (struct nv50_transfer *)ptx; + unsigned flags = 0; + + if (ptx->usage & PIPE_TRANSFER_WRITE) + flags |= PIPE_BUFFER_USAGE_CPU_WRITE; + if (ptx->usage & PIPE_TRANSFER_READ) + flags |= PIPE_BUFFER_USAGE_CPU_READ; + + return pipe_buffer_map(pscreen, tx->buffer, flags); +} + +static void +nv50_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +{ + struct nv50_transfer *tx = (struct nv50_transfer *)ptx; + + pipe_buffer_unmap(pscreen, tx->buffer); +} + +void +nv50_transfer_init_screen_functions(struct pipe_screen *pscreen) +{ + pscreen->get_tex_transfer = nv50_transfer_new; + pscreen->tex_transfer_destroy = nv50_transfer_del; + pscreen->transfer_map = nv50_transfer_map; + pscreen->transfer_unmap = nv50_transfer_unmap; +} diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index 0c970adb03..0749c90691 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -202,7 +202,7 @@ nv50_vbo_validate(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; struct nouveau_stateobj *vtxbuf, *vtxfmt; - int i, vpi = 0; + int i; vtxbuf = so_new(nv50->vtxelt_nr * 4, nv50->vtxelt_nr * 2); vtxfmt = so_new(nv50->vtxelt_nr + 1, 0); @@ -250,5 +250,7 @@ nv50_vbo_validate(struct nv50_context *nv50) so_ref (vtxfmt, &nv50->state.vtxfmt); so_ref (vtxbuf, &nv50->state.vtxbuf); + so_ref (NULL, &vtxbuf); + so_ref (NULL, &vtxfmt); } diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile index e83d943cd8..0e4e115532 100644 --- a/src/gallium/drivers/r300/Makefile +++ b/src/gallium/drivers/r300/Makefile @@ -7,15 +7,17 @@ C_SOURCES = \ r300_chipset.c \ r300_clear.c \ r300_context.c \ + r300_debug.c \ r300_emit.c \ r300_flush.c \ + r300_query.c \ r300_screen.c \ r300_state.c \ + r300_state_derived.c \ + r300_state_invariant.c \ r300_state_shader.c \ r300_surface.c \ r300_swtcl_emit.c \ r300_texture.c include ../../Makefile.template - -symlinks: diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c index 7def62422a..e01a0546b2 100644 --- a/src/gallium/drivers/r300/r300_chipset.c +++ b/src/gallium/drivers/r300/r300_chipset.c @@ -21,7 +21,7 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "r300_chipset.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" /* r300_chipset: A file all to itself for deducing the various properties of * Radeons. */ @@ -30,7 +30,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) { /* Reasonable defaults */ - caps->has_tcl = TRUE; + caps->has_tcl = getenv("RADEON_NO_TCL") ? TRUE : FALSE; caps->is_r500 = FALSE; caps->num_vert_fpus = 4; @@ -204,6 +204,13 @@ void r300_parse_chipset(struct r300_capabilities* caps) caps->has_tcl = FALSE; break; + case 0x793F: + case 0x7941: + case 0x7942: + caps->family = CHIP_FAMILY_RS600; + caps->has_tcl = FALSE; + break; + case 0x796C: case 0x796D: case 0x796E: @@ -343,6 +350,6 @@ void r300_parse_chipset(struct r300_capabilities* caps) break; } - /* Force off TCL for now */ - caps->has_tcl = FALSE; + /* XXX SW TCL is broken so no forcing it off right now + caps->has_tcl = FALSE; */ } diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h index a9cd372ec5..21eebeae60 100644 --- a/src/gallium/drivers/r300/r300_chipset.h +++ b/src/gallium/drivers/r300/r300_chipset.h @@ -64,6 +64,7 @@ enum { CHIP_FAMILY_RC410, CHIP_FAMILY_RS480, CHIP_FAMILY_RS482, + CHIP_FAMILY_RS600, CHIP_FAMILY_RS690, CHIP_FAMILY_RS740, CHIP_FAMILY_RV515, diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 7b605ae87a..b8584702aa 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -22,18 +22,87 @@ #include "r300_context.h" +static boolean r300_draw_range_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count) +{ + struct r300_context* r300 = r300_context(pipe); + int i; + + if (r300->dirty_state) { + r300_emit_dirty_state(r300); + } + + for (i = 0; i < r300->vertex_buffer_count; i++) { + void* buf = pipe_buffer_map(pipe->screen, + r300->vertex_buffers[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(r300->draw, i, buf); + } + + if (indexBuffer) { + void* indices = pipe_buffer_map(pipe->screen, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer_range(r300->draw, indexSize, + minIndex, maxIndex, indices); + } else { + draw_set_mapped_element_buffer(r300->draw, 0, NULL); + } + + draw_set_mapped_constant_buffer(r300->draw, + r300->shader_constants[PIPE_SHADER_VERTEX].constants, + r300->shader_constants[PIPE_SHADER_VERTEX].user_count * + (sizeof(float) * 4)); + + draw_arrays(r300->draw, mode, start, count); + + for (i = 0; i < r300->vertex_buffer_count; i++) { + pipe_buffer_unmap(pipe->screen, r300->vertex_buffers[i].buffer); + draw_set_mapped_vertex_buffer(r300->draw, i, NULL); + } + + if (indexBuffer) { + pipe_buffer_unmap(pipe->screen, indexBuffer); + draw_set_mapped_element_buffer_range(r300->draw, 0, start, + start + count - 1, NULL); + } + + return TRUE; +} + +static boolean r300_draw_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, unsigned mode, + unsigned start, unsigned count) +{ + return r300_draw_range_elements(pipe, indexBuffer, indexSize, 0, ~0, + mode, start, count); +} + +static boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, + unsigned start, unsigned count) +{ + return r300_draw_elements(pipe, NULL, 0, mode, start, count); +} + static void r300_destroy_context(struct pipe_context* context) { struct r300_context* r300 = r300_context(context); draw_destroy(r300->draw); FREE(r300->blend_color_state); + FREE(r300->rs_block); FREE(r300->scissor_state); + FREE(r300->viewport_state); FREE(r300); } struct pipe_context* r300_create_context(struct pipe_screen* screen, - struct pipe_winsys* winsys, struct r300_winsys* r300_winsys) { struct r300_context* r300 = CALLOC_STRUCT(r300_context); @@ -41,26 +110,37 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, if (!r300) return NULL; + /* XXX this could be refactored now? */ r300->winsys = r300_winsys; - r300->context.winsys = winsys; - r300->context.screen = r300_create_screen(winsys, r300_winsys); + + r300->context.winsys = (struct pipe_winsys*)r300_winsys; + r300->context.screen = r300_screen(screen); r300->context.destroy = r300_destroy_context; r300->context.clear = r300_clear; + r300->context.draw_arrays = r300_draw_arrays; + r300->context.draw_elements = r300_draw_elements; + r300->context.draw_range_elements = r300_draw_range_elements; + r300->draw = draw_create(); - /*XXX draw_set_rasterize_stage(r300->draw, r300_draw_swtcl_stage(r300));*/ + draw_set_rasterize_stage(r300->draw, r300_draw_swtcl_stage(r300)); r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state); + r300->rs_block = CALLOC_STRUCT(r300_rs_block); r300->scissor_state = CALLOC_STRUCT(r300_scissor_state); + r300->viewport_state = CALLOC_STRUCT(r300_viewport_state); r300_init_flush_functions(r300); + r300_init_query_functions(r300); + r300_init_surface_functions(r300); r300_init_state_functions(r300); + r300_emit_invariant_state(r300); r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw++; diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 376c57639d..0e5e471d11 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -24,11 +24,13 @@ #define R300_CONTEXT_H #include "draw/draw_context.h" +#include "draw/draw_vertex.h" #include "pipe/p_context.h" #include "tgsi/tgsi_scan.h" #include "util/u_memory.h" #include "r300_clear.h" +#include "r300_query.h" #include "r300_screen.h" #include "r300_winsys.h" @@ -58,8 +60,12 @@ struct r300_dsa_state { }; struct r300_rs_state { + /* XXX icky as fucking hell */ + struct pipe_rasterizer_state rs; + uint32_t vap_control_status; /* R300_VAP_CNTL_STATUS: 0x2140 */ uint32_t point_size; /* R300_GA_POINT_SIZE: 0x421c */ + uint32_t point_minmax; /* R300_GA_POINT_MINMAX: 0x4230 */ uint32_t line_control; /* R300_GA_LINE_CNTL: 0x4234 */ uint32_t depth_scale_front; /* R300_SU_POLY_OFFSET_FRONT_SCALE: 0x42a4 */ uint32_t depth_offset_front;/* R300_SU_POLY_OFFSET_FRONT_OFFSET: 0x42a8 */ @@ -69,6 +75,14 @@ struct r300_rs_state { uint32_t cull_mode; /* R300_SU_CULL_MODE: 0x42b8 */ uint32_t line_stipple_config; /* R300_GA_LINE_STIPPLE_CONFIG: 0x4328 */ uint32_t line_stipple_value; /* R300_GA_LINE_STIPPLE_VALUE: 0x4260 */ + uint32_t color_control; /* R300_GA_COLOR_CONTROL: 0x4278 */ +}; + +struct r300_rs_block { + uint32_t ip[8]; /* R300_RS_IP_[0-7], R500_RS_IP_[0-7] */ + uint32_t count; /* R300_RS_COUNT */ + uint32_t inst_count; /* R300_RS_INST_COUNT */ + uint32_t inst[8]; /* R300_RS_INST_[0-7] */ }; struct r300_sampler_state { @@ -83,23 +97,52 @@ struct r300_scissor_state { }; struct r300_texture_state { + uint32_t format0; /* R300_TX_FORMAT0: 0x4480 */ + uint32_t format1; /* R300_TX_FORMAT1: 0x44c0 */ + uint32_t format2; /* R300_TX_FORMAT2: 0x4500 */ +}; + +struct r300_viewport_state { + float xscale; /* R300_VAP_VPORT_XSCALE: 0x2098 */ + float xoffset; /* R300_VAP_VPORT_XOFFSET: 0x209c */ + float yscale; /* R300_VAP_VPORT_YSCALE: 0x20a0 */ + float yoffset; /* R300_VAP_VPORT_YOFFSET: 0x20a4 */ + float zscale; /* R300_VAP_VPORT_ZSCALE: 0x20a8 */ + float zoffset; /* R300_VAP_VPORT_ZOFFSET: 0x20ac */ + uint32_t vte_control; /* R300_VAP_VTE_CNTL: 0x20b0 */ }; -#define R300_NEW_BLEND 0x000001 -#define R300_NEW_BLEND_COLOR 0x000002 -#define R300_NEW_DSA 0x000004 -#define R300_NEW_FRAMEBUFFERS 0x000008 -#define R300_NEW_FRAGMENT_SHADER 0x000010 -#define R300_NEW_RASTERIZER 0x000020 -#define R300_NEW_SAMPLER 0x000040 -#define R300_NEW_SCISSOR 0x004000 -#define R300_NEW_TEXTURE 0x008000 -#define R300_NEW_VERTEX_SHADER 0x800000 -#define R300_NEW_KITCHEN_SINK 0xffffff +#define R300_NEW_BLEND 0x0000001 +#define R300_NEW_BLEND_COLOR 0x0000002 +#define R300_NEW_CONSTANTS 0x0000004 +#define R300_NEW_DSA 0x0000008 +#define R300_NEW_FRAMEBUFFERS 0x0000010 +#define R300_NEW_FRAGMENT_SHADER 0x0000020 +#define R300_NEW_RASTERIZER 0x0000040 +#define R300_NEW_RS_BLOCK 0x0000080 +#define R300_NEW_SAMPLER 0x0000100 +#define R300_ANY_NEW_SAMPLERS 0x000ff00 +#define R300_NEW_SCISSOR 0x0010000 +#define R300_NEW_TEXTURE 0x0020000 +#define R300_ANY_NEW_TEXTURES 0x1fe0000 +#define R300_NEW_VERTEX_FORMAT 0x2000000 +#define R300_NEW_VERTEX_SHADER 0x4000000 +#define R300_NEW_VIEWPORT 0x8000000 +#define R300_NEW_KITCHEN_SINK 0xfffffff /* The next several objects are not pure Radeon state; they inherit from * various Gallium classes. */ +struct r300_constant_buffer { + /* Buffer of constants */ + /* XXX first number should be raised */ + float constants[8][4]; + /* Number of user-defined constants */ + int user_count; + /* Total number of constants */ + int count; +}; + struct r3xx_fragment_shader { /* Parent class */ struct pipe_shader_state state; @@ -107,16 +150,55 @@ struct r3xx_fragment_shader { /* Has this shader been translated yet? */ boolean translated; + + /* Pixel stack size */ + int stack_size; }; struct r300_fragment_shader { /* Parent class */ struct r3xx_fragment_shader shader; + + /* Number of ALU instructions */ + int alu_instruction_count; + + /* Number of texture instructions */ + int tex_instruction_count; + + /* Number of texture indirections */ + int indirections; + + /* Indirection node offsets */ + int offset0; + int offset1; + int offset2; + int offset3; + + /* Machine instructions */ + struct { + uint32_t alu_rgb_inst; + uint32_t alu_rgb_addr; + uint32_t alu_alpha_inst; + uint32_t alu_alpha_addr; + } instructions[64]; /* XXX magic num */ }; struct r500_fragment_shader { /* Parent class */ struct r3xx_fragment_shader shader; + + /* Number of used instructions */ + int instruction_count; + + /* Machine instructions */ + struct { + uint32_t inst0; + uint32_t inst1; + uint32_t inst2; + uint32_t inst3; + uint32_t inst4; + uint32_t inst5; + } instructions[256]; /*< XXX magic number */ }; struct r300_texture { @@ -126,11 +208,30 @@ struct r300_texture { /* Offsets into the buffer. */ unsigned offset[PIPE_MAX_TEXTURE_LEVELS]; + /* Stride (pitch?) of this texture in bytes */ + unsigned stride; + /* Total size of this texture, in bytes. */ unsigned size; /* Pipe buffer backing this texture. */ struct pipe_buffer* buffer; + + /* Registers carrying texture format data. */ + struct r300_texture_state state; +}; + +struct r300_vertex_format { + /* Parent class */ + struct vertex_info vinfo; + /* R300_VAP_PROG_STREAK_CNTL_[0-7] */ + uint32_t vap_prog_stream_cntl[8]; + /* R300_VAP_PROG_STREAK_CNTL_EXT_[0-7] */ + uint32_t vap_prog_stream_cntl_ext[8]; + /* This is a map of VAP/SW TCL outputs into the GA/RS. + * tab[i] is the location of input i in GA/RS input memory. + * Named tab for historical reasons. */ + int tab[16]; }; struct r300_context { @@ -147,6 +248,8 @@ struct r300_context { struct r300_blend_state* blend_state; /* Blend color state. */ struct r300_blend_color_state* blend_color_state; + /* Shader constants. */ + struct r300_constant_buffer shader_constants[PIPE_SHADER_TYPES]; /* Depth, stencil, and alpha state. */ struct r300_dsa_state* dsa_state; /* Fragment shader. */ @@ -155,6 +258,8 @@ struct r300_context { struct pipe_framebuffer_state framebuffer_state; /* Rasterizer state. */ struct r300_rs_state* rs_state; + /* RS block state. */ + struct r300_rs_block* rs_block; /* Sampler states. */ struct r300_sampler_state* sampler_states[8]; int sampler_count; @@ -162,8 +267,14 @@ struct r300_context { struct r300_scissor_state* scissor_state; /* Texture states. */ struct r300_texture* textures[8]; - struct r300_texture_state* texture_states[8]; int texture_count; + /* Vertex buffers. */ + struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; + int vertex_buffer_count; + /* Vertex information. */ + struct r300_vertex_format vertex_info; + /* Viewport state. */ + struct r300_viewport_state* viewport_state; /* Bitmask of dirty state objects. */ uint32_t dirty_state; /* Flag indicating whether or not the HW is dirty. */ @@ -176,6 +287,7 @@ static struct r300_context* r300_context(struct pipe_context* context) { } /* Context initialization. */ +struct draw_stage* r300_draw_swtcl_stage(struct r300_context* r300); void r300_init_state_functions(struct r300_context* r300); void r300_init_surface_functions(struct r300_context* r300); @@ -183,7 +295,6 @@ void r300_init_surface_functions(struct r300_context* r300); * We'll just step out in that case... */ #ifndef R300_WINSYS_H struct pipe_context* r300_create_context(struct pipe_screen* screen, - struct pipe_winsys* winsys, struct r300_winsys* r300_winsys); #endif diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h index 385b61a096..d8038ff1e1 100644 --- a/src/gallium/drivers/r300/r300_cs.h +++ b/src/gallium/drivers/r300/r300_cs.h @@ -23,27 +23,19 @@ #ifndef R300_CS_H #define R300_CS_H +#include "util/u_math.h" + #include "r300_reg.h" #include "r300_winsys.h" -/* Pack a 32-bit float into a dword. */ -static uint32_t pack_float_32(float f) -{ - union { - float f; - uint32_t u; - } u; - - u.f = f; - return u.u; -} - /* Yes, I know macros are ugly. However, they are much prettier than the code * that they neatly hide away, and don't have the cost of function setup,so * we're going to use them. */ #define MAX_CS_SIZE 64 * 1024 / 4 +#define VERY_VERBOSE_REGISTERS 0 + /* XXX stolen from radeon_drm.h */ #define RADEON_GEM_DOMAIN_CPU 0x1 #define RADEON_GEM_DOMAIN_GTT 0x2 @@ -80,13 +72,14 @@ static uint32_t pack_float_32(float f) } while (0) #define OUT_CS_32F(value) do { \ - cs_winsys->write_cs_dword(cs, pack_float_32(value)); \ + cs_winsys->write_cs_dword(cs, fui(value)); \ cs_count--; \ } while (0) #define OUT_CS_REG(register, value) do { \ - debug_printf("r300: writing 0x%08X to register 0x%04X\n", \ - value, register); \ + if (VERY_VERBOSE_REGISTERS) \ + debug_printf("r300: writing 0x%08X to register 0x%04X\n", \ + value, register); \ assert(register); \ OUT_CS(CP_PACKET0(register, 0)); \ OUT_CS(value); \ @@ -95,8 +88,9 @@ static uint32_t pack_float_32(float f) /* Note: This expects count to be the number of registers, * not the actual packet0 count! */ #define OUT_CS_REG_SEQ(register, count) do { \ - debug_printf("r300: writing register sequence of %d to 0x%04X\n", \ - count, register); \ + if (VERY_VERBOSE_REGISTERS) \ + debug_printf("r300: writing register sequence of %d to 0x%04X\n", \ + count, register); \ assert(register); \ OUT_CS(CP_PACKET0(register, ((count) - 1))); \ } while (0) @@ -119,7 +113,7 @@ static uint32_t pack_float_32(float f) } while (0) #define FLUSH_CS do { \ - debug_printf("r300: FLUSH_CS in %s (%s:%d)\n", __FUNCTION__, __FILE__, \ + debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, __FILE__, \ __LINE__); \ cs_winsys->flush_cs(cs); \ } while (0) diff --git a/src/gallium/drivers/r300/r300_cs_inlines.h b/src/gallium/drivers/r300/r300_cs_inlines.h index 71e6623699..03bb608eb9 100644 --- a/src/gallium/drivers/r300/r300_cs_inlines.h +++ b/src/gallium/drivers/r300/r300_cs_inlines.h @@ -26,12 +26,25 @@ #ifdef R300_CS_H +#define RADEON_ONE_REG_WR (1 << 15) + +#define OUT_CS_ONE_REG(register, count) do { \ + if (VERY_VERBOSE_REGISTERS) \ + debug_printf("r300: writing data sequence of %d to 0x%04X\n", \ + count, register); \ + assert(register); \ + OUT_CS(CP_PACKET0(register, ((count) - 1)) | RADEON_ONE_REG_WR); \ +} while (0) + #define R300_PACIFY do { \ + OUT_CS_REG(RADEON_WAIT_UNTIL, (1 << 14) | (1 << 15) | (1 << 16) | (1 << 17) | \ + (1 << 18)); \ +} while (0) + +#define R300_SCREENDOOR do { \ OUT_CS_REG(R300_SC_SCREENDOOR, 0x0); \ - OUT_CS_REG(RADEON_WAIT_UNTIL, (1 << 15) | (1 << 17) | \ - (1 << 18) | (1 << 31)); \ + R300_PACIFY; \ OUT_CS_REG(R300_SC_SCREENDOOR, 0xffffff); \ } while (0) - #endif /* R300_CS_H */ diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c new file mode 100644 index 0000000000..f657588c72 --- /dev/null +++ b/src/gallium/drivers/r300/r300_debug.c @@ -0,0 +1,218 @@ +/* + * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_debug.h" + +static char* r500_fs_swiz[] = { + " R", + " G", + " B", + " A", + " 0", + ".5", + " 1", + " U", +}; + +static char* r500_fs_op_rgb[] = { + "MAD", + "DP3", + "DP4", + "D2A", + "MIN", + "MAX", + "---", + "CND", + "CMP", + "FRC", + "SOP", + "MDH", + "MDV", +}; + +static char* r500_fs_op_alpha[] = { + "MAD", + " DP", + "MIN", + "MAX", + "---", + "CND", + "CMP", + "FRC", + "EX2", + "LN2", + "RCP", + "RSQ", + "SIN", + "COS", + "MDH", + "MDV", +}; + +static char* r500_fs_mask[] = { + "NONE", + "R ", + " G ", + "RG ", + " B ", + "R B ", + " GB ", + "RGB ", + " A", + "R A", + " G A", + "RG A", + " BA", + "R BA", + " GBA", + "RGBA", +}; + +static char* r500_fs_tex[] = { + " NOP", + " LD", + "TEXKILL", + " PROJ", + "LODBIAS", + " LOD", + " DXDY", +}; + +void r500_fs_dump(struct r500_fragment_shader* fs) +{ + int i; + uint32_t inst; + + for (i = 0; i < fs->instruction_count; i++) { + inst = fs->instructions[i].inst0; + debug_printf("%d: 0: CMN_INST 0x%08x:", i, inst); + switch (inst & 0x3) { + case R500_INST_TYPE_ALU: + debug_printf("ALU "); + break; + case R500_INST_TYPE_OUT: + debug_printf("OUT "); + break; + case R500_INST_TYPE_FC: + debug_printf("FC "); + break; + case R500_INST_TYPE_TEX: + debug_printf("TEX "); + break; + } + debug_printf("%s %s %s %s ", + inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "", + inst & R500_INST_LAST ? "LAST" : "", + inst & R500_INST_NOP ? "NOP" : "", + inst & R500_INST_ALU_WAIT ? "ALU_WAIT" : ""); + debug_printf("wmask: %s omask: %s\n", + r500_fs_mask[(inst >> 11) & 0xf], + r500_fs_mask[(inst >> 15) & 0xf]); + switch (inst & 0x3) { + case R500_INST_TYPE_ALU: + case R500_INST_TYPE_OUT: + inst = fs->instructions[i].inst1; + debug_printf(" 1: RGB_ADDR 0x%08x:", inst); + debug_printf("Addr0: %d%c, Addr1: %d%c, " + "Addr2: %d%c, srcp:%d\n", + inst & 0xff, (inst & (1 << 8)) ? 'c' : 't', + (inst >> 10) & 0xff, (inst & (1 << 18)) ? 'c' : 't', + (inst >> 20) & 0xff, (inst & (1 << 28)) ? 'c' : 't', + (inst >> 30)); + + inst = fs->instructions[i].inst2; + debug_printf(" 2: ALPHA_ADDR 0x%08x:", inst); + debug_printf("Addr0: %d%c, Addr1: %d%c, " + "Addr2: %d%c, srcp:%d\n", + inst & 0xff, (inst & (1 << 8)) ? 'c' : 't', + (inst >> 10) & 0xff, (inst & (1 << 18)) ? 'c' : 't', + (inst >> 20) & 0xff, (inst & (1 << 28)) ? 'c' : 't', + (inst >> 30)); + + inst = fs->instructions[i].inst3; + debug_printf(" 3: RGB_INST 0x%08x:", inst); + debug_printf("rgb_A_src:%d %s/%s/%s %d " + "rgb_B_src:%d %s/%s/%s %d\n", + inst & 0x3, r500_fs_swiz[(inst >> 2) & 0x7], + r500_fs_swiz[(inst >> 5) & 0x7], + r500_fs_swiz[(inst >> 8) & 0x7], + (inst >> 11) & 0x3, (inst >> 13) & 0x3, + r500_fs_swiz[(inst >> 15) & 0x7], + r500_fs_swiz[(inst >> 18) & 0x7], + r500_fs_swiz[(inst >> 21) & 0x7], + (inst >> 24) & 0x3); + + inst = fs->instructions[i].inst4; + debug_printf(" 4: ALPHA_INST 0x%08x:", inst); + debug_printf("%s dest:%d%s alp_A_src:%d %s %d " + "alp_B_src:%d %s %d w:%d\n", + r500_fs_op_alpha[inst & 0xf], (inst >> 4) & 0x7f, + inst & (1<<11) ? "(rel)":"", (inst >> 12) & 0x3, + r500_fs_swiz[(inst >> 14) & 0x7], (inst >> 17) & 0x3, + (inst >> 19) & 0x3, r500_fs_swiz[(inst >> 21) & 0x7], + (inst >> 24) & 0x3, (inst >> 31) & 0x1); + + inst = fs->instructions[i].inst5; + debug_printf(" 5: RGBA_INST 0x%08x:", inst); + debug_printf("%s dest:%d%s rgb_C_src:%d %s/%s/%s %d " + "alp_C_src:%d %s %d\n", + r500_fs_op_rgb[inst & 0xf], (inst >> 4) & 0x7f, + inst & (1 << 11) ? "(rel)":"", (inst >> 12) & 0x3, + r500_fs_swiz[(inst >> 14) & 0x7], + r500_fs_swiz[(inst >> 17) & 0x7], + r500_fs_swiz[(inst >> 20) & 0x7], + (inst >> 23) & 0x3, (inst >> 25) & 0x3, + r500_fs_swiz[(inst >> 27) & 0x7], (inst >> 30) & 0x3); + break; + case R500_INST_TYPE_FC: + /* XXX don't even bother yet */ + break; + case R500_INST_TYPE_TEX: + inst = fs->instructions[i].inst1; + debug_printf(" 1: TEX_INST 0x%08x: id: %d " + "op:%s, %s, %s %s\n", + inst, (inst >> 16) & 0xf, + r500_fs_tex[(inst >> 22) & 0x7], + (inst & (1 << 25)) ? "ACQ" : "", + (inst & (1 << 26)) ? "IGNUNC" : "", + (inst & (1 << 27)) ? "UNSCALED" : "SCALED"); + + inst = fs->instructions[i].inst2; + debug_printf(" 2: TEX_ADDR 0x%08x: " + "src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n", + inst, inst & 0x7f, inst & (1 << 7) ? "(rel)" : "", + r500_fs_swiz[(inst >> 8) & 0x3], + r500_fs_swiz[(inst >> 10) & 0x3], + r500_fs_swiz[(inst >> 12) & 0x3], + r500_fs_swiz[(inst >> 14) & 0x3], + (inst >> 16) & 0x7f, inst & (1 << 23) ? "(rel)" : "", + r500_fs_swiz[(inst >> 24) & 0x3], + r500_fs_swiz[(inst >> 26) & 0x3], + r500_fs_swiz[(inst >> 28) & 0x3], + r500_fs_swiz[(inst >> 30) & 0x3]); + + inst = fs->instructions[i].inst3; + debug_printf(" 3: TEX_DXDY 0x%08x\n", inst); + break; + } + } +} diff --git a/src/gallium/drivers/r300/r300_debug.h b/src/gallium/drivers/r300/r300_debug.h new file mode 100644 index 0000000000..de5d701ed9 --- /dev/null +++ b/src/gallium/drivers/r300/r300_debug.h @@ -0,0 +1,31 @@ +/* + * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_DEBUG_H +#define R300_DEBUG_H + +#include "r300_reg.h" +#include "r300_state_shader.h" + +void r500_fs_dump(struct r500_fragment_shader* fs); + +#endif /* R300_DEBUG_H */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 585a9e729d..a2e771bd1b 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -40,9 +40,9 @@ void r300_emit_blend_state(struct r300_context* r300, void r300_emit_blend_color_state(struct r300_context* r300, struct r300_blend_color_state* bc) { - struct r300_screen* r300screen = - (struct r300_screen*)r300->context.screen; + struct r300_screen* r300screen = r300_screen(r300->context.screen); CS_LOCALS(r300); + if (r300screen->caps->is_r500) { BEGIN_CS(3); OUT_CS_REG_SEQ(R500_RB3D_CONSTANT_COLOR_AR, 2); @@ -59,9 +59,9 @@ void r300_emit_blend_color_state(struct r300_context* r300, void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa) { - struct r300_screen* r300screen = - (struct r300_screen*)r300->context.screen; + struct r300_screen* r300screen = r300_screen(r300->context.screen); CS_LOCALS(r300); + BEGIN_CS(r300screen->caps->is_r500 ? 8 : 8); OUT_CS_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function); /* XXX figure out the r300 counterpart for this */ @@ -79,7 +79,95 @@ void r300_emit_dsa_state(struct r300_context* r300, END_CS; } -/* XXX add pitch, stride, z/stencil buf */ +void r300_emit_fragment_shader(struct r300_context* r300, + struct r300_fragment_shader* fs) +{ + CS_LOCALS(r300); + int i; + + BEGIN_CS(22); + + OUT_CS_REG(R300_US_CONFIG, MAX2(fs->indirections - 1, 0)); + OUT_CS_REG(R300_US_PIXSIZE, fs->shader.stack_size); + /* XXX figure out exactly how big the sizes are on this reg */ + OUT_CS_REG(R300_US_CODE_OFFSET, 0x0); + /* XXX figure these ones out a bit better kthnx */ + OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_3, R300_RGBA_OUT); + + for (i = 0; i < fs->alu_instruction_count; i++) { + OUT_CS_REG(R300_US_ALU_RGB_INST_0 + (4 * i), + fs->instructions[i].alu_rgb_inst); + OUT_CS_REG(R300_US_ALU_RGB_ADDR_0 + (4 * i), + fs->instructions[i].alu_rgb_addr); + OUT_CS_REG(R300_US_ALU_ALPHA_INST_0 + (4 * i), + fs->instructions[i].alu_alpha_inst); + OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0 + (4 * i), + fs->instructions[i].alu_alpha_addr); + } + + END_CS; +} + +void r500_emit_fragment_shader(struct r300_context* r300, + struct r500_fragment_shader* fs) +{ + CS_LOCALS(r300); + struct r300_constant_buffer* constants = + &r300->shader_constants[PIPE_SHADER_FRAGMENT]; + int i; + + BEGIN_CS(9 + (fs->instruction_count * 6) + (constants->count ? 3 : 0) + + (constants->count * 4)); + OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); + OUT_CS_REG(R500_US_PIXSIZE, fs->shader.stack_size); + OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) | + R500_US_CODE_END_ADDR(fs->instruction_count)); + + OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_INSTR); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, fs->instruction_count * 6); + for (i = 0; i < fs->instruction_count; i++) { + OUT_CS(fs->instructions[i].inst0); + OUT_CS(fs->instructions[i].inst1); + OUT_CS(fs->instructions[i].inst2); + OUT_CS(fs->instructions[i].inst3); + OUT_CS(fs->instructions[i].inst4); + OUT_CS(fs->instructions[i].inst5); + } + + if (constants->count) { + OUT_CS_REG(R500_GA_US_VECTOR_INDEX, + R500_GA_US_VECTOR_INDEX_TYPE_CONST); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->count * 4); + for (i = 0; i < constants->count; i++) { + OUT_CS_32F(constants->constants[i][0]); + OUT_CS_32F(constants->constants[i][1]); + OUT_CS_32F(constants->constants[i][2]); + OUT_CS_32F(constants->constants[i][3]); + } + } + + END_CS; +} + +/* Translate pipe_format into US_OUT_FMT. Note that formats are stored from + * C3 to C0. */ +uint32_t translate_out_fmt(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + return R300_US_OUT_FMT_C4_8 | + R300_C0_SEL_B | R300_C1_SEL_G | + R300_C2_SEL_R | R300_C3_SEL_A; + default: + return R300_US_OUT_FMT_UNUSED; + } + return 0; +} + +/* XXX add pitch, stride, clean up */ void r300_emit_fb_state(struct r300_context* r300, struct pipe_framebuffer_state* fb) { @@ -87,23 +175,47 @@ void r300_emit_fb_state(struct r300_context* r300, struct r300_texture* tex; int i; - BEGIN_CS((3 * fb->nr_cbufs) + 6); + BEGIN_CS((6 * fb->nr_cbufs) + (fb->zsbuf ? 6 : 0) + 4); for (i = 0; i < fb->nr_cbufs; i++) { tex = (struct r300_texture*)fb->cbufs[i]->texture; OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + + OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), + translate_out_fmt(fb->cbufs[i]->format)); + } + + if (fb->zsbuf) { + tex = (struct r300_texture*)fb->zsbuf->texture; + OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1); + OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + if (fb->zsbuf->format == PIPE_FORMAT_Z24S8_UNORM) { + OUT_CS_REG(R300_ZB_FORMAT, + R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL); + } else { + OUT_CS_REG(R300_ZB_FORMAT, 0x0); + } } - R300_PACIFY; + + OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, + R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS | + R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D); + OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, + R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | + R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); END_CS; } void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs) { - struct r300_screen* r300screen = - (struct r300_screen*)r300->context.screen; CS_LOCALS(r300); - BEGIN_CS(14); + + BEGIN_CS(20); OUT_CS_REG(R300_VAP_CNTL_STATUS, rs->vap_control_status); + OUT_CS_REG(R300_GA_POINT_SIZE, rs->point_size); + OUT_CS_REG_SEQ(R300_GA_POINT_MINMAX, 2); + OUT_CS(rs->point_minmax); + OUT_CS(rs->line_control); OUT_CS_REG_SEQ(R300_SU_POLY_OFFSET_FRONT_SCALE, 6); OUT_CS(rs->depth_scale_front); OUT_CS(rs->depth_offset_front); @@ -113,43 +225,241 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs) OUT_CS(rs->cull_mode); OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, rs->line_stipple_config); OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, rs->line_stipple_value); + OUT_CS_REG(R300_GA_COLOR_CONTROL, rs->color_control); + END_CS; +} + +void r300_emit_rs_block_state(struct r300_context* r300, + struct r300_rs_block* rs) +{ + struct r300_screen* r300screen = r300_screen(r300->context.screen); + CS_LOCALS(r300); + int i; + + BEGIN_CS(21); + if (r300screen->caps->is_r500) { + OUT_CS_REG_SEQ(R500_RS_IP_0, 8); + } else { + OUT_CS_REG_SEQ(R300_RS_IP_0, 8); + } + for (i = 0; i < 8; i++) { + OUT_CS(rs->ip[i]); + debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); + } + + OUT_CS_REG_SEQ(R300_RS_COUNT, 2); + OUT_CS(rs->count); + OUT_CS(rs->inst_count); + + if (r300screen->caps->is_r500) { + OUT_CS_REG_SEQ(R500_RS_INST_0, 8); + } else { + OUT_CS_REG_SEQ(R300_RS_INST_0, 8); + } + for (i = 0; i < 8; i++) { + OUT_CS(rs->inst[i]); + debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); + } + + debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count, + rs->inst_count); + END_CS; } -static void r300_emit_dirty_state(struct r300_context* r300) +void r300_emit_sampler(struct r300_context* r300, + struct r300_sampler_state* sampler, unsigned offset) { - struct r300_screen* r300screen = - (struct r300_screen*)r300->context.screen; CS_LOCALS(r300); + BEGIN_CS(6); + OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0); + OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); + OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); + END_CS; +} + +void r300_emit_scissor_state(struct r300_context* r300, + struct r300_scissor_state* scissor) +{ + CS_LOCALS(r300); + + BEGIN_CS(3); + OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2); + OUT_CS(scissor->scissor_top_left); + OUT_CS(scissor->scissor_bottom_right); + END_CS; +} + +void r300_emit_texture(struct r300_context* r300, + struct r300_texture* tex, unsigned offset) +{ + CS_LOCALS(r300); + + BEGIN_CS(10); + OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0); + OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1); + OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2); + OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (offset * 4), 1); + OUT_CS_RELOC(tex->buffer, 0, + RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0, 0); + END_CS; +} + +void r300_emit_vertex_format_state(struct r300_context* r300) +{ + CS_LOCALS(r300); + int i; + + BEGIN_CS(26); + OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info.vinfo.size); + + OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL, 2); + OUT_CS(r300->vertex_info.vinfo.hwfmt[0]); + OUT_CS(r300->vertex_info.vinfo.hwfmt[1]); + OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); + OUT_CS(r300->vertex_info.vinfo.hwfmt[2]); + OUT_CS(r300->vertex_info.vinfo.hwfmt[3]); + for (i = 0; i < 4; i++) { + debug_printf("hwfmt%d: 0x%08x\n", i, + r300->vertex_info.vinfo.hwfmt[i]); + } + + OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8); + for (i = 0; i < 8; i++) { + OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]); + debug_printf("prog_stream_cntl%d: 0x%08x\n", i, + r300->vertex_info.vap_prog_stream_cntl[i]); + } + OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8); + for (i = 0; i < 8; i++) { + OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]); + debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, + r300->vertex_info.vap_prog_stream_cntl_ext[i]); + } + END_CS; +} + +void r300_emit_viewport_state(struct r300_context* r300, + struct r300_viewport_state* viewport) +{ + return; + CS_LOCALS(r300); + + BEGIN_CS(7); + OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 7); + OUT_CS_32F(viewport->xscale); + OUT_CS_32F(viewport->xoffset); + OUT_CS_32F(viewport->yscale); + OUT_CS_32F(viewport->yoffset); + OUT_CS_32F(viewport->zscale); + OUT_CS_32F(viewport->zoffset); + OUT_CS(viewport->vte_control); + END_CS; +} + +static void r300_flush_textures(struct r300_context* r300) +{ + CS_LOCALS(r300); + + BEGIN_CS(4); + OUT_CS_REG(R300_TX_INVALTAGS, 0); + OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); + END_CS; +} + +/* Emit all dirty state. */ +void r300_emit_dirty_state(struct r300_context* r300) +{ + struct r300_screen* r300screen = r300_screen(r300->context.screen); + int i; + int dirty_tex = 0; + if (!(r300->dirty_state) && !(r300->dirty_hw)) { return; } + r300_update_derived_state(r300); + /* XXX check size */ if (r300->dirty_state & R300_NEW_BLEND) { r300_emit_blend_state(r300, r300->blend_state); + r300->dirty_state &= ~R300_NEW_BLEND; } if (r300->dirty_state & R300_NEW_BLEND_COLOR) { r300_emit_blend_color_state(r300, r300->blend_color_state); + r300->dirty_state &= ~R300_NEW_BLEND_COLOR; } if (r300->dirty_state & R300_NEW_DSA) { r300_emit_dsa_state(r300, r300->dsa_state); + r300->dirty_state &= ~R300_NEW_DSA; + } + + if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { + if (r300screen->caps->is_r500) { + r500_emit_fragment_shader(r300, + (struct r500_fragment_shader*)r300->fs); + } else { + r300_emit_fragment_shader(r300, + (struct r300_fragment_shader*)r300->fs); + } + r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER; + } + + if (r300->dirty_state & R300_NEW_FRAMEBUFFERS) { + r300_emit_fb_state(r300, &r300->framebuffer_state); + r300->dirty_state &= ~R300_NEW_FRAMEBUFFERS; } if (r300->dirty_state & R300_NEW_RASTERIZER) { r300_emit_rs_state(r300, r300->rs_state); + r300->dirty_state &= ~R300_NEW_RASTERIZER; + } + + if (r300->dirty_state & R300_NEW_RS_BLOCK) { + r300_emit_rs_block_state(r300, r300->rs_block); + r300->dirty_state &= ~R300_NEW_RS_BLOCK; + } + + if (r300->dirty_state & R300_ANY_NEW_SAMPLERS) { + for (i = 0; i < r300->sampler_count; i++) { + if (r300->dirty_state & (R300_NEW_SAMPLER << i)) { + r300_emit_sampler(r300, r300->sampler_states[i], i); + r300->dirty_state &= ~(R300_NEW_SAMPLER << i); + dirty_tex++; + } + } } if (r300->dirty_state & R300_NEW_SCISSOR) { - struct r300_scissor_state* scissor = r300->scissor_state; - /* XXX next two are contiguous regs */ - OUT_CS_REG(R300_SC_SCISSORS_TL, scissor->scissor_top_left); - OUT_CS_REG(R300_SC_SCISSORS_BR, scissor->scissor_bottom_right); + r300_emit_scissor_state(r300, r300->scissor_state); + r300->dirty_state &= ~R300_NEW_SCISSOR; } - r300->dirty_state = 0; + if (r300->dirty_state & R300_ANY_NEW_TEXTURES) { + for (i = 0; i < r300->texture_count; i++) { + if (r300->dirty_state & (R300_NEW_TEXTURE << i)) { + r300_emit_texture(r300, r300->textures[i], i); + r300->dirty_state &= ~(R300_NEW_TEXTURE << i); + dirty_tex++; + } + } + } + + if (r300->dirty_state & R300_NEW_VIEWPORT) { + r300_emit_viewport_state(r300, r300->viewport_state); + r300->dirty_state &= ~R300_NEW_VIEWPORT; + } + + if (dirty_tex) { + r300_flush_textures(r300); + } + + if (r300->dirty_state & R300_NEW_VERTEX_FORMAT) { + r300_emit_vertex_format_state(r300); + r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT; + } } diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index b6e69386f9..4aba1ee08c 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -20,6 +20,11 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#ifndef R300_EMIT_H +#define R300_EMIT_H + +#include "util/u_math.h" + #include "r300_context.h" #include "r300_cs.h" #include "r300_screen.h" @@ -33,4 +38,26 @@ void r300_emit_blend_color_state(struct r300_context* r300, void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa); +void r300_emit_fragment_shader(struct r300_context* r300, + struct r300_fragment_shader* fs); + +void r500_emit_fragment_shader(struct r300_context* r300, + struct r500_fragment_shader* fs); + +void r300_emit_fb_state(struct r300_context* r300, + struct pipe_framebuffer_state* fb); + void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs); + +void r300_emit_rs_block_state(struct r300_context* r300, + struct r300_rs_block* rs); + +void r300_emit_scissor_state(struct r300_context* r300, + struct r300_scissor_state* scissor); + +void r300_emit_vertex_format_state(struct r300_context* r300); + +/* Emit all dirty state. */ +void r300_emit_dirty_state(struct r300_context* r300); + +#endif /* R300_EMIT_H */ diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index 3766f0a0a7..20ca6905ad 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -31,6 +31,7 @@ static void r300_flush(struct pipe_context* pipe, if (r300->dirty_hw) { FLUSH_CS; + r300_emit_invariant_state(r300); r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw = 0; } diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c new file mode 100644 index 0000000000..5f5f4c4dbd --- /dev/null +++ b/src/gallium/drivers/r300/r300_query.c @@ -0,0 +1,111 @@ +/* + * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_query.h" + +static struct pipe_query* r300_create_query(struct pipe_context* pipe, + unsigned query_type) +{ + struct r300_query* q = CALLOC_STRUCT(r300_query); + + q->type = query_type; + assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER); + + /* XXX this is to force winsys to give us a GTT buffer */ + q->buf = pipe->screen->buffer_create(pipe->screen, 64, + PIPE_BUFFER_USAGE_VERTEX, 64); + + return (struct pipe_query*)q; +} + +static void r300_destroy_query(struct pipe_context* pipe, + struct pipe_query* query) +{ + FREE(query); +} + +static void r300_begin_query(struct pipe_context* pipe, + struct pipe_query* query) +{ + struct r300_context* r300 = r300_context(pipe); + struct r300_query* q = (struct r300_query*)query; + CS_LOCALS(r300); + + uint32_t* map = pipe_buffer_map(pipe->screen, q->buf, + PIPE_BUFFER_USAGE_CPU_WRITE); + *map = ~0; + pipe_buffer_unmap(pipe->screen, q->buf); + + BEGIN_CS(2); + OUT_CS_REG(R300_ZB_ZPASS_DATA, 0); + END_CS; +} + +static void r300_end_query(struct pipe_context* pipe, + struct pipe_query* query) +{ + struct r300_context* r300 = r300_context(pipe); + struct r300_query* q = (struct r300_query*)query; + CS_LOCALS(r300); + + BEGIN_CS(4); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(q->buf, 0, 0, RADEON_GEM_DOMAIN_GTT, 0); + END_CS; +} + +static boolean r300_get_query_result(struct pipe_context* pipe, + struct pipe_query* query, + boolean wait, + uint64_t* result) +{ + struct r300_query* q = (struct r300_query*)query; + uint32_t temp; + + if (wait) { + /* Well, we're expected to just sit here and spin, so let's go ahead + * and flush so we can be sure that the card's spinning... */ + /* XXX double-check these params */ + pipe->flush(pipe, 0, NULL); + } + + uint32_t* map = pipe_buffer_map(pipe->screen, q->buf, + PIPE_BUFFER_USAGE_CPU_READ); + temp = *map; + pipe_buffer_unmap(pipe->screen, q->buf); + + if (temp < 0) { + /* Our results haven't been written yet... */ + return FALSE; + } + + *result = temp; + return TRUE; +} + +void r300_init_query_functions(struct r300_context* r300) { + r300->context.create_query = r300_create_query; + r300->context.destroy_query = r300_destroy_query; + r300->context.begin_query = r300_begin_query; + r300->context.end_query = r300_end_query; + r300->context.get_query_result = r300_get_query_result; +} diff --git a/src/gallium/drivers/r300/r300_query.h b/src/gallium/drivers/r300/r300_query.h new file mode 100644 index 0000000000..4f447ea45b --- /dev/null +++ b/src/gallium/drivers/r300/r300_query.h @@ -0,0 +1,44 @@ +/* + * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_QUERY_H +#define R300_QUERY_H + +#include "r300_context.h" +#include "r300_cs.h" +#include "r300_reg.h" + +struct r300_query { + /* The kind of query. Currently only OQ is supported. */ + unsigned type; + /* Buffer object where we want our results to reside. */ + struct pipe_buffer* buf; +}; + +static INLINE struct r300_query* r300_query(struct pipe_query* q) +{ + return (struct r300_query*)q; +} + +void r300_init_query_functions(struct r300_context* r300); + +#endif /* R300_QUERY_H */ diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index dbd0cc28e2..6f3ad970ab 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -64,7 +64,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define R300_SE_VPORT_ZSCALE 0x1DA8 #define R300_SE_VPORT_ZOFFSET 0x1DAC - +#define R300_VAP_PORT_IDX0 0x2040 /* * Vertex Array Processing (VAP) Control */ @@ -139,17 +139,25 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_VAP_OUTPUT_VTX_FMT_1__3_COMPONENTS 3 # define R300_VAP_OUTPUT_VTX_FMT_1__4_COMPONENTS 4 -#define R300_SE_VTE_CNTL 0x20b0 -# define R300_VPORT_X_SCALE_ENA (1 << 0) -# define R300_VPORT_X_OFFSET_ENA (1 << 1) -# define R300_VPORT_Y_SCALE_ENA (1 << 2) -# define R300_VPORT_Y_OFFSET_ENA (1 << 3) -# define R300_VPORT_Z_SCALE_ENA (1 << 4) -# define R300_VPORT_Z_OFFSET_ENA (1 << 5) -# define R300_VTX_XY_FMT (1 << 8) -# define R300_VTX_Z_FMT (1 << 9) -# define R300_VTX_W0_FMT (1 << 10) -# define R300_SERIAL_PROC_ENA (1 << 11) +#define R300_VAP_VPORT_XSCALE 0x2098 +#define R300_VAP_VPORT_XOFFSET 0x209c +#define R300_VAP_VPORT_YSCALE 0x20a0 +#define R300_VAP_VPORT_YOFFSET 0x20a4 +#define R300_VAP_VPORT_ZSCALE 0x20a8 +#define R300_VAP_VPORT_ZOFFSET 0x20ac + +#define R300_VAP_VTE_CNTL 0x20b0 +#define R300_SE_VTE_CNTL R300_VAP_VTE_CNTL +# define R300_VPORT_X_SCALE_ENA (1 << 0) +# define R300_VPORT_X_OFFSET_ENA (1 << 1) +# define R300_VPORT_Y_SCALE_ENA (1 << 2) +# define R300_VPORT_Y_OFFSET_ENA (1 << 3) +# define R300_VPORT_Z_SCALE_ENA (1 << 4) +# define R300_VPORT_Z_OFFSET_ENA (1 << 5) +# define R300_VTX_XY_FMT (1 << 8) +# define R300_VTX_Z_FMT (1 << 9) +# define R300_VTX_W0_FMT (1 << 10) +# define R300_SERIAL_PROC_ENA (1 << 11) #define R300_VAP_VTX_SIZE 0x20b4 @@ -326,6 +334,14 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_WRITE_ENA_Z 4 # define R300_WRITE_ENA_W 8 # define R300_SWIZZLE1_SHIFT 16 + +# define R300_VAP_SWIZZLE_XYZW \ + ((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | \ + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | \ + (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | \ + (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | \ + (0xf << R300_WRITE_ENA_SHIFT)) + #define R300_VAP_PROG_STREAM_CNTL_EXT_1 0x21e4 #define R300_VAP_PROG_STREAM_CNTL_EXT_2 0x21e8 #define R300_VAP_PROG_STREAM_CNTL_EXT_3 0x21ec @@ -732,8 +748,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define R500_RS_IP_TEX_PTR_Q_SHIFT 18 #define R500_RS_IP_COL_PTR_SHIFT 24 #define R500_RS_IP_COL_FMT_SHIFT 27 -# define R500_RS_COL_PTR(x) (x << 24) -# define R500_RS_COL_FMT(x) (x << 27) +# define R500_RS_SEL_S(x) ((x) << 0) +# define R500_RS_SEL_T(x) ((x) << 6) +# define R500_RS_SEL_R(x) ((x) << 12) +# define R500_RS_SEL_Q(x) ((x) << 18) +# define R500_RS_COL_PTR(x) ((x) << 24) +# define R500_RS_COL_FMT(x) ((x) << 27) /* gap */ #define R500_RS_IP_OFFSET_DIS (0 << 31) #define R500_RS_IP_OFFSET_EN (1 << 31) @@ -1019,20 +1039,27 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_THIRD (2 << 16) # define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST (3 << 16) -/** TODO: might be candidate for removal */ -# define R300_RE_SHADE_MODEL_SMOOTH ( \ - R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA0_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_RGB1_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST ) -/** TODO: might be candidate for removal, the GOURAUD stuff also looks buggy to me */ -# define R300_RE_SHADE_MODEL_FLAT ( \ - R300_GA_COLOR_CONTROL_RGB0_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA0_SHADING_FLAT | \ - R300_GA_COLOR_CONTROL_RGB1_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \ - R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST ) +# define R300_SHADE_MODEL_FLAT ( \ + R300_GA_COLOR_CONTROL_RGB0_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_ALPHA0_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_RGB1_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_ALPHA1_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT | \ + R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST ) + +# define R300_SHADE_MODEL_SMOOTH ( \ + R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_ALPHA0_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_RGB1_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \ + R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST ) /* Specifies red & green components of fill color -- S312 format -- Backwards comp. */ #define R300_GA_SOLID_RG 0x427c @@ -1146,6 +1173,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_W_ADDR_MASK 0x0003f000 # define R300_HIRES_DIS (0 << 18) # define R300_HIRES_EN (1 << 18) +# define R300_IT_COUNT(x) ((x) << 0) +# define R300_IC_COUNT(x) ((x) << 7) +# define R300_W_COUNT(x) ((x) << 12) #define R300_RS_INST_COUNT 0x4304 # define R300_RS_INST_COUNT_SHIFT 0 @@ -1175,8 +1205,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_RS_INTERP_SRC_SHIFT 2 /* TODO: check for removal */ # define R300_RS_INTERP_SRC_MASK (7 << 2) /* TODO: check for removal */ # define R300_RS_TEX_PTR(x) (x << 0) -# define R300_RS_COL_PTR(x) (x << 6) -# define R300_RS_COL_FMT(x) (x << 9) +# define R300_RS_COL_PTR(x) ((x) << 6) +# define R300_RS_COL_FMT(x) ((x) << 9) # define R300_RS_COL_FMT_RGBA 0 # define R300_RS_COL_FMT_RGB0 1 # define R300_RS_COL_FMT_RGB1 2 @@ -1186,10 +1216,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_RS_COL_FMT_111A 8 # define R300_RS_COL_FMT_1110 9 # define R300_RS_COL_FMT_1111 10 -# define R300_RS_SEL_S(x) (x << 13) -# define R300_RS_SEL_T(x) (x << 16) -# define R300_RS_SEL_R(x) (x << 19) -# define R300_RS_SEL_Q(x) (x << 22) +# define R300_RS_SEL_S(x) ((x) << 13) +# define R300_RS_SEL_T(x) ((x) << 16) +# define R300_RS_SEL_R(x) ((x) << 19) +# define R300_RS_SEL_Q(x) ((x) << 22) # define R300_RS_SEL_C0 0 # define R300_RS_SEL_C1 1 # define R300_RS_SEL_C2 2 @@ -1216,14 +1246,18 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define R500_RS_INST_14 0x4358 #define R500_RS_INST_15 0x435c #define R500_RS_INST_TEX_ID_SHIFT 0 +# define R500_RS_INST_TEX_ID(x) ((x) << 0) #define R500_RS_INST_TEX_CN_WRITE (1 << 4) #define R500_RS_INST_TEX_ADDR_SHIFT 5 +# define R500_RS_INST_TEX_ADDR(x) ((x) << 0) #define R500_RS_INST_COL_ID_SHIFT 12 +# define R500_RS_INST_COL_ID(x) ((x) << 12) #define R500_RS_INST_COL_CN_NO_WRITE (0 << 16) #define R500_RS_INST_COL_CN_WRITE (1 << 16) #define R500_RS_INST_COL_CN_WRITE_FBUFFER (2 << 16) #define R500_RS_INST_COL_CN_WRITE_BACKFACE (3 << 16) #define R500_RS_INST_COL_ADDR_SHIFT 18 +# define R500_RS_INST_COL_ADDR(x) ((x) << 18) #define R500_RS_INST_TEX_ADJ (1 << 25) #define R500_RS_INST_W_CN (1 << 26) @@ -1240,9 +1274,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define R300_RS_INST_7 0x434C # define R300_RS_INST_TEX_ID(x) ((x) << 0) # define R300_RS_INST_TEX_CN_WRITE (1 << 3) +# define R300_RS_INST_TEX_ADDR(x) ((x) << 6) # define R300_RS_INST_TEX_ADDR_SHIFT 6 # define R300_RS_INST_COL_ID(x) ((x) << 11) # define R300_RS_INST_COL_CN_WRITE (1 << 14) +# define R300_RS_INST_COL_ADDR(x) ((x) << 17) # define R300_RS_INST_COL_ADDR_SHIFT 17 # define R300_RS_INST_TEX_ADJ (1 << 22) # define R300_RS_COL_BIAS_UNUSED_SHIFT 23 @@ -1411,18 +1447,21 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R500_MACRO_SWITCH (1<<22) # define R500_BORDER_FIX (1<<31) -#define R300_TX_SIZE_0 0x4480 +#define R300_TX_FORMAT0_0 0x4480 # define R300_TX_WIDTHMASK_SHIFT 0 # define R300_TX_WIDTHMASK_MASK (2047 << 0) # define R300_TX_HEIGHTMASK_SHIFT 11 # define R300_TX_HEIGHTMASK_MASK (2047 << 11) -# define R300_TX_DEPTHMASK_SHIFT 22 -# define R300_TX_DEPTHMASK_MASK (0xf << 22) +# define R300_TX_DEPTHMASK_SHIFT 22 +# define R300_TX_DEPTHMASK_MASK (0xf << 22) # define R300_TX_MAX_MIP_LEVEL_SHIFT 26 # define R300_TX_MAX_MIP_LEVEL_MASK (0xf << 26) -# define R300_TX_SIZE_PROJECTED (1<<30) -# define R300_TX_SIZE_TXPITCH_EN (1<<31) -#define R300_TX_FORMAT_0 0x44C0 +# define R300_TX_SIZE_PROJECTED (1 << 30) +# define R300_TX_PITCH_EN (1 << 31) +# define R300_TX_WIDTH(x) ((x) << 0) +# define R300_TX_HEIGHT(x) ((x) << 11) + +#define R300_TX_FORMAT1_0 0x44C0 /* The interpretation of the format word by Wladimir van der Laan */ /* The X, Y, Z and W refer to the layout of the components. They are given meanings as R, G, B and Alpha by the swizzle @@ -1708,7 +1747,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_C3_SEL_R (1 << 14) # define R300_C3_SEL_G (2 << 14) # define R300_C3_SEL_B (3 << 14) -# define R300_OUT_SIGN(x) (x << 16) +# define R300_OUT_SIGN(x) ((x) << 16) +# define R500_ROUND_ADJ (1 << 20) /* ALU * The ALU instructions register blocks are enumerated according to the order @@ -1795,6 +1835,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ALU_DSTC_OUTPUT_X (1 << 26) # define R300_ALU_DSTC_OUTPUT_Y (1 << 27) # define R300_ALU_DSTC_OUTPUT_Z (1 << 28) +# define R300_ALU_DSTC_OUTPUT_XYZ (7 << 26) +# define R300_RGB_ADDR0(x) ((x) << 0) +# define R300_RGB_ADDR1(x) ((x) << 6) +# define R300_RGB_ADDR2(x) ((x) << 12) #define R300_US_ALU_ALPHA_ADDR_0 0x47C0 # define R300_ALU_SRC0A_SHIFT 0 @@ -1812,6 +1856,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ALU_DSTA_REG (1 << 23) # define R300_ALU_DSTA_OUTPUT (1 << 24) # define R300_ALU_DSTA_DEPTH (1 << 27) +# define R300_ALPHA_ADDR0(x) ((x) << 0) +# define R300_ALPHA_ADDR1(x) ((x) << 6) +# define R300_ALPHA_ADDR2(x) ((x) << 12) #define R300_US_ALU_RGB_INST_0 0x48C0 # define R300_ALU_ARGC_SRC0C_XYZ 0 @@ -1846,6 +1893,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ALU_ARGC_SRC0CA_WZY 29 # define R300_ALU_ARGC_SRC1CA_WZY 30 # define R300_ALU_ARGC_SRC2CA_WZY 31 +# define R300_RGB_SWIZA(x) ((x) << 0) +# define R300_RGB_SWIZB(x) ((x) << 7) +# define R300_RGB_SWIZC(x) ((x) << 14) # define R300_ALU_ARG0C_SHIFT 0 # define R300_ALU_ARG0C_MASK (31 << 0) @@ -1909,10 +1959,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ALU_ARGA_SRCP_Y 13 # define R300_ALU_ARGA_SRCP_Z 14 # define R300_ALU_ARGA_SRCP_W 15 - # define R300_ALU_ARGA_ZERO 16 # define R300_ALU_ARGA_ONE 17 # define R300_ALU_ARGA_HALF 18 +# define R300_ALPHA_SWIZA(x) ((x) << 0) +# define R300_ALPHA_SWIZB(x) ((x) << 7) +# define R300_ALPHA_SWIZC(x) ((x) << 14) + # define R300_ALU_ARG0A_SHIFT 0 # define R300_ALU_ARG0A_MASK (31 << 0) # define R300_ALU_ARG0A_NOP (0 << 5) @@ -2731,7 +2784,7 @@ enum { # define R500_ALPHA_OP_COS 13 # define R500_ALPHA_OP_MDH 14 # define R500_ALPHA_OP_MDV 15 -# define R500_ALPHA_ADDRD(x) (x << 4) +# define R500_ALPHA_ADDRD(x) ((x) << 4) # define R500_ALPHA_ADDRD_REL (1 << 11) # define R500_ALPHA_SEL_A_SHIFT 12 # define R500_ALPHA_SEL_A_SRC0 (0 << 12) @@ -2775,16 +2828,16 @@ enum { # define R500_ALPHA_OMOD_DIV_4 (5 << 26) # define R500_ALPHA_OMOD_DIV_8 (6 << 26) # define R500_ALPHA_OMOD_DISABLE (7 << 26) -# define R500_ALPHA_TARGET(x) (x << 29) +# define R500_ALPHA_TARGET(x) ((x) << 29) # define R500_ALPHA_W_OMASK (1 << 31) #define R500_US_ALU_ALPHA_ADDR_0 0x9800 -# define R500_ALPHA_ADDR0(x) (x << 0) +# define R500_ALPHA_ADDR0(x) ((x) << 0) # define R500_ALPHA_ADDR0_CONST (1 << 8) # define R500_ALPHA_ADDR0_REL (1 << 9) -# define R500_ALPHA_ADDR1(x) (x << 10) +# define R500_ALPHA_ADDR1(x) ((x) << 10) # define R500_ALPHA_ADDR1_CONST (1 << 18) # define R500_ALPHA_ADDR1_REL (1 << 19) -# define R500_ALPHA_ADDR2(x) (x << 20) +# define R500_ALPHA_ADDR2(x) ((x) << 20) # define R500_ALPHA_ADDR2_CONST (1 << 28) # define R500_ALPHA_ADDR2_REL (1 << 29) # define R500_ALPHA_SRCP_OP_1_MINUS_2A0 (0 << 30) @@ -2805,7 +2858,7 @@ enum { # define R500_ALU_RGBA_OP_SOP (10 << 0) # define R500_ALU_RGBA_OP_MDH (11 << 0) # define R500_ALU_RGBA_OP_MDV (12 << 0) -# define R500_ALU_RGBA_ADDRD(x) (x << 4) +# define R500_ALU_RGBA_ADDRD(x) ((x) << 4) # define R500_ALU_RGBA_ADDRD_REL (1 << 11) # define R500_ALU_RGBA_SEL_C_SHIFT 12 # define R500_ALU_RGBA_SEL_C_SRC0 (0 << 12) @@ -2932,16 +2985,16 @@ enum { # define R500_ALU_RGB_OMOD_DIV_4 (5 << 26) # define R500_ALU_RGB_OMOD_DIV_8 (6 << 26) # define R500_ALU_RGB_OMOD_DISABLE (7 << 26) -# define R500_ALU_RGB_TARGET(x) (x << 29) +# define R500_ALU_RGB_TARGET(x) ((x) << 29) # define R500_ALU_RGB_WMASK (1 << 31) #define R500_US_ALU_RGB_ADDR_0 0x9000 -# define R500_RGB_ADDR0(x) (x << 0) +# define R500_RGB_ADDR0(x) ((x) << 0) # define R500_RGB_ADDR0_CONST (1 << 8) # define R500_RGB_ADDR0_REL (1 << 9) -# define R500_RGB_ADDR1(x) (x << 10) +# define R500_RGB_ADDR1(x) ((x) << 10) # define R500_RGB_ADDR1_CONST (1 << 18) # define R500_RGB_ADDR1_REL (1 << 19) -# define R500_RGB_ADDR2(x) (x << 20) +# define R500_RGB_ADDR2(x) ((x) << 20) # define R500_RGB_ADDR2_CONST (1 << 28) # define R500_RGB_ADDR2_REL (1 << 29) # define R500_RGB_SRCP_OP_1_MINUS_2RGB0 (0 << 30) @@ -2973,6 +3026,7 @@ enum { # define R500_INST_RGB_OMASK_R (1 << 15) # define R500_INST_RGB_OMASK_G (1 << 16) # define R500_INST_RGB_OMASK_B (1 << 17) +# define R500_INST_RGB_OMASK_RGB (7 << 15) # define R500_INST_ALPHA_OMASK (1 << 18) # define R500_INST_RGB_CLAMP (1 << 19) # define R500_INST_ALPHA_CLAMP (1 << 20) @@ -2996,19 +3050,19 @@ enum { /* note that these are 8 bit lengths, despite the offsets, at least for R500 */ #define R500_US_CODE_ADDR 0x4630 -# define R500_US_CODE_START_ADDR(x) (x << 0) -# define R500_US_CODE_END_ADDR(x) (x << 16) +# define R500_US_CODE_START_ADDR(x) ((x) << 0) +# define R500_US_CODE_END_ADDR(x) ((x) << 16) #define R500_US_CODE_OFFSET 0x4638 -# define R500_US_CODE_OFFSET_ADDR(x) (x << 0) +# define R500_US_CODE_OFFSET_ADDR(x) ((x) << 0) #define R500_US_CODE_RANGE 0x4634 -# define R500_US_CODE_RANGE_ADDR(x) (x << 0) -# define R500_US_CODE_RANGE_SIZE(x) (x << 16) +# define R500_US_CODE_RANGE_ADDR(x) ((x) << 0) +# define R500_US_CODE_RANGE_SIZE(x) ((x) << 16) #define R500_US_CONFIG 0x4600 # define R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO (1 << 1) #define R500_US_FC_ADDR_0 0xa000 -# define R500_FC_BOOL_ADDR(x) (x << 0) -# define R500_FC_INT_ADDR(x) (x << 8) -# define R500_FC_JUMP_ADDR(x) (x << 16) +# define R500_FC_BOOL_ADDR(x) ((x) << 0) +# define R500_FC_INT_ADDR(x) ((x) << 8) +# define R500_FC_JUMP_ADDR(x) ((x) << 16) # define R500_FC_JUMP_GLOBAL (1 << 31) #define R500_US_FC_BOOL_CONST 0x4620 # define R500_FC_KBOOL(x) (x) @@ -3029,8 +3083,8 @@ enum { # define R500_FC_A_OP_NONE (0 << 6) # define R500_FC_A_OP_POP (1 << 6) # define R500_FC_A_OP_PUSH (2 << 6) -# define R500_FC_JUMP_FUNC(x) (x << 8) -# define R500_FC_B_POP_CNT(x) (x << 16) +# define R500_FC_JUMP_FUNC(x) ((x) << 8) +# define R500_FC_B_POP_CNT(x) ((x) << 16) # define R500_FC_B_OP0_NONE (0 << 24) # define R500_FC_B_OP0_DECR (1 << 24) # define R500_FC_B_OP0_INCR (2 << 24) @@ -3039,60 +3093,18 @@ enum { # define R500_FC_B_OP1_INCR (2 << 26) # define R500_FC_IGNORE_UNCOVERED (1 << 28) #define R500_US_FC_INT_CONST_0 0x4c00 -# define R500_FC_INT_CONST_KR(x) (x << 0) -# define R500_FC_INT_CONST_KG(x) (x << 8) -# define R500_FC_INT_CONST_KB(x) (x << 16) +# define R500_FC_INT_CONST_KR(x) ((x) << 0) +# define R500_FC_INT_CONST_KG(x) ((x) << 8) +# define R500_FC_INT_CONST_KB(x) ((x) << 16) /* _0 through _15 */ #define R500_US_FORMAT0_0 0x4640 -# define R500_FORMAT_TXWIDTH(x) (x << 0) -# define R500_FORMAT_TXHEIGHT(x) (x << 11) -# define R500_FORMAT_TXDEPTH(x) (x << 22) -/* _0 through _3 */ -#define R500_US_OUT_FMT_0 0x46A4 -# define R500_OUT_FMT_C4_8 (0 << 0) -# define R500_OUT_FMT_C4_10 (1 << 0) -# define R500_OUT_FMT_C4_10_GAMMA (2 << 0) -# define R500_OUT_FMT_C_16 (3 << 0) -# define R500_OUT_FMT_C2_16 (4 << 0) -# define R500_OUT_FMT_C4_16 (5 << 0) -# define R500_OUT_FMT_C_16_MPEG (6 << 0) -# define R500_OUT_FMT_C2_16_MPEG (7 << 0) -# define R500_OUT_FMT_C2_4 (8 << 0) -# define R500_OUT_FMT_C_3_3_2 (9 << 0) -# define R500_OUT_FMT_C_6_5_6 (10 << 0) -# define R500_OUT_FMT_C_11_11_10 (11 << 0) -# define R500_OUT_FMT_C_10_11_11 (12 << 0) -# define R500_OUT_FMT_C_2_10_10_10 (13 << 0) -/* #define R500_OUT_FMT_RESERVED (14 << 0) */ -# define R500_OUT_FMT_UNUSED (15 << 0) -# define R500_OUT_FMT_C_16_FP (16 << 0) -# define R500_OUT_FMT_C2_16_FP (17 << 0) -# define R500_OUT_FMT_C4_16_FP (18 << 0) -# define R500_OUT_FMT_C_32_FP (19 << 0) -# define R500_OUT_FMT_C2_32_FP (20 << 0) -# define R500_OUT_FMT_C4_32_FP (21 << 0) -# define R500_C0_SEL_A (0 << 8) -# define R500_C0_SEL_R (1 << 8) -# define R500_C0_SEL_G (2 << 8) -# define R500_C0_SEL_B (3 << 8) -# define R500_C1_SEL_A (0 << 10) -# define R500_C1_SEL_R (1 << 10) -# define R500_C1_SEL_G (2 << 10) -# define R500_C1_SEL_B (3 << 10) -# define R500_C2_SEL_A (0 << 12) -# define R500_C2_SEL_R (1 << 12) -# define R500_C2_SEL_G (2 << 12) -# define R500_C2_SEL_B (3 << 12) -# define R500_C3_SEL_A (0 << 14) -# define R500_C3_SEL_R (1 << 14) -# define R500_C3_SEL_G (2 << 14) -# define R500_C3_SEL_B (3 << 14) -# define R500_OUT_SIGN(x) (x << 16) -# define R500_ROUND_ADJ (1 << 20) +# define R500_FORMAT_TXWIDTH(x) ((x) << 0) +# define R500_FORMAT_TXHEIGHT(x) ((x) << 11) +# define R500_FORMAT_TXDEPTH(x) ((x) << 22) #define R500_US_PIXSIZE 0x4604 # define R500_PIX_SIZE(x) (x) #define R500_US_TEX_ADDR_0 0x9800 -# define R500_TEX_SRC_ADDR(x) (x << 0) +# define R500_TEX_SRC_ADDR(x) ((x) << 0) # define R500_TEX_SRC_ADDR_REL (1 << 7) # define R500_TEX_SRC_S_SWIZ_R (0 << 8) # define R500_TEX_SRC_S_SWIZ_G (1 << 8) @@ -3110,7 +3122,7 @@ enum { # define R500_TEX_SRC_Q_SWIZ_G (1 << 14) # define R500_TEX_SRC_Q_SWIZ_B (2 << 14) # define R500_TEX_SRC_Q_SWIZ_A (3 << 14) -# define R500_TEX_DST_ADDR(x) (x << 16) +# define R500_TEX_DST_ADDR(x) ((x) << 16) # define R500_TEX_DST_ADDR_REL (1 << 23) # define R500_TEX_DST_R_SWIZ_R (0 << 24) # define R500_TEX_DST_R_SWIZ_G (1 << 24) @@ -3129,7 +3141,7 @@ enum { # define R500_TEX_DST_A_SWIZ_B (2 << 30) # define R500_TEX_DST_A_SWIZ_A (3 << 30) #define R500_US_TEX_ADDR_DXDY_0 0xa000 -# define R500_DX_ADDR(x) (x << 0) +# define R500_DX_ADDR(x) ((x) << 0) # define R500_DX_ADDR_REL (1 << 7) # define R500_DX_S_SWIZ_R (0 << 8) # define R500_DX_S_SWIZ_G (1 << 8) @@ -3147,7 +3159,7 @@ enum { # define R500_DX_Q_SWIZ_G (1 << 14) # define R500_DX_Q_SWIZ_B (2 << 14) # define R500_DX_Q_SWIZ_A (3 << 14) -# define R500_DY_ADDR(x) (x << 16) +# define R500_DY_ADDR(x) ((x) << 16) # define R500_DY_ADDR_REL (1 << 17) # define R500_DY_S_SWIZ_R (0 << 24) # define R500_DY_S_SWIZ_G (1 << 24) @@ -3166,7 +3178,7 @@ enum { # define R500_DY_Q_SWIZ_B (2 << 30) # define R500_DY_Q_SWIZ_A (3 << 30) #define R500_US_TEX_INST_0 0x9000 -# define R500_TEX_ID(x) (x << 16) +# define R500_TEX_ID(x) ((x) << 16) # define R500_TEX_INST_NOP (0 << 22) # define R500_TEX_INST_LD (1 << 22) # define R500_TEX_INST_TEXKILL (2 << 22) @@ -3227,9 +3239,9 @@ enum { #define R300_PACKET3_3D_LOAD_VBPNTR 0x00002F00 #define R300_PACKET3_INDX_BUFFER 0x00003300 -# define R300_EB_UNK1_SHIFT 24 -# define R300_EB_UNK1 (0x80<<24) -# define R300_EB_UNK2 0x0810 +# define R300_INDX_BUFFER_DST_SHIFT 0 +# define R300_INDX_BUFFER_SKIP_SHIFT 16 +# define R300_INDX_BUFFER_ONE_REG_WR (1<<31) /* Same as R300_PACKET3_3D_DRAW_VBUF but without VAP_VTX_FMT */ #define R300_PACKET3_3D_DRAW_VBUF_2 0x00003400 diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index 8ed66a1660..d2c5998c26 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -50,6 +50,7 @@ static const char* chip_families[] = { "RC410", "RS480", "RS482", + "RS600", "RS690", "RS740", "RV515", @@ -100,11 +101,9 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) /* IN THEORY */ return 0; case PIPE_CAP_MAX_RENDER_TARGETS: - /* XXX 4 eventually */ - return 1; + return 4; case PIPE_CAP_OCCLUSION_QUERY: - /* IN THEORY */ - return 0; + return 1; case PIPE_CAP_TEXTURE_SHADOW_MAP: /* IN THEORY */ return 0; @@ -121,7 +120,7 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) * shows why this is silly. Assuming RGBA, 4cpp, we can see that * 4096*4096*4096 = 64.0 GiB exactly, so it's not exactly * practical. However, if at some point a game really wants this, - * then we can remove this limit. */ + * then we can remove or raise this limit. */ if (r300screen->caps->is_r500) { /* 9 == 256x256x256 */ return 9; @@ -142,7 +141,7 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) case PIPE_CAP_TEXTURE_MIRROR_REPEAT: return 1; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: - /* XXX guessing */ + /* XXX guessing (what a terrible guess) */ return 2; default: debug_printf("r300: Implementation error: Bad param %d\n", @@ -175,15 +174,44 @@ static float r300_get_paramf(struct pipe_screen* pscreen, int param) } } -/* XXX moar formats */ -static boolean check_tex_2d_format(enum pipe_format format) +static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500) { switch (format) { + /* Colorbuffer */ + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: case PIPE_FORMAT_A8R8G8B8_UNORM: + /* Colorbuffer or texture */ case PIPE_FORMAT_I8_UNORM: + /* Z buffer */ + case PIPE_FORMAT_Z16_UNORM: + /* Z buffer with stencil */ + case PIPE_FORMAT_Z24S8_UNORM: return TRUE; + + /* XXX These don't even exist + case PIPE_FORMAT_A32R32G32B32: + case PIPE_FORMAT_A16R16G16B16: */ + /* XXX Insert YUV422 packed VYUY and YVYU here */ + /* XXX What the deuce is UV88? (r3xx accel page 14) + debug_printf("r300: Warning: Got unimplemented format: %s in %s\n", + pf_name(format), __FUNCTION__); + return FALSE; */ + + /* XXX Supported yet unimplemented r5xx formats: */ + /* XXX Again, what is UV1010 this time? (r5xx accel page 148) */ + /* XXX Even more that don't exist + case PIPE_FORMAT_A10R10G10B10_UNORM: + case PIPE_FORMAT_A2R10G10B10_UNORM: + case PIPE_FORMAT_I10_UNORM: + debug_printf( + "r300: Warning: Got unimplemented r500 format: %s in %s\n", + pf_name(format), __FUNCTION__); + return FALSE; */ + default: - debug_printf("r300: Warning: Got unknown format: %s, in %s\n", + debug_printf("r300: Warning: Got unsupported format: %s in %s\n", pf_name(format), __FUNCTION__); break; } @@ -200,7 +228,8 @@ static boolean r300_is_format_supported(struct pipe_screen* pscreen, { switch (target) { case PIPE_TEXTURE_2D: - return check_tex_2d_format(format); + return check_tex_2d_format(format, + r300_screen(pscreen)->caps->is_r500); default: debug_printf("r300: Warning: Got unknown format target: %d\n", format); @@ -210,24 +239,84 @@ static boolean r300_is_format_supported(struct pipe_screen* pscreen, return FALSE; } -static void* r300_surface_map(struct pipe_screen* screen, - struct pipe_surface* surface, - unsigned flags) +static struct pipe_transfer* +r300_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, unsigned x, unsigned y, + unsigned w, unsigned h) +{ + struct r300_texture *tex = (struct r300_texture *)texture; + struct r300_transfer *trans; + unsigned offset; /* in bytes */ + + /* XXX Add support for these things */ + if (texture->target == PIPE_TEXTURE_CUBE) { + debug_printf("PIPE_TEXTURE_CUBE is not yet supported.\n"); + /* offset = tex->image_offset[level][face]; */ + } + else if (texture->target == PIPE_TEXTURE_3D) { + debug_printf("PIPE_TEXTURE_3D is not yet supported.\n"); + /* offset = tex->image_offset[level][zslice]; */ + } + else { + offset = tex->offset[level]; + assert(face == 0); + assert(zslice == 0); + } + + trans = CALLOC_STRUCT(r300_transfer); + if (trans) { + pipe_texture_reference(&trans->transfer.texture, texture); + trans->transfer.format = trans->transfer.format; + trans->transfer.width = w; + trans->transfer.height = h; + trans->transfer.block = texture->block; + trans->transfer.nblocksx = texture->nblocksx[level]; + trans->transfer.nblocksy = texture->nblocksy[level]; + trans->transfer.stride = tex->stride; + trans->transfer.usage = usage; + trans->offset = offset; + } + return &trans->transfer; +} + +static void +r300_tex_transfer_destroy(struct pipe_transfer *trans) { - struct r300_texture* tex = (struct r300_texture*)surface->texture; - char* map = pipe_buffer_map(screen, tex->buffer, flags); + pipe_texture_reference(&trans->texture, NULL); + FREE(trans); +} + +static void* r300_transfer_map(struct pipe_screen* screen, + struct pipe_transfer* transfer) +{ + struct r300_texture* tex = (struct r300_texture*)transfer->texture; + char* map; + unsigned flags = 0; + + if (transfer->usage != PIPE_TRANSFER_WRITE) { + flags |= PIPE_BUFFER_USAGE_CPU_READ; + } + if (transfer->usage != PIPE_TRANSFER_READ) { + flags |= PIPE_BUFFER_USAGE_CPU_WRITE; + } + + map = pipe_buffer_map(screen, tex->buffer, flags); if (!map) { return NULL; } - return map + surface->offset; + return map + r300_transfer(transfer)->offset + + transfer->y / transfer->block.height * transfer->stride + + transfer->x / transfer->block.width * transfer->block.size; } -static void r300_surface_unmap(struct pipe_screen* screen, - struct pipe_surface* surface) +static void r300_transfer_unmap(struct pipe_screen* screen, + struct pipe_transfer* transfer) { - struct r300_texture* tex = (struct r300_texture*)surface->texture; + struct r300_texture* tex = (struct r300_texture*)transfer->texture; pipe_buffer_unmap(screen, tex->buffer); } @@ -239,8 +328,7 @@ static void r300_destroy_screen(struct pipe_screen* pscreen) FREE(r300screen); } -struct pipe_screen* r300_create_screen(struct pipe_winsys* winsys, - struct r300_winsys* r300_winsys) +struct pipe_screen* r300_create_screen(struct r300_winsys* r300_winsys) { struct r300_screen* r300screen = CALLOC_STRUCT(r300_screen); struct r300_capabilities* caps = CALLOC_STRUCT(r300_capabilities); @@ -254,15 +342,17 @@ struct pipe_screen* r300_create_screen(struct pipe_winsys* winsys, r300_parse_chipset(caps); r300screen->caps = caps; - r300screen->screen.winsys = winsys; + r300screen->screen.winsys = (struct pipe_winsys*)r300_winsys; r300screen->screen.destroy = r300_destroy_screen; r300screen->screen.get_name = r300_get_name; r300screen->screen.get_vendor = r300_get_vendor; r300screen->screen.get_param = r300_get_param; r300screen->screen.get_paramf = r300_get_paramf; r300screen->screen.is_format_supported = r300_is_format_supported; - r300screen->screen.surface_map = r300_surface_map; - r300screen->screen.surface_unmap = r300_surface_unmap; + r300screen->screen.get_tex_transfer = r300_get_tex_transfer; + r300screen->screen.tex_transfer_destroy = r300_tex_transfer_destroy; + r300screen->screen.transfer_map = r300_transfer_map; + r300screen->screen.transfer_unmap = r300_transfer_unmap; r300_init_screen_texture_functions(&r300screen->screen); u_simple_screen_init(&r300screen->screen); diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h index 2e25f61dbf..3f52dbc3be 100644 --- a/src/gallium/drivers/r300/r300_screen.h +++ b/src/gallium/drivers/r300/r300_screen.h @@ -40,13 +40,27 @@ struct r300_screen { struct r300_capabilities* caps; }; +struct r300_transfer { + /* Parent class */ + struct pipe_transfer transfer; + + /* Offset from start of buffer. */ + unsigned offset; +}; + /* Convenience cast wrapper. */ static struct r300_screen* r300_screen(struct pipe_screen* screen) { return (struct r300_screen*)screen; } +/* Convenience cast wrapper. */ +static INLINE struct r300_transfer* +r300_transfer(struct pipe_transfer* transfer) +{ + return (struct r300_transfer*)transfer; +} + /* Creates a new r300 screen. */ -struct pipe_screen* r300_create_screen(struct pipe_winsys* winsys, - struct r300_winsys* r300_winsys); +struct pipe_screen* r300_create_screen(struct r300_winsys* r300_winsys); #endif /* R300_SCREEN_H */ diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 9392d72342..58bce22fc8 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -22,92 +22,17 @@ #include "util/u_math.h" #include "util/u_pack_color.h" -#include "pipe/p_debug.h" + +#include "util/u_debug.h" +#include "pipe/internal/p_winsys_screen.h" #include "r300_context.h" #include "r300_reg.h" +#include "r300_state_inlines.h" +#include "r300_state_shader.h" /* r300_state: Functions used to intialize state context by translating - * Gallium state objects into semi-native r300 state objects. - * - * XXX break this file up into pieces if it gets too big! */ - -/* Pack a float into a dword. */ -static uint32_t pack_float_32(float f) -{ - union { - float f; - uint32_t u; - } u; - - u.f = f; - return u.u; -} - -static uint32_t translate_blend_function(int blend_func) { - switch (blend_func) { - case PIPE_BLEND_ADD: - return R300_COMB_FCN_ADD_CLAMP; - case PIPE_BLEND_SUBTRACT: - return R300_COMB_FCN_SUB_CLAMP; - case PIPE_BLEND_REVERSE_SUBTRACT: - return R300_COMB_FCN_RSUB_CLAMP; - case PIPE_BLEND_MIN: - return R300_COMB_FCN_MIN; - case PIPE_BLEND_MAX: - return R300_COMB_FCN_MAX; - default: - debug_printf("r300: Unknown blend function %d\n", blend_func); - break; - } - return 0; -} - -/* XXX we can also offer the D3D versions of some of these... */ -static uint32_t translate_blend_factor(int blend_fact) { - switch (blend_fact) { - case PIPE_BLENDFACTOR_ONE: - return R300_BLEND_GL_ONE; - case PIPE_BLENDFACTOR_SRC_COLOR: - return R300_BLEND_GL_SRC_COLOR; - case PIPE_BLENDFACTOR_SRC_ALPHA: - return R300_BLEND_GL_SRC_ALPHA; - case PIPE_BLENDFACTOR_DST_ALPHA: - return R300_BLEND_GL_DST_ALPHA; - case PIPE_BLENDFACTOR_DST_COLOR: - return R300_BLEND_GL_DST_COLOR; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - return R300_BLEND_GL_SRC_ALPHA_SATURATE; - case PIPE_BLENDFACTOR_CONST_COLOR: - return R300_BLEND_GL_CONST_COLOR; - case PIPE_BLENDFACTOR_CONST_ALPHA: - return R300_BLEND_GL_CONST_ALPHA; - /* XXX WTF are these? - case PIPE_BLENDFACTOR_SRC1_COLOR: - case PIPE_BLENDFACTOR_SRC1_ALPHA: */ - case PIPE_BLENDFACTOR_ZERO: - return R300_BLEND_GL_ZERO; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - return R300_BLEND_GL_ONE_MINUS_SRC_COLOR; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - return R300_BLEND_GL_ONE_MINUS_SRC_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - return R300_BLEND_GL_ONE_MINUS_DST_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - return R300_BLEND_GL_ONE_MINUS_DST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - return R300_BLEND_GL_ONE_MINUS_CONST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - return R300_BLEND_GL_ONE_MINUS_CONST_ALPHA; - /* XXX see above - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: */ - default: - debug_printf("r300: Unknown blend factor %d\n", blend_fact); - break; - } - return 0; -} + * Gallium state objects into semi-native r300 state objects. */ /* Create a new blend state based on the CSO blend state. * @@ -123,16 +48,16 @@ static void* r300_create_blend_state(struct pipe_context* pipe, blend->blend_control = R300_ALPHA_BLEND_ENABLE | R300_SEPARATE_ALPHA_ENABLE | R300_READ_ENABLE | - translate_blend_function(state->rgb_func) | - (translate_blend_factor(state->rgb_src_factor) << + r300_translate_blend_function(state->rgb_func) | + (r300_translate_blend_factor(state->rgb_src_factor) << R300_SRC_BLEND_SHIFT) | - (translate_blend_factor(state->rgb_dst_factor) << + (r300_translate_blend_factor(state->rgb_dst_factor) << R300_DST_BLEND_SHIFT); blend->alpha_blend_control = - translate_blend_function(state->alpha_func) | - (translate_blend_factor(state->alpha_src_factor) << + r300_translate_blend_function(state->alpha_func) | + (r300_translate_blend_factor(state->alpha_src_factor) << R300_SRC_BLEND_SHIFT) | - (translate_blend_factor(state->alpha_dst_factor) << + (r300_translate_blend_factor(state->alpha_dst_factor) << R300_DST_BLEND_SHIFT); } @@ -175,21 +100,17 @@ static void r300_set_blend_color(struct pipe_context* pipe, const struct pipe_blend_color* color) { struct r300_context* r300 = r300_context(pipe); - uint32_t r, g, b, a; ubyte ur, ug, ub, ua; - r = util_iround(color->color[0] * 1023.0f); - g = util_iround(color->color[1] * 1023.0f); - b = util_iround(color->color[2] * 1023.0f); - a = util_iround(color->color[3] * 1023.0f); - ur = float_to_ubyte(color->color[0]); ug = float_to_ubyte(color->color[1]); ub = float_to_ubyte(color->color[2]); ua = float_to_ubyte(color->color[3]); - r300->blend_color_state->blend_color = (a << 24) | (r << 16) | (g << 8) | b; + util_pack_color(color->color, PIPE_FORMAT_A8R8G8B8_UNORM, + &r300->blend_color_state->blend_color); + /* XXX this is wrong */ r300->blend_color_state->blend_color_red_alpha = ur | (ua << 16); r300->blend_color_state->blend_color_green_blue = ub | (ug << 16); @@ -210,83 +131,24 @@ static void uint shader, uint index, const struct pipe_constant_buffer* buffer) { - /* XXX */ -} - -static uint32_t translate_depth_stencil_function(int zs_func) { - switch (zs_func) { - case PIPE_FUNC_NEVER: - return R300_ZS_NEVER; - case PIPE_FUNC_LESS: - return R300_ZS_LESS; - case PIPE_FUNC_EQUAL: - return R300_ZS_EQUAL; - case PIPE_FUNC_LEQUAL: - return R300_ZS_LEQUAL; - case PIPE_FUNC_GREATER: - return R300_ZS_GREATER; - case PIPE_FUNC_NOTEQUAL: - return R300_ZS_NOTEQUAL; - case PIPE_FUNC_GEQUAL: - return R300_ZS_GEQUAL; - case PIPE_FUNC_ALWAYS: - return R300_ZS_ALWAYS; - default: - debug_printf("r300: Unknown depth/stencil function %d\n", - zs_func); - break; - } - return 0; -} + struct r300_context* r300 = r300_context(pipe); -static uint32_t translate_stencil_op(int s_op) { - switch (s_op) { - case PIPE_STENCIL_OP_KEEP: - return R300_ZS_KEEP; - case PIPE_STENCIL_OP_ZERO: - return R300_ZS_ZERO; - case PIPE_STENCIL_OP_REPLACE: - return R300_ZS_REPLACE; - case PIPE_STENCIL_OP_INCR: - return R300_ZS_INCR; - case PIPE_STENCIL_OP_DECR: - return R300_ZS_DECR; - case PIPE_STENCIL_OP_INCR_WRAP: - return R300_ZS_INCR_WRAP; - case PIPE_STENCIL_OP_DECR_WRAP: - return R300_ZS_DECR_WRAP; - case PIPE_STENCIL_OP_INVERT: - return R300_ZS_INVERT; - default: - debug_printf("r300: Unknown stencil op %d", s_op); - break; + /* This entire chunk of code seems ever-so-slightly baked. + * It's as if I've got pipe_buffer* matryoshkas... */ + if (buffer && buffer->buffer && buffer->buffer->size) { + void* map = pipe->winsys->buffer_map(pipe->winsys, buffer->buffer, + PIPE_BUFFER_USAGE_CPU_READ); + memcpy(r300->shader_constants[shader].constants, + map, buffer->buffer->size); + pipe->winsys->buffer_unmap(pipe->winsys, buffer->buffer); + + r300->shader_constants[shader].user_count = + buffer->buffer->size / (sizeof(float) * 4); + } else { + r300->shader_constants[shader].user_count = 0; } - return 0; -} -static uint32_t translate_alpha_function(int alpha_func) { - switch (alpha_func) { - case PIPE_FUNC_NEVER: - return R300_FG_ALPHA_FUNC_NEVER; - case PIPE_FUNC_LESS: - return R300_FG_ALPHA_FUNC_LESS; - case PIPE_FUNC_EQUAL: - return R300_FG_ALPHA_FUNC_EQUAL; - case PIPE_FUNC_LEQUAL: - return R300_FG_ALPHA_FUNC_LE; - case PIPE_FUNC_GREATER: - return R300_FG_ALPHA_FUNC_GREATER; - case PIPE_FUNC_NOTEQUAL: - return R300_FG_ALPHA_FUNC_NOTEQUAL; - case PIPE_FUNC_GEQUAL: - return R300_FG_ALPHA_FUNC_GE; - case PIPE_FUNC_ALWAYS: - return R300_FG_ALPHA_FUNC_ALWAYS; - default: - debug_printf("r300: Unknown alpha function %d", alpha_func); - break; - } - return 0; + r300->dirty_state |= R300_NEW_CONSTANTS; } /* Create a new depth, stencil, and alpha state based on the CSO dsa state. @@ -309,7 +171,7 @@ static void* } dsa->z_stencil_control |= - (translate_depth_stencil_function(state->depth.func) << + (r300_translate_depth_stencil_function(state->depth.func) << R300_Z_FUNC_SHIFT); } @@ -317,14 +179,14 @@ static void* if (state->stencil[0].enabled) { dsa->z_buffer_control |= R300_STENCIL_ENABLE; dsa->z_stencil_control |= - (translate_depth_stencil_function(state->stencil[0].func) << - R300_S_FRONT_FUNC_SHIFT) | - (translate_stencil_op(state->stencil[0].fail_op) << - R300_S_FRONT_SFAIL_OP_SHIFT) | - (translate_stencil_op(state->stencil[0].zpass_op) << - R300_S_FRONT_ZPASS_OP_SHIFT) | - (translate_stencil_op(state->stencil[0].zfail_op) << - R300_S_FRONT_ZFAIL_OP_SHIFT); + (r300_translate_depth_stencil_function(state->stencil[0].func) << + R300_S_FRONT_FUNC_SHIFT) | + (r300_translate_stencil_op(state->stencil[0].fail_op) << + R300_S_FRONT_SFAIL_OP_SHIFT) | + (r300_translate_stencil_op(state->stencil[0].zpass_op) << + R300_S_FRONT_ZPASS_OP_SHIFT) | + (r300_translate_stencil_op(state->stencil[0].zfail_op) << + R300_S_FRONT_ZFAIL_OP_SHIFT); dsa->stencil_ref_mask = (state->stencil[0].ref_value) | (state->stencil[0].valuemask << R300_STENCILMASK_SHIFT) | @@ -333,14 +195,14 @@ static void* if (state->stencil[1].enabled) { dsa->z_buffer_control |= R300_STENCIL_FRONT_BACK; dsa->z_stencil_control |= - (translate_depth_stencil_function(state->stencil[1].func) << - R300_S_BACK_FUNC_SHIFT) | - (translate_stencil_op(state->stencil[1].fail_op) << - R300_S_BACK_SFAIL_OP_SHIFT) | - (translate_stencil_op(state->stencil[1].zpass_op) << - R300_S_BACK_ZPASS_OP_SHIFT) | - (translate_stencil_op(state->stencil[1].zfail_op) << - R300_S_BACK_ZFAIL_OP_SHIFT); + (r300_translate_depth_stencil_function(state->stencil[1].func) << + R300_S_BACK_FUNC_SHIFT) | + (r300_translate_stencil_op(state->stencil[1].fail_op) << + R300_S_BACK_SFAIL_OP_SHIFT) | + (r300_translate_stencil_op(state->stencil[1].zpass_op) << + R300_S_BACK_ZPASS_OP_SHIFT) | + (r300_translate_stencil_op(state->stencil[1].zfail_op) << + R300_S_BACK_ZFAIL_OP_SHIFT); dsa->stencil_ref_bf = (state->stencil[1].ref_value) | (state->stencil[1].valuemask << R300_STENCILMASK_SHIFT) | @@ -350,7 +212,8 @@ static void* /* Alpha test setup. */ if (state->alpha.enabled) { - dsa->alpha_function = translate_alpha_function(state->alpha.func) | + dsa->alpha_function = + r300_translate_alpha_function(state->alpha.func) | R300_FG_ALPHA_FUNC_ENABLE; dsa->alpha_reference = CLAMP(state->alpha.ref_value * 1023.0f, 0, 1023); @@ -415,6 +278,8 @@ static void* r300_create_fs_state(struct pipe_context* pipe, /* Copy state directly into shader. */ fs->state = *shader; + tgsi_scan_shader(shader->tokens, &fs->info); + return (void*)fs; } @@ -424,14 +289,18 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) struct r300_context* r300 = r300_context(pipe); struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader; - if (!fs->translated) { + if (fs == NULL) { + r300->fs = NULL; + return; + } else if (!fs->translated) { if (r300_screen(r300->context.screen)->caps->is_r500) { - r500_translate_shader(r300, fs); + r500_translate_fragment_shader(r300, (struct r500_fragment_shader*)fs); } else { - r300_translate_shader(r300, fs); + r300_translate_fragment_shader(r300, (struct r300_fragment_shader*)fs); } } + fs->translated = TRUE; r300->fs = fs; r300->dirty_state |= R300_NEW_FRAGMENT_SHADER; @@ -449,10 +318,6 @@ static void r300_set_polygon_stipple(struct pipe_context* pipe, /* XXX */ } -static INLINE int pack_float_16_6x(float f) { - return ((int)(f * 6.0) & 0xffff); -} - /* Create a new rasterizer state based on the CSO rasterizer state. * * This is a very large chunk of state, and covers most of the graphics @@ -472,6 +337,12 @@ static void* r300_create_rs_state(struct pipe_context* pipe, rs->point_size = pack_float_16_6x(state->point_size) | (pack_float_16_6x(state->point_size) << R300_POINTSIZE_X_SHIFT); + rs->point_minmax = + ((int)(state->point_size_min * 6.0) << + R300_GA_POINT_MINMAX_MIN_SHIFT) | + ((int)(state->point_size_max * 6.0) << + R300_GA_POINT_MINMAX_MAX_SHIFT); + rs->line_control = pack_float_16_6x(state->line_width) | R300_GA_LINE_CNTL_END_TYPE_COMP; @@ -504,20 +375,28 @@ static void* r300_create_rs_state(struct pipe_context* pipe, if (rs->polygon_offset_enable) { rs->depth_offset_front = rs->depth_offset_back = - pack_float_32(state->offset_units); + fui(state->offset_units); rs->depth_scale_front = rs->depth_scale_back = - pack_float_32(state->offset_scale); + fui(state->offset_scale); } if (state->line_stipple_enable) { rs->line_stipple_config = R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_LINE | - (pack_float_32((float)state->line_stipple_factor) & + (fui((float)state->line_stipple_factor) & R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_MASK); /* XXX this might need to be scaled up */ rs->line_stipple_value = state->line_stipple_pattern; } + if (state->flatshade) { + rs->color_control = R300_SHADE_MODEL_FLAT; + } else { + rs->color_control = R300_SHADE_MODEL_SMOOTH; + } + + rs->rs = *state; + return (void*)rs; } @@ -525,8 +404,11 @@ static void* r300_create_rs_state(struct pipe_context* pipe, static void r300_bind_rs_state(struct pipe_context* pipe, void* state) { struct r300_context* r300 = r300_context(pipe); + struct r300_rs_state* rs = (struct r300_rs_state*)state; + + draw_set_rasterizer_state(r300->draw, &rs->rs); - r300->rs_state = (struct r300_rs_state*)state; + r300->rs_state = rs; r300->dirty_state |= R300_NEW_RASTERIZER; } @@ -536,83 +418,6 @@ static void r300_delete_rs_state(struct pipe_context* pipe, void* state) FREE(state); } -static uint32_t translate_wrap(int wrap) { - switch (wrap) { - case PIPE_TEX_WRAP_REPEAT: - return R300_TX_REPEAT; - case PIPE_TEX_WRAP_CLAMP: - return R300_TX_CLAMP; - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - return R300_TX_CLAMP_TO_EDGE; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return R300_TX_CLAMP_TO_BORDER; - case PIPE_TEX_WRAP_MIRROR_REPEAT: - return R300_TX_REPEAT | R300_TX_MIRRORED; - case PIPE_TEX_WRAP_MIRROR_CLAMP: - return R300_TX_CLAMP | R300_TX_MIRRORED; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: - return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED; - default: - debug_printf("r300: Unknown texture wrap %d", wrap); - return 0; - } -} - -static uint32_t translate_tex_filters(int min, int mag, int mip) { - uint32_t retval = 0; - switch (min) { - case PIPE_TEX_FILTER_NEAREST: - retval |= R300_TX_MIN_FILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: - retval |= R300_TX_MIN_FILTER_LINEAR; - case PIPE_TEX_FILTER_ANISO: - retval |= R300_TX_MIN_FILTER_ANISO; - default: - debug_printf("r300: Unknown texture filter %d", min); - break; - } - switch (mag) { - case PIPE_TEX_FILTER_NEAREST: - retval |= R300_TX_MAG_FILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: - retval |= R300_TX_MAG_FILTER_LINEAR; - case PIPE_TEX_FILTER_ANISO: - retval |= R300_TX_MAG_FILTER_ANISO; - default: - debug_printf("r300: Unknown texture filter %d", mag); - break; - } - switch (mip) { - case PIPE_TEX_MIPFILTER_NONE: - retval |= R300_TX_MIN_FILTER_MIP_NONE; - case PIPE_TEX_MIPFILTER_NEAREST: - retval |= R300_TX_MIN_FILTER_MIP_NEAREST; - case PIPE_TEX_MIPFILTER_LINEAR: - retval |= R300_TX_MIN_FILTER_MIP_LINEAR; - default: - debug_printf("r300: Unknown texture filter %d", mip); - break; - } - - return retval; -} - -static uint32_t anisotropy(float max_aniso) { - if (max_aniso >= 16.0f) { - return R300_TX_MAX_ANISO_16_TO_1; - } else if (max_aniso >= 8.0f) { - return R300_TX_MAX_ANISO_8_TO_1; - } else if (max_aniso >= 4.0f) { - return R300_TX_MAX_ANISO_4_TO_1; - } else if (max_aniso >= 2.0f) { - return R300_TX_MAX_ANISO_2_TO_1; - } else { - return R300_TX_MAX_ANISO_1_TO_1; - } -} - static void* r300_create_sampler_state(struct pipe_context* pipe, const struct pipe_sampler_state* state) @@ -622,19 +427,19 @@ static void* int lod_bias; sampler->filter0 |= - (translate_wrap(state->wrap_s) << R300_TX_WRAP_S_SHIFT) | - (translate_wrap(state->wrap_t) << R300_TX_WRAP_T_SHIFT) | - (translate_wrap(state->wrap_r) << R300_TX_WRAP_R_SHIFT); + (r300_translate_wrap(state->wrap_s) << R300_TX_WRAP_S_SHIFT) | + (r300_translate_wrap(state->wrap_t) << R300_TX_WRAP_T_SHIFT) | + (r300_translate_wrap(state->wrap_r) << R300_TX_WRAP_R_SHIFT); - sampler->filter0 |= translate_tex_filters(state->min_img_filter, - state->mag_img_filter, - state->min_mip_filter); + sampler->filter0 |= r300_translate_tex_filters(state->min_img_filter, + state->mag_img_filter, + state->min_mip_filter); lod_bias = CLAMP((int)(state->lod_bias * 32), -(1 << 9), (1 << 9) - 1); sampler->filter1 |= lod_bias << R300_LOD_BIAS_SHIFT; - sampler->filter1 |= anisotropy(state->max_anisotropy); + sampler->filter1 |= r300_anisotropy(state->max_anisotropy); util_pack_color(state->border_color, PIPE_FORMAT_A8R8G8B8_UNORM, &sampler->border_color); @@ -710,20 +515,12 @@ static void r300_set_scissor_state(struct pipe_context* pipe, struct r300_context* r300 = r300_context(pipe); draw_flush(r300->draw); - uint32_t left, top, right, bottom; - - /* So, a bit of info. The scissors are offset by R300_SCISSORS_OFFSET in - * both directions for all values, and can only be 13 bits wide. Why? - * We may never know. */ - left = (state->minx + R300_SCISSORS_OFFSET) & 0x1fff; - top = (state->miny + R300_SCISSORS_OFFSET) & 0x1fff; - right = (state->maxx + R300_SCISSORS_OFFSET) & 0x1fff; - bottom = (state->maxy + R300_SCISSORS_OFFSET) & 0x1fff; - - r300->scissor_state->scissor_top_left = (left << R300_SCISSORS_X_SHIFT) | - (top << R300_SCISSORS_Y_SHIFT); + r300->scissor_state->scissor_top_left = + (state->minx << R300_SCISSORS_X_SHIFT) | + (state->miny << R300_SCISSORS_Y_SHIFT); r300->scissor_state->scissor_bottom_right = - (right << R300_SCISSORS_X_SHIFT) | (bottom << R300_SCISSORS_Y_SHIFT); + (state->maxx << R300_SCISSORS_X_SHIFT) | + (state->maxy << R300_SCISSORS_Y_SHIFT); r300->dirty_state |= R300_NEW_SCISSOR; } @@ -732,8 +529,26 @@ static void r300_set_viewport_state(struct pipe_context* pipe, const struct pipe_viewport_state* state) { struct r300_context* r300 = r300_context(pipe); - /* XXX handing this off to Draw for now */ - draw_set_viewport_state(r300->draw, state); + + r300->viewport_state->xscale = state->scale[0]; + r300->viewport_state->yscale = state->scale[1]; + r300->viewport_state->zscale = state->scale[2]; + + r300->viewport_state->xoffset = state->translate[0]; + r300->viewport_state->yoffset = state->translate[1]; + r300->viewport_state->zoffset = state->translate[2]; + + r300->viewport_state->vte_control = 0; + if (r300_screen(r300->context.screen)->caps->has_tcl) { + /* Do the transform in HW. */ + r300->viewport_state->vte_control |= + R300_VPORT_X_SCALE_ENA | R300_VPORT_X_OFFSET_ENA | + R300_VPORT_Y_SCALE_ENA | R300_VPORT_Y_OFFSET_ENA | + R300_VPORT_Z_SCALE_ENA | R300_VPORT_Z_OFFSET_ENA; + } else { + /* Have Draw do the actual transform. */ + draw_set_viewport_state(r300->draw, state); + } } static void r300_set_vertex_buffers(struct pipe_context* pipe, @@ -741,7 +556,12 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe, const struct pipe_vertex_buffer* buffers) { struct r300_context* r300 = r300_context(pipe); - /* XXX Draw */ + + memcpy(r300->vertex_buffers, buffers, + sizeof(struct pipe_vertex_buffer) * count); + + r300->vertex_buffer_count = count; + draw_flush(r300->draw); draw_set_vertex_buffers(r300->draw, count, buffers); } diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c new file mode 100644 index 0000000000..d761a0302f --- /dev/null +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -0,0 +1,299 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_state_derived.h" + +/* r300_state_derived: Various bits of state which are dependent upon + * currently bound CSO data. */ + +/* Update the vertex_info struct in our r300_context. + * + * The vertex_info struct describes the post-TCL format of vertices. It is + * required for Draw when doing SW TCL, and also for describing the + * dreaded RS block on R300 chipsets. */ +/* XXX this function should be able to handle vert shaders as well as draw */ +static void r300_update_vertex_layout(struct r300_context* r300) +{ + struct r300_vertex_format vformat; + struct vertex_info vinfo; + boolean pos = FALSE, psize = FALSE, fog = FALSE; + int i, texs = 0, cols = 0; + int tab[16]; + + struct tgsi_shader_info* info = &r300->fs->info; + + memset(&vinfo, 0, sizeof(vinfo)); + for (i = 0; i < 16; i++) { + tab[i] = -1; + } + + assert(info->num_inputs <= 16); + + for (i = 0; i < info->num_inputs; i++) { + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + pos = TRUE; + tab[i] = 0; + break; + case TGSI_SEMANTIC_COLOR: + tab[i] = 2 + cols++; + break; + case TGSI_SEMANTIC_PSIZE: + psize = TRUE; + tab[i] = 1; + break; + case TGSI_SEMANTIC_FOG: + fog = TRUE; + /* Fall through... */ + case TGSI_SEMANTIC_GENERIC: + tab[i] = 6 + texs++; + break; + default: + debug_printf("r300: Unknown vertex input %d\n", + info->input_semantic_name[i]); + break; + } + } + + /* Do the actual vertex_info setup. + * + * vertex_info has four uints of hardware-specific data in it. + * vinfo.hwfmt[0] is R300_VAP_VTX_STATE_CNTL + * vinfo.hwfmt[1] is R300_VAP_VSM_VTX_ASSM + * vinfo.hwfmt[2] is R300_VAP_OUTPUT_VTX_FMT_0 + * vinfo.hwfmt[3] is R300_VAP_OUTPUT_VTX_FMT_1 */ + + vinfo.hwfmt[0] = 0x5555; /* XXX this is classic Mesa bonghits */ + + if (!pos) { + debug_printf("r300: Forcing vertex position attribute emit...\n"); + /* Make room for the position attribute + * at the beginning of the tab. */ + for (i = 15; i > 0; i--) { + tab[i] = tab[i-1]; + } + tab[0] = 0; + + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_POS, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_POSITION, 0)); + } else { + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_POSITION, 0)); + } + vinfo.hwfmt[1] |= R300_INPUT_CNTL_POS; + vinfo.hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT; + + if (psize) { + draw_emit_vertex_attr(&vinfo, EMIT_1F_PSIZE, INTERP_POS, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_PSIZE, 0)); + vinfo.hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; + } + + for (i = 0; i < cols; i++) { + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_COLOR, i)); + vinfo.hwfmt[1] |= R300_INPUT_CNTL_COLOR; + vinfo.hwfmt[2] |= (R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i); + } + + for (i = 0; i < texs; i++) { + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i)); + vinfo.hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); + vinfo.hwfmt[3] |= (4 << (3 * i)); + } + + if (fog) { + i++; + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0)); + vinfo.hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); + vinfo.hwfmt[3] |= (4 << (3 * i)); + } + + draw_compute_vertex_size(&vinfo); + + if (memcmp(&r300->vertex_info, &vinfo, sizeof(struct vertex_info))) { + uint32_t temp; + debug_printf("attrib count: %d, fp input count: %d\n", + vinfo.num_attribs, info->num_inputs); + for (i = 0; i < vinfo.num_attribs; i++) { + debug_printf("attrib: offset %d, interp %d, size %d," + " tab %d\n", vinfo.attrib[i].src_index, + vinfo.attrib[i].interp_mode, vinfo.attrib[i].emit, + tab[i]); + } + + for (i = 0; i < vinfo.num_attribs; i++) { + /* Make sure we have a proper destination for our attribute */ + assert(tab[i] != -1); + + temp = translate_vertex_data_type(vinfo.attrib[i].emit) | + (tab[i] << R300_DST_VEC_LOC_SHIFT); + if (i & 1) { + r300->vertex_info.vap_prog_stream_cntl[i >> 1] &= 0x0000ffff; + r300->vertex_info.vap_prog_stream_cntl[i >> 1] |= temp << 16; + } else { + r300->vertex_info.vap_prog_stream_cntl[i >> 1] &= 0xffff0000; + r300->vertex_info.vap_prog_stream_cntl[i >> 1] |= temp; + } + + r300->vertex_info.vap_prog_stream_cntl_ext[i >> 1] |= + (R300_VAP_SWIZZLE_XYZW << (i & 1 ? 16 : 0)); + } + /* Set the last vector. */ + i--; + r300->vertex_info.vap_prog_stream_cntl[i >> 1] |= (R300_LAST_VEC << + (i & 1 ? 16 : 0)); + + memcpy(r300->vertex_info.tab, tab, sizeof(tab)); + memcpy(&r300->vertex_info, &vinfo, sizeof(struct vertex_info)); + r300->dirty_state |= R300_NEW_VERTEX_FORMAT; + } +} + +/* Set up the RS block. This is the part of the chipset that actually does + * the rasterization of vertices into fragments. This is also the part of the + * chipset that locks up if any part of it is even slightly wrong. */ +static void r300_update_rs_block(struct r300_context* r300) +{ + struct r300_rs_block* rs = r300->rs_block; + struct vertex_info* vinfo = &r300->vertex_info.vinfo; + int* tab = r300->vertex_info.tab; + int col_count = 0, fp_offset = 0, i, memory_pos, tex_count = 0; + + memset(rs, 0, sizeof(struct r300_rs_block)); + + if (r300_screen(r300->context.screen)->caps->is_r500) { + for (i = 0; i < vinfo->num_attribs; i++) { + assert(tab[vinfo->attrib[i].src_index] != -1); + memory_pos = tab[vinfo->attrib[i].src_index] * 4; + switch (vinfo->attrib[i].interp_mode) { + case INTERP_LINEAR: + rs->ip[col_count] |= + R500_RS_COL_PTR(memory_pos) | + R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA); + col_count++; + break; + case INTERP_PERSPECTIVE: + rs->ip[tex_count] |= + R500_RS_SEL_S(memory_pos) | + R500_RS_SEL_T(memory_pos + 1) | + R500_RS_SEL_R(memory_pos + 2) | + R500_RS_SEL_Q(memory_pos + 3); + tex_count++; + break; + default: + break; + } + } + + if (col_count == 0) { + rs->ip[0] |= R500_RS_COL_FMT(R300_RS_COL_FMT_0001); + } + + /* Set up at least one texture pointer or RS will not be happy. */ + if (tex_count == 0) { + rs->ip[0] |= + R500_RS_SEL_S(R500_RS_IP_PTR_K0) | + R500_RS_SEL_T(R500_RS_IP_PTR_K0) | + R500_RS_SEL_R(R500_RS_IP_PTR_K0) | + R500_RS_SEL_Q(R500_RS_IP_PTR_K1); + } + + for (i = 0; i < tex_count; i++) { + rs->inst[i] |= R500_RS_INST_TEX_ID(i) | R500_RS_INST_TEX_CN_WRITE | + R500_RS_INST_TEX_ADDR(fp_offset); + fp_offset++; + } + + for (i = 0; i < col_count; i++) { + rs->inst[i] |= R500_RS_INST_COL_ID(i) | R500_RS_INST_COL_CN_WRITE | + R500_RS_INST_COL_ADDR(fp_offset); + fp_offset++; + } + } else { + for (i = 0; i < vinfo->num_attribs; i++) { + memory_pos = tab[vinfo->attrib[i].src_index] * 4; + assert(tab[vinfo->attrib[i].src_index] != -1); + switch (vinfo->attrib[i].interp_mode) { + case INTERP_LINEAR: + rs->ip[col_count] |= + R300_RS_COL_PTR(memory_pos) | + R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA); + col_count++; + break; + case INTERP_PERSPECTIVE: + rs->ip[tex_count] |= + R300_RS_TEX_PTR(memory_pos) | + R300_RS_SEL_S(R300_RS_SEL_C0) | + R300_RS_SEL_T(R300_RS_SEL_C1) | + R300_RS_SEL_R(R300_RS_SEL_C2) | + R300_RS_SEL_Q(R300_RS_SEL_C3); + tex_count++; + break; + default: + break; + } + } + + if (col_count == 0) { + rs->ip[0] |= R300_RS_COL_FMT(R300_RS_COL_FMT_0001); + } + + if (tex_count == 0) { + rs->ip[0] |= + R300_RS_SEL_S(R300_RS_SEL_K0) | + R300_RS_SEL_T(R300_RS_SEL_K0) | + R300_RS_SEL_R(R300_RS_SEL_K0) | + R300_RS_SEL_Q(R300_RS_SEL_K1); + } + + for (i = 0; i < tex_count; i++) { + rs->inst[i] |= R300_RS_INST_TEX_ID(i) | R300_RS_INST_TEX_CN_WRITE | + R300_RS_INST_TEX_ADDR(fp_offset); + fp_offset++; + } + + for (i = 0; i < col_count; i++) { + rs->inst[i] |= R300_RS_INST_COL_ID(i) | R300_RS_INST_COL_CN_WRITE | + R300_RS_INST_COL_ADDR(fp_offset); + fp_offset++; + } + } + + rs->count = (tex_count * 4) | (col_count << R300_IC_COUNT_SHIFT) | + R300_HIRES_EN; + + rs->inst_count = MAX2(MAX2(col_count - 1, tex_count - 1), 0); +} + +void r300_update_derived_state(struct r300_context* r300) +{ + if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { + r300_update_vertex_layout(r300); + } + + if (r300->dirty_state & R300_NEW_VERTEX_FORMAT) { + r300_update_rs_block(r300); + } +} diff --git a/src/gallium/drivers/r300/r300_state_derived.h b/src/gallium/drivers/r300/r300_state_derived.h new file mode 100644 index 0000000000..63ae8eb8d0 --- /dev/null +++ b/src/gallium/drivers/r300/r300_state_derived.h @@ -0,0 +1,34 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_STATE_DERIVED_H +#define R300_STATE_DERIVED_H + +#include "draw/draw_vertex.h" + +#include "r300_context.h" +#include "r300_reg.h" +#include "r300_state_inlines.h" + +void r300_update_derived_state(struct r300_context* r300); + +#endif /* R300_STATE_DERIVED_H */ diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h new file mode 100644 index 0000000000..fd92c71756 --- /dev/null +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -0,0 +1,384 @@ +/* + * Copyright 2009 Joakim Sindholt <opensource@zhasha.com> + * Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_STATE_INLINES_H +#define R300_STATE_INLINES_H + +#include "pipe/p_format.h" + +#include "r300_reg.h" + +/* Some maths. These should probably find their way to u_math, if needed. */ + +static INLINE int pack_float_16_6x(float f) { + return ((int)(f * 6.0) & 0xffff); +} + +/* Blend state. */ + +static INLINE uint32_t r300_translate_blend_function(int blend_func) +{ + switch (blend_func) { + case PIPE_BLEND_ADD: + return R300_COMB_FCN_ADD_CLAMP; + case PIPE_BLEND_SUBTRACT: + return R300_COMB_FCN_SUB_CLAMP; + case PIPE_BLEND_REVERSE_SUBTRACT: + return R300_COMB_FCN_RSUB_CLAMP; + case PIPE_BLEND_MIN: + return R300_COMB_FCN_MIN; + case PIPE_BLEND_MAX: + return R300_COMB_FCN_MAX; + default: + debug_printf("r300: Unknown blend function %d\n", blend_func); + break; + } + return 0; +} + +/* XXX we can also offer the D3D versions of some of these... */ +static INLINE uint32_t r300_translate_blend_factor(int blend_fact) +{ + switch (blend_fact) { + case PIPE_BLENDFACTOR_ONE: + return R300_BLEND_GL_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return R300_BLEND_GL_SRC_COLOR; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return R300_BLEND_GL_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return R300_BLEND_GL_DST_ALPHA; + case PIPE_BLENDFACTOR_DST_COLOR: + return R300_BLEND_GL_DST_COLOR; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return R300_BLEND_GL_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return R300_BLEND_GL_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return R300_BLEND_GL_CONST_ALPHA; + /* XXX WTF are these? + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: */ + case PIPE_BLENDFACTOR_ZERO: + return R300_BLEND_GL_ZERO; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return R300_BLEND_GL_ONE_MINUS_SRC_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return R300_BLEND_GL_ONE_MINUS_SRC_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return R300_BLEND_GL_ONE_MINUS_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return R300_BLEND_GL_ONE_MINUS_DST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return R300_BLEND_GL_ONE_MINUS_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return R300_BLEND_GL_ONE_MINUS_CONST_ALPHA; + /* XXX see above + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: */ + default: + debug_printf("r300: Unknown blend factor %d\n", blend_fact); + break; + } + return 0; +} + +/* DSA state. */ + +static INLINE uint32_t r300_translate_depth_stencil_function(int zs_func) +{ + switch (zs_func) { + case PIPE_FUNC_NEVER: + return R300_ZS_NEVER; + case PIPE_FUNC_LESS: + return R300_ZS_LESS; + case PIPE_FUNC_EQUAL: + return R300_ZS_EQUAL; + case PIPE_FUNC_LEQUAL: + return R300_ZS_LEQUAL; + case PIPE_FUNC_GREATER: + return R300_ZS_GREATER; + case PIPE_FUNC_NOTEQUAL: + return R300_ZS_NOTEQUAL; + case PIPE_FUNC_GEQUAL: + return R300_ZS_GEQUAL; + case PIPE_FUNC_ALWAYS: + return R300_ZS_ALWAYS; + default: + debug_printf("r300: Unknown depth/stencil function %d\n", + zs_func); + break; + } + return 0; +} + +static INLINE uint32_t r300_translate_stencil_op(int s_op) +{ + switch (s_op) { + case PIPE_STENCIL_OP_KEEP: + return R300_ZS_KEEP; + case PIPE_STENCIL_OP_ZERO: + return R300_ZS_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return R300_ZS_REPLACE; + case PIPE_STENCIL_OP_INCR: + return R300_ZS_INCR; + case PIPE_STENCIL_OP_DECR: + return R300_ZS_DECR; + case PIPE_STENCIL_OP_INCR_WRAP: + return R300_ZS_INCR_WRAP; + case PIPE_STENCIL_OP_DECR_WRAP: + return R300_ZS_DECR_WRAP; + case PIPE_STENCIL_OP_INVERT: + return R300_ZS_INVERT; + default: + debug_printf("r300: Unknown stencil op %d", s_op); + break; + } + return 0; +} + +static INLINE uint32_t r300_translate_alpha_function(int alpha_func) +{ + switch (alpha_func) { + case PIPE_FUNC_NEVER: + return R300_FG_ALPHA_FUNC_NEVER; + case PIPE_FUNC_LESS: + return R300_FG_ALPHA_FUNC_LESS; + case PIPE_FUNC_EQUAL: + return R300_FG_ALPHA_FUNC_EQUAL; + case PIPE_FUNC_LEQUAL: + return R300_FG_ALPHA_FUNC_LE; + case PIPE_FUNC_GREATER: + return R300_FG_ALPHA_FUNC_GREATER; + case PIPE_FUNC_NOTEQUAL: + return R300_FG_ALPHA_FUNC_NOTEQUAL; + case PIPE_FUNC_GEQUAL: + return R300_FG_ALPHA_FUNC_GE; + case PIPE_FUNC_ALWAYS: + return R300_FG_ALPHA_FUNC_ALWAYS; + default: + debug_printf("r300: Unknown alpha function %d", alpha_func); + break; + } + return 0; +} + +/* Texture sampler state. */ + +static INLINE uint32_t r300_translate_wrap(int wrap) +{ + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + return R300_TX_REPEAT; + case PIPE_TEX_WRAP_CLAMP: + return R300_TX_CLAMP; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return R300_TX_CLAMP_TO_EDGE; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return R300_TX_CLAMP_TO_BORDER; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return R300_TX_REPEAT | R300_TX_MIRRORED; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + return R300_TX_CLAMP | R300_TX_MIRRORED; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED; + default: + debug_printf("r300: Unknown texture wrap %d", wrap); + return 0; + } +} + +static INLINE uint32_t r300_translate_tex_filters(int min, int mag, int mip) +{ + uint32_t retval = 0; + switch (min) { + case PIPE_TEX_FILTER_NEAREST: + retval |= R300_TX_MIN_FILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + retval |= R300_TX_MIN_FILTER_LINEAR; + break; + case PIPE_TEX_FILTER_ANISO: + retval |= R300_TX_MIN_FILTER_ANISO; + break; + default: + debug_printf("r300: Unknown texture filter %d\n", min); + break; + } + switch (mag) { + case PIPE_TEX_FILTER_NEAREST: + retval |= R300_TX_MAG_FILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + retval |= R300_TX_MAG_FILTER_LINEAR; + break; + case PIPE_TEX_FILTER_ANISO: + retval |= R300_TX_MAG_FILTER_ANISO; + break; + default: + debug_printf("r300: Unknown texture filter %d\n", mag); + break; + } + switch (mip) { + case PIPE_TEX_MIPFILTER_NONE: + retval |= R300_TX_MIN_FILTER_MIP_NONE; + break; + case PIPE_TEX_MIPFILTER_NEAREST: + retval |= R300_TX_MIN_FILTER_MIP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + retval |= R300_TX_MIN_FILTER_MIP_LINEAR; + break; + default: + debug_printf("r300: Unknown texture filter %d\n", mip); + break; + } + + return retval; +} + +static INLINE uint32_t r300_anisotropy(float max_aniso) +{ + if (max_aniso >= 16.0f) { + return R300_TX_MAX_ANISO_16_TO_1; + } else if (max_aniso >= 8.0f) { + return R300_TX_MAX_ANISO_8_TO_1; + } else if (max_aniso >= 4.0f) { + return R300_TX_MAX_ANISO_4_TO_1; + } else if (max_aniso >= 2.0f) { + return R300_TX_MAX_ANISO_2_TO_1; + } else { + return R300_TX_MAX_ANISO_1_TO_1; + } +} + +/* Buffer formats. */ + +static INLINE uint32_t r300_translate_colorformat(enum pipe_format format) +{ + switch (format) { + /* 8-bit buffers */ + case PIPE_FORMAT_I8_UNORM: + return R300_COLOR_FORMAT_I8; + /* 16-bit buffers */ + case PIPE_FORMAT_R5G6B5_UNORM: + return R300_COLOR_FORMAT_RGB565; + case PIPE_FORMAT_A1R5G5B5_UNORM: + return R300_COLOR_FORMAT_ARGB1555; + case PIPE_FORMAT_A4R4G4B4_UNORM: + return R300_COLOR_FORMAT_ARGB4444; + /* 32-bit buffers */ + case PIPE_FORMAT_A8R8G8B8_UNORM: + return R300_COLOR_FORMAT_ARGB8888; + /* XXX Not in pipe_format + case PIPE_FORMAT_A32R32G32B32: + return R300_COLOR_FORMAT_ARGB32323232; + case PIPE_FORMAT_A16R16G16B16: + return R300_COLOR_FORMAT_ARGB16161616; */ + /* XXX Not in pipe_format + case PIPE_FORMAT_A10R10G10B10_UNORM: + return R500_COLOR_FORMAT_ARGB10101010; + case PIPE_FORMAT_A2R10G10B10_UNORM: + return R500_COLOR_FORMAT_ARGB2101010; + case PIPE_FORMAT_I10_UNORM: + return R500_COLOR_FORMAT_I10; */ + default: + debug_printf("r300: Implementation error: " \ + "Got unsupported color format %s in %s\n", + pf_name(format), __FUNCTION__); + break; + } + return 0; +} + +static INLINE uint32_t r300_translate_zsformat(enum pipe_format format) +{ + switch (format) { + /* 16-bit depth, no stencil */ + case PIPE_FORMAT_Z16_UNORM: + return R300_DEPTHFORMAT_16BIT_INT_Z; + /* 24-bit depth, 8-bit stencil */ + case PIPE_FORMAT_Z24S8_UNORM: + return R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL; + default: + debug_printf("r300: Implementation error: " \ + "Got unsupported ZS format %s in %s\n", + pf_name(format), __FUNCTION__); + break; + } + return 0; +} + +/* Non-CSO state. (For now.) */ + +static INLINE uint32_t r300_translate_gb_pipes(int pipe_count) +{ + switch (pipe_count) { + case 1: + return R300_GB_TILE_PIPE_COUNT_RV300; + break; + case 2: + return R300_GB_TILE_PIPE_COUNT_R300; + break; + case 3: + return R300_GB_TILE_PIPE_COUNT_R420_3P; + break; + case 4: + return R300_GB_TILE_PIPE_COUNT_R420; + break; + } + return 0; +} + +static INLINE uint32_t translate_vertex_data_type(int type) { + switch (type) { + case EMIT_1F: + case EMIT_1F_PSIZE: + return R300_DATA_TYPE_FLOAT_1; + break; + case EMIT_2F: + return R300_DATA_TYPE_FLOAT_2; + break; + case EMIT_3F: + return R300_DATA_TYPE_FLOAT_3; + break; + case EMIT_4F: + return R300_DATA_TYPE_FLOAT_4; + break; + case EMIT_4UB: + return R300_DATA_TYPE_BYTE; + break; + default: + debug_printf("r300: Implementation error: " + "Bad vertex data type!\n"); + assert(0); + break; + } + + return 0; +} + +#endif /* R300_STATE_INLINES_H */ diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c new file mode 100644 index 0000000000..3d51a8e65d --- /dev/null +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -0,0 +1,203 @@ +/* + * Copyright 2009 Joakim Sindholt <opensource@zhasha.com> + * Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_state_invariant.h" + +/* Calculate and emit invariant state. This is data that the 3D engine + * will probably want at the beginning of every CS, but it's not currently + * handled by any CSO setup, and in addition it doesn't really change much. + * + * Note that eventually this should be empty, but it's useful for development + * and general unduplication of code. */ +void r300_emit_invariant_state(struct r300_context* r300) +{ + struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; + CS_LOCALS(r300); + + BEGIN_CS(24 + (caps->has_tcl ? 2: 0)); + + /* Various GB enables */ + OUT_CS_REG(R300_GB_ENABLE, R300_GB_POINT_STUFF_ENABLE | + R300_GB_LINE_STUFF_ENABLE | R300_GB_TRIANGLE_STUFF_ENABLE); + /* Subpixel multisampling for AA */ + OUT_CS_REG(R300_GB_MSPOS0, 0x66666666); + OUT_CS_REG(R300_GB_MSPOS1, 0x66666666); + /* GB tile config and pipe setup */ + OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_DISABLE | + r300_translate_gb_pipes(caps->num_frag_pipes)); + /* Source of fog depth */ + OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W); + /* AA enable */ + OUT_CS_REG(R300_GB_AA_CONFIG, 0x0); + /* GA errata fixes. */ + if (caps->is_r500) { + OUT_CS_REG(R300_GA_ENHANCE, + R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL | + R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE | + R500_GA_ENHANCE_REG_READWRITE_ENABLE | + R500_GA_ENHANCE_REG_NOSTALL_ENABLE); + } else { + OUT_CS_REG(R300_GA_ENHANCE, + R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL | + R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE); + } + + /* Fog block. */ + OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000000); + OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x00000000); + OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x00000000); + OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x00000000); + OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000); + + /* TCL-only stuff */ + if (caps->has_tcl) { + /* Amount of time to wait for vertex fetches in PVS */ + OUT_CS_REG(VAP_PVS_VTX_TIMEOUT_REG, 0xffff); + } + + END_CS; + + /* XXX unsorted stuff from surface_fill */ + BEGIN_CS(99 + (caps->has_tcl ? 26 : 0)); + /* Flush PVS. */ + OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); + + OUT_CS_REG(R300_SE_VTE_CNTL, R300_VPORT_X_SCALE_ENA | + R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | + R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | + R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT); + /* Max and min vertex index clamp. */ + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, 0xFFFFFF); + OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, 0x0); + /* XXX endian */ + if (caps->has_tcl) { + OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP); + OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE | + R300_PS_UCP_MODE_CLIP_AS_TRIFAN); + OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4); + OUT_CS_32F(1.0); + OUT_CS_32F(1.0); + OUT_CS_32F(1.0); + OUT_CS_32F(1.0); + } else { + OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP | + R300_VAP_TCL_BYPASS); + } + /* XXX magic number not in r300_reg */ + OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA); + /* XXX point tex stuffing */ + OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1); + OUT_CS_32F(0.0); + OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1); + OUT_CS_32F(1.0); + OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 | + (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT)); + /* XXX this big chunk should be refactored into rs_state */ + OUT_CS_REG(R300_GA_LINE_S0, 0x00000000); + OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000); + OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000); + OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000); + OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000); + OUT_CS_REG(R300_GA_ROUND_MODE, 0x00000001); + OUT_CS_REG(R300_GA_OFFSET, 0x00000000); + OUT_CS_REG(R300_GA_FOG_SCALE, 0x3DBF1412); + OUT_CS_REG(R300_GA_FOG_OFFSET, 0x00000000); + OUT_CS_REG(R300_SU_TEX_WRAP, 0x00000000); + OUT_CS_REG(R300_SU_DEPTH_SCALE, 0x4B7FFFFF); + OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000); + OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C); + OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525); + OUT_CS_REG(R300_RB3D_CCTL, 0x00000000); + OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F); + OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000); + OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); + OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF); + OUT_CS_REG(R300_ZB_FORMAT, 0x00000002); + OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003); + OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000); + OUT_CS_REG(R300_ZB_DEPTHCLEARVALUE, 0x00000000); + OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0x00000000); + OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000); + if (caps->has_tcl) { + OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, + (R300_DATA_TYPE_FLOAT_4 << R300_DATA_TYPE_0_SHIFT) | + ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | + R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)); + } else { + OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, + (R300_DATA_TYPE_FLOAT_4 << R300_DATA_TYPE_0_SHIFT) | + ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | + R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)); + } + OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_EXT_0, + (R300_VAP_SWIZZLE_XYZW << R300_SWIZZLE0_SHIFT) | + (R300_VAP_SWIZZLE_XYZW << R300_SWIZZLE1_SHIFT)); + OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1); + OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405); + OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F); + /* Vertex size. */ + OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8); + OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA); + OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_0, 0x00000003); + OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x00000000); + OUT_CS_REG(R300_TX_ENABLE, 0x0); + + /* XXX */ + OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa); + + OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4); + OUT_CS(R300_C0_SEL_B | R300_C1_SEL_G | R300_C2_SEL_R | R300_C3_SEL_A); + OUT_CS(R300_US_OUT_FMT_UNUSED); + OUT_CS(R300_US_OUT_FMT_UNUSED); + OUT_CS(R300_US_OUT_FMT_UNUSED); + OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0); + /* XXX these magic numbers should be explained when + * this becomes a cached state object */ + if (caps->has_tcl) { + OUT_CS_REG(R300_VAP_CNTL, 0xA | + (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) | + (0xB << R300_VF_MAX_VTX_NUM_SHIFT) | + (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT)); + OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, 0x00100000); + OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x00000000); + OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, 0x00000001); + /* XXX translate these back into normal instructions */ + OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x1); + OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0x0); + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, 8); + OUT_CS(0x00F00203); + OUT_CS(0x00D10001); + OUT_CS(0x01248001); + OUT_CS(0x00000000); + OUT_CS(0x00F02203); + OUT_CS(0x00D10021); + OUT_CS(0x01248021); + OUT_CS(0x00000000); + } else { + OUT_CS_REG(R300_VAP_CNTL, 0xA | + (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) | + (0x5 << R300_VF_MAX_VTX_NUM_SHIFT) | + (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT)); + } + END_CS; +} diff --git a/src/gallium/drivers/r300/r300_state_invariant.h b/src/gallium/drivers/r300/r300_state_invariant.h new file mode 100644 index 0000000000..8204bf9588 --- /dev/null +++ b/src/gallium/drivers/r300/r300_state_invariant.h @@ -0,0 +1,33 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_STATE_INVARIANT_H +#define R300_STATE_INVARIANT_H + +#include "r300_context.h" +#include "r300_cs.h" +#include "r300_reg.h" +#include "r300_state_inlines.h" + +void r300_emit_invariant_state(struct r300_context* r300); + +#endif /* R300_STATE_INVARIANT_H */ diff --git a/src/gallium/drivers/r300/r300_state_shader.c b/src/gallium/drivers/r300/r300_state_shader.c index e87172128f..20b83bd15b 100644 --- a/src/gallium/drivers/r300/r300_state_shader.c +++ b/src/gallium/drivers/r300/r300_state_shader.c @@ -22,12 +22,557 @@ #include "r300_state_shader.h" -void r300_translate_shader(struct r300_context* r300, - struct r300_fragment_shader* fs) +static void r300_copy_passthrough_shader(struct r300_fragment_shader* fs) { + struct r300_fragment_shader* pt = &r300_passthrough_fragment_shader; + fs->shader.stack_size = pt->shader.stack_size; + fs->alu_instruction_count = pt->alu_instruction_count; + fs->tex_instruction_count = pt->tex_instruction_count; + fs->indirections = pt->indirections; + fs->instructions[0] = pt->instructions[0]; } -void r500_translate_shader(struct r300_context* r300, - struct r500_fragment_shader* fs) +static void r500_copy_passthrough_shader(struct r500_fragment_shader* fs) { + struct r500_fragment_shader* pt = &r500_passthrough_fragment_shader; + fs->shader.stack_size = pt->shader.stack_size; + fs->instruction_count = pt->instruction_count; + fs->instructions[0] = pt->instructions[0]; +} + +static void r300_fs_declare(struct r300_fs_asm* assembler, + struct tgsi_full_declaration* decl) +{ + switch (decl->Declaration.File) { + case TGSI_FILE_INPUT: + switch (decl->Semantic.SemanticName) { + case TGSI_SEMANTIC_COLOR: + assembler->color_count++; + break; + case TGSI_SEMANTIC_GENERIC: + assembler->tex_count++; + break; + default: + debug_printf("r300: fs: Bad semantic declaration %d\n", + decl->Semantic.SemanticName); + break; + } + break; + case TGSI_FILE_OUTPUT: + case TGSI_FILE_CONSTANT: + break; + case TGSI_FILE_TEMPORARY: + assembler->temp_count++; + break; + default: + debug_printf("r300: fs: Bad file %d\n", decl->Declaration.File); + break; + } + + assembler->temp_offset = assembler->color_count + assembler->tex_count; +} + +static INLINE unsigned r300_fs_src(struct r300_fs_asm* assembler, + struct tgsi_src_register* src) +{ + switch (src->File) { + case TGSI_FILE_NULL: + return 0; + case TGSI_FILE_INPUT: + /* XXX may be wrong */ + return src->Index; + break; + case TGSI_FILE_TEMPORARY: + return src->Index + assembler->temp_offset; + break; + case TGSI_FILE_IMMEDIATE: + return (src->Index + assembler->imm_offset) | (1 << 8); + break; + case TGSI_FILE_CONSTANT: + /* XXX magic */ + return src->Index | (1 << 8); + break; + default: + debug_printf("r300: fs: Unimplemented src %d\n", src->File); + break; + } + return 0; +} + +static INLINE unsigned r300_fs_dst(struct r300_fs_asm* assembler, + struct tgsi_dst_register* dst) +{ + switch (dst->File) { + case TGSI_FILE_NULL: + /* This happens during KIL instructions. */ + return 0; + break; + case TGSI_FILE_OUTPUT: + return 0; + break; + case TGSI_FILE_TEMPORARY: + return dst->Index + assembler->temp_offset; + break; + default: + debug_printf("r300: fs: Unimplemented dst %d\n", dst->File); + break; + } + return 0; +} + +static INLINE unsigned r500_fix_swiz(unsigned s) +{ + /* For historical reasons, the swizzle values x, y, z, w, and 0 are + * equivalent to the actual machine code, but 1 is not. Thus, we just + * adjust it a bit... */ + if (s == TGSI_EXTSWIZZLE_ONE) { + return R500_SWIZZLE_ONE; + } else { + return s; + } +} + +static uint32_t r500_rgba_swiz(struct tgsi_full_src_register* reg) +{ + if (reg->SrcRegister.Extended) { + return r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleX) | + (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleY) << 3) | + (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleZ) << 6) | + (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleW) << 9); + } else { + return reg->SrcRegister.SwizzleX | + (reg->SrcRegister.SwizzleY << 3) | + (reg->SrcRegister.SwizzleZ << 6) | + (reg->SrcRegister.SwizzleW << 9); + } +} + +static uint32_t r500_strq_swiz(struct tgsi_full_src_register* reg) +{ + return reg->SrcRegister.SwizzleX | + (reg->SrcRegister.SwizzleY << 2) | + (reg->SrcRegister.SwizzleZ << 4) | + (reg->SrcRegister.SwizzleW << 6); +} + +static INLINE uint32_t r500_rgb_swiz(struct tgsi_full_src_register* reg) +{ + /* Only the first 9 bits... */ + return (r500_rgba_swiz(reg) & 0x1ff) | + (reg->SrcRegister.Negate ? (1 << 9) : 0) | + (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0); +} + +static INLINE uint32_t r500_alpha_swiz(struct tgsi_full_src_register* reg) +{ + /* Only the last 3 bits... */ + return (r500_rgba_swiz(reg) >> 9) | + (reg->SrcRegister.Negate ? (1 << 9) : 0) | + (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0); +} + +static INLINE uint32_t r500_rgba_op(unsigned op) +{ + switch (op) { + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + return R500_ALU_RGBA_OP_SOP; + case TGSI_OPCODE_FRC: + return R500_ALU_RGBA_OP_FRC; + case TGSI_OPCODE_DP3: + return R500_ALU_RGBA_OP_DP3; + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + return R500_ALU_RGBA_OP_DP4; + case TGSI_OPCODE_ABS: + case TGSI_OPCODE_CMP: + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: + return R500_ALU_RGBA_OP_CMP; + case TGSI_OPCODE_ADD: + case TGSI_OPCODE_MAD: + case TGSI_OPCODE_MUL: + case TGSI_OPCODE_SUB: + return R500_ALU_RGBA_OP_MAD; + default: + return 0; + } +} + +static INLINE uint32_t r500_alpha_op(unsigned op) +{ + switch (op) { + case TGSI_OPCODE_EX2: + return R500_ALPHA_OP_EX2; + case TGSI_OPCODE_LG2: + return R500_ALPHA_OP_LN2; + case TGSI_OPCODE_RCP: + return R500_ALPHA_OP_RCP; + case TGSI_OPCODE_RSQ: + return R500_ALPHA_OP_RSQ; + case TGSI_OPCODE_FRC: + return R500_ALPHA_OP_FRC; + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + return R500_ALPHA_OP_DP; + case TGSI_OPCODE_ABS: + case TGSI_OPCODE_CMP: + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: + return R500_ALPHA_OP_CMP; + case TGSI_OPCODE_ADD: + case TGSI_OPCODE_MAD: + case TGSI_OPCODE_MUL: + case TGSI_OPCODE_SUB: + return R500_ALPHA_OP_MAD; + default: + return 0; + } +} + +static INLINE uint32_t r500_tex_op(unsigned op) +{ + switch (op) { + case TGSI_OPCODE_KIL: + return R500_TEX_INST_TEXKILL; + case TGSI_OPCODE_TEX: + return R500_TEX_INST_LD; + case TGSI_OPCODE_TXB: + return R500_TEX_INST_LODBIAS; + case TGSI_OPCODE_TXP: + return R500_TEX_INST_PROJ; + default: + return 0; + } +} + +/* Setup an ALU operation. */ +static INLINE void r500_emit_alu(struct r500_fragment_shader* fs, + struct r300_fs_asm* assembler, + struct tgsi_full_dst_register* dst) +{ + int i = fs->instruction_count; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + fs->instructions[i].inst0 = R500_INST_TYPE_OUT | + R500_ALU_OMASK(dst->DstRegister.WriteMask); + } else { + fs->instructions[i].inst0 = R500_INST_TYPE_ALU | + R500_ALU_WMASK(dst->DstRegister.WriteMask); + } + + fs->instructions[i].inst0 |= R500_INST_TEX_SEM_WAIT; + + fs->instructions[i].inst4 = + R500_ALPHA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister)); + fs->instructions[i].inst5 = + R500_ALU_RGBA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister)); +} + +static INLINE void r500_emit_maths(struct r500_fragment_shader* fs, + struct r300_fs_asm* assembler, + struct tgsi_full_src_register* src, + struct tgsi_full_dst_register* dst, + unsigned op, + unsigned count) +{ + int i = fs->instruction_count; + + r500_emit_alu(fs, assembler, dst); + + switch (count) { + case 3: + fs->instructions[i].inst1 = + R500_RGB_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister)); + fs->instructions[i].inst2 = + R500_ALPHA_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister)); + fs->instructions[i].inst5 |= + R500_ALU_RGBA_SEL_C_SRC2 | + R500_SWIZ_RGBA_C(r500_rgb_swiz(&src[2])) | + R500_ALU_RGBA_ALPHA_SEL_C_SRC2 | + R500_SWIZ_ALPHA_C(r500_alpha_swiz(&src[2])); + case 2: + fs->instructions[i].inst1 |= + R500_RGB_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister)); + fs->instructions[i].inst2 |= + R500_ALPHA_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister)); + fs->instructions[i].inst3 = + R500_ALU_RGB_SEL_B_SRC1 | + R500_SWIZ_RGB_B(r500_rgb_swiz(&src[1])); + fs->instructions[i].inst4 |= + R500_SWIZ_ALPHA_B(r500_alpha_swiz(&src[1])) | + R500_ALPHA_SEL_B_SRC1; + case 1: + case 0: + default: + fs->instructions[i].inst1 |= + R500_RGB_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister)); + fs->instructions[i].inst2 |= + R500_ALPHA_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister)); + fs->instructions[i].inst3 |= + R500_ALU_RGB_SEL_A_SRC0 | + R500_SWIZ_RGB_A(r500_rgb_swiz(&src[0])); + fs->instructions[i].inst4 |= + R500_SWIZ_ALPHA_A(r500_alpha_swiz(&src[0])) | + R500_ALPHA_SEL_A_SRC0; + break; + } + + fs->instructions[i].inst4 |= r500_alpha_op(op); + fs->instructions[i].inst5 |= r500_rgba_op(op); + + fs->instruction_count++; +} + +static INLINE void r500_emit_tex(struct r500_fragment_shader* fs, + struct r300_fs_asm* assembler, + struct tgsi_full_src_register* src, + struct tgsi_full_dst_register* dst, + uint32_t op) +{ + int i = fs->instruction_count; + + fs->instructions[i].inst0 = R500_INST_TYPE_TEX | + R500_TEX_WMASK(dst->DstRegister.WriteMask) | + R500_INST_TEX_SEM_WAIT; + fs->instructions[i].inst1 = R500_TEX_ID(0) | + R500_TEX_SEM_ACQUIRE | //R500_TEX_IGNORE_UNCOVERED | + r500_tex_op(op); + fs->instructions[i].inst2 = + R500_TEX_SRC_ADDR(r300_fs_src(assembler, &src->SrcRegister)) | + R500_SWIZ_TEX_STRQ(r500_strq_swiz(src)) | + R500_TEX_DST_ADDR(r300_fs_dst(assembler, &dst->DstRegister)) | + R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G | + R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + fs->instructions[i].inst2 |= + R500_TEX_DST_ADDR(assembler->temp_count + + assembler->temp_offset); + + fs->instruction_count++; + + /* Setup and emit a MOV. */ + src[0].SrcRegister.Index = assembler->temp_count; + src[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + + src[1] = src[0]; + src[2] = r500_constant_zero; + r500_emit_maths(fs, assembler, src, dst, TGSI_OPCODE_MOV, 3); + } else { + fs->instruction_count++; + } +} + +static void r500_fs_instruction(struct r500_fragment_shader* fs, + struct r300_fs_asm* assembler, + struct tgsi_full_instruction* inst) +{ + int i; + /* Switch between opcodes. When possible, prefer using the official + * AMD/ATI names for opcodes, please, as it facilitates using the + * documentation. */ + switch (inst->Instruction.Opcode) { + /* The simple scalar ops. */ + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + /* Copy red swizzle to alpha for src0 */ + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX; + inst->FullSrcRegisters[0].SrcRegister.SwizzleW = + inst->FullSrcRegisters[0].SrcRegister.SwizzleX; + /* Fall through */ + case TGSI_OPCODE_FRC: + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 1); + break; + + /* The dot products. */ + case TGSI_OPCODE_DPH: + /* Set alpha swizzle to one for src0 */ + if (!inst->FullSrcRegisters[0].SrcRegister.Extended) { + inst->FullSrcRegisters[0].SrcRegister.Extended = TRUE; + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX = + inst->FullSrcRegisters[0].SrcRegister.SwizzleX; + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleY = + inst->FullSrcRegisters[0].SrcRegister.SwizzleY; + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleZ = + inst->FullSrcRegisters[0].SrcRegister.SwizzleZ; + } + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = + TGSI_EXTSWIZZLE_ONE; + /* Fall through */ + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 2); + break; + + /* Simple three-source operations. */ + case TGSI_OPCODE_CMP: + /* Swap src0 and src2 */ + inst->FullSrcRegisters[3] = inst->FullSrcRegisters[2]; + inst->FullSrcRegisters[2] = inst->FullSrcRegisters[0]; + inst->FullSrcRegisters[0] = inst->FullSrcRegisters[3]; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); + break; + + /* The MAD variants. */ + case TGSI_OPCODE_SUB: + /* Just like ADD, but flip the negation on src1 first */ + inst->FullSrcRegisters[1].SrcRegister.Negate = + !inst->FullSrcRegisters[1].SrcRegister.Negate; + /* Fall through */ + case TGSI_OPCODE_ADD: + /* Force src0 to one, move all registers over */ + inst->FullSrcRegisters[2] = inst->FullSrcRegisters[1]; + inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0]; + inst->FullSrcRegisters[0] = r500_constant_one; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); + break; + case TGSI_OPCODE_MUL: + /* Force our src2 to zero */ + inst->FullSrcRegisters[2] = r500_constant_zero; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); + break; + case TGSI_OPCODE_MAD: + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); + break; + + /* The MOV variants. */ + case TGSI_OPCODE_ABS: + /* Set absolute value modifiers. */ + inst->FullSrcRegisters[0].SrcRegisterExtMod.Absolute = TRUE; + /* Fall through */ + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: + /* src0 -> src1 and src2 forced to zero */ + inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0]; + inst->FullSrcRegisters[2] = r500_constant_zero; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); + break; + + /* The texture instruction set. */ + case TGSI_OPCODE_KIL: + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXB: + case TGSI_OPCODE_TXP: + r500_emit_tex(fs, assembler, &inst->FullSrcRegisters[0], + &inst->FullDstRegisters[0], inst->Instruction.Opcode); + break; + + /* This is the end. My only friend, the end. */ + case TGSI_OPCODE_END: + break; + default: + debug_printf("r300: fs: Bad opcode %d\n", + inst->Instruction.Opcode); + break; + } + + /* Clamp, if saturation flags are set. */ + if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE) { + fs->instructions[fs->instruction_count - 1].inst0 |= + R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP; + } +} + +static void r500_fs_finalize(struct r500_fragment_shader* fs, + struct r300_fs_asm* assembler) +{ + fs->shader.stack_size = assembler->temp_count + assembler->temp_offset; + + /* XXX should this just go with OPCODE_END? */ + fs->instructions[fs->instruction_count - 1].inst0 |= + R500_INST_LAST; +} + +void r300_translate_fragment_shader(struct r300_context* r300, + struct r300_fragment_shader* fs) +{ + struct tgsi_parse_context parser; + + tgsi_parse_init(&parser, fs->shader.state.tokens); + + while (!tgsi_parse_end_of_tokens(&parser)) { + tgsi_parse_token(&parser); + } + + r300_copy_passthrough_shader(fs); +} + +void r500_translate_fragment_shader(struct r300_context* r300, + struct r500_fragment_shader* fs) +{ + struct tgsi_parse_context parser; + int i; + struct r300_constant_buffer* consts = + &r300->shader_constants[PIPE_SHADER_FRAGMENT]; + + struct r300_fs_asm* assembler = CALLOC_STRUCT(r300_fs_asm); + if (assembler == NULL) { + return; + } + /* Setup starting offset for immediates. */ + assembler->imm_offset = consts->user_count; + + tgsi_parse_init(&parser, fs->shader.state.tokens); + + while (!tgsi_parse_end_of_tokens(&parser)) { + tgsi_parse_token(&parser); + + /* This is seriously the lamest way to create fragment programs ever. + * I blame TGSI. */ + switch (parser.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* Allocated registers sitting at the beginning + * of the program. */ + r300_fs_declare(assembler, &parser.FullToken.FullDeclaration); + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + debug_printf("r300: Emitting immediate to constant buffer, " + "position %d\n", + assembler->imm_offset + assembler->imm_count); + /* I am not amused by the length of these. */ + for (i = 0; i < 4; i++) { + consts->constants[assembler->imm_offset + + assembler->imm_count][i] = + parser.FullToken.FullImmediate.u.ImmediateFloat32[i] + .Float; + } + assembler->imm_count++; + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + r500_fs_instruction(fs, assembler, + &parser.FullToken.FullInstruction); + break; + } + + } + + debug_printf("r300: %d texs and %d colors, first free reg is %d\n", + assembler->tex_count, assembler->color_count, + assembler->tex_count + assembler->color_count); + + consts->count = consts->user_count + assembler->imm_count; + debug_printf("r300: %d total constants, " + "%d from user and %d from immediates\n", consts->count, + consts->user_count, assembler->imm_count); + r500_fs_finalize(fs, assembler); + + tgsi_dump(fs->shader.state.tokens); + r500_fs_dump(fs); + + tgsi_parse_free(&parser); + FREE(assembler); } diff --git a/src/gallium/drivers/r300/r300_state_shader.h b/src/gallium/drivers/r300/r300_state_shader.h index a20bd4276c..06c0bb7378 100644 --- a/src/gallium/drivers/r300/r300_state_shader.h +++ b/src/gallium/drivers/r300/r300_state_shader.h @@ -23,13 +23,142 @@ #ifndef R300_STATE_SHADER_H #define R300_STATE_SHADER_H +#include "tgsi/tgsi_parse.h" + #include "r300_context.h" +#include "r300_debug.h" +#include "r300_reg.h" #include "r300_screen.h" -void r300_translate_shader(struct r300_context* r300, +/* XXX this all should find its way back to r300_reg */ +/* Swizzle tools */ +#define R500_SWIZZLE_ZERO 4 +#define R500_SWIZZLE_HALF 5 +#define R500_SWIZZLE_ONE 6 +#define R500_SWIZ_RGB_ZERO ((4 << 0) | (4 << 3) | (4 << 6)) +#define R500_SWIZ_RGB_ONE ((6 << 0) | (6 << 3) | (6 << 6)) +#define R500_SWIZ_RGB_RGB ((0 << 0) | (1 << 3) | (2 << 6)) +#define R500_SWIZ_MOD_NEG 1 +#define R500_SWIZ_MOD_ABS 2 +#define R500_SWIZ_MOD_NEG_ABS 3 +/* Swizzles for inst2 */ +#define R500_SWIZ_TEX_STRQ(x) ((x) << 8) +#define R500_SWIZ_TEX_RGBA(x) ((x) << 24) +/* Swizzles for inst3 */ +#define R500_SWIZ_RGB_A(x) ((x) << 2) +#define R500_SWIZ_RGB_B(x) ((x) << 15) +/* Swizzles for inst4 */ +#define R500_SWIZ_ALPHA_A(x) ((x) << 14) +#define R500_SWIZ_ALPHA_B(x) ((x) << 21) +/* Swizzle for inst5 */ +#define R500_SWIZ_RGBA_C(x) ((x) << 14) +#define R500_SWIZ_ALPHA_C(x) ((x) << 27) +/* Writemasks */ +#define R500_TEX_WMASK(x) ((x) << 11) +#define R500_ALU_WMASK(x) ((x) << 11) +#define R500_ALU_OMASK(x) ((x) << 15) + +/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're + * not using enough of it. */ +static const struct tgsi_full_src_register r500_constant_zero = { + .SrcRegister.Extended = TRUE, + .SrcRegister.File = TGSI_FILE_NULL, + .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO, + .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO, + .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO, + .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO, +}; + +static const struct tgsi_full_src_register r500_constant_one = { + .SrcRegister.Extended = TRUE, + .SrcRegister.File = TGSI_FILE_NULL, + .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE, + .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE, + .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE, + .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE, +}; + +/* Temporary struct used to hold assembly state while putting together + * fragment programs. */ +struct r300_fs_asm { + /* Pipe context. */ + struct r300_context* r300; + /* Number of colors. */ + unsigned color_count; + /* Number of texcoords. */ + unsigned tex_count; + /* Offset for temporary registers. Inputs and temporaries have no + * distinguishing markings, so inputs start at 0 and the first usable + * temporary register is after all inputs. */ + unsigned temp_offset; + /* Number of requested temporary registers. */ + unsigned temp_count; + /* Offset for immediate constants. Neither R300 nor R500 can do four + * inline constants per source, so instead we copy immediates into the + * constant buffer. */ + unsigned imm_offset; + /* Number of immediate constants. */ + unsigned imm_count; +}; + +void r300_translate_fragment_shader(struct r300_context* r300, struct r300_fragment_shader* fs); -void r500_translate_shader(struct r300_context* r300, +void r500_translate_fragment_shader(struct r300_context* r300, struct r500_fragment_shader* fs); +static const struct r300_fragment_shader r300_passthrough_fragment_shader = { + /* XXX This is the emission code. TODO: decode + OUT_CS_REG(R300_US_CONFIG, 0); + OUT_CS_REG(R300_US_CODE_OFFSET, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); + OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000); +*/ + .alu_instruction_count = 1, + .tex_instruction_count = 0, + .indirections = 1, + .shader.stack_size = 2, + + .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) | + R300_RGB_SWIZB(R300_ALU_ARGC_ONE) | + R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) | + R300_ALU_OUTC_MAD, + .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) | + R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ, + .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) | + R300_ALPHA_SWIZB(R300_ALU_ARGA_ONE) | + R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) | + R300_ALU_OUTA_MAD, + .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) | + R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT, +}; + +static const struct r500_fragment_shader r500_passthrough_fragment_shader = { + .shader.stack_size = 0, + .instruction_count = 1, + .instructions[0].inst0 = R500_INST_TYPE_OUT | + R500_INST_TEX_SEM_WAIT | R500_INST_LAST | + R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK | + R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, + .instructions[0].inst1 = + R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST | + R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST, + .instructions[0].inst2 = + R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST | + R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST, + .instructions[0].inst3 = + R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R | + R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B | + R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R | + R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B, + .instructions[0].inst4 = + R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A, + .instructions[0].inst5 = + R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 | + R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 | + R500_ALU_RGBA_A_SWIZ_0, +}; + #endif /* R300_STATE_SHADER_H */ diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c index 1e1f96a7f9..2cc0677e52 100644 --- a/src/gallium/drivers/r300/r300_surface.c +++ b/src/gallium/drivers/r300/r300_surface.c @@ -1,5 +1,6 @@ /* * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -32,321 +33,161 @@ static void r300_surface_fill(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); CS_LOCALS(r300); - struct r300_capabilities* caps = ((struct r300_screen*)pipe->screen)->caps; + struct r300_capabilities* caps = r300_screen(pipe->screen)->caps; struct r300_texture* tex = (struct r300_texture*)dest->texture; int i; - float r, g, b, a; + float r, g, b, a, depth; + unsigned pixpitch = tex->stride / tex->tex.block.size; + r = (float)((color >> 16) & 0xff) / 255.0f; g = (float)((color >> 8) & 0xff) / 255.0f; b = (float)((color >> 0) & 0xff) / 255.0f; debug_printf("r300: Filling surface %p at (%d,%d)," - " dimensions %dx%d (stride %d), color 0x%x\n", - dest, x, y, w, h, dest->stride, color); + " dimensions %dx%d (pixel pitch %d), color 0x%x\n", + dest, x, y, w, h, pixpitch, color); /* Fallback? */ - if (0) { + if (tex->tex.format != PIPE_FORMAT_A8R8G8B8_UNORM) { debug_printf("r300: Falling back on surface clear..."); - void* map = pipe->screen->surface_map(pipe->screen, dest, - PIPE_BUFFER_USAGE_CPU_WRITE); - pipe_fill_rect(map, &dest->block, &dest->stride, x, y, w, h, color); - pipe->screen->surface_unmap(pipe->screen, dest); + util_surface_fill(pipe, dest, x, y, w, h, color); return; } -BEGIN_CS((caps->is_r500) ? 309 : 280); -R300_PACIFY; -OUT_CS_REG(R300_TX_INVALTAGS, 0x0); -R300_PACIFY; -/* Flush PVS. */ -OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); - -OUT_CS_REG(R300_SE_VTE_CNTL, R300_VPORT_X_SCALE_ENA | - R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | - R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | - R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT); -/* Vertex size. */ -OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8); -/* Max and min vertex index clamp. */ -OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, 0xFFFFFF); -OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, 0x0); -/* XXX endian */ -OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP); -OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, 0x0); -/* XXX magic number not in r300_reg */ -OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA); -OUT_CS_REG(R300_VAP_CLIP_CNTL, 0x0); -OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4); -OUT_CS_32F(1.0); -OUT_CS_32F(1.0); -OUT_CS_32F(1.0); -OUT_CS_32F(1.0); -/* XXX is this too long? */ -OUT_CS_REG(VAP_PVS_VTX_TIMEOUT_REG, 0xFFFF); -OUT_CS_REG(R300_GB_ENABLE, R300_GB_POINT_STUFF_ENABLE | - R300_GB_LINE_STUFF_ENABLE | R300_GB_TRIANGLE_STUFF_ENABLE); -/* XXX more magic numbers */ -OUT_CS_REG(R300_GB_MSPOS0, 0x66666666); -OUT_CS_REG(R300_GB_MSPOS1, 0x66666666); -/* XXX why doesn't classic Mesa write the number of pipes, too? */ -OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_ENABLE | R300_GB_TILE_SIZE_16); -OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W); -OUT_CS_REG(R300_GB_AA_CONFIG, 0x0); -/* XXX point tex stuffing */ -OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1); -OUT_CS_32F(0.0); -OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1); -OUT_CS_32F(1.0); -OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 | - (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT)); -/* XXX should this be related to the actual point size? */ -OUT_CS_REG(R300_GA_POINT_MINMAX, 0x6 | - (0x1800 << R300_GA_POINT_MINMAX_MAX_SHIFT)); -/* XXX this big chunk should be refactored into rs_state */ -OUT_CS_REG(R300_GA_LINE_CNTL, 0x00030006); -OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, 0x3BAAAAAB); -OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, 0x00000000); -OUT_CS_REG(R300_GA_LINE_S0, 0x00000000); -OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000); -OUT_CS_REG(R300_GA_ENHANCE, 0x00000002); -OUT_CS_REG(R300_GA_COLOR_CONTROL, 0x0003AAAA); -OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000); -OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000); -OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000); -OUT_CS_REG(R300_GA_ROUND_MODE, 0x00000001); -OUT_CS_REG(R300_GA_OFFSET, 0x00000000); -OUT_CS_REG(R300_GA_FOG_SCALE, 0x3DBF1412); -OUT_CS_REG(R300_GA_FOG_OFFSET, 0x00000000); -OUT_CS_REG(R300_SU_TEX_WRAP, 0x00000000); -OUT_CS_REG(R300_SU_POLY_OFFSET_FRONT_SCALE, 0x00000000); -OUT_CS_REG(R300_SU_POLY_OFFSET_FRONT_OFFSET, 0x00000000); -OUT_CS_REG(R300_SU_POLY_OFFSET_BACK_SCALE, 0x00000000); -OUT_CS_REG(R300_SU_POLY_OFFSET_BACK_OFFSET, 0x00000000); -OUT_CS_REG(R300_SU_POLY_OFFSET_ENABLE, 0x00000000); -OUT_CS_REG(R300_SU_CULL_MODE, 0x00000000); -OUT_CS_REG(R300_SU_DEPTH_SCALE, 0x4B7FFFFF); -OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000); -OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C); -OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525); -OUT_CS_REG(R300_SC_SCREENDOOR, 0x00FFFFFF); -OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000002); -OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x00000000); -OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x00000000); -OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x00000000); -OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000); -OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000); -OUT_CS_REG(R300_RB3D_CCTL, 0x00000000); -OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F); - -/* XXX: Oh the wonderful unknown */ -OUT_CS_REG_SEQ(0x4E54, 8); -for (i = 0; i < 8; i++) - OUT_CS(0x00000000); -OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000); -OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); -OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF); -OUT_CS_REG(R300_ZB_FORMAT, 0x00000002); -OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003); -OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000); -OUT_CS_REG(R300_ZB_DEPTHCLEARVALUE, 0x00000000); -OUT_CS_REG(0x4F30, 0x00000000); -OUT_CS_REG(0x4F34, 0x00000000); -OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0x00000000); -OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000); -R300_PACIFY; -if (caps->has_tcl) { - OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, - (R300_DATA_TYPE_FLOAT_4 << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | - R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)); -} else { - OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, - (R300_DATA_TYPE_FLOAT_4 << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | - R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)); -} -OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000000); -OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_EXT_0, 0xF688F688); -OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1); -OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405); -OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F); -OUT_CS_REG(R300_VAP_VTX_SIZE, 0x00000008); -OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA); -OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_0, 0x00000003); -OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x00000000); -OUT_CS_REG(R300_TX_ENABLE, 0x0); -/* XXX viewport setup */ -OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6); -OUT_CS_32F(1.0); -OUT_CS_32F((float)x); -OUT_CS_32F(1.0); -OUT_CS_32F((float)y); -OUT_CS_32F(1.0); -OUT_CS_32F(0.0); - -if (caps->has_tcl) { - OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE | - R300_PS_UCP_MODE_CLIP_AS_TRIFAN); -} - -OUT_CS_REG(R300_GA_POINT_SIZE, ((h * 6) & R300_POINTSIZE_Y_MASK) | - ((w * 6) << R300_POINTSIZE_X_SHIFT)); - -/* XXX RS block and fp setup */ -if (caps->is_r500) { - OUT_CS_REG_SEQ(R500_RS_IP_0, 8); - for (i = 0; i < 8; i++) { - /* I like the operator macros more than the shift macros... */ - OUT_CS((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) | - (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT)); + r300_emit_blend_state(r300, &blend_clear_state); + r300_emit_blend_color_state(r300, &blend_color_clear_state); + r300_emit_dsa_state(r300, &dsa_clear_state); + r300_emit_rs_state(r300, &rs_clear_state); + + /* Fragment shader setup */ + if (caps->is_r500) { + r500_emit_fragment_shader(r300, &r500_passthrough_fragment_shader); + r300_emit_rs_block_state(r300, &r500_rs_block_clear_state); + } else { + r300_emit_fragment_shader(r300, &r300_passthrough_fragment_shader); + r300_emit_rs_block_state(r300, &r300_rs_block_clear_state); } - /* XXX */ - OUT_CS_REG_SEQ(R300_RS_COUNT, 2); - OUT_CS((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_CS(0x0); - OUT_CS_REG(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE); - OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); - OUT_CS_REG(R500_US_PIXSIZE, 0x00000000); - OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) | - R500_US_CODE_END_ADDR(1)); - OUT_CS_REG(R500_US_CODE_RANGE, R500_US_CODE_RANGE_ADDR(0) | - R500_US_CODE_RANGE_SIZE(1)); - OUT_CS_REG(R500_US_CODE_OFFSET, R500_US_CODE_OFFSET_ADDR(0)); - R300_PACIFY; - OUT_CS_REG(R500_GA_US_VECTOR_INDEX, - 0 | R500_GA_US_VECTOR_INDEX_TYPE_INSTR); - OUT_CS_REG(R500_GA_US_VECTOR_DATA, - R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT | R500_INST_LAST | - R500_INST_RGB_OMASK_R | R500_INST_RGB_OMASK_G | R500_INST_RGB_OMASK_B | - R500_INST_ALPHA_OMASK | R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP); - OUT_CS_REG(R500_GA_US_VECTOR_DATA, - R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST | - R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST); - OUT_CS_REG(R500_GA_US_VECTOR_DATA, - R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST | - R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST); - OUT_CS_REG(R500_GA_US_VECTOR_DATA, - R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R | - R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B | - R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R | - R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B); - OUT_CS_REG(R500_GA_US_VECTOR_DATA, - R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A); - OUT_CS_REG(R500_GA_US_VECTOR_DATA, - R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 | - R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 | - R500_ALU_RGBA_A_SWIZ_0); -} else { - OUT_CS_REG_SEQ(R300_RS_IP_0, 8); - for (i = 0; i < 8; i++) { - OUT_CS(R300_RS_SEL_T(R300_RS_SEL_K0) | - R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1)); - } - /* XXX */ - OUT_CS_REG_SEQ(R300_RS_COUNT, 2); - OUT_CS((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_CS(1); - OUT_CS_REG(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE); - - /* XXX magic numbers */ - OUT_CS_REG(R300_US_CONFIG, 0); - OUT_CS_REG(R300_US_PIXSIZE, 2); - OUT_CS_REG(R300_US_CODE_OFFSET, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000); - OUT_CS_REG(R300_US_ALU_RGB_INST_0, 0x50A80); - OUT_CS_REG(R300_US_ALU_RGB_ADDR_0, 0x1C000000); - OUT_CS_REG(R300_US_ALU_ALPHA_INST_0, 0x40889); - OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0, 0x1000000); - OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4); - OUT_CS(R300_C0_SEL_B | R300_C1_SEL_G | R300_C2_SEL_R | R300_C3_SEL_A); - OUT_CS(R300_US_OUT_FMT_UNUSED); - OUT_CS(R300_US_OUT_FMT_UNUSED); - OUT_CS(R300_US_OUT_FMT_UNUSED); - OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0); -} -/* XXX these magic numbers should be explained when - * this becomes a cached state object */ -if (caps->has_tcl) { - OUT_CS_REG(R300_VAP_CNTL, 0xA | - (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (0xB << R300_VF_MAX_VTX_NUM_SHIFT) | - (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT)); - OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, 0x00100000); - OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x00000000); - OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, 0x00000001); - R300_PACIFY; - /* XXX translate these back into normal instructions */ - OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x1); - OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0x0); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xF00203); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xD10001); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x1248001); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x0); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xF02203); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xD10021); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x1248021); - OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x0); -} else { - OUT_CS_REG(R300_VAP_CNTL, 0xA | - (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (0x5 << R300_VF_MAX_VTX_NUM_SHIFT) | - (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT)); + BEGIN_CS(36); + + /* Viewport setup */ + OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6); + OUT_CS_32F(1.0); + OUT_CS_32F((float)x); + OUT_CS_32F(1.0); + OUT_CS_32F((float)y); + OUT_CS_32F(1.0); + OUT_CS_32F(0.0); + + /* Pixel scissors */ + OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2); + OUT_CS((x << R300_SCISSORS_X_SHIFT) | (y << R300_SCISSORS_Y_SHIFT)); + OUT_CS((w << R300_SCISSORS_X_SHIFT) | (h << R300_SCISSORS_Y_SHIFT)); + + /* The size of the point we're about to draw, in sixths of pixels */ + OUT_CS_REG(R300_GA_POINT_SIZE, + ((h * 6) & R300_POINTSIZE_Y_MASK) | + ((w * 6) << R300_POINTSIZE_X_SHIFT)); + + /* Flush colorbuffer and blend caches. */ + OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, + R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D | + R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_SIGNAL); + OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, + R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | + R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); + + OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1); + OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_CS_REG(R300_RB3D_COLORPITCH0, pixpitch | + r300_translate_colorformat(tex->tex.format)); + OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F); + /* XXX Packet3 */ + OUT_CS(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8)); + OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING | + (1 << R300_PRIM_NUM_VERTICES_SHIFT)); + OUT_CS_32F(w / 2.0); + OUT_CS_32F(h / 2.0); + /* XXX this should be the depth value to clear to */ + OUT_CS_32F(1.0); + OUT_CS_32F(1.0); + OUT_CS_32F(r); + OUT_CS_32F(g); + OUT_CS_32F(b); + OUT_CS_32F(1.0); + + /* XXX figure out why this is 0xA and not 0x2 */ + OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA); + /* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, + R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | + R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */ + + END_CS; + + r300->dirty_hw++; } -R300_PACIFY; -END_CS; -r300_emit_blend_state(r300, &blend_clear_state); -r300_emit_blend_color_state(r300, &blend_color_clear_state); -r300_emit_dsa_state(r300, &dsa_clear_state); - -BEGIN_CS(36); -R300_PACIFY; -/* Flush colorbuffer and blend caches. */ -OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, - R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D | - R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_SIGNAL); -OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, - R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | - R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); - -OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1); -OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); -/* XXX this should not be so rigid and it still doesn't work right */ -OUT_CS_REG(R300_RB3D_COLORPITCH0, (dest->stride >> 2) | R300_COLOR_FORMAT_ARGB8888); -OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F); -/* XXX Packet3 */ -OUT_CS(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8)); -OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING | -(1 << R300_PRIM_NUM_VERTICES_SHIFT)); -OUT_CS_32F(w / 2.0); -OUT_CS_32F(h / 2.0); -/* XXX this should be the depth value to clear to */ -OUT_CS_32F(1.0); -OUT_CS_32F(1.0); -OUT_CS_32F(r); -OUT_CS_32F(g); -OUT_CS_32F(b); -OUT_CS_32F(1.0); - -/* XXX figure out why this is 0xA and not 0x2 */ -OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA); -/* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, - R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | - R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */ -R300_PACIFY; - -END_CS; -FLUSH_CS; - - r300->dirty_state = R300_NEW_KITCHEN_SINK; +static void r300_surface_copy(struct pipe_context* pipe, + struct pipe_surface* dest, + unsigned destx, unsigned desty, + struct pipe_surface* src, + unsigned srcx, unsigned srcy, + unsigned w, unsigned h) +{ + struct r300_context* r300 = r300_context(pipe); + CS_LOCALS(r300); + struct r300_texture* srctex = (struct r300_texture*)src->texture; + struct r300_texture* desttex = (struct r300_texture*)dest->texture; + + unsigned pixpitch = srctex->stride / srctex->tex.block.size; + debug_printf("r300: Copying surface %p at (%d,%d) to %p at (%d, %d)," + " dimensions %dx%d (pixel pitch %d)\n", + src, srcx, srcy, dest, destx, desty, w, h, pixpitch); + + if (TRUE) { + debug_printf("r300: Falling back on surface_copy\n"); + return util_surface_copy(pipe, FALSE, dest, destx, desty, src, + srcx, srcy, w, h); + } +#if 0 + BEGIN_CS(); + OUT_CS_REG(RADEON_DEFAULT_SC_BOTTOM_RIGHT,(RADEON_DEFAULT_SC_RIGHT_MAX | + RADEON_DEFAULT_SC_BOTTOM_MAX)); + OUT_ACCEL_REG(RADEON_DP_GUI_MASTER_CNTL, (RADEON_GMC_DST_PITCH_OFFSET_CNTL | + RADEON_GMC_SRC_PITCH_OFFSET_CNTL | + RADEON_GMC_BRUSH_NONE | + (datatype << 8) | + RADEON_GMC_SRC_DATATYPE_COLOR | + RADEON_ROP[rop].rop | + RADEON_DP_SRC_SOURCE_MEMORY | + RADEON_GMC_CLR_CMP_CNTL_DIS)); + OUT_CS_REG(RADEON_DP_BRUSH_FRGD_CLR, 0xffffffff); + OUT_CS_REG(RADEON_DP_BRUSH_BKGD_CLR, 0x0); + OUT_CS_REG(RADEON_DP_SRC_FRGD_CLR, 0xffffffff); + OUT_CS_REG(RADEON_DP_SRC_BKGD_CLR, 0x0); + OUT_ACCEL_REG(RADEON_DP_WRITE_MASK, planemask); + OUT_ACCEL_REG(RADEON_DP_CNTL, ((info->accel_state->xdir >= 0 ? RADEON_DST_X_LEFT_TO_RIGHT : 0) | + (info->accel_state->ydir >= 0 ? RADEON_DST_Y_TOP_TO_BOTTOM : 0)); +); + + OUT_CS_REG_SEQ(RADEON_DST_PITCH_OFFSET, 1); + OUT_CS_RELOC(desttex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + + OUT_CS_REG_SEQ(RADEON_SRC_PITCH_OFFSET, 1); + OUT_CS_RELOC(srctex->buffer, 0, + RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0, 0); + + OUT_CS_REG(RADEON_SRC_Y_X, (srcy << 16) | srcx); + OUT_CS_REG(RADEON_DST_Y_X, (desty << 16) | destx); + OUT_CS_REG(RADEON_DST_HEIGHT_WIDTH, (h << 16) | w); + OUT_CS_REG(RADEON_DSTCACHE_CTLSTAT, RADEON_RB2D_DC_FLUSH_ALL); + OUT_CS_REG(RADEON_WAIT_UNTIL, + RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_DMA_GUI_IDLE); + END_CS; +#endif } void r300_init_surface_functions(struct r300_context* r300) { r300->context.surface_fill = r300_surface_fill; + r300->context.surface_copy = r300_surface_copy; } diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h index e1d53116a1..b75b3ab84c 100644 --- a/src/gallium/drivers/r300/r300_surface.h +++ b/src/gallium/drivers/r300/r300_surface.h @@ -31,6 +31,8 @@ #include "r300_context.h" #include "r300_cs.h" #include "r300_emit.h" +#include "r300_state_shader.h" +#include "r300_state_inlines.h" const struct r300_blend_state blend_clear_state = { .blend_control = 0x0, @@ -55,4 +57,38 @@ const struct r300_dsa_state dsa_clear_state = { .stencil_ref_bf = 0x0, }; +const struct r300_rs_state rs_clear_state = { + .point_minmax = 0x36000006, + .line_control = 0x00030006, + .depth_scale_front = 0x0, + .depth_offset_front = 0x0, + .depth_scale_back = 0x0, + .depth_offset_back = 0x0, + .polygon_offset_enable = 0x0, + .cull_mode = 0x0, + .line_stipple_config = 0x3BAAAAAB, + .line_stipple_value = 0x0, + .color_control = R300_SHADE_MODEL_FLAT, +}; + +const struct r300_rs_block r300_rs_block_clear_state = { + .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) | + R500_RS_SEL_T(R300_RS_SEL_K0) | + R500_RS_SEL_R(R300_RS_SEL_K0) | + R500_RS_SEL_Q(R300_RS_SEL_K1), + .inst[0] = R300_RS_INST_COL_CN_WRITE, + .count = R300_IT_COUNT(0) | R300_IC_COUNT(1) | R300_HIRES_EN, + .inst_count = 0, +}; + +const struct r300_rs_block r500_rs_block_clear_state = { + .ip[0] = R500_RS_SEL_S(R500_RS_IP_PTR_K0) | + R500_RS_SEL_T(R500_RS_IP_PTR_K0) | + R500_RS_SEL_R(R500_RS_IP_PTR_K0) | + R500_RS_SEL_Q(R500_RS_IP_PTR_K1), + .inst[0] = R500_RS_INST_COL_CN_WRITE, + .count = R300_IT_COUNT(0) | R300_IC_COUNT(1) | R300_HIRES_EN, + .inst_count = 0, +}; + #endif /* R300_SURFACE_H */ diff --git a/src/gallium/drivers/r300/r300_swtcl_emit.c b/src/gallium/drivers/r300/r300_swtcl_emit.c index f6e98d23e9..3db09514c6 100644 --- a/src/gallium/drivers/r300/r300_swtcl_emit.c +++ b/src/gallium/drivers/r300/r300_swtcl_emit.c @@ -21,109 +21,315 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "draw/draw_pipe.h" +#include "draw/draw_vbuf.h" #include "util/u_memory.h" #include "r300_cs.h" #include "r300_context.h" #include "r300_reg.h" +#include "r300_state_derived.h" -/* r300_swtcl_emit: Primitive vertex emission using an immediate - * vertex buffer and no HW TCL. */ +/* r300_swtcl_emit: Vertex and index buffer primitive emission. No HW TCL. */ -struct swtcl_stage { +struct r300_swtcl_render { /* Parent class */ - struct draw_stage draw; - + struct vbuf_render base; + + /* Pipe context */ struct r300_context* r300; + + /* Vertex information */ + size_t vertex_size; + unsigned prim; + unsigned hwprim; + + /* VBO */ + struct pipe_buffer* vbo; + size_t vbo_size; + size_t vbo_offset; + void* vbo_map; + size_t vbo_alloc_size; + size_t vbo_max_used; }; -static INLINE struct swtcl_stage* swtcl_stage(struct draw_stage* draw) { - return (struct swtcl_stage*)draw; +static INLINE struct r300_swtcl_render* +r300_swtcl_render(struct vbuf_render* render) +{ + return (struct r300_swtcl_render*)render; } -static void r300_emit_vertex(struct r300_context* r300, - const struct vertex_header* vertex) +static const struct vertex_info* +r300_swtcl_render_get_vertex_info(struct vbuf_render* render) { - /* XXX */ + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + struct r300_context* r300 = r300render->r300; + + r300_update_derived_state(r300); + + return &r300->vertex_info; } -static INLINE void r300_emit_prim(struct draw_stage* draw, - struct prim_header* prim, - unsigned hwprim, - unsigned count) +static boolean r300_swtcl_render_allocate_vertices(struct vbuf_render* render, + ushort vertex_size, + ushort count) { - struct r300_context* r300 = swtcl_stage(draw)->r300; - CS_LOCALS(r300); - int i; + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + struct r300_context* r300 = r300render->r300; + struct pipe_screen* screen = r300->context.screen; + size_t size = (size_t)vertex_size * (size_t)count; - r300_emit_dirty_state(r300); + if (r300render->vbo) { + pipe_buffer_reference(&r300render->vbo, NULL); + } + + r300render->vbo_size = MAX2(size, r300render->vbo_alloc_size); + r300render->vbo_offset = 0; + r300render->vbo = pipe_buffer_create(screen, + 64, + PIPE_BUFFER_USAGE_VERTEX, + r300render->vbo_size); - /* XXX should be count * vtx size */ - BEGIN_CS(2 + count + 6); - OUT_CS(CP_PACKET3(R200_3D_DRAW_IMMD_2, count)); - OUT_CS(hwprim | R300_PRIM_WALK_RING | - (count << R300_PRIM_NUM_VERTICES_SHIFT)); + r300render->vertex_size = vertex_size; - for (i = 0; i < count; i++) { - r300_emit_vertex(r300, prim->v[i]); + if (r300render->vbo) { + return TRUE; + } else { + return FALSE; } - R300_PACIFY; - END_CS; } -/* Just as an aside... - * - * Radeons can do many more primitives: - * - Line strip - * - Triangle fan - * - Triangle strip - * - Line loop - * - Quads - * - Quad strip - * - Polygons - * - * The following were just the only ones in Draw. */ +static void* r300_swtcl_render_map_vertices(struct vbuf_render* render) +{ + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + struct pipe_screen* screen = r300render->r300->context.screen; -static void r300_emit_point(struct draw_stage* draw, struct prim_header* prim) + r300render->vbo_map = pipe_buffer_map(screen, r300render->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); + + return (unsigned char*)r300render->vbo_map + r300render->vbo_offset; +} + +static void r300_swtcl_render_unmap_vertices(struct vbuf_render* render, + ushort min, + ushort max) { - r300_emit_prim(draw, prim, R300_PRIM_TYPE_POINT, 1); + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + struct pipe_screen* screen = r300render->r300->context.screen; + + r300render->vbo_max_used = MAX2(r300render->vbo_max_used, + r300render->vertex_size * (max + 1)); + + pipe_buffer_unmap(screen, r300render->vbo); } -static void r300_emit_line(struct draw_stage* draw, struct prim_header* prim) +static void r300_swtcl_render_release_vertices(struct vbuf_render* render) { - r300_emit_prim(draw, prim, R300_PRIM_TYPE_LINE, 2); + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + + pipe_buffer_reference(&r300render->vbo, NULL); } -static void r300_emit_tri(struct draw_stage* draw, struct prim_header* prim) +static boolean r300_swtcl_render_set_primitive(struct vbuf_render* render, + unsigned prim) { - r300_emit_prim(draw, prim, R300_PRIM_TYPE_TRI_LIST, 3); + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + r300render->prim = prim; + + switch (prim) { + case PIPE_PRIM_POINTS: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_POINTS; + break; + case PIPE_PRIM_LINES: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_LINES; + break; + case PIPE_PRIM_LINE_LOOP: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_LINE_LOOP; + break; + case PIPE_PRIM_LINE_STRIP: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_LINE_STRIP; + break; + case PIPE_PRIM_TRIANGLES: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_TRIANGLES; + break; + case PIPE_PRIM_TRIANGLE_STRIP: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP; + break; + case PIPE_PRIM_TRIANGLE_FAN: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN; + break; + case PIPE_PRIM_QUADS: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_QUADS; + break; + case PIPE_PRIM_QUAD_STRIP: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_QUAD_STRIP; + break; + case PIPE_PRIM_POLYGON: + r300render->hwprim = R300_VAP_VF_CNTL__PRIM_POLYGON; + break; + default: + return FALSE; + break; + } + + return TRUE; } -static void r300_swtcl_flush(struct draw_stage* draw, unsigned flags) +static void prepare_render(struct r300_swtcl_render* render, unsigned count) { + struct r300_context* r300 = render->r300; + int i; + + CS_LOCALS(r300); + + /* Make sure that all possible state is emitted. */ + r300_emit_dirty_state(r300); + + debug_printf("r300: Preparing vertex buffer %p for render, " + "vertex size %d, vertex count %d\n", render->vbo, + r300->vertex_info.vinfo.size, count); + /* Set the pointer to our vertex buffer. The emitted values are this: + * PACKET3 [3D_LOAD_VBPNTR] + * COUNT [1] + * FORMAT [size | stride << 8] + * OFFSET [0] + * VBPNTR [relocated BO] + */ + BEGIN_CS(7); + OUT_CS(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 3)); + OUT_CS(1); + OUT_CS(r300->vertex_info.vinfo.size | + (r300->vertex_info.vinfo.size << 8)); + OUT_CS(render->vbo_offset); + OUT_CS_RELOC(render->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_CS; } -static void r300_reset_stipple(struct draw_stage* draw) +static void r300_swtcl_render_draw_arrays(struct vbuf_render* render, + unsigned start, + unsigned count) { - /* XXX */ + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + struct r300_context* r300 = r300render->r300; + struct pipe_screen* screen = r300->context.screen; + + CS_LOCALS(r300); + + r300render->vbo_offset = start; + + prepare_render(r300render, count); + + debug_printf("r300: Doing vbuf render, count %d\n", count); + + BEGIN_CS(2); + OUT_CS(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0)); + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) | + r300render->hwprim); + END_CS; +} + +static void r300_swtcl_render_draw(struct vbuf_render* render, + const ushort* indices, + uint count) +{ + struct r300_swtcl_render* r300render = r300_swtcl_render(render); + struct r300_context* r300 = r300render->r300; + struct pipe_screen* screen = r300->context.screen; + struct pipe_buffer* index_buffer; + void* index_map; + + CS_LOCALS(r300); + + count /= 4; + + prepare_render(r300render, count); + + /* Send our indices into an index buffer. */ + index_buffer = pipe_buffer_create(screen, 64, PIPE_BUFFER_USAGE_VERTEX, + count * 4); + if (!index_buffer) { + return; + } + + index_map = pipe_buffer_map(screen, index_buffer, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(index_map, indices, count * 4); + pipe_buffer_unmap(screen, index_buffer); + + debug_printf("r300: Doing indexbuf render, count %d\n", count); +#if 0 + BEGIN_CS(5); + OUT_CS(CP_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0)); + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | + r300render->hwprim | R300_VAP_VF_CNTL__INDEX_SIZE_32bit); + + OUT_CS(CP_PACKET3(R300_PACKET3_INDX_BUFFER, 2)); + OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2)); + OUT_CS_RELOC(index_buffer, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_CS; +#endif +} + +static void r300_swtcl_render_destroy(struct vbuf_render* render) +{ + FREE(render); } -static void r300_swtcl_destroy(struct draw_stage* draw) +static struct vbuf_render* r300_swtcl_render_create(struct r300_context* r300) { - FREE(draw); + struct r300_swtcl_render* r300render = CALLOC_STRUCT(r300_swtcl_render); + struct pipe_screen* screen = r300->context.screen; + + r300render->r300 = r300; + + /* XXX find real numbers plz */ + r300render->base.max_vertex_buffer_bytes = 128 * 1024; + r300render->base.max_indices = 16 * 1024; + + r300render->base.get_vertex_info = r300_swtcl_render_get_vertex_info; + r300render->base.allocate_vertices = r300_swtcl_render_allocate_vertices; + r300render->base.map_vertices = r300_swtcl_render_map_vertices; + r300render->base.unmap_vertices = r300_swtcl_render_unmap_vertices; + r300render->base.set_primitive = r300_swtcl_render_set_primitive; + r300render->base.draw = r300_swtcl_render_draw; + r300render->base.draw_arrays = r300_swtcl_render_draw_arrays; + r300render->base.release_vertices = r300_swtcl_render_release_vertices; + r300render->base.destroy = r300_swtcl_render_destroy; + + /* XXX bonghits ahead + r300render->vbo_alloc_size = 128 * 4096; + r300render->vbo_size = r300render->vbo_alloc_size; + r300render->vbo_offset = 0; + r300render->vbo = pipe_buffer_create(screen, + 64, + PIPE_BUFFER_USAGE_VERTEX, + r300render->vbo_size); + r300render->vbo_map = pipe_buffer_map(screen, + r300render->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); + pipe_buffer_unmap(screen, r300render->vbo); */ + + return &r300render->base; } struct draw_stage* r300_draw_swtcl_stage(struct r300_context* r300) { - struct swtcl_stage* swtcl = CALLOC_STRUCT(swtcl_stage); + struct vbuf_render* render; + struct draw_stage* stage; + + render = r300_swtcl_render_create(r300); + + if (!render) { + return NULL; + } + + stage = draw_vbuf_stage(r300->draw, render); + + if (!stage) { + render->destroy(render); + return NULL; + } - swtcl->r300 = r300; - swtcl->draw.point = r300_emit_point; - swtcl->draw.line = r300_emit_line; - swtcl->draw.tri = r300_emit_tri; - swtcl->draw.flush = r300_swtcl_flush; - swtcl->draw.reset_stipple_counter = r300_reset_stipple; - swtcl->draw.destroy = r300_swtcl_destroy; + draw_set_render(r300->draw, render); - return &swtcl->draw; + return stage; } diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index ff812c09f8..6cdea3d285 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -27,12 +27,37 @@ static int minify(int i) return MAX2(1, i >> 1); } +static void r300_setup_texture_state(struct r300_texture* tex, + unsigned width, + unsigned height, + unsigned pitch) +{ + struct r300_texture_state* state = &tex->state; + + state->format0 = R300_TX_WIDTH((width - 1) & 0x7ff) | + R300_TX_HEIGHT((height - 1) & 0x7ff) | R300_TX_PITCH_EN; + + /* XXX */ + state->format1 = R300_TX_FORMAT_A8R8G8B8; + + state->format2 = pitch - 1; + + /* XXX + if (width > 2048) { + state->pitch |= R300_TXWIDTH_11; + } + if (height > 2048) { + state->pitch |= R300_TXHEIGHT_11; + } */ +} + static void r300_setup_miptree(struct r300_texture* tex) { struct pipe_texture* base = &tex->tex; int stride, size, offset; + int i; - for (int i = 0; i <= base->last_level; i++) { + for (i = 0; i <= base->last_level; i++) { if (i > 0) { base->width[i] = minify(base->width[i-1]); base->height[i] = minify(base->height[i-1]); @@ -43,13 +68,16 @@ static void r300_setup_miptree(struct r300_texture* tex) base->nblocksy[i] = pf_get_nblocksy(&base->block, base->width[i]); /* Radeons enjoy things in multiples of 32. */ - /* XXX NPOT -> 64, not 32 */ + /* XXX this can be 32 when POT */ stride = (base->nblocksx[i] * base->block.size + 63) & ~63; size = stride * base->nblocksy[i] * base->depth[i]; - /* XXX 64 for NPOT */ tex->offset[i] = (tex->size + 63) & ~63; tex->size = tex->offset[i] + size; + + if (i == 0) { + tex->stride = stride; + } } } @@ -67,12 +95,16 @@ static struct pipe_texture* } tex->tex = *template; - tex->tex.refcount = 1; + pipe_reference_init(&tex->tex.reference, 1); tex->tex.screen = screen; r300_setup_miptree(tex); - tex->buffer = screen->buffer_create(screen, 63, + /* XXX */ + r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0], + tex->tex.width[0]); + + tex->buffer = screen->buffer_create(screen, 64, PIPE_BUFFER_USAGE_PIXEL, tex->size); @@ -84,24 +116,13 @@ static struct pipe_texture* return (struct pipe_texture*)tex; } -static void r300_texture_release(struct pipe_screen* screen, - struct pipe_texture** texture) +static void r300_texture_destroy(struct pipe_texture* texture) { - if (!*texture) { - return; - } - - (*texture)->refcount--; - - if ((*texture)->refcount <= 0) { - struct r300_texture* tex = (struct r300_texture*)*texture; + struct r300_texture* tex = (struct r300_texture*)texture; - pipe_buffer_reference(screen, &tex->buffer, NULL); + pipe_buffer_reference(&tex->buffer, NULL); - FREE(tex); - } - - *texture = NULL; + FREE(tex); } static struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen, @@ -119,17 +140,11 @@ static struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen, offset = tex->offset[level]; if (surface) { - surface->refcount = 1; + pipe_reference_init(&surface->reference, 1); pipe_texture_reference(&surface->texture, texture); surface->format = texture->format; surface->width = texture->width[level]; surface->height = texture->height[level]; - surface->block = texture->block; - surface->nblocksx = texture->nblocksx[level]; - surface->nblocksy = texture->nblocksy[level]; - /* XXX save the actual stride instead plz kthnxbai */ - surface->stride = - (texture->nblocksx[level] * texture->block.size + 63) & ~63; surface->offset = offset; surface->usage = flags; surface->status = PIPE_SURFACE_STATUS_DEFINED; @@ -138,19 +153,10 @@ static struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen, return surface; } -static void r300_tex_surface_release(struct pipe_screen* screen, - struct pipe_surface** surface) +static void r300_tex_surface_destroy(struct pipe_surface* s) { - struct pipe_surface* s = *surface; - - s->refcount--; - - if (s->refcount <= 0) { - pipe_texture_reference(&s->texture, NULL); - FREE(s); - } - - *surface = NULL; + pipe_texture_reference(&s->texture, NULL); + FREE(s); } static struct pipe_texture* @@ -173,12 +179,15 @@ static struct pipe_texture* } tex->tex = *base; - tex->tex.refcount = 1; + pipe_reference_init(&tex->tex.reference, 1); tex->tex.screen = screen; - /* XXX tex->stride = *stride; */ + tex->stride = *stride; + + r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0], + tex->stride); - pipe_buffer_reference(screen, &tex->buffer, buffer); + pipe_buffer_reference(&tex->buffer, buffer); return (struct pipe_texture*)tex; } @@ -186,8 +195,26 @@ static struct pipe_texture* void r300_init_screen_texture_functions(struct pipe_screen* screen) { screen->texture_create = r300_texture_create; - screen->texture_release = r300_texture_release; + screen->texture_destroy = r300_texture_destroy; screen->get_tex_surface = r300_get_tex_surface; - screen->tex_surface_release = r300_tex_surface_release; + screen->tex_surface_destroy = r300_tex_surface_destroy; screen->texture_blanket = r300_texture_blanket; } + +boolean r300_get_texture_buffer(struct pipe_texture* texture, + struct pipe_buffer** buffer, + unsigned* stride) +{ + struct r300_texture* tex = (struct r300_texture*)texture; + if (!tex) { + return FALSE; + } + + pipe_buffer_reference(buffer, tex->buffer); + + if (stride) { + *stride = tex->stride; + } + + return TRUE; +} diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h index 7964229a94..98fb5c9a08 100644 --- a/src/gallium/drivers/r300/r300_texture.h +++ b/src/gallium/drivers/r300/r300_texture.h @@ -28,7 +28,16 @@ #include "util/u_math.h" #include "r300_context.h" +#include "r300_reg.h" void r300_init_screen_texture_functions(struct pipe_screen* screen); +#ifndef R300_WINSYS_H + +boolean r300_get_texture_buffer(struct pipe_texture* texture, + struct pipe_buffer** buffer, + unsigned* stride); + +#endif /* R300_WINSYS_H */ + #endif /* R300_TEXTURE_H */ diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h index 5a3a212892..baa95282c3 100644 --- a/src/gallium/drivers/r300/r300_winsys.h +++ b/src/gallium/drivers/r300/r300_winsys.h @@ -33,10 +33,16 @@ extern "C" { #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/internal/p_winsys_screen.h" struct radeon_cs; struct r300_winsys { + /* Parent class */ + struct pipe_winsys base; + + /* Opaque Radeon-specific winsys object. */ + void* radeon_winsys; /* PCI ID */ uint32_t pci_id; @@ -47,10 +53,6 @@ struct r300_winsys { /* CS object. This is very much like Intel's batchbuffer. * Fill it full of dwords and relocs and then submit. * Repeat as needed. */ - /* Note: Unlike Mesa's version of this, we don't keep a copy of the CSM - * that was used to create this CS. Is this a good idea? */ - /* Note: The pipe driver doesn't know how to use this. This is purely - * for the winsys. */ struct radeon_cs* cs; /* Check to see if there's room for commands. */ @@ -84,9 +86,12 @@ struct r300_winsys { }; struct pipe_context* r300_create_context(struct pipe_screen* screen, - struct pipe_winsys* winsys, struct r300_winsys* r300_winsys); +boolean r300_get_texture_buffer(struct pipe_texture* texture, + struct pipe_buffer** buffer, + unsigned* stride); + #ifdef __cplusplus } #endif diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index 120bdfd9dd..516e3992fd 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -14,7 +14,7 @@ C_SOURCES = \ sp_draw_arrays.c \ sp_prim_setup.c \ sp_prim_vbuf.c \ - sp_quad.c \ + sp_quad_pipe.c \ sp_quad_alpha_test.c \ sp_quad_blend.c \ sp_quad_colormask.c \ @@ -42,6 +42,3 @@ C_SOURCES = \ sp_surface.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index c1f7daa8ab..f8720638a7 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -17,7 +17,7 @@ softpipe = env.ConvenienceLibrary( 'sp_setup.c', 'sp_quad_alpha_test.c', 'sp_quad_blend.c', - 'sp_quad.c', + 'sp_quad_pipe.c', 'sp_quad_colormask.c', 'sp_quad_coverage.c', 'sp_quad_depth_test.c', diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index c2d882a819..06ace27d14 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -32,7 +32,6 @@ #include "draw/draw_context.h" #include "pipe/p_defines.h" -#include "pipe/p_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "sp_clear.h" @@ -53,15 +52,15 @@ * Map any drawing surfaces which aren't already mapped */ void -softpipe_map_surfaces(struct softpipe_context *sp) +softpipe_map_transfers(struct softpipe_context *sp) { unsigned i; for (i = 0; i < sp->framebuffer.nr_cbufs; i++) { - sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); + sp_tile_cache_map_transfers(sp->cbuf_cache[i]); } - sp_tile_cache_map_surfaces(sp->zsbuf_cache); + sp_tile_cache_map_transfers(sp->zsbuf_cache); } @@ -69,7 +68,7 @@ softpipe_map_surfaces(struct softpipe_context *sp) * Unmap any mapped drawing surfaces */ void -softpipe_unmap_surfaces(struct softpipe_context *sp) +softpipe_unmap_transfers(struct softpipe_context *sp) { uint i; @@ -78,16 +77,15 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) sp_flush_tile_cache(sp, sp->zsbuf_cache); for (i = 0; i < sp->framebuffer.nr_cbufs; i++) { - sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); + sp_tile_cache_unmap_transfers(sp->cbuf_cache[i]); } - sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); + sp_tile_cache_unmap_transfers(sp->zsbuf_cache); } static void softpipe_destroy( struct pipe_context *pipe ) { struct softpipe_context *softpipe = softpipe_context( pipe ); - struct pipe_screen *screen = pipe->screen; uint i; if (softpipe->draw) @@ -116,7 +114,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) for (i = 0; i < Elements(softpipe->constants); i++) { if (softpipe->constants[i].buffer) { - pipe_buffer_reference(screen, &softpipe->constants[i].buffer, NULL); + pipe_buffer_reference(&softpipe->constants[i].buffer, NULL); } } diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index e2451c6ecb..59d6df8f2d 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -32,11 +32,10 @@ #define SP_CONTEXT_H #include "pipe/p_context.h" -#include "pipe/p_defines.h" #include "draw/draw_vertex.h" -#include "sp_quad.h" +#include "sp_quad_pipe.h" #include "sp_tex_sample.h" @@ -51,7 +50,6 @@ */ #define SP_NUM_QUAD_THREADS 1 -struct softpipe_winsys; struct softpipe_vbuf_render; struct draw_context; struct draw_stage; @@ -63,15 +61,15 @@ struct sp_vertex_shader; struct softpipe_context { struct pipe_context pipe; /**< base class */ - /* The most recent drawing state as set by the driver: - */ - const struct pipe_blend_state *blend; + /** Constant state objects */ + const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_alpha_state *depth_stencil; + const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct sp_fragment_shader *fs; const struct sp_vertex_shader *vs; + /** Other rendering state */ struct pipe_blend_color blend_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; @@ -82,23 +80,20 @@ struct softpipe_context { struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; - unsigned dirty; unsigned num_samplers; unsigned num_textures; unsigned num_vertex_elements; unsigned num_vertex_buffers; - boolean no_rast; + unsigned dirty; /**< Mask of SP_NEW_x flags */ /* Counter for occlusion queries. Note this supports overlapping * queries. */ uint64_t occlusion_count; - /* - * Mapped vertex buffers - */ + /** Mapped vertex buffers */ ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS]; /** Mapped constant buffers */ @@ -108,16 +103,11 @@ struct softpipe_context { struct vertex_info vertex_info; struct vertex_info vertex_info_vbuf; + /** Which vertex shader output slot contains point size */ int psize_slot; unsigned reduced_api_prim; /**< PIPE_PRIM_POINTS, _LINES or _TRIANGLES */ -#if 0 - /* Stipple derived state: - */ - ubyte stipple_masks[16][16]; -#endif - /** Derived from scissor and surface bounds: */ struct pipe_scissor_state cliprect; @@ -159,8 +149,9 @@ struct softpipe_context { struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; - int use_sse : 1; - int dump_fs : 1; + unsigned use_sse : 1; + unsigned dump_fs : 1; + unsigned no_rast : 1; }; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 7e3a25e34b..f117096bf7 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -134,7 +134,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe, if (sp->dirty) softpipe_update_derived( sp ); - softpipe_map_surfaces(sp); + softpipe_map_transfers(sp); softpipe_map_constant_buffers(sp); /* diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index c21faf57f3..035f4b963e 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -70,7 +70,7 @@ softpipe_flush( struct pipe_context *pipe, * that's called before swapbuffers because we don't always want * to unmap surfaces when flushing. */ - softpipe_unmap_surfaces(softpipe); + softpipe_unmap_transfers(softpipe); } /* Enable to dump BMPs of the color/depth buffers each frame */ diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 453b0373f0..0c14d92864 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -29,13 +29,12 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_fs.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "util/u_memory.h" -#include "pipe/p_inlines.h" #include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 34adac5226..f33b3e3285 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -37,7 +37,6 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "util/u_memory.h" -#include "pipe/p_inlines.h" #include "tgsi/tgsi_sse2.h" #if 0 diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 9a273c8764..366abe2ed4 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -29,13 +29,12 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_fs.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "util/u_memory.h" -#include "pipe/p_inlines.h" #include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_sse2.h" @@ -146,7 +145,7 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, return NULL; } - shader->base.shader = *templ; + shader->base.shader.tokens = NULL; /* don't hold reference to templ->tokens */ shader->base.prepare = fs_sse_prepare; shader->base.run = fs_sse_run; shader->base.delete = fs_sse_delete; diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h deleted file mode 100644 index 4a42cb3c19..0000000000 --- a/src/gallium/drivers/softpipe/sp_headers.h +++ /dev/null @@ -1,95 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell <keith@tungstengraphics.com> - */ - -#ifndef SP_HEADERS_H -#define SP_HEADERS_H - -#include "pipe/p_state.h" -#include "tgsi/tgsi_exec.h" - -#define PRIM_POINT 1 -#define PRIM_LINE 2 -#define PRIM_TRI 3 - - -/* The rasterizer generates 2x2 quads of fragment and feeds them to - * the current fp_machine (see below). - * Remember that Y=0=top with Y increasing down the window. - */ -#define QUAD_TOP_LEFT 0 -#define QUAD_TOP_RIGHT 1 -#define QUAD_BOTTOM_LEFT 2 -#define QUAD_BOTTOM_RIGHT 3 - -#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) -#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) -#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) -#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) -#define MASK_ALL 0xf - - -/** - * Encodes everything we need to know about a 2x2 pixel block. Uses - * "Channel-Serial" or "SoA" layout. - */ -struct quad_header_input -{ - int x0; - int y0; - float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ - unsigned facing:1; /**< Front (0) or back (1) facing? */ - unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ -}; - -struct quad_header_inout -{ - unsigned mask:4; -}; - -struct quad_header_output -{ - /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */ - float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE]; - float depth[QUAD_SIZE]; -}; - -struct quad_header { - struct quad_header_input input; - struct quad_header_inout inout; - struct quad_header_output output; - - const struct tgsi_interp_coef *coef; - const struct tgsi_interp_coef *posCoef; - - unsigned nr_attrs; -}; - -#endif /* SP_HEADERS_H */ - diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 9cd5784e5b..eef6e5806c 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -60,6 +60,8 @@ struct softpipe_vbuf_render struct softpipe_context *softpipe; uint prim; uint vertex_size; + uint nr_vertices; + uint vertex_buffer_size; void *vertex_buffer; }; @@ -80,26 +82,76 @@ sp_vbuf_get_vertex_info(struct vbuf_render *vbr) } -static void * +static boolean sp_vbuf_allocate_vertices(struct vbuf_render *vbr, - ushort vertex_size, ushort nr_vertices) + ushort vertex_size, ushort nr_vertices) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - assert(!cvbr->vertex_buffer); - cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); + unsigned size = vertex_size * nr_vertices; + + if (cvbr->vertex_buffer_size < size) { + align_free(cvbr->vertex_buffer); + cvbr->vertex_buffer = align_malloc(size, 16); + cvbr->vertex_buffer_size = size; + } + cvbr->vertex_size = vertex_size; - return cvbr->vertex_buffer; + cvbr->nr_vertices = nr_vertices; + + return cvbr->vertex_buffer != NULL; } - static void -sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, - unsigned vertex_size, unsigned vertices_used) +sp_vbuf_release_vertices(struct vbuf_render *vbr) +{ +#if 0 + { + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + const struct vertex_info *info = + softpipe_get_vbuf_vertex_info(cvbr->softpipe); + const float *vtx = (const float *) cvbr->vertex_buffer; + uint i, j; + debug_printf("%s (vtx_size = %u, vtx_used = %u)\n", + __FUNCTION__, cvbr->vertex_size, cvbr->nr_vertices); + for (i = 0; i < cvbr->nr_vertices; i++) { + for (j = 0; j < info->num_attribs; j++) { + uint k; + switch (info->attrib[j].emit) { + case EMIT_4F: k = 4; break; + case EMIT_3F: k = 3; break; + case EMIT_2F: k = 2; break; + case EMIT_1F: k = 1; break; + default: assert(0); + } + debug_printf("Vert %u attr %u: ", i, j); + while (k-- > 0) { + debug_printf("%g ", vtx[0]); + vtx++; + } + debug_printf("\n"); + } + } + } +#endif + + /* keep the old allocation for next time */ +} + +static void * +sp_vbuf_map_vertices(struct vbuf_render *vbr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - align_free(vertices); - assert(vertices == cvbr->vertex_buffer); - cvbr->vertex_buffer = NULL; + return cvbr->vertex_buffer; +} + +static void +sp_vbuf_unmap_vertices(struct vbuf_render *vbr, + ushort min_index, + ushort max_index ) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + assert( cvbr->vertex_buffer_size >= (max_index+1) * cvbr->vertex_size ); + /* do nothing */ } @@ -115,8 +167,6 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) setup_prepare( setup_ctx ); - - cvbr->prim = prim; return TRUE; @@ -394,6 +444,8 @@ sp_init_vbuf(struct softpipe_context *sp) sp->vbuf_render->base.get_vertex_info = sp_vbuf_get_vertex_info; sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices; + sp->vbuf_render->base.map_vertices = sp_vbuf_map_vertices; + sp->vbuf_render->base.unmap_vertices = sp_vbuf_unmap_vertices; sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive; sp->vbuf_render->base.draw = sp_vbuf_draw; sp->vbuf_render->base.draw_arrays = sp_vbuf_draw_arrays; diff --git a/src/gallium/drivers/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h index 08513cb95f..bd6c6cb912 100644 --- a/src/gallium/drivers/softpipe/sp_quad.h +++ b/src/gallium/drivers/softpipe/sp_quad.h @@ -31,39 +31,76 @@ #ifndef SP_QUAD_H #define SP_QUAD_H +#include "pipe/p_state.h" +#include "tgsi/tgsi_exec.h" -struct softpipe_context; -struct quad_header; +#define QUAD_PRIM_POINT 1 +#define QUAD_PRIM_LINE 2 +#define QUAD_PRIM_TRI 3 -struct quad_stage { - struct softpipe_context *softpipe; - struct quad_stage *next; +/* The rasterizer generates 2x2 quads of fragment and feeds them to + * the current fp_machine (see below). + * Remember that Y=0=top with Y increasing down the window. + */ +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 + +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) +#define MASK_ALL 0xf + + +/** + * Quad stage inputs (pos, coverage, front/back face, etc) + */ +struct quad_header_input +{ + int x0, y0; /**< quad window pos, always even */ + float coverage[QUAD_SIZE]; /**< fragment coverage for antialiasing */ + unsigned facing:1; /**< Front (0) or back (1) facing? */ + unsigned prim:2; /**< QUAD_PRIM_POINT, LINE, TRI */ +}; - void (*begin)(struct quad_stage *qs); - /** the stage action */ - void (*run)(struct quad_stage *qs, struct quad_header *quad); +/** + * Quad stage inputs/outputs. + */ +struct quad_header_inout +{ + unsigned mask:4; +}; + - void (*destroy)(struct quad_stage *qs); +/** + * Quad stage outputs (color & depth). + */ +struct quad_header_output +{ + /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */ + float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE]; + float depth[QUAD_SIZE]; }; -struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); +/** + * Encodes everything we need to know about a 2x2 pixel block. Uses + * "Channel-Serial" or "SoA" layout. + */ +struct quad_header { + struct quad_header_input input; + struct quad_header_inout inout; + struct quad_header_output output; -void sp_build_quad_pipeline(struct softpipe_context *sp); + const struct tgsi_interp_coef *coef; + const struct tgsi_interp_coef *posCoef; -void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); + unsigned nr_attrs; +}; #endif /* SP_QUAD_H */ diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c index 85c9f037a3..0845bae0e6 100644 --- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -4,8 +4,8 @@ */ #include "sp_context.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" #include "pipe/p_defines.h" #include "util/u_memory.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index fb1d430a4f..e134e44337 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -34,10 +34,10 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "sp_surface.h" #include "sp_tile_cache.h" -#include "sp_quad.h" +#include "sp_quad_pipe.h" #define VEC4_COPY(DST, SRC) \ diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c index d7d6a6974d..953d8516b9 100644 --- a/src/gallium/drivers/softpipe/sp_quad_bufloop.c +++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c @@ -1,9 +1,9 @@ #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" /** diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c index 563c2fc739..dc90e5d5e9 100644 --- a/src/gallium/drivers/softpipe/sp_quad_colormask.c +++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c @@ -34,9 +34,9 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" #include "sp_tile_cache.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c index c27fd1482d..4aeee85870 100644 --- a/src/gallium/drivers/softpipe/sp_quad_coverage.c +++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c @@ -35,8 +35,8 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" /** @@ -46,10 +46,11 @@ static void coverage_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; + const uint prim = quad->input.prim; - if ((softpipe->rasterizer->poly_smooth && quad->input.prim == PRIM_TRI) || - (softpipe->rasterizer->line_smooth && quad->input.prim == PRIM_LINE) || - (softpipe->rasterizer->point_smooth && quad->input.prim == PRIM_POINT)) { + if ((softpipe->rasterizer->poly_smooth && prim == QUAD_PRIM_TRI) || + (softpipe->rasterizer->line_smooth && prim == QUAD_PRIM_LINE) || + (softpipe->rasterizer->point_smooth && prim == QUAD_PRIM_POINT)) { uint cbuf; /* loop over colorbuffer outputs */ diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index 523bd3e080..d463930bae 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -32,9 +32,9 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" #include "sp_tile_cache.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c index 6e2dde304e..496fd39ed1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_earlyz.c +++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c @@ -31,8 +31,8 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" /** diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 5dacbbe55f..adca5df73d 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -43,8 +43,8 @@ #include "sp_context.h" #include "sp_state.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" #include "sp_texture.h" #include "sp_tex_sample.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c index 169bd82876..dfa7ff3b1d 100644 --- a/src/gallium/drivers/softpipe/sp_quad_occlusion.c +++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c @@ -35,9 +35,9 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" static unsigned count_bits( unsigned val ) { diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c index a37c8b4c39..92d5f9f3c1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_output.c +++ b/src/gallium/drivers/softpipe/sp_quad_output.c @@ -27,9 +27,9 @@ #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" #include "sp_tile_cache.h" diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad_pipe.c index 892ef87ee9..892ef87ee9 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad_pipe.c diff --git a/src/gallium/drivers/softpipe/sp_quad_pipe.h b/src/gallium/drivers/softpipe/sp_quad_pipe.h new file mode 100644 index 0000000000..0e40586ffc --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_pipe.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell <keith@tungstengraphics.com> + */ + +#ifndef SP_QUAD_PIPE_H +#define SP_QUAD_PIPE_H + + +struct softpipe_context; +struct quad_header; + + +/** + * Fragment processing is performed on 2x2 blocks of pixels called "quads". + * Quad processing is performed with a pipeline of stages represented by + * this type. + */ +struct quad_stage { + struct softpipe_context *softpipe; + + struct quad_stage *next; + + void (*begin)(struct quad_stage *qs); + + /** the stage action */ + void (*run)(struct quad_stage *qs, struct quad_header *quad); + + void (*destroy)(struct quad_stage *qs); +}; + + +struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); + +void sp_build_quad_pipeline(struct softpipe_context *sp); + +void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); + +#endif /* SP_QUAD_PIPE_H */ diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c index 7495515764..5e9d447737 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stencil.c +++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c @@ -5,10 +5,10 @@ #include "sp_context.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "sp_surface.h" #include "sp_tile_cache.h" -#include "sp_quad.h" +#include "sp_quad_pipe.h" #include "pipe/p_defines.h" #include "util/u_memory.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c index ccf37f6be5..05e862f097 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stipple.c +++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c @@ -4,8 +4,8 @@ */ #include "sp_context.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" #include "pipe/p_defines.h" #include "util/u_memory.h" @@ -19,11 +19,13 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) static const uint bit31 = 1 << 31; static const uint bit30 = 1 << 30; - if (quad->input.prim == PRIM_TRI) { + if (quad->input.prim == QUAD_PRIM_TRI) { struct softpipe_context *softpipe = qs->softpipe; /* need to invert Y to index into OpenGL's stipple pattern */ int y0, y1; uint stipple0, stipple1; + const int col0 = quad->input.x0 % 32; + if (softpipe->rasterizer->origin_lower_left) { y0 = softpipe->framebuffer.height - 1 - quad->input.y0; y1 = y0 - 1; @@ -32,12 +34,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) y0 = quad->input.y0; y1 = y0 + 1; } + stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; -#if 1 - { - const int col0 = quad->input.x0 % 32; + /* turn off quad mask bits that fail the stipple test */ if ((stipple0 & (bit31 >> col0)) == 0) quad->inout.mask &= ~MASK_TOP_LEFT; @@ -49,19 +50,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) if ((stipple1 & (bit30 >> col0)) == 0) quad->inout.mask &= ~MASK_BOTTOM_RIGHT; - } -#else - /* We'd like to use this code, but we'd need to redefine - * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0), - * and similarly for the BOTTOM bits. But that may have undesirable - * side effects elsewhere. - */ - const int col0 = 30 - (quad->input.x0 % 32); - quad->inout.mask &= (((stipple0 >> col0) & 0x3) | - (((stipple1 >> col0) & 0x3) << 2)); -#endif - if (!quad->inout.mask) + + if (!quad->inout.mask) { + /* all fragments failed stipple test, end of quad pipeline */ return; + } } qs->next->run(qs->next, quad); diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c index b0d8e01426..93dab236d6 100644 --- a/src/gallium/drivers/softpipe/sp_query.c +++ b/src/gallium/drivers/softpipe/sp_query.c @@ -31,7 +31,6 @@ #include "draw/draw_context.h" #include "pipe/p_defines.h" -#include "pipe/p_inlines.h" #include "util/u_memory.h" #include "sp_context.h" #include "sp_query.h" diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index b1adb9cb7a..0925653b5d 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -32,13 +32,12 @@ * \author Brian Paul */ -#include "sp_setup.h" - #include "sp_context.h" -#include "sp_headers.h" +#include "sp_prim_setup.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" +#include "sp_setup.h" #include "sp_state.h" -#include "sp_prim_setup.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vertex.h" @@ -265,17 +264,20 @@ is_inf_or_nan(float x) } -static boolean cull_tri( struct setup_context *setup, - float det ) +/** + * Do triangle cull test using tri determinant (sign indicates orientation) + * \return true if triangle is to be culled. + */ +static INLINE boolean +cull_tri(const struct setup_context *setup, float det) { - if (det != 0) - { + if (det != 0) { /* if (det < 0 then Z points toward camera and triangle is * counter-clockwise winding. */ unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; - - if ((winding & setup->winding) == 0) + + if ((winding & setup->winding) == 0) return FALSE; } @@ -968,7 +970,7 @@ void setup_tri( struct setup_context *setup, setup_tri_coefficients( setup ); setup_tri_edges( setup ); - setup->quad.input.prim = PRIM_TRI; + setup->quad.input.prim = QUAD_PRIM_TRI; setup->span.y = 0; setup->span.y_flags = 0; @@ -1009,7 +1011,7 @@ void setup_tri( struct setup_context *setup, * for a line. */ static void -line_linear_coeff(struct setup_context *setup, +line_linear_coeff(const struct setup_context *setup, struct tgsi_interp_coef *coef, uint vertSlot, uint i) { @@ -1029,9 +1031,9 @@ line_linear_coeff(struct setup_context *setup, * for a line. */ static void -line_persp_coeff(struct setup_context *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) +line_persp_coeff(const struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { /* XXX double-check/verify this arithmetic */ const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3]; @@ -1206,7 +1208,7 @@ setup_line(struct setup_context *setup, setup->quad.input.x0 = setup->quad.input.y0 = -1; setup->quad.inout.mask = 0x0; - setup->quad.input.prim = PRIM_LINE; + setup->quad.input.prim = QUAD_PRIM_LINE; /* XXX temporary: set coverage to 1.0 so the line appears * if AA mode happens to be enabled. */ @@ -1266,7 +1268,7 @@ setup_line(struct setup_context *setup, static void -point_persp_coeff(struct setup_context *setup, +point_persp_coeff(const struct setup_context *setup, const float (*vert)[4], struct tgsi_interp_coef *coef, uint vertSlot, uint i) @@ -1361,7 +1363,7 @@ setup_point( struct setup_context *setup, } } - setup->quad.input.prim = PRIM_POINT; + setup->quad.input.prim = QUAD_PRIM_POINT; if (halfSize <= 0.5 && !round) { /* special case for 1-pixel points */ diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 3eff41ffa5..9776e978e3 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -85,7 +85,7 @@ struct sp_fragment_shader { /** Subclass of pipe_shader_state */ struct sp_vertex_shader { - struct pipe_shader_state shader; /* Note: this field not actually used */ + struct pipe_shader_state shader; struct draw_vertex_shader *draw_data; }; @@ -184,10 +184,10 @@ softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); void -softpipe_map_surfaces(struct softpipe_context *sp); +softpipe_map_transfers(struct softpipe_context *sp); void -softpipe_unmap_surfaces(struct softpipe_context *sp); +softpipe_unmap_transfers(struct softpipe_context *sp); void softpipe_map_texture_surfaces(struct softpipe_context *sp); diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 4d01a9dbe1..4330c20393 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -31,12 +31,12 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" -#include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "tgsi/tgsi_dump.h" #include "tgsi/tgsi_scan.h" +#include "tgsi/tgsi_parse.h" void * @@ -98,17 +98,28 @@ softpipe_create_vs_state(struct pipe_context *pipe, struct sp_vertex_shader *state; state = CALLOC_STRUCT(sp_vertex_shader); - if (state == NULL ) { - return NULL; - } + if (state == NULL ) + goto fail; + + /* copy shader tokens, the ones passed in will go away. + */ + state->shader.tokens = tgsi_dup_tokens(templ->tokens); + if (state->shader.tokens == NULL) + goto fail; state->draw_data = draw_create_vertex_shader(softpipe->draw, templ); - if (state->draw_data == NULL) { - FREE( state ); - return NULL; - } + if (state->draw_data == NULL) + goto fail; return state; + +fail: + if (state) { + FREE( (void *)state->shader.tokens ); + FREE( state->draw_data ); + FREE( state ); + } + return NULL; } @@ -146,14 +157,12 @@ softpipe_set_constant_buffer(struct pipe_context *pipe, const struct pipe_constant_buffer *buf) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct pipe_screen *screen = pipe->screen; assert(shader < PIPE_SHADER_TYPES); assert(index == 0); /* note: reference counting */ - pipe_buffer_reference(screen, - &softpipe->constants[shader].buffer, + pipe_buffer_reference(&softpipe->constants[shader].buffer, buf ? buf->buffer : NULL); softpipe->dirty |= SP_NEW_CONSTANTS; diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 99a28c0d7e..cb517b02e4 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -30,7 +30,6 @@ */ #include "util/u_memory.h" -#include "pipe/p_inlines.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index 1493c65884..7c06d864a7 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -27,7 +27,6 @@ /* Authors: Keith Whitwell <keith@tungstengraphics.com> */ -#include "pipe/p_inlines.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 6ade732698..ef04843f17 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -29,10 +29,21 @@ #include "sp_context.h" +static void +sp_surface_copy(struct pipe_context *pipe, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + util_surface_copy(pipe, FALSE, + dest, destx, desty, + src, srcx, srcy, + width, height); +} void sp_init_surface_functions(struct softpipe_context *sp) { - sp->pipe.surface_copy = util_surface_copy; + sp->pipe.surface_copy = sp_surface_copy; sp->pipe.surface_fill = util_surface_fill; } diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 32aa5025e4..adbd0cb7f0 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -34,7 +34,7 @@ */ #include "sp_context.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "sp_surface.h" #include "sp_texture.h" #include "sp_tex_sample.h" diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 3eed0d0d29..48b2c22af4 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -59,7 +59,6 @@ static boolean softpipe_texture_layout(struct pipe_screen *screen, struct softpipe_texture * spt) { - struct pipe_winsys *ws = screen->winsys; struct pipe_texture *pt = &spt->base; unsigned level; unsigned width = pt->width[0]; @@ -87,9 +86,9 @@ softpipe_texture_layout(struct pipe_screen *screen, depth = minify(depth); } - spt->buffer = ws->buffer_create(ws, 32, - PIPE_BUFFER_USAGE_PIXEL, - buffer_size); + spt->buffer = screen->buffer_create(screen, 32, + PIPE_BUFFER_USAGE_PIXEL, + buffer_size); return spt->buffer != NULL; } @@ -98,19 +97,18 @@ static boolean softpipe_displaytarget_layout(struct pipe_screen *screen, struct softpipe_texture * spt) { - struct pipe_winsys *ws = screen->winsys; unsigned usage = (PIPE_BUFFER_USAGE_CPU_READ_WRITE | PIPE_BUFFER_USAGE_GPU_READ_WRITE); spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); - spt->buffer = ws->surface_buffer_create( ws, - spt->base.width[0], - spt->base.height[0], - spt->base.format, - usage, - &spt->stride[0]); + spt->buffer = screen->surface_buffer_create( screen, + spt->base.width[0], + spt->base.height[0], + spt->base.format, + usage, + &spt->stride[0]); return spt->buffer != NULL; } @@ -128,7 +126,7 @@ softpipe_texture_create(struct pipe_screen *screen, return NULL; spt->base = *templat; - spt->base.refcount = 1; + pipe_reference_init(&spt->base.reference, 1); spt->base.screen = screen; if (spt->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { @@ -140,7 +138,7 @@ softpipe_texture_create(struct pipe_screen *screen, goto fail; } - assert(spt->base.refcount == 1); + assert(p_atomic_read(&spt->base.reference.count) == 1); return &spt->base; fail: @@ -170,32 +168,25 @@ softpipe_texture_blanket(struct pipe_screen * screen, return NULL; spt->base = *base; - spt->base.refcount = 1; + pipe_reference_init(&spt->base.reference, 1); spt->base.screen = screen; spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); spt->stride[0] = stride[0]; - pipe_buffer_reference(screen, &spt->buffer, buffer); + pipe_buffer_reference(&spt->buffer, buffer); return &spt->base; } static void -softpipe_texture_release(struct pipe_screen *screen, - struct pipe_texture **pt) +softpipe_texture_destroy(struct pipe_texture *pt) { - if (!*pt) - return; - - if (--(*pt)->refcount <= 0) { - struct softpipe_texture *spt = softpipe_texture(*pt); + struct softpipe_texture *spt = softpipe_texture(pt); - pipe_buffer_reference(screen, &spt->buffer, NULL); - FREE(spt); - } - *pt = NULL; + pipe_buffer_reference(&spt->buffer, NULL); + FREE(spt); } @@ -212,15 +203,11 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps = CALLOC_STRUCT(pipe_surface); if (ps) { - ps->refcount = 1; + pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->block = pt->block; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = spt->stride[level]; ps->offset = spt->level_offset[level]; ps->usage = usage; @@ -247,10 +234,11 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps->level = level; ps->zslice = zslice; - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - ps->nblocksy * - ps->stride; + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset += face * pt->nblocksy[level] * spt->stride[level]; + } + else if (pt->target == PIPE_TEXTURE_3D) { + ps->offset += zslice * pt->nblocksy[level] * spt->stride[level]; } else { assert(face == 0); @@ -262,38 +250,99 @@ softpipe_get_tex_surface(struct pipe_screen *screen, static void -softpipe_tex_surface_release(struct pipe_screen *screen, - struct pipe_surface **s) +softpipe_tex_surface_destroy(struct pipe_surface *surf) { - struct pipe_surface *surf = *s; /* Effectively do the texture_update work here - if texture images * needed post-processing to put them into hardware layout, this is * where it would happen. For softpipe, nothing to do. */ assert(surf->texture); - if (--surf->refcount == 0) { - pipe_texture_reference(&surf->texture, NULL); - FREE(surf); + pipe_texture_reference(&surf->texture, NULL); + FREE(surf); +} + + +static struct pipe_transfer * +softpipe_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct softpipe_texture *sptex = softpipe_texture(texture); + struct softpipe_transfer *spt; + + assert(texture); + assert(level <= texture->last_level); + + spt = CALLOC_STRUCT(softpipe_transfer); + if (spt) { + struct pipe_transfer *pt = &spt->base; + pipe_texture_reference(&pt->texture, texture); + pt->format = texture->format; + pt->block = texture->block; + pt->x = x; + pt->y = y; + pt->width = w; + pt->height = h; + pt->nblocksx = texture->nblocksx[level]; + pt->nblocksy = texture->nblocksy[level]; + pt->stride = sptex->stride[level]; + pt->usage = usage; + pt->face = face; + pt->level = level; + pt->zslice = zslice; + + spt->offset = sptex->level_offset[level]; + + if (texture->target == PIPE_TEXTURE_CUBE) { + spt->offset += face * pt->nblocksy * pt->stride; + } + else if (texture->target == PIPE_TEXTURE_3D) { + spt->offset += zslice * pt->nblocksy * pt->stride; + } + else { + assert(face == 0); + assert(zslice == 0); + } + return pt; } - *s = NULL; + return NULL; +} + + +static void +softpipe_tex_transfer_destroy(struct pipe_transfer *transfer) +{ + /* Effectively do the texture_update work here - if texture images + * needed post-processing to put them into hardware layout, this is + * where it would happen. For softpipe, nothing to do. + */ + assert (transfer->texture); + pipe_texture_reference(&transfer->texture, NULL); + FREE(transfer); } static void * -softpipe_surface_map( struct pipe_screen *screen, - struct pipe_surface *surface, - unsigned flags ) +softpipe_transfer_map( struct pipe_screen *screen, + struct pipe_transfer *transfer ) { ubyte *map; struct softpipe_texture *spt; + unsigned flags = 0; - if (flags & ~surface->usage) { - assert(0); - return NULL; + assert(transfer->texture); + spt = softpipe_texture(transfer->texture); + + if (transfer->usage != PIPE_TRANSFER_READ) { + flags |= PIPE_BUFFER_USAGE_CPU_WRITE; + } + + if (transfer->usage != PIPE_TRANSFER_WRITE) { + flags |= PIPE_BUFFER_USAGE_CPU_READ; } - assert(surface->texture); - spt = softpipe_texture(surface->texture); map = pipe_buffer_map(screen, spt->buffer, flags); if (map == NULL) return NULL; @@ -301,8 +350,7 @@ softpipe_surface_map( struct pipe_screen *screen, /* May want to different things here depending on read/write nature * of the map: */ - if (surface->texture && - (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) + if (transfer->texture && transfer->usage != PIPE_TRANSFER_READ) { /* Do something to notify sharing contexts of a texture change. * In softpipe, that would mean flushing the texture cache. @@ -310,18 +358,20 @@ softpipe_surface_map( struct pipe_screen *screen, softpipe_screen(screen)->timestamp++; } - return map + surface->offset; + return map + softpipe_transfer(transfer)->offset + + transfer->y / transfer->block.height * transfer->stride + + transfer->x / transfer->block.width * transfer->block.size; } static void -softpipe_surface_unmap(struct pipe_screen *screen, - struct pipe_surface *surface) +softpipe_transfer_unmap(struct pipe_screen *screen, + struct pipe_transfer *transfer) { struct softpipe_texture *spt; - assert(surface->texture); - spt = softpipe_texture(surface->texture); + assert(transfer->texture); + spt = softpipe_texture(transfer->texture); pipe_buffer_unmap( screen, spt->buffer ); } @@ -338,11 +388,13 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen) { screen->texture_create = softpipe_texture_create; screen->texture_blanket = softpipe_texture_blanket; - screen->texture_release = softpipe_texture_release; + screen->texture_destroy = softpipe_texture_destroy; screen->get_tex_surface = softpipe_get_tex_surface; - screen->tex_surface_release = softpipe_tex_surface_release; + screen->tex_surface_destroy = softpipe_tex_surface_destroy; - screen->surface_map = softpipe_surface_map; - screen->surface_unmap = softpipe_surface_unmap; + screen->get_tex_transfer = softpipe_get_tex_transfer; + screen->tex_transfer_destroy = softpipe_tex_transfer_destroy; + screen->transfer_map = softpipe_transfer_map; + screen->transfer_unmap = softpipe_transfer_unmap; } diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index c1636920cd..893aa7d11d 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -51,14 +51,27 @@ struct softpipe_texture boolean modified; }; +struct softpipe_transfer +{ + struct pipe_transfer base; + + unsigned long offset; +}; + -/** cast wrapper */ +/** cast wrappers */ static INLINE struct softpipe_texture * softpipe_texture(struct pipe_texture *pt) { return (struct softpipe_texture *) pt; } +static INLINE struct softpipe_transfer * +softpipe_transfer(struct pipe_transfer *pt) +{ + return (struct softpipe_transfer *) pt; +} + extern void softpipe_init_texture_funcs( struct softpipe_context *softpipe ); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index ab76009375..69292753f1 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -26,7 +26,7 @@ **************************************************************************/ /** - * Framebuffer/surface tile caching. + * Texture tile caching. * * Author: * Brian Paul @@ -40,7 +40,7 @@ #include "sp_texture.h" #include "sp_tile_cache.h" -#define NUM_ENTRIES 32 +#define NUM_ENTRIES 50 /** XXX move these */ @@ -52,7 +52,8 @@ struct softpipe_tile_cache { struct pipe_screen *screen; struct pipe_surface *surface; /**< the surface we're caching */ - void *surface_map; + struct pipe_transfer *transfer; + void *transfer_map; struct pipe_texture *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; @@ -60,8 +61,8 @@ struct softpipe_tile_cache uint clear_val; boolean depth_stencil; /** Is the surface a depth/stencil format? */ - struct pipe_surface *tex_surf; - void *tex_surf_map; + struct pipe_transfer *tex_trans; + void *tex_trans_map; int tex_face, tex_level, tex_z; struct softpipe_cached_tile tile; /**< scratch tile for clears */ @@ -131,16 +132,19 @@ sp_create_tile_cache( struct pipe_screen *screen ) void sp_destroy_tile_cache(struct softpipe_tile_cache *tc) { + struct pipe_screen *screen; uint pos; for (pos = 0; pos < NUM_ENTRIES; pos++) { /*assert(tc->entries[pos].x < 0);*/ } - if (tc->surface) { - pipe_surface_reference(&tc->surface, NULL); + if (tc->transfer) { + screen = tc->transfer->texture->screen; + screen->tex_transfer_destroy(tc->transfer); } - if (tc->tex_surf) { - pipe_surface_reference(&tc->tex_surf, NULL); + if (tc->tex_trans) { + screen = tc->tex_trans->texture->screen; + screen->tex_transfer_destroy(tc->tex_trans); } FREE( tc ); @@ -156,18 +160,30 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, { assert(!tc->texture); - if (tc->surface_map) { - tc->screen->surface_unmap(tc->screen, tc->surface); - tc->surface_map = NULL; + if (tc->transfer) { + struct pipe_screen *screen = tc->transfer->texture->screen; + + if (ps == tc->surface) + return; + + if (tc->transfer_map) { + screen->transfer_unmap(screen, tc->transfer); + tc->transfer_map = NULL; + } + + screen->tex_transfer_destroy(tc->transfer); + tc->transfer = NULL; } - pipe_surface_reference(&tc->surface, ps); + tc->surface = ps; - if (tc->surface) { - if (tc->surface_map) /* XXX: this is always NULL!? */ - tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); + if (ps) { + struct pipe_screen *screen = ps->texture->screen; + + tc->transfer = screen->get_tex_transfer(screen, ps->texture, ps->face, + ps->level, ps->zslice, + PIPE_TRANSFER_READ_WRITE, + 0, 0, ps->width, ps->height); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_X8Z24_UNORM || @@ -181,7 +197,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, /** - * Return the surface being cached. + * Return the transfer being cached. */ struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) @@ -191,30 +207,27 @@ sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) void -sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) +sp_tile_cache_map_transfers(struct softpipe_tile_cache *tc) { - if (tc->surface && !tc->surface_map) - tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, - PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ); - - if (tc->tex_surf && !tc->tex_surf_map) - tc->tex_surf_map = tc->screen->surface_map(tc->screen, tc->tex_surf, - PIPE_BUFFER_USAGE_CPU_READ); + if (tc->transfer && !tc->transfer_map) + tc->transfer_map = tc->screen->transfer_map(tc->screen, tc->transfer); + + if (tc->tex_trans && !tc->tex_trans_map) + tc->tex_trans_map = tc->screen->transfer_map(tc->screen, tc->tex_trans); } void -sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) +sp_tile_cache_unmap_transfers(struct softpipe_tile_cache *tc) { - if (tc->surface_map) { - tc->screen->surface_unmap(tc->screen, tc->surface); - tc->surface_map = NULL; + if (tc->transfer_map) { + tc->screen->transfer_unmap(tc->screen, tc->transfer); + tc->transfer_map = NULL; } - if (tc->tex_surf_map) { - tc->screen->surface_unmap(tc->screen, tc->tex_surf); - tc->tex_surf_map = NULL; + if (tc->tex_trans_map) { + tc->screen->transfer_unmap(tc->screen, tc->tex_trans); + tc->tex_trans_map = NULL; } } @@ -229,15 +242,21 @@ sp_tile_cache_set_texture(struct pipe_context *pipe, { uint i; - assert(!tc->surface); + assert(!tc->transfer); pipe_texture_reference(&tc->texture, texture); - if (tc->tex_surf_map) { - tc->screen->surface_unmap(tc->screen, tc->tex_surf); - tc->tex_surf_map = NULL; + if (tc->tex_trans) { + struct pipe_screen *screen = tc->tex_trans->texture->screen; + + if (tc->tex_trans_map) { + screen->transfer_unmap(screen, tc->tex_trans); + tc->tex_trans_map = NULL; + } + + screen->tex_transfer_destroy(tc->tex_trans); + tc->tex_trans = NULL; } - pipe_surface_reference(&tc->tex_surf, NULL); /* mark as entries as invalid/empty */ /* XXX we should try to avoid this when the teximage hasn't changed */ @@ -328,20 +347,20 @@ static void sp_tile_cache_flush_clear(struct pipe_context *pipe, struct softpipe_tile_cache *tc) { - struct pipe_surface *ps = tc->surface; - const uint w = tc->surface->width; - const uint h = tc->surface->height; + struct pipe_transfer *pt = tc->transfer; + const uint w = tc->transfer->width; + const uint h = tc->transfer->height; uint x, y; uint numCleared = 0; /* clear the scratch tile to the clear value */ - clear_tile(&tc->tile, ps->format, tc->clear_val); + clear_tile(&tc->tile, pt->format, tc->clear_val); /* push the tile to all positions marked as clear */ for (y = 0; y < h; y += TILE_SIZE) { for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { - pipe_put_tile_raw(ps, + pipe_put_tile_raw(pt, x, y, TILE_SIZE, TILE_SIZE, tc->tile.data.color32, 0/*STRIDE*/); @@ -359,28 +378,28 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, /** - * Flush the tile cache: write all dirty tiles back to the surface. + * Flush the tile cache: write all dirty tiles back to the transfer. * any tiles "flagged" as cleared will be "really" cleared. */ void sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc) { - struct pipe_surface *ps = tc->surface; + struct pipe_transfer *pt = tc->transfer; int inuse = 0, pos; - if (ps) { - /* caching a drawing surface */ + if (pt) { + /* caching a drawing transfer */ for (pos = 0; pos < NUM_ENTRIES; pos++) { struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { if (tc->depth_stencil) { - pipe_put_tile_raw(ps, + pipe_put_tile_raw(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_put_tile_rgba(ps, + pipe_put_tile_rgba(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -415,7 +434,7 @@ struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y) { - struct pipe_surface *ps = tc->surface; + struct pipe_transfer *pt = tc->transfer; /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); @@ -431,12 +450,12 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (tile->x != -1) { /* put dirty tile back in framebuffer */ if (tc->depth_stencil) { - pipe_put_tile_raw(ps, + pipe_put_tile_raw(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_put_tile_rgba(ps, + pipe_put_tile_rgba(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -448,22 +467,22 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (is_clear_flag_set(tc->clear_flags, x, y)) { /* don't get tile from framebuffer, just clear it */ if (tc->depth_stencil) { - clear_tile(tile, ps->format, tc->clear_val); + clear_tile(tile, pt->format, tc->clear_val); } else { - clear_tile_rgba(tile, ps->format, tc->clear_color); + clear_tile_rgba(tile, pt->format, tc->clear_color); } clear_clear_flag(tc->clear_flags, x, y); } else { - /* get new tile data from surface */ + /* get new tile data from transfer */ if (tc->depth_stencil) { - pipe_get_tile_raw(ps, + pipe_get_tile_raw(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_get_tile_rgba(ps, + pipe_get_tile_rgba(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -484,7 +503,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, static INLINE uint tex_cache_pos(int x, int y, int z, int face, int level) { - uint entry = x + y * 5 + z * 4 + face + level; + uint entry = x + y * 9 + z * 3 + face + level * 7; return entry % NUM_ENTRIES; } @@ -510,8 +529,12 @@ sp_get_cached_tile_tex(struct softpipe_context *sp, if (tc->texture) { struct softpipe_texture *spt = softpipe_texture(tc->texture); if (spt->modified) { - /* texture was modified, force a cache reload */ - tile->x = -1; + /* texture was modified, invalidate all cached tiles */ + uint p; + for (p = 0; p < NUM_ENTRIES; p++) { + tile = tc->entries + p; + tile->x = -1; + } spt->modified = FALSE; } } @@ -523,28 +546,40 @@ sp_get_cached_tile_tex(struct softpipe_context *sp, level != tile->level) { /* cache miss */ - /* check if we need to get a new surface */ - if (!tc->tex_surf || +#if 0 + printf("miss at %u x=%d y=%d z=%d face=%d level=%d\n", pos, + x/TILE_SIZE, y/TILE_SIZE, z, face, level); +#endif + /* check if we need to get a new transfer */ + if (!tc->tex_trans || tc->tex_face != face || tc->tex_level != level || tc->tex_z != z) { - /* get new surface (view into texture) */ + /* get new transfer (view into texture) */ + + if (tc->tex_trans) { + if (tc->tex_trans_map) { + tc->screen->transfer_unmap(tc->screen, tc->tex_trans); + tc->tex_trans_map = NULL; + } - if (tc->tex_surf_map) - tc->screen->surface_unmap(tc->screen, tc->tex_surf); + screen->tex_transfer_destroy(tc->tex_trans); + tc->tex_trans = NULL; + } - tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z, - PIPE_BUFFER_USAGE_CPU_READ); - tc->tex_surf_map = screen->surface_map(screen, tc->tex_surf, - PIPE_BUFFER_USAGE_CPU_READ); + tc->tex_trans = screen->get_tex_transfer(screen, tc->texture, face, level, z, + PIPE_TRANSFER_READ, 0, 0, + tc->texture->width[level], + tc->texture->height[level]); + tc->tex_trans_map = screen->transfer_map(screen, tc->tex_trans); tc->tex_face = face; tc->tex_level = level; tc->tex_z = z; } - /* get tile from the surface (view into texture) */ - pipe_get_tile_rgba(tc->tex_surf, + /* get tile from the transfer (view into texture) */ + pipe_get_tile_rgba(tc->tex_trans, tile_x, tile_y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); tile->x = tile_x; @@ -571,7 +606,7 @@ sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) tc->clear_val = clearValue; - switch (tc->surface->format) { + switch (tc->transfer->format) { case PIPE_FORMAT_R8G8B8A8_UNORM: r = (clearValue >> 24) & 0xff; g = (clearValue >> 16) & 0xff; diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h index a66bb50bcc..9ac3fdda94 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.h +++ b/src/gallium/drivers/softpipe/sp_tile_cache.h @@ -74,10 +74,10 @@ extern struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc); +sp_tile_cache_map_transfers(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc); +sp_tile_cache_unmap_transfers(struct softpipe_tile_cache *tc); extern void sp_tile_cache_set_texture(struct pipe_context *pipe, diff --git a/src/gallium/drivers/trace/Makefile b/src/gallium/drivers/trace/Makefile index 3859b8acb0..e087db169a 100644 --- a/src/gallium/drivers/trace/Makefile +++ b/src/gallium/drivers/trace/Makefile @@ -4,15 +4,11 @@ include $(TOP)/configs/current LIBNAME = trace C_SOURCES = \ + tr_buffer.c \ tr_context.c \ tr_dump.c \ tr_screen.c \ tr_state.c \ - tr_texture.c \ - tr_winsys.c - + tr_texture.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript index 0a6bfb8f4c..4215215d1a 100644 --- a/src/gallium/drivers/trace/SConscript +++ b/src/gallium/drivers/trace/SConscript @@ -5,12 +5,12 @@ env = env.Clone() trace = env.ConvenienceLibrary( target = 'trace', source = [ + 'tr_buffer.c', 'tr_context.c', 'tr_dump.c', 'tr_screen.c', 'tr_state.c', 'tr_texture.c', - 'tr_winsys.c', ]) -Export('trace')
\ No newline at end of file +Export('trace') diff --git a/src/gallium/drivers/trace/tr_buffer.c b/src/gallium/drivers/trace/tr_buffer.c new file mode 100644 index 0000000000..6ffce1660e --- /dev/null +++ b/src/gallium/drivers/trace/tr_buffer.c @@ -0,0 +1,70 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "util/u_memory.h" + +#include "tr_buffer.h" + + +struct pipe_buffer * +trace_buffer_create(struct trace_screen *tr_scr, + struct pipe_buffer *buffer) +{ + struct trace_buffer *tr_buf; + + if(!buffer) + goto error; + + assert(buffer->screen == tr_scr->screen); + + tr_buf = CALLOC_STRUCT(trace_buffer); + if(!tr_buf) + goto error; + + memcpy(&tr_buf->base, buffer, sizeof(struct pipe_buffer)); + + pipe_reference_init(&tr_buf->base.reference, 1); + tr_buf->base.screen = &tr_scr->base; + tr_buf->buffer = buffer; + + return &tr_buf->base; + +error: + pipe_buffer_reference(&buffer, NULL); + return NULL; +} + + +void +trace_buffer_destroy(struct trace_screen *tr_scr, + struct pipe_buffer *buffer) +{ + struct trace_buffer *tr_buf = trace_buffer(tr_scr, buffer); + pipe_buffer_reference(&tr_buf->buffer, NULL); + FREE(tr_buf); +} diff --git a/src/gallium/winsys/drm/intel/dri/intel_reg.h b/src/gallium/drivers/trace/tr_buffer.h index 4f33bee438..e9e4d354da 100644 --- a/src/gallium/winsys/drm/intel/dri/intel_reg.h +++ b/src/gallium/drivers/trace/tr_buffer.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2009 VMware, Inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,29 +25,46 @@ * **************************************************************************/ +#ifndef TR_BUFFER_H_ +#define TR_BUFFER_H_ -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 +#include "tr_screen.h" -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) +struct trace_buffer +{ + struct pipe_buffer base; -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + struct pipe_buffer *buffer; -#define MI_BATCH_BUFFER_END (0xA<<23) + void *map; + boolean range_flushed; +}; + + +static INLINE struct trace_buffer * +trace_buffer(struct trace_screen *tr_scr, + struct pipe_buffer *buffer) +{ + if(!buffer) + return NULL; + assert(tr_scr); + assert(buffer->screen == &tr_scr->base); + return (struct trace_buffer *)buffer; +} + + +struct pipe_buffer * +trace_buffer_create(struct trace_screen *tr_scr, + struct pipe_buffer *buffer); + +void +trace_buffer_destroy(struct trace_screen *tr_scr, + struct pipe_buffer *buffer); #endif diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index ec8be27077..c894972904 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -30,48 +30,66 @@ #include "tr_dump.h" #include "tr_state.h" +#include "tr_buffer.h" #include "tr_screen.h" #include "tr_texture.h" -#include "tr_winsys.h" #include "tr_context.h" -static INLINE struct pipe_texture * +static INLINE struct pipe_buffer * +trace_buffer_unwrap(struct trace_context *tr_ctx, + struct pipe_buffer *buffer) +{ + struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); + struct trace_buffer *tr_buf; + + if(!buffer) + return NULL; + + tr_buf = trace_buffer(tr_scr, buffer); + + assert(tr_buf->buffer); + assert(tr_buf->buffer->screen == tr_scr->screen); + return tr_buf->buffer; +} + + +static INLINE struct pipe_texture * trace_texture_unwrap(struct trace_context *tr_ctx, struct pipe_texture *texture) { - struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); + struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); struct trace_texture *tr_tex; - + if(!texture) return NULL; - + tr_tex = trace_texture(tr_scr, texture); - + assert(tr_tex->texture); assert(tr_tex->texture->screen == tr_scr->screen); return tr_tex->texture; } -static INLINE struct pipe_surface * +static INLINE struct pipe_surface * trace_surface_unwrap(struct trace_context *tr_ctx, struct pipe_surface *surface) { - struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); + struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); struct trace_texture *tr_tex; struct trace_surface *tr_surf; - + if(!surface) return NULL; assert(surface->texture); if(!surface->texture) return surface; - + tr_tex = trace_texture(tr_scr, surface->texture); tr_surf = trace_surface(tr_tex, surface); - + assert(tr_surf->surface); assert(tr_surf->surface->texture->screen == tr_scr->screen); return tr_surf->surface; @@ -86,7 +104,7 @@ trace_context_set_edgeflags(struct pipe_context *_pipe, struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin("pipe_context", "set_edgeflags"); - + trace_dump_arg(ptr, pipe); /* FIXME: we don't know how big this array is */ trace_dump_arg(ptr, bitfield); @@ -115,24 +133,27 @@ trace_context_draw_arrays(struct pipe_context *_pipe, result = pipe->draw_arrays(pipe, mode, start, count);; trace_dump_ret(bool, result); - + trace_dump_call_end(); - + return result; } static INLINE boolean trace_context_draw_elements(struct pipe_context *_pipe, - struct pipe_buffer *indexBuffer, + struct pipe_buffer *_indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { + struct trace_screen *tr_scr = trace_screen(_pipe->screen); struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _indexBuffer); struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_buffer *indexBuffer = tr_buf->buffer; boolean result; - trace_winsys_user_buffer_update(_pipe->winsys, indexBuffer); + trace_screen_user_buffer_update(_pipe->screen, indexBuffer); trace_dump_call_begin("pipe_context", "draw_elements"); @@ -146,28 +167,31 @@ trace_context_draw_elements(struct pipe_context *_pipe, result = pipe->draw_elements(pipe, indexBuffer, indexSize, mode, start, count);; trace_dump_ret(bool, result); - + trace_dump_call_end(); - + return result; } static INLINE boolean trace_context_draw_range_elements(struct pipe_context *_pipe, - struct pipe_buffer *indexBuffer, + struct pipe_buffer *_indexBuffer, unsigned indexSize, unsigned minIndex, unsigned maxIndex, - unsigned mode, - unsigned start, + unsigned mode, + unsigned start, unsigned count) { + struct trace_screen *tr_scr = trace_screen(_pipe->screen); struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _indexBuffer); struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_buffer *indexBuffer = tr_buf->buffer; boolean result; - trace_winsys_user_buffer_update(_pipe->winsys, indexBuffer); + trace_screen_user_buffer_update(_pipe->screen, indexBuffer); trace_dump_call_begin("pipe_context", "draw_range_elements"); @@ -180,15 +204,15 @@ trace_context_draw_range_elements(struct pipe_context *_pipe, trace_dump_arg(uint, start); trace_dump_arg(uint, count); - result = pipe->draw_range_elements(pipe, - indexBuffer, - indexSize, minIndex, maxIndex, + result = pipe->draw_range_elements(pipe, + indexBuffer, + indexSize, minIndex, maxIndex, mode, start, count); - + trace_dump_ret(bool, result); - + trace_dump_call_end(); - + return result; } @@ -209,9 +233,9 @@ trace_context_create_query(struct pipe_context *_pipe, result = pipe->create_query(pipe, query_type);; trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + return result; } @@ -235,7 +259,7 @@ trace_context_destroy_query(struct pipe_context *_pipe, static INLINE void -trace_context_begin_query(struct pipe_context *_pipe, +trace_context_begin_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -253,7 +277,7 @@ trace_context_begin_query(struct pipe_context *_pipe, static INLINE void -trace_context_end_query(struct pipe_context *_pipe, +trace_context_end_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -271,7 +295,7 @@ trace_context_end_query(struct pipe_context *_pipe, static INLINE boolean -trace_context_get_query_result(struct pipe_context *_pipe, +trace_context_get_query_result(struct pipe_context *_pipe, struct pipe_query *query, boolean wait, uint64_t *presult) @@ -290,9 +314,9 @@ trace_context_get_query_result(struct pipe_context *_pipe, trace_dump_arg(uint, result); trace_dump_ret(bool, _result); - + trace_dump_call_end(); - + return _result; } @@ -315,13 +339,13 @@ trace_context_create_blend_state(struct pipe_context *_pipe, trace_dump_ret(ptr, result); trace_dump_call_end(); - + return result; } static INLINE void -trace_context_bind_blend_state(struct pipe_context *_pipe, +trace_context_bind_blend_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -339,7 +363,7 @@ trace_context_bind_blend_state(struct pipe_context *_pipe, static INLINE void -trace_context_delete_blend_state(struct pipe_context *_pipe, +trace_context_delete_blend_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -372,15 +396,15 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, result = pipe->create_sampler_state(pipe, state);; trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + return result; } static INLINE void -trace_context_bind_sampler_states(struct pipe_context *_pipe, +trace_context_bind_sampler_states(struct pipe_context *_pipe, unsigned num_states, void **states) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -399,7 +423,7 @@ trace_context_bind_sampler_states(struct pipe_context *_pipe, static INLINE void -trace_context_delete_sampler_state(struct pipe_context *_pipe, +trace_context_delete_sampler_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -432,15 +456,15 @@ trace_context_create_rasterizer_state(struct pipe_context *_pipe, result = pipe->create_rasterizer_state(pipe, state);; trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + return result; } static INLINE void -trace_context_bind_rasterizer_state(struct pipe_context *_pipe, +trace_context_bind_rasterizer_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -458,7 +482,7 @@ trace_context_bind_rasterizer_state(struct pipe_context *_pipe, static INLINE void -trace_context_delete_rasterizer_state(struct pipe_context *_pipe, +trace_context_delete_rasterizer_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -489,17 +513,17 @@ trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(depth_stencil_alpha_state, state); - + trace_dump_ret(ptr, result); trace_dump_call_end(); - + return result; } static INLINE void -trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, +trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -517,7 +541,7 @@ trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, static INLINE void -trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, +trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -550,15 +574,15 @@ trace_context_create_fs_state(struct pipe_context *_pipe, result = pipe->create_fs_state(pipe, state);; trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + return result; } static INLINE void -trace_context_bind_fs_state(struct pipe_context *_pipe, +trace_context_bind_fs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -576,7 +600,7 @@ trace_context_bind_fs_state(struct pipe_context *_pipe, static INLINE void -trace_context_delete_fs_state(struct pipe_context *_pipe, +trace_context_delete_fs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -609,15 +633,15 @@ trace_context_create_vs_state(struct pipe_context *_pipe, result = pipe->create_vs_state(pipe, state);; trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + return result; } static INLINE void -trace_context_bind_vs_state(struct pipe_context *_pipe, +trace_context_bind_vs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -635,7 +659,7 @@ trace_context_bind_vs_state(struct pipe_context *_pipe, static INLINE void -trace_context_delete_vs_state(struct pipe_context *_pipe, +trace_context_delete_vs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); @@ -696,8 +720,9 @@ trace_context_set_constant_buffer(struct pipe_context *_pipe, struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; - trace_winsys_user_buffer_update(_pipe->winsys, (struct pipe_buffer *)buffer); - + if (buffer) + trace_screen_user_buffer_update(_pipe->screen, buffer->buffer); + trace_dump_call_begin("pipe_context", "set_constant_buffer"); trace_dump_arg(ptr, pipe); @@ -705,7 +730,13 @@ trace_context_set_constant_buffer(struct pipe_context *_pipe, trace_dump_arg(uint, index); trace_dump_arg(constant_buffer, buffer); - pipe->set_constant_buffer(pipe, shader, index, buffer);; + if (buffer) { + struct pipe_constant_buffer _buffer; + _buffer.buffer = trace_buffer_unwrap(tr_ctx, buffer->buffer); + pipe->set_constant_buffer(pipe, shader, index, &_buffer); + } else { + pipe->set_constant_buffer(pipe, shader, index, buffer); + } trace_dump_call_end(); } @@ -719,7 +750,7 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, struct pipe_context *pipe = tr_ctx->pipe; struct pipe_framebuffer_state unwrapped_state; unsigned i; - + /* Unwrap the input state */ memcpy(&unwrapped_state, state, sizeof(unwrapped_state)); for(i = 0; i < state->nr_cbufs; ++i) @@ -728,7 +759,7 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, unwrapped_state.cbufs[i] = NULL; unwrapped_state.zsbuf = trace_surface_unwrap(tr_ctx, state->zsbuf); state = &unwrapped_state; - + trace_dump_call_begin("pipe_context", "set_framebuffer_state"); trace_dump_arg(ptr, pipe); @@ -803,7 +834,7 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe, struct pipe_context *pipe = tr_ctx->pipe; struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS]; unsigned i; - + for(i = 0; i < num_textures; ++i) unwrapped_textures[i] = trace_texture_unwrap(tr_ctx, textures[i]); textures = unwrapped_textures; @@ -830,18 +861,27 @@ trace_context_set_vertex_buffers(struct pipe_context *_pipe, unsigned i; for(i = 0; i < num_buffers; ++i) - trace_winsys_user_buffer_update(_pipe->winsys, buffers[i].buffer); + trace_screen_user_buffer_update(_pipe->screen, buffers[i].buffer); trace_dump_call_begin("pipe_context", "set_vertex_buffers"); trace_dump_arg(ptr, pipe); trace_dump_arg(uint, num_buffers); - + trace_dump_arg_begin("buffers"); trace_dump_struct_array(vertex_buffer, buffers, num_buffers); trace_dump_arg_end(); - pipe->set_vertex_buffers(pipe, num_buffers, buffers);; + if (num_buffers) { + struct pipe_vertex_buffer *_buffers = malloc(num_buffers * sizeof(*_buffers)); + memcpy(_buffers, buffers, num_buffers * sizeof(*_buffers)); + for (i = 0; i < num_buffers; i++) + _buffers[i].buffer = trace_buffer_unwrap(tr_ctx, buffers[i].buffer); + pipe->set_vertex_buffers(pipe, num_buffers, _buffers); + free(_buffers); + } else { + pipe->set_vertex_buffers(pipe, num_buffers, NULL); + } trace_dump_call_end(); } @@ -872,7 +912,6 @@ trace_context_set_vertex_elements(struct pipe_context *_pipe, static INLINE void trace_context_surface_copy(struct pipe_context *_pipe, - boolean do_flip, struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, @@ -884,11 +923,10 @@ trace_context_surface_copy(struct pipe_context *_pipe, dest = trace_surface_unwrap(tr_ctx, dest); src = trace_surface_unwrap(tr_ctx, src); - + trace_dump_call_begin("pipe_context", "surface_copy"); trace_dump_arg(ptr, pipe); - trace_dump_arg(bool, do_flip); trace_dump_arg(ptr, dest); trace_dump_arg(uint, destx); trace_dump_arg(uint, desty); @@ -898,10 +936,10 @@ trace_context_surface_copy(struct pipe_context *_pipe, trace_dump_arg(uint, width); trace_dump_arg(uint, height); - pipe->surface_copy(pipe, do_flip, - dest, destx, desty, + pipe->surface_copy(pipe, + dest, destx, desty, src, srcx, srcy, width, height); - + trace_dump_call_end(); } @@ -934,7 +972,7 @@ trace_context_surface_fill(struct pipe_context *_pipe, static INLINE void -trace_context_clear(struct pipe_context *_pipe, +trace_context_clear(struct pipe_context *_pipe, struct pipe_surface *surface, unsigned clearValue) { @@ -988,7 +1026,7 @@ trace_context_destroy(struct pipe_context *_pipe) trace_dump_arg(ptr, pipe); pipe->destroy(pipe); - + trace_dump_call_end(); FREE(tr_ctx); @@ -996,23 +1034,25 @@ trace_context_destroy(struct pipe_context *_pipe) struct pipe_context * -trace_context_create(struct pipe_screen *screen, +trace_context_create(struct pipe_screen *_screen, struct pipe_context *pipe) { + struct trace_screen *tr_scr = trace_screen(_screen); struct trace_context *tr_ctx; - + struct pipe_screen *screen = tr_scr->screen; + if(!pipe) goto error1; - + if(!trace_dump_enabled()) goto error1; - + tr_ctx = CALLOC_STRUCT(trace_context); if(!tr_ctx) goto error1; - tr_ctx->base.winsys = screen->winsys; - tr_ctx->base.screen = screen; + tr_ctx->base.winsys = _screen->winsys; + tr_ctx->base.screen = _screen; tr_ctx->base.destroy = trace_context_destroy; tr_ctx->base.set_edgeflags = trace_context_set_edgeflags; tr_ctx->base.draw_arrays = trace_context_draw_arrays; @@ -1057,16 +1097,14 @@ trace_context_create(struct pipe_screen *screen, tr_ctx->base.flush = trace_context_flush; tr_ctx->pipe = pipe; - + trace_dump_call_begin("", "pipe_context_create"); - trace_dump_arg_begin("screen"); - trace_dump_ptr(pipe->screen); - trace_dump_arg_end(); + trace_dump_arg(ptr, screen); trace_dump_ret(ptr, pipe); trace_dump_call_end(); return &tr_ctx->base; - + error1: return pipe; } diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 7831900ec2..d02b22a069 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -30,7 +30,7 @@ #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_context.h" @@ -38,11 +38,11 @@ extern "C" { #endif - + struct trace_context { struct pipe_context base; - + struct pipe_context *pipe; }; diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index a0ead0ded3..6837c94542 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -29,35 +29,38 @@ /** * @file * Trace dumping functions. - * + * * For now we just use standard XML for dumping the trace calls, as this is - * simple to write, parse, and visually inspect, but the actual representation - * is abstracted out of this file, so that we can switch to a binary + * simple to write, parse, and visually inspect, but the actual representation + * is abstracted out of this file, so that we can switch to a binary * representation if/when it becomes justified. - * - * @author Jose Fonseca <jrfonseca@tungstengraphics.com> + * + * @author Jose Fonseca <jrfonseca@tungstengraphics.com> */ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #include <stdlib.h> #endif #include "pipe/p_compiler.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" #include "util/u_string.h" #include "util/u_stream.h" #include "tr_dump.h" +#include "tr_screen.h" +#include "tr_texture.h" +#include "tr_buffer.h" static struct util_stream *stream = NULL; static unsigned refcount = 0; -static INLINE void +static INLINE void trace_dump_write(const char *buf, size_t size) { if(stream) @@ -65,14 +68,14 @@ trace_dump_write(const char *buf, size_t size) } -static INLINE void +static INLINE void trace_dump_writes(const char *s) { trace_dump_write(s, strlen(s)); } -static INLINE void +static INLINE void trace_dump_writef(const char *format, ...) { static char buf[1024]; @@ -85,8 +88,8 @@ trace_dump_writef(const char *format, ...) } -static INLINE void -trace_dump_escape(const char *str) +static INLINE void +trace_dump_escape(const char *str) { const unsigned char *p = (const unsigned char *)str; unsigned char c; @@ -109,7 +112,7 @@ trace_dump_escape(const char *str) } -static INLINE void +static INLINE void trace_dump_indent(unsigned level) { unsigned i; @@ -118,14 +121,14 @@ trace_dump_indent(unsigned level) } -static INLINE void -trace_dump_newline(void) +static INLINE void +trace_dump_newline(void) { trace_dump_writes("\n"); } -static INLINE void +static INLINE void trace_dump_tag(const char *name) { trace_dump_writes("<"); @@ -134,7 +137,7 @@ trace_dump_tag(const char *name) } -static INLINE void +static INLINE void trace_dump_tag_begin(const char *name) { trace_dump_writes("<"); @@ -142,8 +145,8 @@ trace_dump_tag_begin(const char *name) trace_dump_writes(">"); } -static INLINE void -trace_dump_tag_begin1(const char *name, +static INLINE void +trace_dump_tag_begin1(const char *name, const char *attr1, const char *value1) { trace_dump_writes("<"); @@ -156,8 +159,8 @@ trace_dump_tag_begin1(const char *name, } -static INLINE void -trace_dump_tag_begin2(const char *name, +static INLINE void +trace_dump_tag_begin2(const char *name, const char *attr1, const char *value1, const char *attr2, const char *value2) { @@ -175,8 +178,8 @@ trace_dump_tag_begin2(const char *name, } -static INLINE void -trace_dump_tag_begin3(const char *name, +static INLINE void +trace_dump_tag_begin3(const char *name, const char *attr1, const char *value1, const char *attr2, const char *value2, const char *attr3, const char *value3) @@ -207,7 +210,7 @@ trace_dump_tag_end(const char *name) trace_dump_writes(">"); } -static void +static void trace_dump_trace_close(void) { if(stream) { @@ -221,30 +224,30 @@ trace_dump_trace_close(void) boolean trace_dump_trace_begin() { const char *filename; - + filename = debug_get_option("GALLIUM_TRACE", NULL); if(!filename) return FALSE; - + if(!stream) { - + stream = util_stream_create(filename, 0); if(!stream) return FALSE; - + trace_dump_writes("<?xml version='1.0' encoding='UTF-8'?>\n"); trace_dump_writes("<?xml-stylesheet type='text/xsl' href='trace.xsl'?>\n"); trace_dump_writes("<trace version='0.1'>\n"); - -#if defined(PIPE_OS_LINUX) - /* Linux applications rarely cleanup GL / Gallium resources so catch - * application exit here */ + +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) + /* Linux applications rarely cleanup GL / Gallium resources so catch + * application exit here */ atexit(trace_dump_trace_close); #endif } - + ++refcount; - + return TRUE; } @@ -402,3 +405,49 @@ void trace_dump_ptr(const void *value) else trace_dump_null(); } + +void trace_dump_buffer_ptr(struct pipe_buffer *_buffer) +{ + if (_buffer) { + struct trace_screen *tr_scr = trace_screen(_buffer->screen); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + trace_dump_ptr(tr_buf->buffer); + } else { + trace_dump_null(); + } +} + +void trace_dump_texture_ptr(struct pipe_texture *_texture) +{ + if (_texture) { + struct trace_screen *tr_scr = trace_screen(_texture->screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _texture); + trace_dump_ptr(tr_tex->texture); + } else { + trace_dump_null(); + } +} + +void trace_dump_surface_ptr(struct pipe_surface *_surface) +{ + if (_surface) { + struct trace_screen *tr_scr = trace_screen(_surface->texture->screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _surface->texture); + struct trace_surface *tr_surf = trace_surface(tr_tex, _surface); + trace_dump_ptr(tr_surf->surface); + } else { + trace_dump_null(); + } +} + +void trace_dump_transfer_ptr(struct pipe_transfer *_transfer) +{ + if (_transfer) { + struct trace_screen *tr_scr = trace_screen(_transfer->texture->screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _transfer->texture); + struct trace_transfer *tr_tran = trace_transfer(tr_tex, _transfer); + trace_dump_ptr(tr_tran->transfer); + } else { + trace_dump_null(); + } +} diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 76a53731b3..26409f26c6 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -37,6 +37,11 @@ #include "pipe/p_compiler.h" +struct pipe_buffer; +struct pipe_texture; +struct pipe_surface; +struct pipe_transfer; + boolean trace_dump_trace_begin(void); boolean trace_dump_enabled(void); void trace_dump_trace_end(void); @@ -63,10 +68,14 @@ void trace_dump_member_begin(const char *name); void trace_dump_member_end(void); void trace_dump_null(void); void trace_dump_ptr(const void *value); - +/* will turn a wrapped object into the real one and dump ptr */ +void trace_dump_buffer_ptr(struct pipe_buffer *_buffer); +void trace_dump_texture_ptr(struct pipe_texture *_texture); +void trace_dump_surface_ptr(struct pipe_surface *_surface); +void trace_dump_transfer_ptr(struct pipe_transfer *_transfer); /* - * Code saving macros. + * Code saving macros. */ #define trace_dump_arg(_type, _arg) \ diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 8789f86b1a..954576d721 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -27,12 +27,14 @@ #include "util/u_memory.h" +#include "tr_buffer.h" #include "tr_dump.h" #include "tr_state.h" -#include "tr_winsys.h" #include "tr_texture.h" #include "tr_screen.h" +#include "pipe/p_inlines.h" + static const char * trace_screen_get_name(struct pipe_screen *_screen) @@ -40,17 +42,17 @@ trace_screen_get_name(struct pipe_screen *_screen) struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; const char *result; - + trace_dump_call_begin("pipe_screen", "get_name"); - + trace_dump_arg(ptr, screen); result = screen->get_name(screen); - + trace_dump_ret(string, result); - + trace_dump_call_end(); - + return result; } @@ -61,80 +63,80 @@ trace_screen_get_vendor(struct pipe_screen *_screen) struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; const char *result; - + trace_dump_call_begin("pipe_screen", "get_vendor"); - + trace_dump_arg(ptr, screen); - + result = screen->get_vendor(screen); - + trace_dump_ret(string, result); - + trace_dump_call_end(); - + return result; } -static int -trace_screen_get_param(struct pipe_screen *_screen, +static int +trace_screen_get_param(struct pipe_screen *_screen, int param) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; int result; - + trace_dump_call_begin("pipe_screen", "get_param"); - + trace_dump_arg(ptr, screen); trace_dump_arg(int, param); result = screen->get_param(screen, param); - + trace_dump_ret(int, result); - + trace_dump_call_end(); - + return result; } -static float -trace_screen_get_paramf(struct pipe_screen *_screen, +static float +trace_screen_get_paramf(struct pipe_screen *_screen, int param) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; float result; - + trace_dump_call_begin("pipe_screen", "get_paramf"); - + trace_dump_arg(ptr, screen); trace_dump_arg(int, param); result = screen->get_paramf(screen, param); - + trace_dump_ret(float, result); - + trace_dump_call_end(); - + return result; } -static boolean +static boolean trace_screen_is_format_supported(struct pipe_screen *_screen, enum pipe_format format, enum pipe_texture_target target, - unsigned tex_usage, + unsigned tex_usage, unsigned geom_flags) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; boolean result; - + trace_dump_call_begin("pipe_screen", "is_format_supported"); - + trace_dump_arg(ptr, screen); trace_dump_arg(format, format); trace_dump_arg(int, target); @@ -142,15 +144,45 @@ trace_screen_is_format_supported(struct pipe_screen *_screen, trace_dump_arg(uint, geom_flags); result = screen->is_format_supported(screen, format, target, tex_usage, geom_flags); - + trace_dump_ret(bool, result); - + trace_dump_call_end(); - + return result; } +static void +trace_screen_flush_frontbuffer(struct pipe_screen *_screen, + struct pipe_surface *_surface, + void *context_private) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _surface->texture); + struct trace_surface *tr_surf = trace_surface(tr_tex, _surface); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_surface *surface = tr_surf->surface; + + trace_dump_call_begin("pipe_screen", "flush_frontbuffer"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, surface); + /* XXX: hide, as there is nothing we can do with this + trace_dump_arg(ptr, context_private); + */ + + screen->flush_frontbuffer(screen, surface, context_private); + + trace_dump_call_end(); +} + + +/******************************************************************** + * texture + */ + + static struct pipe_texture * trace_screen_texture_create(struct pipe_screen *_screen, const struct pipe_texture *templat) @@ -158,20 +190,20 @@ trace_screen_texture_create(struct pipe_screen *_screen, struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; struct pipe_texture *result; - + trace_dump_call_begin("pipe_screen", "texture_create"); trace_dump_arg(ptr, screen); trace_dump_arg(template, templat); result = screen->texture_create(screen, templat); - + trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + result = trace_texture_create(tr_scr, result); - + return result; } @@ -195,71 +227,60 @@ trace_screen_texture_blanket(struct pipe_screen *_screen, trace_dump_arg(ptr, buffer); result = screen->texture_blanket(screen, templat, ppitch, buffer); - + trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + result = trace_texture_create(tr_scr, result); - + return result; } -static void -trace_screen_texture_release(struct pipe_screen *_screen, - struct pipe_texture **ptexture) +static void +trace_screen_texture_destroy(struct pipe_texture *_texture) { - struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_screen *tr_scr = trace_screen(_texture->screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _texture); struct pipe_screen *screen = tr_scr->screen; - struct trace_texture *tr_tex; - struct pipe_texture *texture; - - assert(ptexture); - if(*ptexture) { - tr_tex = trace_texture(tr_scr, *ptexture); - texture = tr_tex->texture; - assert(texture->screen == screen); - } - else - texture = NULL; - - if (*ptexture) { - if (!--(*ptexture)->refcount) { - trace_dump_call_begin("pipe_screen", "texture_destroy"); - - trace_dump_arg(ptr, screen); - trace_dump_arg(ptr, texture); - - trace_texture_destroy(tr_scr, *ptexture); - - trace_dump_call_end(); - } - - *ptexture = NULL; - } + struct pipe_texture *texture = tr_tex->texture; + + assert(texture->screen == screen); + + trace_dump_call_begin("pipe_screen", "texture_destroy"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, texture); + + trace_dump_call_end(); + + trace_texture_destroy(tr_scr, _texture); } +/******************************************************************** + * surface + */ + + static struct pipe_surface * trace_screen_get_tex_surface(struct pipe_screen *_screen, - struct pipe_texture *texture, + struct pipe_texture *_texture, unsigned face, unsigned level, unsigned zslice, unsigned usage) { struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _texture); struct pipe_screen *screen = tr_scr->screen; - struct trace_texture *tr_tex; - struct pipe_surface *result; - - assert(texture); - tr_tex = trace_texture(tr_scr, texture); - texture = tr_tex->texture; + struct pipe_texture *texture = tr_tex->texture; + struct pipe_surface *result = NULL; + assert(texture->screen == screen); - + trace_dump_call_begin("pipe_screen", "get_tex_surface"); - + trace_dump_arg(ptr, screen); trace_dump_arg(ptr, texture); trace_dump_arg(uint, face); @@ -270,133 +291,538 @@ trace_screen_get_tex_surface(struct pipe_screen *_screen, result = screen->get_tex_surface(screen, texture, face, level, zslice, usage); trace_dump_ret(ptr, result); - + trace_dump_call_end(); - + result = trace_surface_create(tr_tex, result); return result; } -static void -trace_screen_tex_surface_release(struct pipe_screen *_screen, - struct pipe_surface **psurface) +static void +trace_screen_tex_surface_destroy(struct pipe_surface *_surface) +{ + struct trace_screen *tr_scr = trace_screen(_surface->texture->screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _surface->texture); + struct trace_surface *tr_surf = trace_surface(tr_tex, _surface); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_surface *surface = tr_surf->surface; + + trace_dump_call_begin("pipe_screen", "tex_surface_destroy"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, surface); + + trace_dump_call_end(); + + trace_surface_destroy(tr_tex, _surface); +} + + +/******************************************************************** + * transfer + */ + + +static struct pipe_transfer * +trace_screen_get_tex_transfer(struct pipe_screen *_screen, + struct pipe_texture *_texture, + unsigned face, unsigned level, + unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) { struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _texture); struct pipe_screen *screen = tr_scr->screen; - struct trace_texture *tr_tex; - struct trace_surface *tr_surf; - struct pipe_surface *surface; - - assert(psurface); - if(*psurface) { - tr_tex = trace_texture(tr_scr, (*psurface)->texture); - tr_surf = trace_surface(tr_tex, *psurface); - surface = tr_surf->surface; - } - else - surface = NULL; - - if (*psurface) { - if (!--(*psurface)->refcount) { - trace_dump_call_begin("pipe_screen", "tex_surface_destroy"); - - trace_dump_arg(ptr, screen); - trace_dump_arg(ptr, surface); + struct pipe_texture *texture = tr_tex->texture; + struct pipe_transfer *result = NULL; - trace_surface_destroy(tr_tex, *psurface); + assert(texture->screen == screen); - trace_dump_call_end(); - } - - *psurface = NULL; - } + trace_dump_call_begin("pipe_screen", "get_tex_transfer"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, texture); + trace_dump_arg(uint, face); + trace_dump_arg(uint, level); + trace_dump_arg(uint, zslice); + trace_dump_arg(uint, usage); + + trace_dump_arg(uint, x); + trace_dump_arg(uint, y); + trace_dump_arg(uint, w); + trace_dump_arg(uint, h); + + result = screen->get_tex_transfer(screen, texture, face, level, zslice, usage, + x, y, w, h); + + trace_dump_ret(ptr, result); + + trace_dump_call_end(); + + result = trace_transfer_create(tr_tex, result); + + return result; +} + + +static void +trace_screen_tex_transfer_destroy(struct pipe_transfer *_transfer) +{ + struct trace_screen *tr_scr = trace_screen(_transfer->texture->screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _transfer->texture); + struct trace_transfer *tr_tran = trace_transfer(tr_tex, _transfer); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_transfer *transfer = tr_tran->transfer; + + trace_dump_call_begin("pipe_screen", "tex_transfer_destroy"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, transfer); + + trace_dump_call_end(); + + trace_transfer_destroy(tr_tex, _transfer); } static void * -trace_screen_surface_map(struct pipe_screen *_screen, - struct pipe_surface *surface, - unsigned flags) +trace_screen_transfer_map(struct pipe_screen *_screen, + struct pipe_transfer *_transfer) { struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _transfer->texture); + struct trace_transfer *tr_trans = trace_transfer(tr_tex, _transfer); struct pipe_screen *screen = tr_scr->screen; - struct trace_texture *tr_tex; - struct trace_surface *tr_surf; + struct pipe_transfer *transfer = tr_trans->transfer; void *map; - - tr_tex = trace_texture(tr_scr, surface->texture); - tr_surf = trace_surface(tr_tex, surface); - surface = tr_surf->surface; - map = screen->surface_map(screen, surface, flags); + map = screen->transfer_map(screen, transfer); if(map) { - if(flags & PIPE_BUFFER_USAGE_CPU_WRITE) { - assert(!tr_surf->map); - tr_surf->map = map; + if(transfer->usage != PIPE_TRANSFER_READ) { + assert(!tr_trans->map); + tr_trans->map = map; } } - + return map; } -static void -trace_screen_surface_unmap(struct pipe_screen *_screen, - struct pipe_surface *surface) +static void +trace_screen_transfer_unmap(struct pipe_screen *_screen, + struct pipe_transfer *_transfer) { struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_texture *tr_tex = trace_texture(tr_scr, _transfer->texture); + struct trace_transfer *tr_trans = trace_transfer(tr_tex, _transfer); struct pipe_screen *screen = tr_scr->screen; - struct trace_texture *tr_tex; - struct trace_surface *tr_surf; - - tr_tex = trace_texture(tr_scr, surface->texture); - tr_surf = trace_surface(tr_tex, surface); - surface = tr_surf->surface; - - if(tr_surf->map) { - size_t size = surface->nblocksy * surface->stride; - - trace_dump_call_begin("pipe_winsys", "surface_write"); - + struct pipe_transfer *transfer = tr_trans->transfer; + + if(tr_trans->map) { + size_t size = transfer->nblocksy * transfer->stride; + + trace_dump_call_begin("pipe_screen", "transfer_write"); + trace_dump_arg(ptr, screen); - - trace_dump_arg(ptr, surface); - - trace_dump_arg_begin("data"); - trace_dump_bytes(tr_surf->map, size); - trace_dump_arg_end(); + + trace_dump_arg(ptr, transfer); trace_dump_arg_begin("stride"); - trace_dump_uint(surface->stride); + trace_dump_uint(transfer->stride); + trace_dump_arg_end(); + + trace_dump_arg_begin("data"); + trace_dump_bytes(tr_trans->map, size); trace_dump_arg_end(); trace_dump_arg_begin("size"); trace_dump_uint(size); trace_dump_arg_end(); - + trace_dump_call_end(); - tr_surf->map = NULL; + tr_trans->map = NULL; + } + + screen->transfer_unmap(screen, transfer); +} + + +/******************************************************************** + * buffer + */ + + +static struct pipe_buffer * +trace_screen_surface_buffer_create(struct pipe_screen *_screen, + unsigned width, unsigned height, + enum pipe_format format, + unsigned usage, + unsigned *pstride) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + unsigned stride; + struct pipe_buffer *result; + + trace_dump_call_begin("pipe_screen", "surface_buffer_create"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(uint, width); + trace_dump_arg(uint, height); + trace_dump_arg(format, format); + trace_dump_arg(uint, usage); + + result = screen->surface_buffer_create(screen, + width, height, + format, + usage, + pstride); + + stride = *pstride; + + trace_dump_arg(uint, stride); + + trace_dump_ret(ptr, result); + + trace_dump_call_end(); + + return trace_buffer_create(tr_scr, result); +} + + +static struct pipe_buffer * +trace_screen_buffer_create(struct pipe_screen *_screen, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *result; + + trace_dump_call_begin("pipe_screen", "buffer_create"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(uint, alignment); + trace_dump_arg(uint, usage); + trace_dump_arg(uint, size); + + result = screen->buffer_create(screen, alignment, usage, size); + + trace_dump_ret(ptr, result); + + trace_dump_call_end(); + + /* Zero the buffer to avoid dumping uninitialized memory */ + if(result->usage & PIPE_BUFFER_USAGE_CPU_WRITE) { + void *map; + map = pipe_buffer_map(screen, result, PIPE_BUFFER_USAGE_CPU_WRITE); + if(map) { + memset(map, 0, result->size); + screen->buffer_unmap(screen, result); + } + } + + return trace_buffer_create(tr_scr, result); +} + + +static struct pipe_buffer * +trace_screen_user_buffer_create(struct pipe_screen *_screen, + void *data, + unsigned size) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *result; + + trace_dump_call_begin("pipe_screen", "user_buffer_create"); + + trace_dump_arg(ptr, screen); + trace_dump_arg_begin("data"); + trace_dump_bytes(data, size); + trace_dump_arg_end(); + trace_dump_arg(uint, size); + + result = screen->user_buffer_create(screen, data, size); + + trace_dump_ret(ptr, result); + + trace_dump_call_end(); + + if(result) { + assert(!(result->usage & TRACE_BUFFER_USAGE_USER)); + result->usage |= TRACE_BUFFER_USAGE_USER; + } + + return trace_buffer_create(tr_scr, result); +} + + +/** + * This function is used to track if data has been changed on a user buffer + * without map/unmap being called. + */ +void +trace_screen_user_buffer_update(struct pipe_screen *_screen, + struct pipe_buffer *_buffer) +{ +#if 0 + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + const void *map; + + if(buffer && buffer->usage & TRACE_BUFFER_USAGE_USER) { + map = screen->buffer_map(screen, buffer, PIPE_BUFFER_USAGE_CPU_READ); + if(map) { + trace_dump_call_begin("pipe_winsys", "buffer_write"); + + trace_dump_arg(ptr, screen); + + trace_dump_arg(ptr, buffer); + + trace_dump_arg_begin("data"); + trace_dump_bytes(map, buffer->size); + trace_dump_arg_end(); + + trace_dump_arg_begin("size"); + trace_dump_uint(buffer->size); + trace_dump_arg_end(); + + trace_dump_call_end(); + + screen->buffer_unmap(screen, buffer); + } + } +#endif +} + + +static void * +trace_screen_buffer_map(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned usage) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *buffer = tr_buf->buffer; + void *map; + + assert(screen->buffer_map); + map = screen->buffer_map(screen, buffer, usage); + if(map) { + if(usage & PIPE_BUFFER_USAGE_CPU_WRITE) { + tr_buf->map = map; + } + } + + return map; +} + + +static void * +trace_screen_buffer_map_range(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned offset, + unsigned length, + unsigned usage) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *buffer = tr_buf->buffer; + void *map; + + assert(screen->buffer_map_range); + map = screen->buffer_map_range(screen, buffer, offset, length, usage); + if(map) { + if(usage & PIPE_BUFFER_USAGE_CPU_WRITE) { + tr_buf->map = map; + } } - screen->surface_unmap(screen, surface); + return map; +} + + +static void +buffer_write(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned offset, + const char *map, + unsigned size) +{ + assert(map); + + trace_dump_call_begin("pipe_screen", "buffer_write"); + + trace_dump_arg(ptr, screen); + + trace_dump_arg(ptr, buffer); + + trace_dump_arg(uint, offset); + + trace_dump_arg_begin("data"); + trace_dump_bytes(map + offset, size); + trace_dump_arg_end(); + + trace_dump_arg(uint, size); + + trace_dump_call_end(); + +} + + +static void +trace_screen_buffer_flush_mapped_range(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned offset, + unsigned length) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *buffer = tr_buf->buffer; + + assert(tr_buf->map); + buffer_write(screen, buffer, offset, tr_buf->map, length); + tr_buf->range_flushed = TRUE; + screen->buffer_flush_mapped_range(screen, buffer, offset, length); +} + + +static void +trace_screen_buffer_unmap(struct pipe_screen *_screen, + struct pipe_buffer *_buffer) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *buffer = tr_buf->buffer; + + if (tr_buf->map && !tr_buf->range_flushed) + buffer_write(screen, buffer, tr_buf->map, 0, buffer->size); + tr_buf->map = NULL; + tr_buf->range_flushed = FALSE; + screen->buffer_unmap(screen, buffer); +} + + +static void +trace_screen_buffer_destroy(struct pipe_buffer *_buffer) +{ + struct trace_screen *tr_scr = trace_screen(_buffer->screen); + struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *buffer = tr_buf->buffer; + + trace_dump_call_begin("pipe_screen", "buffer_destroy"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, buffer); + + trace_dump_call_end(); + + trace_buffer_destroy(tr_scr, _buffer); +} + + +/******************************************************************** + * fence + */ + + +static void +trace_screen_fence_reference(struct pipe_screen *_screen, + struct pipe_fence_handle **dst, + struct pipe_fence_handle *src) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + + trace_dump_call_begin("pipe_screen", "fence_reference"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, dst); + trace_dump_arg(ptr, src); + + screen->fence_reference(screen, dst, src); + + trace_dump_call_end(); +} + + +static int +trace_screen_fence_signalled(struct pipe_screen *_screen, + struct pipe_fence_handle *fence, + unsigned flag) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + int result; + + trace_dump_call_begin("pipe_screen", "fence_signalled"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, fence); + trace_dump_arg(uint, flag); + + result = screen->fence_signalled(screen, fence, flag); + + trace_dump_ret(int, result); + + trace_dump_call_end(); + + return result; } +static int +trace_screen_fence_finish(struct pipe_screen *_screen, + struct pipe_fence_handle *fence, + unsigned flag) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct pipe_screen *screen = tr_scr->screen; + int result; + + trace_dump_call_begin("pipe_screen", "fence_finish"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, fence); + trace_dump_arg(uint, flag); + + result = screen->fence_finish(screen, fence, flag); + + trace_dump_ret(int, result); + + trace_dump_call_end(); + + return result; +} + + +/******************************************************************** + * screen + */ + static void trace_screen_destroy(struct pipe_screen *_screen) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; - + trace_dump_call_begin("pipe_screen", "destroy"); - + trace_dump_arg(ptr, screen); screen->destroy(screen); - + trace_dump_call_end(); trace_dump_trace_end(); @@ -410,21 +836,27 @@ trace_screen_create(struct pipe_screen *screen) { struct trace_screen *tr_scr; struct pipe_winsys *winsys; - + if(!screen) goto error1; if(!trace_dump_trace_begin()) goto error1; + trace_dump_call_begin("", "pipe_screen_create"); + tr_scr = CALLOC_STRUCT(trace_screen); if(!tr_scr) goto error2; +#if 0 winsys = trace_winsys_create(screen->winsys); if(!winsys) goto error3; - +#else + winsys = screen->winsys; +#endif + tr_scr->base.winsys = winsys; tr_scr->base.destroy = trace_screen_destroy; tr_scr->base.get_name = trace_screen_get_name; @@ -434,26 +866,43 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->base.is_format_supported = trace_screen_is_format_supported; tr_scr->base.texture_create = trace_screen_texture_create; tr_scr->base.texture_blanket = trace_screen_texture_blanket; - tr_scr->base.texture_release = trace_screen_texture_release; + tr_scr->base.texture_destroy = trace_screen_texture_destroy; tr_scr->base.get_tex_surface = trace_screen_get_tex_surface; - tr_scr->base.tex_surface_release = trace_screen_tex_surface_release; - tr_scr->base.surface_map = trace_screen_surface_map; - tr_scr->base.surface_unmap = trace_screen_surface_unmap; - + tr_scr->base.tex_surface_destroy = trace_screen_tex_surface_destroy; + tr_scr->base.get_tex_transfer = trace_screen_get_tex_transfer; + tr_scr->base.tex_transfer_destroy = trace_screen_tex_transfer_destroy; + tr_scr->base.transfer_map = trace_screen_transfer_map; + tr_scr->base.transfer_unmap = trace_screen_transfer_unmap; + tr_scr->base.buffer_create = trace_screen_buffer_create; + tr_scr->base.user_buffer_create = trace_screen_user_buffer_create; + tr_scr->base.surface_buffer_create = trace_screen_surface_buffer_create; + if (screen->buffer_map) + tr_scr->base.buffer_map = trace_screen_buffer_map; + if (screen->buffer_map_range) + tr_scr->base.buffer_map_range = trace_screen_buffer_map_range; + if (screen->buffer_flush_mapped_range) + tr_scr->base.buffer_flush_mapped_range = trace_screen_buffer_flush_mapped_range; + if (screen->buffer_unmap) + tr_scr->base.buffer_unmap = trace_screen_buffer_unmap; + tr_scr->base.buffer_destroy = trace_screen_buffer_destroy; + tr_scr->base.fence_reference = trace_screen_fence_reference; + tr_scr->base.fence_signalled = trace_screen_fence_signalled; + tr_scr->base.fence_finish = trace_screen_fence_finish; + tr_scr->base.flush_frontbuffer = trace_screen_flush_frontbuffer; tr_scr->screen = screen; - trace_dump_call_begin("", "pipe_screen_create"); - trace_dump_arg_begin("winsys"); - trace_dump_ptr(screen->winsys); - trace_dump_arg_end(); trace_dump_ret(ptr, screen); trace_dump_call_end(); return &tr_scr->base; +#if 0 error3: FREE(tr_scr); +#endif error2: + trace_dump_ret(ptr, screen); + trace_dump_call_end(); trace_dump_trace_end(); error1: return screen; diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 93fefdb9a5..8c65516b50 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -36,11 +36,19 @@ extern "C" { #endif - + +/** + * It often happens that new data is written directly to the user buffers + * without mapping/unmapping. This flag marks user buffers, so that their + * contents can be dumpped before being used by the pipe context. + */ +#define TRACE_BUFFER_USAGE_USER (1 << 31) + + struct trace_screen { struct pipe_screen base; - + struct pipe_screen *screen; }; @@ -53,6 +61,11 @@ struct pipe_screen * trace_screen_create(struct pipe_screen *screen); +void +trace_screen_user_buffer_update(struct pipe_screen *screen, + struct pipe_buffer *buffer); + + #ifdef __cplusplus } #endif diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 524f2d6194..b6a1ce0d62 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -50,6 +50,14 @@ void trace_dump_block(const struct pipe_format_block *block) } +static void trace_dump_reference(const struct pipe_reference *reference) +{ + trace_dump_struct_begin("pipe_reference"); + trace_dump_member(int, &reference->count, count); + trace_dump_struct_end(); +} + + void trace_dump_template(const struct pipe_texture *templat) { if(!templat) { @@ -58,10 +66,10 @@ void trace_dump_template(const struct pipe_texture *templat) } trace_dump_struct_begin("pipe_texture"); - + trace_dump_member(int, templat, target); trace_dump_member(format, templat, format); - + trace_dump_member_begin("width"); trace_dump_array(uint, templat->width, 1); trace_dump_member_end(); @@ -77,10 +85,10 @@ void trace_dump_template(const struct pipe_texture *templat) trace_dump_member_begin("block"); trace_dump_block(&templat->block); trace_dump_member_end(); - + trace_dump_member(uint, templat, last_level); trace_dump_member(uint, templat, tex_usage); - + trace_dump_struct_end(); } @@ -114,8 +122,7 @@ void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) trace_dump_member(uint, state, line_stipple_factor); trace_dump_member(uint, state, line_stipple_pattern); trace_dump_member(bool, state, line_last_pixel); - trace_dump_member(bool, state, bypass_clipping); - trace_dump_member(bool, state, bypass_vs); + trace_dump_member(bool, state, bypass_vs_clip_and_viewport); trace_dump_member(bool, state, origin_lower_left); trace_dump_member(bool, state, flatshade_first); trace_dump_member(bool, state, gl_rasterization_rules); @@ -126,9 +133,9 @@ void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) trace_dump_member(float, state, point_size_max); trace_dump_member(float, state, offset_units); trace_dump_member(float, state, offset_scale); - + trace_dump_member_array(uint, state, sprite_coord_mode); - + trace_dump_struct_end(); } @@ -144,10 +151,10 @@ void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) trace_dump_member_begin("stipple"); trace_dump_array(uint, - state->stipple, + state->stipple, Elements(state->stipple)); trace_dump_member_end(); - + trace_dump_struct_end(); } @@ -163,7 +170,7 @@ void trace_dump_viewport_state(const struct pipe_viewport_state *state) trace_dump_member_array(float, state, scale); trace_dump_member_array(float, state, translate); - + trace_dump_struct_end(); } @@ -189,7 +196,7 @@ void trace_dump_scissor_state(const struct pipe_scissor_state *state) void trace_dump_clip_state(const struct pipe_clip_state *state) { unsigned i; - + if(!state) { trace_dump_null(); return; @@ -222,7 +229,7 @@ void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) trace_dump_struct_begin("pipe_constant_buffer"); - trace_dump_member(ptr, state, buffer); + trace_dump_member(buffer_ptr, state, buffer); trace_dump_struct_end(); } @@ -238,7 +245,7 @@ void trace_dump_shader_state(const struct pipe_shader_state *state) } tgsi_dump_str(state->tokens, 0, str, sizeof(str)); - + trace_dump_struct_begin("pipe_shader_state"); trace_dump_member_begin("tokens"); @@ -252,7 +259,7 @@ void trace_dump_shader_state(const struct pipe_shader_state *state) void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state) { unsigned i; - + if(!state) { trace_dump_null(); return; @@ -268,7 +275,7 @@ void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_ trace_dump_member(bool, &state->depth, occlusion_count); trace_dump_struct_end(); trace_dump_member_end(); - + trace_dump_member_begin("stencil"); trace_dump_array_begin(); for(i = 0; i < Elements(state->stencil); ++i) { @@ -397,22 +404,47 @@ void trace_dump_surface(const struct pipe_surface *state) trace_dump_struct_begin("pipe_surface"); + trace_dump_reference(&state->reference); + trace_dump_member(format, state, format); trace_dump_member(uint, state, status); trace_dump_member(uint, state, clear_value); trace_dump_member(uint, state, width); trace_dump_member(uint, state, height); + trace_dump_member(uint, state, layout); + trace_dump_member(uint, state, offset); + trace_dump_member(uint, state, usage); + + trace_dump_member(ptr, state, texture); + trace_dump_member(uint, state, face); + trace_dump_member(uint, state, level); + trace_dump_member(uint, state, zslice); + + trace_dump_struct_end(); +} + + +void trace_dump_transfer(const struct pipe_transfer *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_transfer"); + + trace_dump_member(format, state, format); + trace_dump_member(uint, state, width); + trace_dump_member(uint, state, height); + trace_dump_member_begin("block"); trace_dump_block(&state->block); trace_dump_member_end(); - + trace_dump_member(uint, state, nblocksx); trace_dump_member(uint, state, nblocksy); trace_dump_member(uint, state, stride); - trace_dump_member(uint, state, layout); - trace_dump_member(uint, state, offset); - trace_dump_member(uint, state, refcount); trace_dump_member(uint, state, usage); trace_dump_member(ptr, state, texture); @@ -436,7 +468,7 @@ void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) trace_dump_member(uint, state, stride); trace_dump_member(uint, state, max_index); trace_dump_member(uint, state, buffer_offset); - trace_dump_member(ptr, state, buffer); + trace_dump_member(buffer_ptr, state, buffer); trace_dump_struct_end(); } @@ -455,7 +487,7 @@ void trace_dump_vertex_element(const struct pipe_vertex_element *state) trace_dump_member(uint, state, vertex_buffer_index); trace_dump_member(uint, state, nr_components); - + trace_dump_member(format, state, src_format); trace_dump_struct_end(); diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h index 5ae533dc66..513ed0ac98 100644 --- a/src/gallium/drivers/trace/tr_state.h +++ b/src/gallium/drivers/trace/tr_state.h @@ -68,6 +68,8 @@ void trace_dump_sampler_state(const struct pipe_sampler_state *state); void trace_dump_surface(const struct pipe_surface *state); +void trace_dump_transfer(const struct pipe_transfer *state); + void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state); void trace_dump_vertex_element(const struct pipe_vertex_element *state); diff --git a/src/gallium/drivers/trace/tr_texture.c b/src/gallium/drivers/trace/tr_texture.c index 1cc4f0bd43..7b392f0728 100644 --- a/src/gallium/drivers/trace/tr_texture.c +++ b/src/gallium/drivers/trace/tr_texture.c @@ -25,7 +25,6 @@ * **************************************************************************/ -#include "pipe/p_inlines.h" #include "util/u_hash_table.h" #include "util/u_memory.h" @@ -34,26 +33,28 @@ struct pipe_texture * -trace_texture_create(struct trace_screen *tr_scr, +trace_texture_create(struct trace_screen *tr_scr, struct pipe_texture *texture) { struct trace_texture *tr_tex; - + if(!texture) goto error; - + assert(texture->screen == tr_scr->screen); - + tr_tex = CALLOC_STRUCT(trace_texture); if(!tr_tex) goto error; - + memcpy(&tr_tex->base, texture, sizeof(struct pipe_texture)); + + pipe_reference_init(&tr_tex->base.reference, 1); tr_tex->base.screen = &tr_scr->base; tr_tex->texture = texture; - + return &tr_tex->base; - + error: pipe_texture_reference(&texture, NULL); return NULL; @@ -61,38 +62,39 @@ error: void -trace_texture_destroy(struct trace_screen *tr_scr, +trace_texture_destroy(struct trace_screen *tr_scr, struct pipe_texture *texture) { - struct trace_texture *tr_tex = trace_texture(tr_scr, texture); + struct trace_texture *tr_tex = trace_texture(tr_scr, texture); pipe_texture_reference(&tr_tex->texture, NULL); FREE(tr_tex); } struct pipe_surface * -trace_surface_create(struct trace_texture *tr_tex, +trace_surface_create(struct trace_texture *tr_tex, struct pipe_surface *surface) { struct trace_surface *tr_surf; - + if(!surface) goto error; - + assert(surface->texture == tr_tex->texture); - + tr_surf = CALLOC_STRUCT(trace_surface); if(!tr_surf) goto error; - + memcpy(&tr_surf->base, surface, sizeof(struct pipe_surface)); - + + pipe_reference_init(&tr_surf->base.reference, 1); tr_surf->base.texture = NULL; pipe_texture_reference(&tr_surf->base.texture, &tr_tex->base); tr_surf->surface = surface; return &tr_surf->base; - + error: pipe_surface_reference(&surface, NULL); return NULL; @@ -100,7 +102,7 @@ error: void -trace_surface_destroy(struct trace_texture *tr_tex, +trace_surface_destroy(struct trace_texture *tr_tex, struct pipe_surface *surface) { struct trace_surface *tr_surf = trace_surface(tr_tex, surface); @@ -109,3 +111,45 @@ trace_surface_destroy(struct trace_texture *tr_tex, FREE(tr_surf); } + +struct pipe_transfer * +trace_transfer_create(struct trace_texture *tr_tex, + struct pipe_transfer *transfer) +{ + struct trace_transfer *tr_trans; + + if(!transfer) + goto error; + + assert(transfer->texture == tr_tex->texture); + + tr_trans = CALLOC_STRUCT(trace_transfer); + if(!tr_trans) + goto error; + + memcpy(&tr_trans->base, transfer, sizeof(struct pipe_transfer)); + + tr_trans->base.texture = NULL; + pipe_texture_reference(&tr_trans->base.texture, &tr_tex->base); + tr_trans->transfer = transfer; + assert(tr_trans->base.texture == &tr_tex->base); + + return &tr_trans->base; + +error: + transfer->texture->screen->tex_transfer_destroy(transfer); + return NULL; +} + + +void +trace_transfer_destroy(struct trace_texture *tr_tex, + struct pipe_transfer *transfer) +{ + struct trace_transfer *tr_trans = trace_transfer(tr_tex, transfer); + struct pipe_screen *screen = tr_trans->transfer->texture->screen; + pipe_texture_reference(&tr_trans->base.texture, NULL); + screen->tex_transfer_destroy(tr_trans->transfer); + FREE(tr_trans); +} + diff --git a/src/gallium/drivers/trace/tr_texture.h b/src/gallium/drivers/trace/tr_texture.h index 9e72edb8a3..9c21bc7d27 100644 --- a/src/gallium/drivers/trace/tr_texture.h +++ b/src/gallium/drivers/trace/tr_texture.h @@ -48,48 +48,78 @@ struct trace_surface struct pipe_surface base; struct pipe_surface *surface; - +}; + + +struct trace_transfer +{ + struct pipe_transfer base; + + struct pipe_transfer *transfer; + void *map; }; static INLINE struct trace_texture * -trace_texture(struct trace_screen *tr_scr, +trace_texture(struct trace_screen *tr_scr, struct pipe_texture *texture) { if(!texture) return NULL; + assert(tr_scr); assert(texture->screen == &tr_scr->base); return (struct trace_texture *)texture; } static INLINE struct trace_surface * -trace_surface(struct trace_texture *tr_tex, +trace_surface(struct trace_texture *tr_tex, struct pipe_surface *surface) { if(!surface) return NULL; + assert(tr_tex); assert(surface->texture == &tr_tex->base); return (struct trace_surface *)surface; } +static INLINE struct trace_transfer * +trace_transfer(struct trace_texture *tr_tex, + struct pipe_transfer *transfer) +{ + if(!transfer) + return NULL; + assert(tr_tex); + assert(transfer->texture == &tr_tex->base); + return (struct trace_transfer *)transfer; +} + + struct pipe_texture * -trace_texture_create(struct trace_screen *tr_scr, +trace_texture_create(struct trace_screen *tr_scr, struct pipe_texture *texture); void -trace_texture_destroy(struct trace_screen *tr_scr, +trace_texture_destroy(struct trace_screen *tr_scr, struct pipe_texture *texture); struct pipe_surface * -trace_surface_create(struct trace_texture *tr_tex, +trace_surface_create(struct trace_texture *tr_tex, struct pipe_surface *surface); void trace_surface_destroy(struct trace_texture *tr_tex, struct pipe_surface *surface); +struct pipe_transfer * +trace_transfer_create(struct trace_texture *tr_tex, + struct pipe_transfer *transfer); + +void +trace_transfer_destroy(struct trace_texture *tr_tex, + struct pipe_transfer *transfer); + #endif /* TR_TEXTURE_H_ */ diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c deleted file mode 100644 index c4148fe810..0000000000 --- a/src/gallium/drivers/trace/tr_winsys.c +++ /dev/null @@ -1,450 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "util/u_memory.h" -#include "util/u_hash_table.h" - -#include "tr_dump.h" -#include "tr_state.h" -#include "tr_screen.h" -#include "tr_texture.h" -#include "tr_winsys.h" - - -static unsigned trace_buffer_hash(void *buffer) -{ - return (unsigned)(uintptr_t)buffer; -} - - -static int trace_buffer_compare(void *buffer1, void *buffer2) -{ - return (char *)buffer2 - (char *)buffer1; -} - - -static const char * -trace_winsys_get_name(struct pipe_winsys *_winsys) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - const char *result; - - trace_dump_call_begin("pipe_winsys", "get_name"); - - trace_dump_arg(ptr, winsys); - - result = winsys->get_name(winsys); - - trace_dump_ret(string, result); - - trace_dump_call_end(); - - return result; -} - - -static void -trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys, - struct pipe_surface *surface, - void *context_private) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - - assert(surface); - if(surface->texture) { - struct trace_screen *tr_scr = trace_screen(surface->texture->screen); - struct trace_texture *tr_tex = trace_texture(tr_scr, surface->texture); - struct trace_surface *tr_surf = trace_surface(tr_tex, surface); - surface = tr_surf->surface; - } - - trace_dump_call_begin("pipe_winsys", "flush_frontbuffer"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(ptr, surface); - /* XXX: hide, as there is nothing we can do with this - trace_dump_arg(ptr, context_private); - */ - - winsys->flush_frontbuffer(winsys, surface, context_private); - - trace_dump_call_end(); -} - - -static struct pipe_buffer * -trace_winsys_surface_buffer_create(struct pipe_winsys *_winsys, - unsigned width, unsigned height, - enum pipe_format format, - unsigned usage, - unsigned *pstride) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - unsigned stride; - struct pipe_buffer *result; - - trace_dump_call_begin("pipe_winsys", "surface_buffer_create"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(uint, width); - trace_dump_arg(uint, height); - trace_dump_arg(format, format); - trace_dump_arg(uint, usage); - - result = winsys->surface_buffer_create(winsys, - width, height, - format, - usage, - pstride); - - stride = *pstride; - - trace_dump_arg(uint, stride); - - trace_dump_ret(ptr, result); - - trace_dump_call_end(); - - return result; -} - - -static struct pipe_buffer * -trace_winsys_buffer_create(struct pipe_winsys *_winsys, - unsigned alignment, - unsigned usage, - unsigned size) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - struct pipe_buffer *buffer; - - trace_dump_call_begin("pipe_winsys", "buffer_create"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(uint, alignment); - trace_dump_arg(uint, usage); - trace_dump_arg(uint, size); - - buffer = winsys->buffer_create(winsys, alignment, usage, size); - - trace_dump_ret(ptr, buffer); - - trace_dump_call_end(); - - /* Zero the buffer to avoid dumping uninitialized memory */ - if(buffer->usage & PIPE_BUFFER_USAGE_CPU_WRITE) { - void *map; - map = winsys->buffer_map(winsys, buffer, PIPE_BUFFER_USAGE_CPU_WRITE); - if(map) { - memset(map, 0, buffer->size); - winsys->buffer_unmap(winsys, buffer); - } - } - - return buffer; -} - - -static struct pipe_buffer * -trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, - void *data, - unsigned size) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - struct pipe_buffer *result; - - trace_dump_call_begin("pipe_winsys", "user_buffer_create"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg_begin("data"); - trace_dump_bytes(data, size); - trace_dump_arg_end(); - trace_dump_arg(uint, size); - - result = winsys->user_buffer_create(winsys, data, size); - - trace_dump_ret(ptr, result); - - trace_dump_call_end(); - - /* XXX: Mark the user buffers. (we should wrap pipe_buffers, but is is - * impossible to do so while texture-less surfaces are still around */ - if(result) { - assert(!(result->usage & TRACE_BUFFER_USAGE_USER)); - result->usage |= TRACE_BUFFER_USAGE_USER; - } - - return result; -} - - -void -trace_winsys_user_buffer_update(struct pipe_winsys *_winsys, - struct pipe_buffer *buffer) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - const void *map; - - if(buffer && buffer->usage & TRACE_BUFFER_USAGE_USER) { - map = winsys->buffer_map(winsys, buffer, PIPE_BUFFER_USAGE_CPU_READ); - if(map) { - trace_dump_call_begin("pipe_winsys", "buffer_write"); - - trace_dump_arg(ptr, winsys); - - trace_dump_arg(ptr, buffer); - - trace_dump_arg_begin("data"); - trace_dump_bytes(map, buffer->size); - trace_dump_arg_end(); - - trace_dump_arg_begin("size"); - trace_dump_uint(buffer->size); - trace_dump_arg_end(); - - trace_dump_call_end(); - - winsys->buffer_unmap(winsys, buffer); - } - } -} - - -static void * -trace_winsys_buffer_map(struct pipe_winsys *_winsys, - struct pipe_buffer *buffer, - unsigned usage) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - void *map; - - map = winsys->buffer_map(winsys, buffer, usage); - if(map) { - if(usage & PIPE_BUFFER_USAGE_CPU_WRITE) { - assert(!hash_table_get(tr_ws->buffer_maps, buffer)); - hash_table_set(tr_ws->buffer_maps, buffer, map); - } - } - - return map; -} - - -static void -trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, - struct pipe_buffer *buffer) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - const void *map; - - map = hash_table_get(tr_ws->buffer_maps, buffer); - if(map) { - trace_dump_call_begin("pipe_winsys", "buffer_write"); - - trace_dump_arg(ptr, winsys); - - trace_dump_arg(ptr, buffer); - - trace_dump_arg_begin("data"); - trace_dump_bytes(map, buffer->size); - trace_dump_arg_end(); - - trace_dump_arg_begin("size"); - trace_dump_uint(buffer->size); - trace_dump_arg_end(); - - trace_dump_call_end(); - - hash_table_remove(tr_ws->buffer_maps, buffer); - } - - winsys->buffer_unmap(winsys, buffer); -} - - -static void -trace_winsys_buffer_destroy(struct pipe_winsys *_winsys, - struct pipe_buffer *buffer) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - - trace_dump_call_begin("pipe_winsys", "buffer_destroy"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(ptr, buffer); - - winsys->buffer_destroy(winsys, buffer); - - trace_dump_call_end(); -} - - -static void -trace_winsys_fence_reference(struct pipe_winsys *_winsys, - struct pipe_fence_handle **pdst, - struct pipe_fence_handle *src) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - struct pipe_fence_handle *dst = *pdst; - - trace_dump_call_begin("pipe_winsys", "fence_reference"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(ptr, dst); - trace_dump_arg(ptr, src); - - winsys->fence_reference(winsys, pdst, src); - - trace_dump_call_end(); -} - - -static int -trace_winsys_fence_signalled(struct pipe_winsys *_winsys, - struct pipe_fence_handle *fence, - unsigned flag) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - int result; - - trace_dump_call_begin("pipe_winsys", "fence_signalled"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(ptr, fence); - trace_dump_arg(uint, flag); - - result = winsys->fence_signalled(winsys, fence, flag); - - trace_dump_ret(int, result); - - trace_dump_call_end(); - - return result; -} - - -static int -trace_winsys_fence_finish(struct pipe_winsys *_winsys, - struct pipe_fence_handle *fence, - unsigned flag) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - int result; - - trace_dump_call_begin("pipe_winsys", "fence_finish"); - - trace_dump_arg(ptr, winsys); - trace_dump_arg(ptr, fence); - trace_dump_arg(uint, flag); - - result = winsys->fence_finish(winsys, fence, flag); - - trace_dump_ret(int, result); - - trace_dump_call_end(); - - return result; -} - - -static void -trace_winsys_destroy(struct pipe_winsys *_winsys) -{ - struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct pipe_winsys *winsys = tr_ws->winsys; - - trace_dump_call_begin("pipe_winsys", "destroy"); - - trace_dump_arg(ptr, winsys); - - /* - winsys->destroy(winsys); - */ - - trace_dump_call_end(); - - hash_table_destroy(tr_ws->buffer_maps); - - FREE(tr_ws); -} - - -struct pipe_winsys * -trace_winsys_create(struct pipe_winsys *winsys) -{ - struct trace_winsys *tr_ws; - - if(!winsys) - goto error1; - - tr_ws = CALLOC_STRUCT(trace_winsys); - if(!tr_ws) - goto error1; - - tr_ws->base.destroy = trace_winsys_destroy; - tr_ws->base.get_name = trace_winsys_get_name; - tr_ws->base.flush_frontbuffer = trace_winsys_flush_frontbuffer; - tr_ws->base.surface_buffer_create = trace_winsys_surface_buffer_create; - tr_ws->base.buffer_create = trace_winsys_buffer_create; - tr_ws->base.user_buffer_create = trace_winsys_user_buffer_create; - tr_ws->base.buffer_map = trace_winsys_buffer_map; - tr_ws->base.buffer_unmap = trace_winsys_buffer_unmap; - tr_ws->base.buffer_destroy = trace_winsys_buffer_destroy; - tr_ws->base.fence_reference = trace_winsys_fence_reference; - tr_ws->base.fence_signalled = trace_winsys_fence_signalled; - tr_ws->base.fence_finish = trace_winsys_fence_finish; - - tr_ws->winsys = winsys; - - tr_ws->buffer_maps = hash_table_create(trace_buffer_hash, - trace_buffer_compare); - if(!tr_ws->buffer_maps) - goto error2; - - trace_dump_call_begin("", "pipe_winsys_create"); - trace_dump_ret(ptr, winsys); - trace_dump_call_end(); - - return &tr_ws->base; - -error2: - FREE(tr_ws); -error1: - return winsys; -} diff --git a/src/gallium/include/pipe/internal/p_winsys_screen.h b/src/gallium/include/pipe/internal/p_winsys_screen.h index ee835578b2..f4a29e63c7 100644 --- a/src/gallium/include/pipe/internal/p_winsys_screen.h +++ b/src/gallium/include/pipe/internal/p_winsys_screen.h @@ -68,6 +68,11 @@ struct pipe_winsys const char *(*get_name)( struct pipe_winsys *ws ); /** + * Do any special operations to ensure buffer size is correct + */ + void (*update_buffer)( struct pipe_winsys *ws, + void *context_private ); + /** * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. */ @@ -149,8 +154,7 @@ struct pipe_winsys void (*buffer_unmap)( struct pipe_winsys *ws, struct pipe_buffer *buf ); - void (*buffer_destroy)( struct pipe_winsys *ws, - struct pipe_buffer *buf ); + void (*buffer_destroy)( struct pipe_buffer *buf ); /** Set ptr = fence, with reference counting */ diff --git a/src/gallium/include/pipe/p_atomic.h b/src/gallium/include/pipe/p_atomic.h new file mode 100644 index 0000000000..f2fe083efa --- /dev/null +++ b/src/gallium/include/pipe/p_atomic.h @@ -0,0 +1,382 @@ +/** + * Many similar implementations exist. See for example libwsbm + * or the linux kernel include/atomic.h + * + * No copyright claimed on this file. + * + */ + +#ifndef P_ATOMIC_H +#define P_ATOMIC_H + +#include "p_compiler.h" +#include "p_defines.h" + +#ifdef __cplusplus +extern "C" { +#endif + + +/* Favor OS-provided implementations. + */ +#define PIPE_ATOMIC_OS_UNLOCKED \ + (defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || \ + defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)) + +#define PIPE_ATOMIC_OS_MS_INTERLOCK \ + (!defined(PIPE_CC_GCC) && \ + !PIPE_ATOMIC_OS_UNLOCKED && \ + defined(PIPE_SUBSYSTEM_WINDOWS_USER)) + +#define PIPE_ATOMIC_OS_PROVIDED \ + (PIPE_ATOMIC_OS_UNLOCKED || \ + PIPE_ATOMIC_OS_MS_INTERLOCK) + +/* Where no OS-provided implementation is available, fall back to + * either locally coded assembly or ultimately a mutex-based + * implementation: + */ +#define PIPE_ATOMIC_ASM_GCC_X86 \ + (!PIPE_ATOMIC_OS_PROVIDED && \ + defined(PIPE_CC_GCC) && \ + defined(PIPE_ARCH_X86)) + +/* KW: this was originally used when x86 asm wasn't available. + * Maintain that logic here. + */ +#define PIPE_ATOMIC_GCC_INTRINISIC \ + (!PIPE_ATOMIC_OS_PROVIDED && \ + !PIPE_ATOMIC_ASM_GCC_X86 && \ + defined(PIPE_CC_GCC)) + +#define PIPE_ATOMIC_ASM_MSVC_X86 \ + (!PIPE_ATOMIC_OS_PROVIDED && \ + defined(PIPE_CC_MSVC) && \ + defined(PIPE_ARCH_X86)) + +#define PIPE_ATOMIC_ASM \ + (PIPE_ATOMIC_ASM_GCC_X86 || \ + PIPE_ATOMIC_ASM_GCC_INTRINSIC || \ + PIPE_ATOMIC_ASM_MSVC_X86) + + +/* Where no OS-provided or locally-coded assembly implemenation is + * available, use pipe_mutex: + */ +#define PIPE_ATOMIC_MUTEX \ + (!PIPE_ATOMIC_OS_PROVIDED && \ + !PIPE_ATOMIC_ASM) + + + +#if (PIPE_ATOMIC_ASM_GCC_X86) + +#define PIPE_ATOMIC "GCC x86 assembly" + +struct pipe_atomic { + int32_t count; +}; + +#define p_atomic_set(_v, _i) ((_v)->count = (_i)) +#define p_atomic_read(_v) ((_v)->count) + + +static INLINE boolean +p_atomic_dec_zero(struct pipe_atomic *v) +{ + unsigned char c; + + __asm__ __volatile__("lock; decl %0; sete %1":"+m"(v->count), "=qm"(c) + ::"memory"); + + return c != 0; +} + +static INLINE void +p_atomic_inc(struct pipe_atomic *v) +{ + __asm__ __volatile__("lock; incl %0":"+m"(v->count)); +} + +static INLINE void +p_atomic_dec(struct pipe_atomic *v) +{ + __asm__ __volatile__("lock; decl %0":"+m"(v->count)); +} + +static INLINE int32_t +p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) +{ + return __sync_val_compare_and_swap(&v->count, old, _new); +} +#endif + + + +/* Implementation using GCC-provided synchronization intrinsics + */ +#if (PIPE_ATOMIC_ASM_GCC_INTRINSIC) + +#define PIPE_ATOMIC "GCC Sync Intrinsics" + +struct pipe_atomic { + int32_t count; +}; + +#define p_atomic_set(_v, _i) ((_v)->count = (_i)) +#define p_atomic_read(_v) ((_v)->count) + + +static INLINE boolean +p_atomic_dec_zero(struct pipe_atomic *v) +{ + return (__sync_sub_and_fetch(&v->count, 1) == 0); +} + +static INLINE void +p_atomic_inc(struct pipe_atomic *v) +{ + (void) __sync_add_and_fetch(&v->count, 1); +} + +static INLINE void +p_atomic_dec(struct pipe_atomic *v) +{ + (void) __sync_sub_and_fetch(&v->count, 1); +} + +static INLINE int32_t +p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) +{ + return __sync_val_compare_and_swap(&v->count, old, _new); +} +#endif + + + +/* Unlocked version for single threaded environments, such as some + * windows kernel modules. + */ +#if (PIPE_ATOMIC_OS_UNLOCKED) + +#define PIPE_ATOMIC "Unlocked" + +struct pipe_atomic +{ + int32_t count; +}; + +#define p_atomic_set(_v, _i) ((_v)->count = (_i)) +#define p_atomic_read(_v) ((_v)->count) +#define p_atomic_dec_zero(_v) ((boolean) --(_v)->count) +#define p_atomic_inc(_v) ((void) (_v)->count++) +#define p_atomic_dec(_v) ((void) (_v)->count--) +#define p_atomic_cmpxchg(_v, old, _new) ((_v)->count == old ? (_v)->count = (_new) : (_v)->count) + +#endif + + +/* Locally coded assembly for MSVC on x86: + */ +#if (PIPE_ATOMIC_ASM_MSVC_X86) + +#define PIPE_ATOMIC "MSVC x86 assembly" + +struct pipe_atomic +{ + int32_t count; +}; + +#define p_atomic_set(_v, _i) ((_v)->count = (_i)) +#define p_atomic_read(_v) ((_v)->count) + +static INLINE boolean +p_atomic_dec_zero(struct pipe_atomic *v) +{ + int32_t *pcount = &v->count; + unsigned char c; + + __asm { + mov eax, [pcount] + lock dec dword ptr [eax] + sete byte ptr [c] + } + + return c != 0; +} + +static INLINE void +p_atomic_inc(struct pipe_atomic *v) +{ + int32_t *pcount = &v->count; + + __asm { + mov eax, [pcount] + lock inc dword ptr [eax] + } +} + +static INLINE void +p_atomic_dec(struct pipe_atomic *v) +{ + int32_t *pcount = &v->count; + + __asm { + mov eax, [pcount] + lock dec dword ptr [eax] + } +} + +static INLINE int32_t +p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) +{ + int32_t *pcount = &v->count; + int32_t orig; + + __asm { + mov ecx, [pcount] + mov eax, [old] + mov edx, [_new] + lock cmpxchg [ecx], edx + mov [orig], eax + } + + return orig; +} +#endif + + +#if (PIPE_ATOMIC_OS_MS_INTERLOCK) + +#define PIPE_ATOMIC "MS userspace interlocks" + +#include <windows.h> + +struct pipe_atomic +{ + long count; +}; + +#define p_atomic_set(_v, _i) ((_v)->count = (_i)) +#define p_atomic_read(_v) ((_v)->count) + +static INLINE boolean +p_atomic_dec_zero(struct pipe_atomic *v) +{ + return InterlockedDecrement(&v->count); +} + +static INLINE void +p_atomic_inc(struct pipe_atomic *v) +{ + InterlockedIncrement(&v->count); +} + +static INLINE void +p_atomic_dec(struct pipe_atomic *v) +{ + InterlockedDecrement(&v->count); +} + +static INLINE int32_t +p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) +{ + return InterlockedCompareExchange(&v->count, _new, old); +} + +#endif + + + +#if (PIPE_ATOMIC_MUTEX) + +#define PIPE_ATOMIC "mutex-based fallback" + +#include "pipe/p_thread.h" + +/** + * This implementation should really not be used. + * Add an assembly port instead. It may abort and + * doesn't destroy used mutexes. + */ + +struct pipe_atomic { + pipe_mutex mutex; + int32_t count; +}; + +static INLINE void +p_atomic_set(struct pipe_atomic *v, int32_t i) +{ + pipe_mutex_init(v->mutex); + pipe_mutex_lock(v->mutex); + v->count = i; + pipe_mutex_unlock(v->mutex); +} + +static INLINE int32_t +p_atomic_read(struct pipe_atomic *v) +{ + int32_t ret; + + pipe_mutex_lock(v->mutex); + ret = v->count; + pipe_mutex_unlock(v->mutex); + return ret; +} + +static INLINE void +p_atomic_inc(struct pipe_atomic *v) +{ + pipe_mutex_lock(v->mutex); + ++v->count; + pipe_mutex_unlock(v->mutex); +} + +static INLINE void +p_atomic_dec(struct pipe_atomic *v) +{ + pipe_mutex_lock(v->mutex); + --v->count; + pipe_mutex_unlock(v->mutex); +} + +static INLINE boolean +p_atomic_dec_zero(struct pipe_atomic *v) +{ + boolean ret; + + pipe_mutex_lock(v->mutex); + ret = (--v->count == 0); + pipe_mutex_unlock(v->mutex); + return ret; +} + +static INLINE int32_t +p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) +{ + int32_t ret; + + pipe_mutex_lock(v->mutex); + ret = v->count; + if (ret == old) + v->count = _new; + pipe_mutex_unlock(v->mutex); + + return ret; +} + +#endif + + +#ifndef PIPE_ATOMIC +#error "No pipe_atomic implementation selected" +#endif + + + +#ifdef __cplusplus +} +#endif + +#endif /* P_ATOMIC_H */ diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h index 05cbd2fc4d..7f7657031d 100644 --- a/src/gallium/include/pipe/p_config.h +++ b/src/gallium/include/pipe/p_config.h @@ -111,6 +111,10 @@ #define PIPE_OS_LINUX #endif +#if defined(__FreeBSD__) +#define PIPE_OS_BSD +#endif + #if defined(_WIN32) || defined(WIN32) #define PIPE_OS_WINDOWS #endif @@ -122,9 +126,9 @@ * NOTE: There is no way to auto-detect most of these. */ -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #define PIPE_SUBSYSTEM_DRI -#endif /* PIPE_OS_LINUX */ +#endif /* PIPE_OS_LINUX || PIPE_OS_BSD */ #if defined(PIPE_OS_WINDOWS) #if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 9454cc87db..2452bf3522 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -193,7 +193,6 @@ struct pipe_context { */ /*@{*/ void (*surface_copy)(struct pipe_context *pipe, - boolean do_flip,/**< flip surface contents vertically */ struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, /* don't make this const - diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 4f0b301f31..52d443970b 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -194,6 +194,16 @@ enum pipe_texture_target { /** + * Transfer object usage flags + */ +enum pipe_transfer_usage { + PIPE_TRANSFER_READ, + PIPE_TRANSFER_WRITE, + PIPE_TRANSFER_READ_WRITE //< Read/modify/write +}; + + +/** * Buffer usage flags */ #define PIPE_BUFFER_USAGE_CPU_READ (1 << 0) @@ -205,6 +215,7 @@ enum pipe_texture_target { #define PIPE_BUFFER_USAGE_INDEX (1 << 6) #define PIPE_BUFFER_USAGE_CONSTANT (1 << 7) #define PIPE_BUFFER_USAGE_DISCARD (1 << 8) +#define PIPE_BUFFER_USAGE_DONTBLOCK (1 << 9) /** Pipe driver custom usage flags should be greater or equal to this value */ #define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) @@ -247,6 +258,7 @@ enum pipe_texture_target { #define PIPE_PRIM_QUADS 7 #define PIPE_PRIM_QUAD_STRIP 8 #define PIPE_PRIM_POLYGON 9 +#define PIPE_PRIM_MAX 10 /** diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index b42f98ceba..3f65a60436 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -30,8 +30,9 @@ #define PIPE_FORMAT_H #include "p_compiler.h" -#include "p_debug.h" +/* FIXME: remove these header dependencies */ +#include "util/u_debug.h" #include "util/u_string.h" #ifdef __cplusplus diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index 1219c817b4..1232c87968 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -38,95 +38,6 @@ extern "C" { #endif -/* XXX: these are a kludge. will fix when all surfaces are views into - * textures, and free-floating winsys surfaces go away. - */ -static INLINE void * -pipe_surface_map( struct pipe_surface *surf, unsigned flags ) -{ - struct pipe_screen *screen; - assert(surf->texture); - screen = surf->texture->screen; - return screen->surface_map( screen, surf, flags ); -} - -static INLINE void -pipe_surface_unmap( struct pipe_surface *surf ) -{ - struct pipe_screen *screen; - assert(surf->texture); - screen = surf->texture->screen; - screen->surface_unmap( screen, surf ); -} - - - -/** - * Set 'ptr' to point to 'surf' and update reference counting. - * The old thing pointed to, if any, will be unreferenced first. - * 'surf' may be NULL. - */ -static INLINE void -pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) -{ - /* bump the refcount first */ - if (surf) { - assert(surf->refcount); - surf->refcount++; - } - - if (*ptr) { - struct pipe_screen *screen; - assert((*ptr)->refcount); - assert((*ptr)->texture); - screen = (*ptr)->texture->screen; - screen->tex_surface_release( screen, ptr ); - assert(!*ptr); - } - - *ptr = surf; -} - - -/** - * \sa pipe_surface_reference - */ -static INLINE void -pipe_texture_reference(struct pipe_texture **ptr, - struct pipe_texture *pt) -{ - assert(ptr); - - if (pt) { - assert(pt->refcount); - pt->refcount++; - } - - if (*ptr) { - struct pipe_screen *screen = (*ptr)->screen; - assert(screen); - assert((*ptr)->refcount); - screen->texture_release(screen, ptr); - - assert(!*ptr); - } - - *ptr = pt; -} - - -static INLINE void -pipe_texture_release(struct pipe_texture **ptr) -{ - struct pipe_screen *screen; - assert(ptr); - screen = (*ptr)->screen; - assert((*ptr)->refcount); - screen->texture_release(screen, ptr); - *ptr = NULL; -} - - /** * Convenience wrappers for screen buffer functions. */ @@ -149,7 +60,13 @@ pipe_buffer_map(struct pipe_screen *screen, struct pipe_buffer *buf, unsigned usage) { - return screen->buffer_map(screen, buf, usage); + if(screen->buffer_map_range) { + unsigned offset = 0; + unsigned length = buf->size; + return screen->buffer_map_range(screen, buf, offset, length, usage); + } + else + return screen->buffer_map(screen, buf, usage); } static INLINE void @@ -159,26 +76,74 @@ pipe_buffer_unmap(struct pipe_screen *screen, screen->buffer_unmap(screen, buf); } -/* XXX: thread safety issues! - */ +static INLINE void * +pipe_buffer_map_range(struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned offset, + unsigned length, + unsigned usage) +{ + assert(offset < buf->size); + assert(offset + length <= buf->size); + assert(length); + if(screen->buffer_map_range) + return screen->buffer_map_range(screen, buf, offset, length, usage); + else + return screen->buffer_map(screen, buf, usage); +} + static INLINE void -pipe_buffer_reference(struct pipe_screen *screen, - struct pipe_buffer **ptr, - struct pipe_buffer *buf) +pipe_buffer_flush_mapped_range(struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned offset, + unsigned length) { - if (buf) { - assert(buf->refcount); - buf->refcount++; - } + assert(offset < buf->size); + assert(offset + length <= buf->size); + assert(length); + if(screen->buffer_flush_mapped_range) + screen->buffer_flush_mapped_range(screen, buf, offset, length); +} - if (*ptr) { - assert((*ptr)->refcount); - if(--(*ptr)->refcount == 0) { - screen->buffer_destroy( screen, *ptr ); - } +static INLINE void +pipe_buffer_write(struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned offset, unsigned size, + const void *data) +{ + uint8_t *map; + + assert(offset < buf->size); + assert(offset + size <= buf->size); + assert(size); + + map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(map); + if(map) { + memcpy(map + offset, data, size); + pipe_buffer_flush_mapped_range(screen, buf, offset, size); + pipe_buffer_unmap(screen, buf); } +} - *ptr = buf; +static INLINE void +pipe_buffer_read(struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned offset, unsigned size, + void *data) +{ + uint8_t *map; + + assert(offset < buf->size); + assert(offset + size <= buf->size); + assert(size); + + map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_READ); + assert(map); + if(map) { + memcpy(data, map + offset, size); + pipe_buffer_unmap(screen, buf); + } } diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/include/pipe/p_refcnt.h index 0fd2a40556..60844e40a5 100644 --- a/src/gallium/drivers/trace/tr_winsys.h +++ b/src/gallium/include/pipe/p_refcnt.h @@ -1,8 +1,8 @@ /************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * + * Copyright 2009 VMware, Inc. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,55 +22,67 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ -#ifndef TR_WINSYS_H_ -#define TR_WINSYS_H_ - +#ifndef P_REFCNT_H +#define P_REFCNT_H -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" -#include "pipe/internal/p_winsys_screen.h" - -/** - * It often happens that new data is written directly to the user buffers - * without mapping/unmapping. This flag marks user buffers, so that their - * contents can be dumpped before being used by the pipe context. - */ -#define TRACE_BUFFER_USAGE_USER (1 << 31) +#include "p_defines.h" +#include "p_atomic.h" -struct hash_table; +#ifdef __cplusplus +extern "C" { +#endif -struct trace_winsys +struct pipe_reference { - struct pipe_winsys base; - - struct pipe_winsys *winsys; - - struct hash_table *buffer_maps; + struct pipe_atomic count; }; -static INLINE struct trace_winsys * -trace_winsys(struct pipe_winsys *winsys) +static INLINE void +pipe_reference_init(struct pipe_reference *reference, unsigned count) { - assert(winsys); - return (struct trace_winsys *)winsys; + p_atomic_set(&reference->count, count); } +/** + * Set 'ptr' to point to 'reference' and update reference counting. + * The old thing pointed to, if any, will be unreferenced first. + * 'reference' may be NULL. + * + * XXX: thread safety issues! + */ +static INLINE bool +pipe_reference(struct pipe_reference **ptr, struct pipe_reference *reference) +{ + bool destroy = FALSE; + + /* bump the reference.count first */ + if (reference) { + assert(p_atomic_read(&reference->count) != 0); + p_atomic_inc(&reference->count); + } -struct pipe_winsys * -trace_winsys_create(struct pipe_winsys *winsys); + if (*ptr) { + assert(p_atomic_read(&(*ptr)->count) != 0); + if (p_atomic_dec_zero(&(*ptr)->count)) { + destroy = TRUE; + } + } + *ptr = reference; -void -trace_winsys_user_buffer_update(struct pipe_winsys *winsys, - struct pipe_buffer *buffer); + return destroy; +} +#ifdef __cplusplus +} +#endif -#endif /* TR_WINSYS_H_ */ +#endif /* P_REFCNT_H */ diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index 17d1548253..ed3a026023 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -55,6 +55,7 @@ struct pipe_winsys; struct pipe_buffer; + /** * Gallium screen/adapter context. Basically everything * hardware-specific that doesn't actually require a rendering @@ -112,8 +113,7 @@ struct pipe_screen { const unsigned *stride, struct pipe_buffer *buffer); - void (*texture_release)(struct pipe_screen *, - struct pipe_texture **pt); + void (*texture_destroy)(struct pipe_texture *pt); /** Get a surface which is a "view" into a texture */ struct pipe_surface *(*get_tex_surface)(struct pipe_screen *, @@ -122,18 +122,25 @@ struct pipe_screen { unsigned zslice, unsigned usage ); - /* Surfaces allocated by the above must be released here: - */ - void (*tex_surface_release)( struct pipe_screen *, - struct pipe_surface ** ); + void (*tex_surface_destroy)(struct pipe_surface *); - void *(*surface_map)( struct pipe_screen *, - struct pipe_surface *surface, - unsigned flags ); + /** Get a transfer object for transferring data to/from a texture */ + struct pipe_transfer *(*get_tex_transfer)(struct pipe_screen *, + struct pipe_texture *texture, + unsigned face, unsigned level, + unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, + unsigned w, unsigned h); + + void (*tex_transfer_destroy)(struct pipe_transfer *); + + void *(*transfer_map)( struct pipe_screen *, + struct pipe_transfer *transfer ); - void (*surface_unmap)( struct pipe_screen *, - struct pipe_surface *surface ); + void (*transfer_unmap)( struct pipe_screen *, + struct pipe_transfer *transfer ); /** @@ -195,12 +202,37 @@ struct pipe_screen { void *(*buffer_map)( struct pipe_screen *screen, struct pipe_buffer *buf, unsigned usage ); + /** + * Map a subrange of the buffer data store into the client's address space. + * + * The returned pointer is always relative to buffer start, regardless of + * the specified range. This is different from the ARB_map_buffer_range + * semantics because we don't forbid multiple mappings of the same buffer + * (yet). + */ + void *(*buffer_map_range)( struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned offset, + unsigned length, + unsigned usage); + + /** + * Notify a range that was actually written into. + * + * The range is relative to the buffer start, regardless of the range + * specified to buffer_map_range. This is different from the + * ARB_map_buffer_range semantics because we don't forbid multiple mappings + * of the same buffer (yet). + */ + void (*buffer_flush_mapped_range)( struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned offset, + unsigned length); void (*buffer_unmap)( struct pipe_screen *screen, - struct pipe_buffer *buf ); + struct pipe_buffer *buf ); - void (*buffer_destroy)( struct pipe_screen *screen, - struct pipe_buffer *buf ); + void (*buffer_destroy)( struct pipe_buffer *buf ); /** diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 35df70e7b7..4dafdd6f0a 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -191,8 +191,8 @@ struct tgsi_immediate_float32 /* * GL_EXT_vertex_shader */ -#define TGSI_OPCODE_INDEX 22 -#define TGSI_OPCODE_NEGATE 23 +#define TGSI_OPCODE_INDEX 22 /* considered for removal */ +#define TGSI_OPCODE_NEGATE 23 /* considered for removal */ #define TGSI_OPCODE_MADD TGSI_OPCODE_MAD #define TGSI_OPCODE_FRAC 24 #define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE @@ -206,7 +206,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP #define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ #define TGSI_OPCODE_CROSSPRODUCT 31 -#define TGSI_OPCODE_MULTIPLYMATRIX 32 +#define TGSI_OPCODE_MULTIPLYMATRIX 32 /* considered for removal */ /* * GL_NV_vertex_program1_1 diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 9dc541630c..aad41fab11 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -43,6 +43,8 @@ #include "p_compiler.h" #include "p_defines.h" #include "p_format.h" +#include "p_refcnt.h" +#include "p_screen.h" #ifdef __cplusplus @@ -64,23 +66,20 @@ extern "C" { /* fwd decls */ -struct pipe_screen; struct pipe_surface; - /** * The driver will certainly subclass this to include actual memory * management information. */ struct pipe_buffer { + struct pipe_reference reference; + struct pipe_screen *screen; unsigned alignment; unsigned usage; unsigned size; - - /** Reference count */ - unsigned refcount; }; @@ -109,9 +108,15 @@ struct pipe_rasterizer_state unsigned line_stipple_factor:8; /**< [1..256] actually */ unsigned line_stipple_pattern:16; unsigned line_last_pixel:1; - unsigned bypass_clipping:1; - unsigned bypass_vs:1; /**< Skip the vertex shader. Note that the shader is - still needed though, to indicate inputs/outputs */ + + /** + * Vertex coordinates are pre-transformed to screen space. Skip + * the vertex shader, clipping and viewport processing. Note that + * a vertex shader is still needed though, to indicate the mapping + * from vertex elements to fragment shader input semantics. + */ + unsigned bypass_vs_clip_and_viewport:1; + unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ unsigned flatshade_first:1; /**< take color attribute from the first vertex of a primitive */ unsigned gl_rasterization_rules:1; /**< enable tweaks for GL rasterization? */ @@ -275,18 +280,14 @@ struct pipe_sampler_state */ struct pipe_surface { + struct pipe_reference reference; enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned status; /**< PIPE_SURFACE_STATUS_x */ unsigned clear_value; /**< XXX may be temporary */ unsigned width; /**< logical width in pixels */ unsigned height; /**< logical height in pixels */ - struct pipe_format_block block; - unsigned nblocksx; /**< allocated width in blocks */ - unsigned nblocksy; /**< allocated height in blocks */ - unsigned stride; /**< stride in bytes between rows of blocks */ unsigned layout; /**< PIPE_SURFACE_LAYOUT_x */ unsigned offset; /**< offset from start of buffer, in bytes */ - unsigned refcount; unsigned usage; /**< PIPE_BUFFER_USAGE_* */ struct pipe_texture *texture; /**< texture into which this is a view */ @@ -297,10 +298,35 @@ struct pipe_surface /** + * Transfer object. For data transfer to/from a texture. + */ +struct pipe_transfer +{ + enum pipe_format format; /**< PIPE_FORMAT_x */ + unsigned x; /**< x offset from start of texture image */ + unsigned y; /**< y offset from start of texture image */ + unsigned width; /**< logical width in pixels */ + unsigned height; /**< logical height in pixels */ + struct pipe_format_block block; + unsigned nblocksx; /**< allocated width in blocks */ + unsigned nblocksy; /**< allocated height in blocks */ + unsigned stride; /**< stride in bytes between rows of blocks */ + unsigned usage; /**< PIPE_TRANSFER_* */ + + struct pipe_texture *texture; /**< texture to transfer to/from */ + unsigned face; + unsigned level; + unsigned zslice; +}; + + +/** * Texture object. */ struct pipe_texture { + struct pipe_reference reference; + enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ enum pipe_format format; /**< PIPE_FORMAT_x */ @@ -319,10 +345,6 @@ struct pipe_texture unsigned tex_usage; /* PIPE_TEXTURE_USAGE_* */ - /* These are also refcounted: - */ - unsigned refcount; - struct pipe_screen *screen; /**< screen that this texture belongs to */ }; @@ -359,6 +381,35 @@ struct pipe_vertex_element }; +/* Reference counting helper functions */ +static INLINE void +pipe_buffer_reference(struct pipe_buffer **ptr, struct pipe_buffer *buf) +{ + struct pipe_buffer *old_buf = *ptr; + + if (pipe_reference((struct pipe_reference **)ptr, &buf->reference)) + old_buf->screen->buffer_destroy(old_buf); +} + +static INLINE void +pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) +{ + struct pipe_surface *old_surf = *ptr; + + if (pipe_reference((struct pipe_reference **)ptr, &surf->reference)) + old_surf->texture->screen->tex_surface_destroy(old_surf); +} + +static INLINE void +pipe_texture_reference(struct pipe_texture **ptr, struct pipe_texture *tex) +{ + struct pipe_texture *old_tex = *ptr; + + if (pipe_reference((struct pipe_reference **)ptr, &tex->reference)) + old_tex->screen->texture_destroy(old_tex); +} + + #ifdef __cplusplus } #endif diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h index 8af3cd958b..a9cd77541d 100644 --- a/src/gallium/include/pipe/p_thread.h +++ b/src/gallium/include/pipe/p_thread.h @@ -38,7 +38,7 @@ #include "pipe/p_compiler.h" -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) #include <pthread.h> /* POSIX threads headers */ #include <stdio.h> /* for perror() */ @@ -73,7 +73,7 @@ typedef pthread_cond_t pipe_condvar; static pipe_mutex mutex = PTHREAD_MUTEX_INITIALIZER #define pipe_mutex_init(mutex) \ - pthread_mutex_init(&(mutex), NULL) + (void) pthread_mutex_init(&(mutex), NULL) #define pipe_mutex_destroy(mutex) \ pthread_mutex_destroy(&(mutex)) @@ -134,20 +134,20 @@ static INLINE int pipe_thread_destroy( pipe_thread thread ) typedef CRITICAL_SECTION pipe_mutex; -#define pipe_static_mutex(name) \ - /*static*/ pipe_mutex name = {0,0,0,0,0,0} +#define pipe_static_mutex(mutex) \ + /*static*/ pipe_mutex mutex = {0,0,0,0,0,0} -#define pipe_mutex_init(name) \ - InitializeCriticalSection(&name) +#define pipe_mutex_init(mutex) \ + InitializeCriticalSection(&mutex) -#define pipe_mutex_destroy(name) \ - DeleteCriticalSection(&name) +#define pipe_mutex_destroy(mutex) \ + DeleteCriticalSection(&mutex) -#define pipe_mutex_lock(name) \ - EnterCriticalSection(&name) +#define pipe_mutex_lock(mutex) \ + EnterCriticalSection(&mutex) -#define pipe_mutex_unlock(name) \ - LeaveCriticalSection(&name) +#define pipe_mutex_unlock(mutex) \ + LeaveCriticalSection(&mutex) /* XXX: dummy definitions, make it compile */ @@ -210,7 +210,7 @@ typedef unsigned pipe_condvar; */ typedef struct { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) pthread_key_t key; #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) DWORD key; @@ -225,7 +225,7 @@ typedef struct { static INLINE void pipe_tsd_init(pipe_tsd *tsd) { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) if (pthread_key_create(&tsd->key, NULL/*free*/) != 0) { perror("pthread_key_create(): failed to allocate key for thread specific data"); exit(-1); @@ -242,7 +242,7 @@ pipe_tsd_get(pipe_tsd *tsd) if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) { pipe_tsd_init(tsd); } -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) return pthread_getspecific(tsd->key); #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) assert(0); @@ -259,7 +259,7 @@ pipe_tsd_set(pipe_tsd *tsd, void *value) if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) { pipe_tsd_init(tsd); } -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) if (pthread_setspecific(tsd->key, value) != 0) { perror("pthread_set_specific() failed"); exit(-1); diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h index 54480fa047..435435da29 100644 --- a/src/gallium/include/state_tracker/drm_api.h +++ b/src/gallium/include/state_tracker/drm_api.h @@ -2,14 +2,18 @@ #ifndef _DRM_API_H_ #define _DRM_API_H_ +#include "pipe/p_compiler.h" + struct pipe_screen; struct pipe_winsys; +struct pipe_buffer; struct pipe_context; +struct pipe_texture; struct drm_api { /** - * Special buffer function + * Special buffer functions */ /*@{*/ struct pipe_screen* (*create_screen)(int drmFB, int pciID); @@ -17,17 +21,27 @@ struct drm_api /*@}*/ /** - * Special buffer function + * Special buffer functions */ /*@{*/ - struct pipe_buffer* (*buffer_from_handle)(struct pipe_winsys *winsys, const char *name, unsigned handle); - unsigned (*handle_from_buffer)(struct pipe_winsys *winsys, struct pipe_buffer *buffer); + boolean (*buffer_from_texture)(struct pipe_texture *texture, + struct pipe_buffer **buffer, + unsigned *stride); + struct pipe_buffer* (*buffer_from_handle)(struct pipe_screen *screen, + const char *name, + unsigned handle); + boolean (*handle_from_buffer)(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle); + boolean (*global_handle_from_buffer)(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle); /*@}*/ }; /** * A driver needs to export this symbol */ -extern struct drm_api drm_api_hocks; +extern struct drm_api drm_api_hooks; #endif diff --git a/src/gallium/state_trackers/dri2/Makefile b/src/gallium/state_trackers/dri2/Makefile new file mode 100644 index 0000000000..47750e997e --- /dev/null +++ b/src/gallium/state_trackers/dri2/Makefile @@ -0,0 +1,28 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = dri2drm + +LIBRARY_INCLUDES = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -I$(TOP)/src/mesa/main \ + $(shell pkg-config --cflags-only-I libdrm) + + +C_SOURCES = \ + dri_context.c \ + dri_screen.c \ + dri_drawable.c \ + dri_extensions.c + +# $(TOP)/src/mesa/drivers/dri/common/utils.c \ + $(TOP)/src/mesa/drivers/dri/common/vblank.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_util.c \ + $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c \ + $(TOP)/src/mesa/drivers/common/driverfuncs.c \ + $(TOP)/src/mesa/drivers/dri/common/texmem.c \ + $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c + +include ../../Makefile.template diff --git a/src/gallium/state_trackers/glx/dri/dri_context.c b/src/gallium/state_trackers/dri2/dri_context.c index 9424e18bee..92c26ac70f 100644 --- a/src/gallium/state_trackers/glx/dri/dri_context.c +++ b/src/gallium/state_trackers/dri2/dri_context.c @@ -18,22 +18,29 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ - +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ #include "dri_screen.h" -#include "dri_context.h" -#include "dri_winsys.h" +#include "dri_drawable.h" + + +#include "state_tracker/drm_api.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "pipe/p_context.h" +#include "dri_context.h" + #include "util/u_memory.h" @@ -59,33 +66,33 @@ dri_create_context(const __GLcontextModes *visual, ctx->cPriv = cPriv; ctx->sPriv = sPriv; - driParseConfigFiles(&ctx->optionCache, + driParseConfigFiles(&ctx->optionCache, &screen->optionCache, - sPriv->myNum, + sPriv->myNum, "dri"); - - ctx->pipe = screen->pipe_screen->create_context(screen->pipe_screen, - screen->pipe_winsys, - hw_winsys ); + + ctx->pipe = drm_api_hooks.create_context(screen->pipe_screen); + if (ctx->pipe == NULL) goto fail; - ctx->pipe->priv = ctx; /* I guess */ + /* used in dri_flush_frontbuffer */ + ctx->pipe->priv = ctx; ctx->st = st_create_context(ctx->pipe, visual, st_share); if (ctx->st == NULL) goto fail; - dri_init_extensions( ctx ); + dri_init_extensions(ctx); return GL_TRUE; fail: if (ctx && ctx->st) - st_destroy_context( ctx->st ); + st_destroy_context(ctx->st); if (ctx && ctx->pipe) - ctx->pipe->destroy( ctx->pipe ); + ctx->pipe->destroy(ctx->pipe); FREE(ctx); return FALSE; @@ -97,14 +104,13 @@ dri_destroy_context(__DRIcontextPrivate *cPriv) { struct dri_context *ctx = dri_context(cPriv); struct dri_screen *screen = dri_screen(cPriv->driScreenPriv); - struct pipe_winsys *winsys = screen->winsys; /* No particular reason to wait for command completion before * destroying a context, but it is probably worthwhile flushing it * to avoid having to add code elsewhere to cope with flushing a * partially destroyed context. */ - st_flush(ctx->st); + st_flush(ctx->st, 0, NULL); if (screen->dummyContext == ctx) screen->dummyContext = NULL; @@ -143,26 +149,22 @@ dri_make_current(__DRIcontextPrivate *cPriv, */ screen->dummyContext = ctx; - st_make_current( ctx->st, - draw->stfb, - read->stfb ); + st_make_current(ctx->st, + draw->stfb, + read->stfb); + /* used in dri_flush_frontbuffer */ ctx->dPriv = driDrawPriv; - /* Update window sizes if necessary: - */ - if (draw->stamp != driDrawPriv->lastStamp) { - dri_update_window_size( draw ); - } - - if (read->stamp != driReadPriv->lastStamp) { - dri_update_window_size( read ); - } - - } - else { + if (driDrawPriv) + dri_get_buffers(driDrawPriv); + if (driDrawPriv != driReadPriv && driReadPriv) + dri_get_buffers(driReadPriv); + } else { st_make_current(NULL, NULL, NULL); } return GL_TRUE; } + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/glx/dri/dri_context.h b/src/gallium/state_trackers/dri2/dri_context.h index 4e6a305abb..e910472700 100644 --- a/src/gallium/state_trackers/glx/dri/dri_context.h +++ b/src/gallium/state_trackers/dri2/dri_context.h @@ -1,6 +1,7 @@ /************************************************************************** * - * Copyright (C) 2009 VMware, Inc. All Rights Reserved. + * Copyright (C) 2009 VMware, Inc. + * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -17,12 +18,16 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ #ifndef DRI_CONTEXT_H #define DRI_CONTEXT_H @@ -35,22 +40,21 @@ struct pipe_context; struct pipe_fence; struct st_context; +struct dri_drawable; struct dri_context { + /* dri */ + __DRIscreenPrivate *sPriv; __DRIcontextPrivate *cPriv; __DRIdrawablePrivate *dPriv; + driOptionCache optionCache; + + /* gallium */ struct st_context *st; struct pipe_context *pipe; - - boolean locked; - - /** - * Configuration cache - */ - driOptionCache optionCache; }; @@ -60,13 +64,14 @@ dri_context(__DRIcontextPrivate *driContextPriv) return (struct dri_context *) driContextPriv->driverPrivate; } + /*********************************************************************** * dri_context.c */ -void +void dri_destroy_context(__DRIcontextPrivate * driContextPriv); -boolean +boolean dri_unbind_context(__DRIcontextPrivate * driContextPriv); boolean @@ -80,16 +85,12 @@ dri_create_context(const __GLcontextModes * visual, void *sharedContextPrivate); - /*********************************************************************** - * dri_lock.c + * dri_extensions.c */ -void dri_lock_hardware( struct dri_context *context, - struct dri_drawable *drawable ); - -void dri_unlock_hardware( struct dri_context *dri ); -boolean dri_is_locked( struct dri_context *dri ); - - +void +dri_init_extensions(struct dri_context *ctx); #endif + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/dri2/dri_drawable.c b/src/gallium/state_trackers/dri2/dri_drawable.c new file mode 100644 index 0000000000..2e3f4099e2 --- /dev/null +++ b/src/gallium/state_trackers/dri2/dri_drawable.c @@ -0,0 +1,325 @@ +/************************************************************************** + * + * Copyright 2009, VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ + +#include "dri_screen.h" +#include "dri_context.h" +#include "dri_drawable.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" +#include "state_tracker/drm_api.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" + +#include "util/u_memory.h" + + +static void +dri_copy_to_front(__DRIdrawablePrivate *dPriv, + struct pipe_surface *from, + int x, int y, unsigned w, unsigned h) +{ + /* TODO send a message to the Xserver to copy to the real front buffer */ +} + + +static struct pipe_surface * +dri_surface_from_handle(struct pipe_screen *screen, + unsigned handle, + enum pipe_format format, + unsigned width, + unsigned height, + unsigned pitch) +{ + struct pipe_surface *surface = NULL; + struct pipe_texture *texture = NULL; + struct pipe_texture templat; + struct pipe_buffer *buf = NULL; + + buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle); + if (!buf) + return NULL; + + memset(&templat, 0, sizeof(templat)); + templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; + templat.target = PIPE_TEXTURE_2D; + templat.last_level = 0; + templat.depth[0] = 1; + templat.format = format; + templat.width[0] = width; + templat.height[0] = height; + pf_get_block(templat.format, &templat.block); + + texture = screen->texture_blanket(screen, + &templat, + &pitch, + buf); + + /* we don't need the buffer from this point on */ + pipe_buffer_reference(&buf, NULL); + + if (!texture) + return NULL; + + surface = screen->get_tex_surface(screen, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* we don't need the texture from this point on */ + pipe_texture_reference(&texture, NULL); + return surface; +} + + +/** + * This will be called a drawable is known to have been resized. + */ +void +dri_get_buffers(__DRIdrawablePrivate *dPriv) +{ + struct dri_drawable *drawable = dri_drawable(dPriv); + struct pipe_surface *surface = NULL; + struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen; + __DRIbuffer *buffers = NULL; + __DRIscreen *dri_screen = drawable->sPriv; + __DRIdrawable *dri_drawable = drawable->dPriv; + boolean have_depth = FALSE; + int i, count; + + buffers = (*dri_screen->dri2.loader->getBuffers)(dri_drawable, + &dri_drawable->w, + &dri_drawable->h, + drawable->attachments, + drawable->num_attachments, + &count, + dri_drawable->loaderPrivate); + + if (buffers == NULL) { + return; + } + + /* set one cliprect to cover the whole dri_drawable */ + dri_drawable->x = 0; + dri_drawable->y = 0; + dri_drawable->backX = 0; + dri_drawable->backY = 0; + dri_drawable->numClipRects = 1; + dri_drawable->pClipRects[0].x1 = 0; + dri_drawable->pClipRects[0].y1 = 0; + dri_drawable->pClipRects[0].x2 = dri_drawable->w; + dri_drawable->pClipRects[0].y2 = dri_drawable->h; + dri_drawable->numBackClipRects = 1; + dri_drawable->pBackClipRects[0].x1 = 0; + dri_drawable->pBackClipRects[0].y1 = 0; + dri_drawable->pBackClipRects[0].x2 = dri_drawable->w; + dri_drawable->pBackClipRects[0].y2 = dri_drawable->h; + + for (i = 0; i < count; i++) { + enum pipe_format format = 0; + int index = 0; + + switch (buffers[i].attachment) { + case __DRI_BUFFER_FRONT_LEFT: + index = ST_SURFACE_FRONT_LEFT; + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + case __DRI_BUFFER_FAKE_FRONT_LEFT: + index = ST_SURFACE_FRONT_LEFT; + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + case __DRI_BUFFER_BACK_LEFT: + index = ST_SURFACE_BACK_LEFT; + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + case __DRI_BUFFER_DEPTH: + index = ST_SURFACE_DEPTH; + format = PIPE_FORMAT_Z24S8_UNORM; + break; + case __DRI_BUFFER_STENCIL: + index = ST_SURFACE_DEPTH; + format = PIPE_FORMAT_Z24S8_UNORM; + break; + case __DRI_BUFFER_ACCUM: + default: + assert(0); + } + assert(buffers[i].cpp == 4); + + if (index == ST_SURFACE_DEPTH) { + if (have_depth) + continue; + else + have_depth = TRUE; + } + + surface = dri_surface_from_handle(screen, + buffers[i].name, + format, + dri_drawable->w, + dri_drawable->h, + buffers[i].pitch); + + st_set_framebuffer_surface(drawable->stfb, index, surface); + pipe_surface_reference(&surface, NULL); + } + /* this needed, or else the state tracker fails to pick the new buffers */ + st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h); +} + + +void +dri_flush_frontbuffer(struct pipe_screen *screen, + struct pipe_surface *surf, + void *context_private) +{ + struct dri_context *ctx = (struct dri_context *)context_private; + dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height); +} + + +void +dri_swap_buffers(__DRIdrawablePrivate * dPriv) +{ + /* not needed for dri2 */ + assert(0); +} + + +void +dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + /* not needed for dri2 */ + assert(0); +} + + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +boolean +dri_create_buffer(__DRIscreenPrivate *sPriv, + __DRIdrawablePrivate *dPriv, + const __GLcontextModes *visual, + boolean isPixmap) +{ + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct dri_screen *screen = sPriv->private; + struct dri_drawable *drawable = NULL; + struct pipe_screen *pscreen = screen->pipe_screen; + int i; + + if (isPixmap) + goto fail; /* not implemented */ + + drawable = CALLOC_STRUCT(dri_drawable); + if (drawable == NULL) + goto fail; + + /* XXX: todo: use the pipe_screen queries to figure out which + * render targets are supportable. + */ + assert(visual->redBits == 8); + assert(visual->depthBits == 24 || visual->depthBits == 0); + assert(visual->stencilBits == 8 || visual->stencilBits == 0); + + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits) { + if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) + depthFormat = PIPE_FORMAT_Z24S8_UNORM; + else + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + } else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits) { + if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) + stencilFormat = PIPE_FORMAT_Z24S8_UNORM; + else + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + } else + stencilFormat = PIPE_FORMAT_NONE; + + drawable->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + dPriv->w, + dPriv->h, + (void*) drawable); + if (drawable->stfb == NULL) + goto fail; + + drawable->sPriv = sPriv; + drawable->dPriv = dPriv; + dPriv->driverPrivate = (void *) drawable; + + /* setup dri2 buffers information */ + i = 0; + drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT; +#if 0 + /* TODO incase of double buffer visual, delay fake creation */ + drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT; +#endif + if (visual->doubleBufferMode) + drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT; + if (visual->depthBits) + drawable->attachments[i++] = __DRI_BUFFER_DEPTH; + if (visual->stencilBits) + drawable->attachments[i++] = __DRI_BUFFER_STENCIL; + drawable->num_attachments = i; + + return GL_TRUE; +fail: + FREE(drawable); + return GL_FALSE; +} + + +void +dri_destroy_buffer(__DRIdrawablePrivate *dPriv) +{ + struct dri_drawable *drawable = dri_drawable(dPriv); + + st_unreference_framebuffer(drawable->stfb); + + FREE(drawable); +} + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/glx/dri/dri_drawable.h b/src/gallium/state_trackers/dri2/dri_drawable.h index 1001bb8c57..185c657b35 100644 --- a/src/gallium/state_trackers/glx/dri/dri_drawable.h +++ b/src/gallium/state_trackers/dri2/dri_drawable.h @@ -18,7 +18,7 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. @@ -37,12 +37,14 @@ struct st_framebuffer; struct dri_drawable { + /* dri */ __DRIdrawablePrivate *dPriv; - unsigned stamp; + __DRIscreenPrivate *sPriv; - struct pipe_fence *last_swap_fence; - struct pipe_fence *first_swap_fence; + unsigned attachments[8]; + unsigned num_attachments; + /* gallium */ struct st_framebuffer *stfb; }; @@ -57,17 +59,31 @@ dri_drawable(__DRIdrawablePrivate * driDrawPriv) /*********************************************************************** * dri_drawable.c */ - -void +boolean +dri_create_buffer(__DRIscreenPrivate *sPriv, + __DRIdrawablePrivate *dPriv, + const __GLcontextModes *visual, + boolean isPixmap); + +void +dri_flush_frontbuffer(struct pipe_screen *screen, + struct pipe_surface *surf, + void *context_private); + +void dri_swap_buffers(__DRIdrawablePrivate * dPriv); -void +void dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, - int x, int y, + int x, int y, int w, int h); -void -dri_update_window_size(__DRIdrawablePrivate *dPriv); +void +dri_get_buffers(__DRIdrawablePrivate * dPriv); +void +dri_destroy_buffer(__DRIdrawablePrivate *dPriv); #endif + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/glx/dri/dri_extensions.c b/src/gallium/state_trackers/dri2/dri_extensions.c index 126faf7601..732d1e89b0 100644 --- a/src/gallium/state_trackers/glx/dri/dri_extensions.c +++ b/src/gallium/state_trackers/dri2/dri_extensions.c @@ -18,17 +18,23 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ - - +#include "dri_screen.h" +#include "dri_context.h" +#include "state_tracker/st_context.h" #define need_GL_ARB_multisample +#define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters #define need_GL_ARB_texture_compression #define need_GL_ARB_vertex_buffer_object @@ -53,6 +59,7 @@ const struct dri_extension card_extensions[] = { {"GL_ARB_multisample", GL_ARB_multisample_functions}, {"GL_ARB_multitexture", NULL}, + {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, {"GL_ARB_texture_border_clamp", NULL}, {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, @@ -96,13 +103,17 @@ const struct dri_extension card_extensions[] = { }; - -void -dri_init_extensions( void ) +void +dri_init_extensions(struct dri_context *ctx) { /* The card_extensions list should be pruned according to the - * capabilities of the pipe_screen. This is actually something + * capabilities of the pipe_screen. This is actually something * that can/should be done inside st_create_context(). */ - driInitExtensions( ctx->st->ctx, card_extensions, GL_TRUE ); + if (ctx) + driInitExtensions(ctx->st->ctx, card_extensions, GL_TRUE); + else + driInitExtensions(NULL, card_extensions, GL_FALSE); } + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/glx/dri/dri_screen.c b/src/gallium/state_trackers/dri2/dri_screen.c index f7119b949a..ab5878a4bc 100644 --- a/src/gallium/state_trackers/glx/dri/dri_screen.c +++ b/src/gallium/state_trackers/dri2/dri_screen.c @@ -18,23 +18,29 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ #include "utils.h" #include "vblank.h" #include "xmlpool.h" -#include "dri_context.h" #include "dri_screen.h" +#include "dri_context.h" +#include "dri_drawable.h" #include "pipe/p_context.h" #include "pipe/p_screen.h" #include "pipe/p_inlines.h" +#include "state_tracker/drm_api.h" #include "state_tracker/st_public.h" #include "state_tracker/st_cb_fbo.h" @@ -44,19 +50,15 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY -// DRI_CONF_FORCE_S3TC_ENABLE(false) + /*DRI_CONF_FORCE_S3TC_ENABLE(false)*/ DRI_CONF_ALLOW_LARGE_TEXTURES(1) DRI_CONF_SECTION_END DRI_CONF_END; -const uint __driNConfigOptions = 3; - -static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; - -extern const struct dri_extension card_extensions[]; +const uint __driNConfigOptions = 3; -static const __DRIextension *driScreenExtensions[] = { +static const __DRIextension *dri_screen_extensions[] = { &driReadDrawableExtension, &driCopySubBufferExtension.base, &driSwapControlExtension.base, @@ -66,46 +68,40 @@ static const __DRIextension *driScreenExtensions[] = { }; - - -static const char * -dri_get_name( struct pipe_winsys *winsys ) +static void +dri_get_drm_minor(struct dri_screen *screen) { - return "dri"; + /* TODO get the real minor */ + screen->minor = 0; } - static void -dri_destroy_screen(__DRIscreenPrivate * sPriv) +dri_get_device_id(struct dri_screen *screen) { - struct dri_screen *screen = dri_screen(sPriv); - - screen->pipe_screen->destroy( screen->pipe_screen ); - screen->pipe_winsys->destroy( screen->pipe_winsys ); - FREE(screen); - sPriv->private = NULL; -} + char path[512]; + FILE *file; + /* + * There must be a better way to get the deviceID. + * XXX this only works on Linux. + */ + snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", screen->minor); + file = fopen(path, "r"); + if (!file) { + return; + } -/** - * Get information about previous buffer swaps. - */ -static int -dri_get_swap_info(__DRIdrawablePrivate * dPriv, - __DRIswapInfo * sInfo) -{ - if (dPriv == NULL || - dPriv->driverPrivate == NULL || - sInfo == NULL) - return -1; - else - return 0; + fgets(path, sizeof(path), file); + sscanf(path, "%x", &screen->deviceID); + fclose(file); } + static const __DRIconfig ** dri_fill_in_modes(__DRIscreenPrivate *psp, - unsigned pixel_bits ) + unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, GLboolean have_back_buffer) { __DRIconfig **configs; __GLcontextModes *m; @@ -115,17 +111,19 @@ dri_fill_in_modes(__DRIscreenPrivate *psp, uint8_t msaa_samples_array[1]; unsigned depth_buffer_factor; unsigned back_buffer_factor; + unsigned msaa_samples_factor; GLenum fb_format; GLenum fb_type; int i; static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML + GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; + /* TODO probe the hardware of what is supports */ depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; + depth_bits_array[1] = 24; + depth_bits_array[2] = 24; stencil_bits_array[0] = 0; /* no depth or stencil */ stencil_bits_array[1] = 0; /* z24x8 */ @@ -134,9 +132,10 @@ dri_fill_in_modes(__DRIscreenPrivate *psp, msaa_samples_array[0] = 0; depth_buffer_factor = 3; - back_buffer_factor = 1; + back_buffer_factor = 3; + msaa_samples_factor = 1; - num_modes = depth_buffer_factor * back_buffer_factor * 4; + num_modes = depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4; if (pixel_bits == 16) { fb_format = GL_RGB; @@ -148,85 +147,92 @@ dri_fill_in_modes(__DRIscreenPrivate *psp, } configs = driCreateConfigs(fb_format, fb_type, - depth_bits_array, - stencil_bits_array, depth_buffer_factor, + depth_bits_array, + stencil_bits_array, depth_buffer_factor, back_buffer_modes, back_buffer_factor, - msaa_samples_array, 1); + msaa_samples_array, msaa_samples_factor); if (configs == NULL) { debug_printf("%s: driCreateConfigs failed\n", __FUNCTION__); return NULL; } - return configs; + for (i = 0; configs[i]; i++) { + m = &configs[i]->modes; + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return (const const __DRIconfig **) configs; } +/** + * Get information about previous buffer swaps. + */ +int +dri_get_swap_info(__DRIdrawablePrivate * dPriv, + __DRIswapInfo * sInfo) +{ + if (dPriv == NULL || + dPriv->driverPrivate == NULL || + sInfo == NULL) + return -1; + else + return 0; +} -/* This is the driver specific part of the createNewScreen entry point. - * - * Returns the __GLcontextModes supported by this driver. + +/** + * NULL stub for old dri loaders */ -static const __DRIconfig **dri_init_screen(__DRIscreenPrivate *sPriv) +const __DRIconfig ** +dri_init_screen(__DRIscreenPrivate *sPriv) { - static const __DRIversion ddx_expected = { 1, 6, 0 }; /* hw query */ - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 5, 0 }; /* hw query */ - struct dri_screen *screen; + return NULL; +} - if (!driCheckDriDdxDrmVersions2("dri", - &sPriv->dri_version, &dri_expected, - &sPriv->ddx_version, &ddx_expected, - &sPriv->drm_version, &drm_expected)) { - return NULL; - } - /* Set up dispatch table to cope with all known extensions: - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); +/** + * This is the driver specific part of the createNewScreen entry point. + * + * Returns the __GLcontextModes supported by this driver. + */ +const __DRIconfig ** +dri_init_screen2(__DRIscreenPrivate *sPriv) +{ + struct dri_screen *screen; + /* Set up dispatch table to cope with all known extensions */ + dri_init_extensions(NULL); screen = CALLOC_STRUCT(dri_screen); if (!screen) goto fail; screen->sPriv = sPriv; + screen->fd = sPriv->fd; + dri_get_drm_minor(screen); + dri_get_device_id(screen); sPriv->private = (void *) screen; + sPriv->extensions = dri_screen_extensions; - /* Search the registered winsys' for one that likes this sPriv. - * This is required in situations where multiple devices speak to - * the same DDX and are built into the same binary. - * - * Note that cases like Intel i915 vs i965 doesn't fall into this - * category because they are built into separate binaries. - * - * Nonetheless, it's healthy to keep that level of detail out of - * this state_tracker. - */ - for (i = 0; - i < dri1_winsys_count && - screen->st_winsys == NULL; - i++) - { - screen->dri_winsys = - dri_winsys[i]->check_dri_privates( sPriv->pDevPriv, - sPriv->pSAREA - /* versions, etc?? */)); + screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, screen->deviceID); + if (!screen->pipe_screen) { + debug_printf("%s: failed to create pipe_screen\n", __FUNCTION__); + goto fail; } - + + /* We need to hook in here */ + screen->pipe_screen->flush_frontbuffer = dri_flush_frontbuffer; driParseOptionInfo(&screen->optionCache, - __driConfigOptions, + __driConfigOptions, __driNConfigOptions); - - /* Plug our info back into the __DRIscreenPrivate: - */ - sPriv->private = (void *) screen; - sPriv->extensions = driScreenExtensions; - - return dri_fill_in_modes(sPriv, - dri_priv->cpp * 8, + return dri_fill_in_modes(sPriv, + 4 * 8, 24, 8, 1); @@ -235,21 +241,32 @@ fail: } +void +dri_destroy_screen(__DRIscreenPrivate * sPriv) +{ + struct dri_screen *screen = dri_screen(sPriv); + + screen->pipe_screen->destroy(screen->pipe_screen); + FREE(screen); + sPriv->private = NULL; +} -const struct __DriverAPIRec driDriverAPI = { - .InitScreen = dri_init_screen, - .DestroyScreen = dri_destroy_screen, - .CreateContext = dri_create_context, - .DestroyContext = dri_destroy_context, - .CreateBuffer = dri_create_buffer, - .DestroyBuffer = dri_destroy_buffer, - .SwapBuffers = dri_swap_buffers, - .MakeCurrent = dri_make_current, - .UnbindContext = dri_unbind_context, - .GetSwapInfo = dri_get_swap_info, - .GetDrawableMSC = driDrawableGetMSC32, - .WaitForMSC = driWaitForMSC32, - .CopySubBuffer = dri_copy_sub_buffer, - //.InitScreen2 = dri_init_screen2, +PUBLIC const struct __DriverAPIRec driDriverAPI = { + .InitScreen = dri_init_screen, /* not supported but exported */ + .DestroyScreen = dri_destroy_screen, + .CreateContext = dri_create_context, + .DestroyContext = dri_destroy_context, + .CreateBuffer = dri_create_buffer, + .DestroyBuffer = dri_destroy_buffer, + .SwapBuffers = dri_swap_buffers, /* not supported but exported */ + .MakeCurrent = dri_make_current, + .UnbindContext = dri_unbind_context, + .GetSwapInfo = dri_get_swap_info, + .GetDrawableMSC = driDrawableGetMSC32, + .WaitForMSC = driWaitForMSC32, + .CopySubBuffer = dri_copy_sub_buffer, /* not supported but exported */ + .InitScreen2 = dri_init_screen2, }; + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/glx/dri/dri_screen.h b/src/gallium/state_trackers/dri2/dri_screen.h index 12ed86d22a..fe2676d0be 100644 --- a/src/gallium/state_trackers/glx/dri/dri_screen.h +++ b/src/gallium/state_trackers/dri2/dri_screen.h @@ -18,12 +18,16 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ #ifndef DRI_SCREEN_H #define DRI_SCREEN_H @@ -35,27 +39,8 @@ struct dri_screen { - __DRIScreenPrivate *sPriv; - struct pipe_winsys *pipe_winsys; - struct pipe_screen *pipe_screen; - - struct { - /* Need a pipe_surface pointer to do client-side swapbuffers: - */ - unsigned long buffer_handle; - struct pipe_surface *surface; - struct pipe_texture *texture; - - int pitch; /* row stride, in bytes */ - int width; - int height; - int size; - int cpp; /* for front and back buffers */ - } front; - - int deviceID; - int drmMinor; - + /* dri */ + __DRIscreenPrivate *sPriv; /** * Configuration cache with default values for all contexts @@ -67,8 +52,16 @@ struct dri_screen * which we need a rendering context, but none is currently bound. */ struct dri_context *dummyContext; -}; + /* drm */ + int deviceID; + int fd; + int minor; + + /* gallium */ + struct pipe_winsys *pipe_winsys; + struct pipe_screen *pipe_screen; +}; /** cast wrapper */ @@ -79,5 +72,19 @@ dri_screen(__DRIscreenPrivate *sPriv) } +/*********************************************************************** + * dri_screen.c + */ +const __DRIconfig ** +dri_init_screen2(__DRIscreenPrivate *sPriv); + +void +dri_destroy_screen(__DRIscreenPrivate * sPriv); + +int +dri_get_swap_info(__DRIdrawablePrivate * dPriv, + __DRIswapInfo * sInfo); #endif + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/egl/Makefile b/src/gallium/state_trackers/egl/Makefile index c9124b1c90..692a3c8b76 100644 --- a/src/gallium/state_trackers/egl/Makefile +++ b/src/gallium/state_trackers/egl/Makefile @@ -6,15 +6,15 @@ TOP = ../../../.. include ${TOP}/configs/current -CFLAGS += -g -Wall -Werror=implicit-function-declaration -fPIC \ +CFLAGS := \ -I${GALLIUMDIR}/include \ -I${GALLIUMDIR}/auxiliary \ -I${TOP}/src/mesa/drivers/dri/common \ -I${TOP}/src/mesa \ -I$(TOP)/include \ -I$(TOP)/src/egl/main \ - $(shell pkg-config --cflags pixman-1 xorg-server) \ - ${LIBDRM_CFLAGS} + ${LIBDRM_CFLAGS} \ + ${CFLAGS} ############################################# diff --git a/src/gallium/state_trackers/egl/egl_context.c b/src/gallium/state_trackers/egl/egl_context.c index 8564972b91..36548fae26 100644 --- a/src/gallium/state_trackers/egl/egl_context.c +++ b/src/gallium/state_trackers/egl/egl_context.c @@ -115,7 +115,7 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext _eglInitContext(drv, dpy, &ctx->base, config, attrib_list); - ctx->pipe = drm_api_hocks.create_context(dev->screen); + ctx->pipe = drm_api_hooks.create_context(dev->screen); if (!ctx->pipe) goto err_pipe; diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c index 091d437d81..e6e80b985a 100644 --- a/src/gallium/state_trackers/egl/egl_surface.c +++ b/src/gallium/state_trackers/egl/egl_surface.c @@ -16,11 +16,11 @@ * Util functions */ -static struct drm_mode_modeinfo * +static drmModeModeInfoPtr drm_find_mode(drmModeConnectorPtr connector, _EGLMode *mode) { int i; - struct drm_mode_modeinfo *m = NULL; + drmModeModeInfoPtr m = NULL; for (i = 0; i < connector->count_modes; i++) { m = &connector->modes[i]; @@ -132,7 +132,7 @@ drm_create_texture(_EGLDriver *drv, scrn->front.width = w; scrn->front.height = h; scrn->front.pitch = pitch; - scrn->front.handle = drm_api_hocks.handle_from_buffer(dev->winsys, scrn->buffer); + drm_api_hooks.handle_from_buffer(screen, scrn->buffer, &scrn->front.handle); if (0) goto err_handle; @@ -143,7 +143,7 @@ err_handle: err_surf: pipe_texture_reference(&texture, NULL); err_tex: - pipe_buffer_reference(screen, &buf, NULL); + pipe_buffer_reference(&buf, NULL); err_buf: return; } @@ -173,7 +173,7 @@ drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) pipe_surface_reference(&screen->surface, NULL); pipe_texture_reference(&screen->tex, NULL); - pipe_buffer_reference(dev->screen, &screen->buffer, NULL); + pipe_buffer_reference(&screen->buffer, NULL); screen->shown = 0; } @@ -271,7 +271,6 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, struct drm_device *dev = (struct drm_device *)drv; struct drm_surface *surf = lookup_drm_surface(surface); struct drm_screen *scrn = lookup_drm_screen(dpy, screen); - struct pipe_context *pipe; _EGLMode *mode = _eglLookupMode(dpy, m); int ret; unsigned int i, k; @@ -349,7 +348,7 @@ err_fb: err_bo: pipe_surface_reference(&scrn->surface, NULL); pipe_texture_reference(&scrn->tex, NULL); - pipe_buffer_reference(dev->screen, &scrn->buffer, NULL); + pipe_buffer_reference(&scrn->buffer, NULL); return EGL_FALSE; } @@ -392,7 +391,6 @@ drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) if (surf->screen) { surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL); surf->user->pipe->surface_copy(surf->user->pipe, - 0, surf->screen->surface, 0, 0, back_surf, diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c index dec82c3a00..a22ef381b9 100644 --- a/src/gallium/state_trackers/egl/egl_tracker.c +++ b/src/gallium/state_trackers/egl/egl_tracker.c @@ -92,7 +92,7 @@ drm_update_res(struct drm_device *dev) static void drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector) { - struct drm_mode_modeinfo *m; + drmModeModeInfoPtr m = NULL; int i; for (i = 0; i < connector->count_modes; i++) { @@ -121,7 +121,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) dev->drmFD = fd; drm_get_device_id(dev); - dev->screen = drm_api_hocks.create_screen(dev->drmFD, dev->deviceID); + dev->screen = drm_api_hooks.create_screen(dev->drmFD, dev->deviceID); if (!dev->screen) goto err_screen; dev->winsys = dev->screen->winsys; diff --git a/src/gallium/state_trackers/egl/egl_tracker.h b/src/gallium/state_trackers/egl/egl_tracker.h index 0b4dd9797d..908bab5f9b 100644 --- a/src/gallium/state_trackers/egl/egl_tracker.h +++ b/src/gallium/state_trackers/egl/egl_tracker.h @@ -128,7 +128,7 @@ struct drm_screen /*drmModeCrtcPtr crtc;*/ uint32_t crtcID; - struct drm_mode_modeinfo *mode; + drmModeModeInfoPtr mode; }; diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile index cddfca54fe..f9f4d6be3c 100644 --- a/src/gallium/state_trackers/g3dvl/Makefile +++ b/src/gallium/state_trackers/g3dvl/Makefile @@ -3,7 +3,7 @@ OBJECTS = vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o vl_r16snorm_mc_buf.o GALLIUMDIR = ../.. -CFLAGS += -g -Wall -Werror=implicit-function-declaration -fPIC \ +CFLAGS += -g -Wall -Werror-implicit-function-declaration -fPIC \ -I${GALLIUMDIR}/include \ -I${GALLIUMDIR}/auxiliary \ -I${GALLIUMDIR}/winsys/g3dvl \ diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c index 187a13a560..b61b49a2f8 100644 --- a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c +++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c @@ -237,10 +237,10 @@ static int vlDestroy pipe->delete_fs_state(pipe, basic_csc->fragment_shader); for (i = 0; i < 2; ++i) - pipe_buffer_reference(pipe->screen, &basic_csc->vertex_bufs[i].buffer, NULL); + pipe_buffer_reference(&basic_csc->vertex_bufs[i].buffer, NULL); - pipe_buffer_reference(pipe->screen, &basic_csc->vs_const_buf.buffer, NULL); - pipe_buffer_reference(pipe->screen, &basic_csc->fs_const_buf.buffer, NULL); + pipe_buffer_reference(&basic_csc->vs_const_buf.buffer, NULL); + pipe_buffer_reference(&basic_csc->fs_const_buf.buffer, NULL); FREE(basic_csc); @@ -371,7 +371,7 @@ static int vlCreateVertexShader unsigned int ti; unsigned int i; - assert(context); + assert(csc); pipe = csc->pipe; tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token)); @@ -458,7 +458,7 @@ static int vlCreateFragmentShader unsigned int ti; unsigned int i; - assert(context); + assert(csc); pipe = csc->pipe; tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token)); diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c index 65ddb9f01e..1d8ad0b046 100644 --- a/src/gallium/state_trackers/g3dvl/vl_context.c +++ b/src/gallium/state_trackers/g3dvl/vl_context.c @@ -41,9 +41,7 @@ static int vlInitCommon(struct vlContext *context) rast.line_stipple_factor = 0; rast.line_stipple_pattern = 0; rast.line_last_pixel = 0; - /* Don't need clipping, but viewport mapping done here */ - rast.bypass_clipping = 0; - rast.bypass_vs = 0; + rast.bypass_vs_clip_and_viewport = 0; rast.origin_lower_left = 0; rast.line_width = 1; rast.point_smooth = 0; diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c index 2176bb86d8..eb8270ecad 100644 --- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c @@ -66,7 +66,7 @@ struct vlR16SnormBufferedMC struct vlVertex2f zero_block[3]; unsigned int num_macroblocks; struct vlMpeg2MacroBlock *macroblocks; - struct pipe_surface *tex_surface[3]; + struct pipe_transfer *tex_transfer[3]; short *texels[3]; struct pipe_context *pipe; @@ -187,7 +187,7 @@ static inline int vlGrabBlocks assert(mc); assert(blocks); - tex_pitch = mc->tex_surface[0]->stride / mc->tex_surface[0]->block.size; + tex_pitch = mc->tex_transfer[0]->stride / mc->tex_transfer[0]->block.size; texels = mc->texels[0] + mbpy * tex_pitch + mbpx; for (y = 0; y < 2; ++y) @@ -235,7 +235,7 @@ static inline int vlGrabBlocks for (tb = 0; tb < 2; ++tb) { - tex_pitch = mc->tex_surface[tb + 1]->stride / mc->tex_surface[tb + 1]->block.size; + tex_pitch = mc->tex_transfer[tb + 1]->stride / mc->tex_transfer[tb + 1]->block.size; texels = mc->texels[tb + 1] + mbpy * tex_pitch + mbpx; if ((coded_block_pattern >> (1 - tb)) & 1) @@ -635,8 +635,8 @@ static int vlFlush for (i = 0; i < 3; ++i) { - pipe_surface_unmap(mc->tex_surface[i]); - pipe_surface_reference(&mc->tex_surface[i], NULL); + pipe->screen->transfer_unmap(pipe->screen, mc->tex_transfer[i]); + pipe->screen->tex_transfer_destroy(mc->tex_transfer[i]); } mc->render_target.cbufs[0] = pipe->screen->get_tex_surface @@ -809,14 +809,16 @@ static int vlRenderMacroBlocksMpeg2R16SnormBuffered for (i = 0; i < 3; ++i) { - mc->tex_surface[i] = mc->pipe->screen->get_tex_surface + mc->tex_transfer[i] = mc->pipe->screen->get_tex_transfer ( mc->pipe->screen, mc->textures.all[i], - 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD + 0, 0, 0, PIPE_TRANSFER_WRITE, 0, 0, + surface->texture->width[0], + surface->texture->height[0] ); - mc->texels[i] = pipe_surface_map(mc->tex_surface[i], PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD); + mc->texels[i] = mc->pipe->screen->transfer_map(mc->pipe->screen, mc->tex_transfer[i]); } } @@ -854,7 +856,7 @@ static int vlDestroy pipe->delete_sampler_state(pipe, mc->samplers.all[i]); for (i = 0; i < 3; ++i) - pipe_buffer_reference(pipe->screen, &mc->vertex_bufs.all[i].buffer, NULL); + pipe_buffer_reference(&mc->vertex_bufs.all[i].buffer, NULL); /* Textures 3 & 4 are not created directly, no need to release them here */ for (i = 0; i < 3; ++i) @@ -871,8 +873,8 @@ static int vlDestroy pipe->delete_fs_state(pipe, mc->b_fs[i]); } - pipe_buffer_reference(pipe->screen, &mc->vs_const_buf.buffer, NULL); - pipe_buffer_reference(pipe->screen, &mc->fs_const_buf.buffer, NULL); + pipe_buffer_reference(&mc->vs_const_buf.buffer, NULL); + pipe_buffer_reference(&mc->fs_const_buf.buffer, NULL); FREE(mc->macroblocks); FREE(mc); diff --git a/src/gallium/state_trackers/glx/dri/dri_drawable.c b/src/gallium/state_trackers/glx/dri/dri_drawable.c deleted file mode 100644 index b712acda88..0000000000 --- a/src/gallium/state_trackers/glx/dri/dri_drawable.c +++ /dev/null @@ -1,363 +0,0 @@ -/************************************************************************** - * - * Copyright 2009, VMware, Inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "dri_screen.h" -#include "dri_context.h" -#include "dri_swapbuffers.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" - - -static void -blit_swapbuffers(__DRIdrawablePrivate *dPriv, - __DRIcontextPrivate *cPriv, - struct pipe_surface *src, - const drm_clip_rect_t *rect) -{ - struct dri_screen *screen = dri_screen(dPriv->driScreenPriv); - struct dri_drawable *fb = dri_drawable(dPriv); - struct dri_context *context = dri_context(cPriv); - - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - - struct pipe_surface *dest = fb->front_surface; - const int backWidth = fb->stfb->Base.Width; - const int backHeight = fb->stfb->Base.Height; - int i; - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - (pbox->x2 - pbox->x1) > dest->width || - (pbox->y2 - pbox->y1) > dest->height) - continue; - - box = *pbox; - - if (rect) { - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > backWidth) - box.x2 = backWidth + box.x1; - if (box.y2 - box.y1 > backHeight) - box.y2 = backHeight + box.y1; - - debug_printf("%s: box %d,%d-%d,%d\n", __FUNCTION__, - box.x1, box.y1, box.x2, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - ctx->st->pipe->surface_copy( ctx->st->pipe, - FALSE, - dest, - box.x1, box.y1, - src, - sbox.x1, sbox.y1, - box.x2 - box.x1, - box.y2 - box.y1 ); - } -} - -/** - * Display a colorbuffer surface in an X window. - * Used for SwapBuffers and flushing front buffer rendering. - * - * \param dPriv the window/drawable to display into - * \param surf the surface to display - * \param rect optional subrect of surface to display (may be NULL). - */ -static void -dri_display_surface(__DRIdrawablePrivate *dPriv, - struct pipe_surface *source, - const drm_clip_rect_t *rect) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - struct dri_screen *screen = dri_screen(dPriv->driScreenPriv); - struct dri_context *context = screen->dummy_context; - struct pipe_winsys *winsys = screen->winsys; - - if (!context) - return; - - if (drawable->last_swap_fence) { - winsys->fence_finish( winsys, - drawable->last_swap_fence, - 0 ); - - winsys->fence_reference( winsys, - &drawable->last_swap_fence, - NULL ); - } - - drawable->last_swap_fence = drawable->first_swap_fence; - drawable->first_swap_fence = NULL; - - /* Call lock_hardware to update dPriv cliprects. - */ - dri_lock_hardware(context, drawable); - { - if (dPriv->numClipRects) { - blit_swapbuffers( context, dPriv, source, rect ); - } - } - dri_unlock_hardware(context); - - if (drawble->stamp != drawable->dPriv->lastStamp) { - dri_update_window_size( dpriv ); - } -} - - - -/** - * This will be called a drawable is known to have moved/resized. - */ -void -dri_update_window_size(__DRIdrawablePrivate *dPriv) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - st_resize_framebuffer(drawable->stfb, dPriv->w, dPriv->h); - drawable->stamp = dPriv->lastStamp; -} - - - -void -dri_swap_buffers(__DRIdrawablePrivate * dPriv) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - struct pipe_surface *back_surf; - - assert(drawable); - assert(drawable->stfb); - - back_surf = st_get_framebuffer_surface(drawable->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - st_notify_swapbuffers(drawable->stfb); - dri_display_surface(dPriv, back_surf, NULL); - st_notify_swapbuffers_complete(drawable->stfb); - } -} - - -/** - * Called via glXCopySubBufferMESA() to copy a subrect of the back - * buffer to the front buffer/screen. - */ -void -dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - struct pipe_surface *back_surf; - - assert(drawable); - assert(drawable->stfb); - - back_surf = st_get_framebuffer_surface(drawable->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - drm_clip_rect_t rect; - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - - st_notify_swapbuffers(drawable->stfb); - dri_display_surface(dPriv, back_surf, &rect); - } -} - - - -/* - * The state tracker keeps track of whether the fake frontbuffer has - * been touched by any rendering since the last time we copied its - * contents to the real frontbuffer. Our task is easy: - */ -static void -dri_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - struct dri_context *dri = (struct dri_context *) context_private; - __DRIdrawablePrivate *dPriv = dri->driDrawable; - - dri_display_surface(dPriv, surf, NULL); -} - - - -/* Need to create a surface which wraps the front surface to support - * client-side swapbuffers. - */ -static void -dri_create_front_surface(struct dri_screen *screen, - struct pipe_winsys *winsys, - unsigned handle) -{ - struct pipe_screen *pipe_screen = screen->pipe_screen; - struct pipe_texture *texture; - struct pipe_texture templat; - struct pipe_surface *surface; - struct pipe_buffer *buffer; - unsigned pitch; - - assert(screen->front.cpp == 4); - -// buffer = dri_buffer_from_handle(screen->winsys, -// "front", handle); - - if (!buffer) - return; - - screen->front.buffer = dri_bo(buffer); - - memset(&templat, 0, sizeof(templat)); - templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; - templat.target = PIPE_TEXTURE_2D; - templat.last_level = 0; - templat.depth[0] = 1; - templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; - templat.width[0] = screen->front.width; - templat.height[0] = screen->front.height; - pf_get_block(templat.format, &templat.block); - pitch = screen->front.pitch; - - texture = pipe_screen->texture_blanket(pipe_screen, - &templat, - &pitch, - buffer); - - /* Unref the buffer we don't need it anyways */ - pipe_buffer_reference(screen, &buffer, NULL); - - surface = pipe_screen->get_tex_surface(pipe_screen, - texture, - 0, - 0, - 0, - PIPE_BUFFER_USAGE_GPU_WRITE); - - screen->front.texture = texture; - screen->front.surface = surface; -} - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static boolean -dri_create_buffer(__DRIscreenPrivate *sPriv, - __DRIdrawablePrivate *dPriv, - const __GLcontextModes *visual, - boolean isPixmap) -{ - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct dri_drawable *drawable; - - if (isPixmap) - goto fail; /* not implemented */ - - drawable = CALLOC_STRUCT(dri_drawable); - if (drawable == NULL) - goto fail; - - /* XXX: todo: use the pipe_screen queries to figure out which - * render targets are supportable. - */ - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) { - if (visual->stencilBits == 8) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_X8Z24_UNORM; - } - - drawable->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - dPriv->w, - dPriv->h, - (void*) drawable); - if (drawable->stfb == NULL) - goto fail; - - dPriv->driverPrivate = (void *) drawable; - return GL_TRUE; - -fail: - FREE(drawable); - return GL_FALSE; -} - -static void -dri_destroy_buffer(__DRIdrawablePrivate *dPriv) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - - /* No particular need to wait on fences before dereferencing them: - */ - winsys->fence_reference( winsys, &ctx->last_swap_fence, NULL ); - winsys->fence_reference( winsys, &ctx->first_swap_fence, NULL ); - - st_unreference_framebuffer(drawable->stfb); - - FREE(drawable); -} - diff --git a/src/gallium/state_trackers/glx/dri/dri_lock.c b/src/gallium/state_trackers/glx/dri/dri_lock.c deleted file mode 100644 index b272ab55f3..0000000000 --- a/src/gallium/state_trackers/glx/dri/dri_lock.c +++ /dev/null @@ -1,90 +0,0 @@ -/************************************************************************** - * - * Copyright 2009 VMware, Inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_thread.h" -#include "dri_context.h" -#include "xf86drm.h" - -pipe_static_mutex( lockMutex ); - -static void -dri_contended_lock(struct dri_context *ctx) -{ - __DRIdrawablePrivate *dPriv = ctx->dPriv; - __DRIcontextPrivate *cPriv = ctx->cPriv; - __DRIscreenPrivate *sPriv = cPriv->driScreenPriv; - - drmGetLock(sPriv->fd, cPriv->hHWContext, 0); - - /* Perform round trip communication with server (including dropping - * and retaking the above lock) to update window dimensions: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); -} - - -/* Lock the hardware and validate our state. - */ -void dri_lock_hardware( struct dri_context *ctx ) -{ - __DRIcontextPrivate *cPriv = ctx->cPriv; - __DRIscreenPrivate *sPriv = cPriv->driScreenPriv; - char __ret = 0; - - pipe_mutex_lock(lockMutex); - assert(!ctx->locked); - - DRM_CAS((drmLock *) &sPriv->pSAREA->lock, - cPriv->hHWContext, - (DRM_LOCK_HELD | cPriv->hHWContext), - __ret); - - if (__ret) - dri_contended_lock( ctx ); - - ctx->locked = TRUE; -} - - -/* Unlock the hardware using the global current context - */ -void dri_unlock_hardware( struct dri_context *ctx ) -{ - __DRIcontextPrivate *cPriv = ctx->cPriv; - __DRIscreenPrivate *sPriv = cPriv->driScreenPriv; - - assert(ctx->locked); - ctx->locked = FALSE; - - DRM_UNLOCK(sPriv->fd, - (drmLock *) &sPriv->pSAREA->lock, - cPriv->hHWContext); - - pipe_mutex_unlock(lockMutex); -} diff --git a/src/gallium/state_trackers/glx/xlib/Makefile b/src/gallium/state_trackers/glx/xlib/Makefile index 1b63db1f0e..6d10b090aa 100644 --- a/src/gallium/state_trackers/glx/xlib/Makefile +++ b/src/gallium/state_trackers/glx/xlib/Makefile @@ -3,23 +3,15 @@ include $(TOP)/configs/current LIBNAME = xlib - -DRIVER_INCLUDES = \ +LIBRARY_INCLUDES = \ -I$(TOP)/include \ -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/gallium/auxiliary + -I$(TOP)/src/mesa/main C_SOURCES = \ - glxapi.c \ - fakeglx.c \ - fakeglx_fonts.c \ + glxapi.c \ + fakeglx.c \ + fakeglx_fonts.c \ xm_api.c - include ../../../Makefile.template - -symlinks: - diff --git a/src/gallium/state_trackers/glx/xlib/SConscript b/src/gallium/state_trackers/glx/xlib/SConscript index 14cdad69cb..0dbe341397 100644 --- a/src/gallium/state_trackers/glx/xlib/SConscript +++ b/src/gallium/state_trackers/glx/xlib/SConscript @@ -5,8 +5,7 @@ Import('*') if env['platform'] == 'linux' \ and 'mesa' in env['statetrackers'] \ - and ('softpipe' or 'i915simple' or 'trace') in env['drivers'] \ - and not env['dri']: + and ('softpipe' or 'i915simple' or 'trace') in env['drivers']: env = env.Clone() @@ -15,6 +14,8 @@ if env['platform'] == 'linux' \ '#/src/mesa/main', ]) + env.Append(CPPDEFINES = ['USE_XSHM']) + st_xlib = env.ConvenienceLibrary( target = 'st_xlib', source = [ 'glxapi.c', diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.c b/src/gallium/state_trackers/glx/xlib/glxapi.c index 1ff04804f1..c2cb34d7cf 100644 --- a/src/gallium/state_trackers/glx/xlib/glxapi.c +++ b/src/gallium/state_trackers/glx/xlib/glxapi.c @@ -34,13 +34,21 @@ #include <stdlib.h> #include <stdio.h> #include <string.h> -#include "main/glheader.h" #include "glapi/glapi.h" #include "glxapi.h" #include "fakeglx.h" #include "pipe/p_thread.h" +#if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 303 +# define PUBLIC __attribute__((visibility("default"))) +# define USED __attribute__((used)) +#else +# define PUBLIC +# define USED +#endif + + struct display_dispatch { Display *Dpy; struct _glxapi_table *Table; diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c index 33dc044ad5..a3d1651653 100644 --- a/src/gallium/state_trackers/glx/xlib/xm_api.c +++ b/src/gallium/state_trackers/glx/xlib/xm_api.c @@ -458,7 +458,7 @@ xmesa_free_buffer(XMesaBuffer buffer) XDestroyImage(buffer->tempImage); /* Unreference. If count = zero we'll really delete the buffer */ - _mesa_unreference_framebuffer(&fb); + _mesa_reference_framebuffer(&fb, NULL); XFreeGC(b->xm_visual->display, b->gc); @@ -763,7 +763,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) c->xm_visual = v; c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ - + c->xm_read_buffer = NULL; + /* XXX: create once per Xlib Display. */ screen = driver.create_pipe_screen(); @@ -1037,27 +1038,25 @@ PUBLIC GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, XMesaBuffer readBuffer ) { + XMesaContext old_ctx = XMesaGetCurrentContext(); + + if (old_ctx && old_ctx != c) { + XMesaFlush(old_ctx); + old_ctx->xm_buffer = NULL; + old_ctx->xm_read_buffer = NULL; + } + if (c) { if (!drawBuffer || !readBuffer) return GL_FALSE; /* must specify buffers! */ -#if 0 - /* XXX restore this optimization */ - if (&(c->mesa) == _mesa_get_current_context() - && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer - && c->mesa.ReadBuffer == &readBuffer->mesa_buffer - && xmesa_buffer(c->mesa.DrawBuffer)->wasCurrent) { - /* same context and buffer, do nothing */ - return GL_TRUE; - } -#endif + if (c == old_ctx && + c->xm_buffer == drawBuffer && + c->xm_read_buffer == readBuffer) + return GL_TRUE; c->xm_buffer = drawBuffer; - - /* Call this periodically to detect when the user has begun using - * GL rendering from multiple threads. - */ - _glapi_check_multithread(); + c->xm_read_buffer = readBuffer; st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); @@ -1071,6 +1070,7 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, else { /* Detach */ st_make_current( NULL, NULL, NULL ); + } return GL_TRUE; } @@ -1143,7 +1143,6 @@ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) return; pipe->surface_copy(pipe, - FALSE, surf_front, x, y, /* dest */ surf_back, x, y, /* src */ width, height); diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.h b/src/gallium/state_trackers/glx/xlib/xm_api.h index 2b8302d174..bdd434cd36 100644 --- a/src/gallium/state_trackers/glx/xlib/xm_api.h +++ b/src/gallium/state_trackers/glx/xlib/xm_api.h @@ -295,6 +295,7 @@ struct xmesa_context { struct st_context *st; XMesaVisual xm_visual; /** pixel format info */ XMesaBuffer xm_buffer; /** current drawbuffer */ + XMesaBuffer xm_read_buffer; /** current readbuffer */ }; diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index f4c4b36ea7..79e68de1df 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -42,13 +42,15 @@ #include "pipe/p_screen.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" -#include "util/u_memory.h" -#include "pipe/p_shader_tokens.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/internal/p_winsys_screen.h" #include "cso_cache/cso_context.h" -#include "util/u_draw_quad.h" -#include "util/u_tile.h" -#include "tgsi/tgsi_text.h" -#include "tgsi/tgsi_dump.h" +#include "util/u_draw_quad.h" +#include "util/u_tile.h" +#include "util/u_math.h" +#include "util/u_memory.h" +#include "tgsi/tgsi_text.h" +#include "tgsi/tgsi_dump.h" #include "st_device.h" #include "st_sample.h" @@ -93,7 +95,7 @@ %include "p_compiler.i" -%include "pipe/p_defines.h"; +%include "p_defines.h"; %include "p_format.i" %include "p_device.i" diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index 7b8b64b592..1fdbdf98b2 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -121,7 +121,6 @@ struct st_context { struct pipe_constant_buffer state; memset(&state, 0, sizeof(state)); state.buffer = buffer ? buffer->buffer : NULL; - state.size = buffer->buffer->size; $self->pipe->set_constant_buffer($self->pipe, shader, index, &state); } @@ -161,7 +160,7 @@ struct st_context { struct pipe_vertex_buffer state; memset(&state, 0, sizeof(state)); - state.pitch = pitch; + state.stride = pitch; state.max_index = max_index; state.buffer_offset = buffer_offset; state.buffer = buffer ? buffer->buffer : NULL; @@ -248,7 +247,7 @@ struct st_context { util_draw_vertex_buffer(pipe, vbuf, 0, prim, num_verts, num_attribs); error2: - pipe_buffer_reference(screen, &vbuf, NULL); + pipe_buffer_reference(&vbuf, NULL); error1: ; } @@ -266,13 +265,12 @@ error1: * Surface functions */ - void surface_copy(int do_flip, - struct pipe_surface *dest, + void surface_copy(struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - $self->pipe->surface_copy($self->pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height); + $self->pipe->surface_copy($self->pipe, dest, destx, desty, src, srcx, srcy, width, height); } void surface_fill(struct pipe_surface *dst, diff --git a/src/gallium/state_trackers/python/p_state.i b/src/gallium/state_trackers/python/p_state.i index 7f5760b3b6..110b3d5da4 100644 --- a/src/gallium/state_trackers/python/p_state.i +++ b/src/gallium/state_trackers/python/p_state.i @@ -37,6 +37,7 @@ %ignore winsys; %ignore pipe_vertex_buffer::buffer; +%include "pipe/p_compiler.h"; %include "pipe/p_state.h"; diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i index 08ba0ebe4d..1e64fc8e41 100644 --- a/src/gallium/state_trackers/python/p_texture.i +++ b/src/gallium/state_trackers/python/p_texture.i @@ -121,6 +121,50 @@ pipe_put_tile_rgba($self, x, y, w, h, rgba); } + %cstring_output_allocate_size(char **STRING, int *LENGTH, free(*$1)); + void + get_tile_rgba8(unsigned x, unsigned y, unsigned w, unsigned h, char **STRING, int *LENGTH) + { + unsigned surface_usage; + float *rgba; + unsigned char *rgba8; + unsigned i, j, k; + + *LENGTH = 0; + *STRING = NULL; + + if (!$self) + return; + + *LENGTH = h*w*4; + *STRING = (char *) malloc(*LENGTH); + if(!*STRING) + return; + + rgba = malloc(w*4*sizeof(float)); + if(!rgba) + return; + + rgba8 = (unsigned char *) *STRING; + + /* XXX: force mappable surface */ + surface_usage = $self->usage; + $self->usage |= PIPE_BUFFER_USAGE_CPU_READ; + + for(j = 0; j < h; ++j) { + pipe_get_tile_rgba($self, + x, y + j, w, 1, + rgba); + for(i = 0; i < w; ++i) + for(k = 0; k <4; ++k) + rgba8[j*w*4 + i*4 + k] = float_to_ubyte(rgba[i*4 + k]); + } + + $self->usage = surface_usage; + + free(rgba); + } + void get_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, unsigned *z) { pipe_get_tile_z($self, x, y, w, h, z); @@ -181,7 +225,7 @@ struct st_buffer { unsigned __len__(void) { - assert($self->buffer->refcount); + assert(p_atomic_read(&$self->buffer->reference.count) > 0); return $self->buffer->size; } @@ -191,7 +235,7 @@ struct st_buffer { struct pipe_screen *screen = $self->st_dev->screen; const char *map; - assert($self->buffer->refcount); + assert(p_atomic_read(&$self->buffer->reference.count) > 0); *LENGTH = $self->buffer->size; *STRING = (char *) malloc($self->buffer->size); @@ -211,7 +255,7 @@ struct st_buffer { struct pipe_screen *screen = $self->st_dev->screen; char *map; - assert($self->buffer->refcount); + assert(p_atomic_read(&$self->buffer->reference.count) > 0); if(offset > $self->buffer->size) { PyErr_SetString(PyExc_ValueError, "offset must be smaller than buffer size"); diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index f418f80d7b..e6999a2211 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -35,21 +35,19 @@ import model import parser +try: + from struct import unpack_from +except ImportError: + def unpack_from(fmt, buf, offset=0): + size = struct.calcsize(fmt) + return struct.unpack(fmt, buf[offset:offset + size]) + + def make_image(surface): - pixels = gallium.FloatArray(surface.height*surface.width*4) - surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + data = surface.get_tile_rgba8(0, 0, surface.width, surface.height) import Image - outimage = Image.new( - mode='RGB', - size=(surface.width, surface.height), - color=(0,0,0)) - outpixels = outimage.load() - for y in range(0, surface.height): - for x in range(0, surface.width): - offset = (y*surface.width + x)*4 - r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] - outpixels[x, y] = r, g, b + outimage = Image.fromstring('RGBA', (surface.width, surface.height), data, "raw", 'RGBA', 0, 1) return outimage def save_image(filename, surface): @@ -77,6 +75,7 @@ def show_image(surface): root.mainloop() +verbose = 1 class Struct: @@ -297,6 +296,7 @@ class Context(Object): self.zsbuf = None self.vbufs = [] self.velems = [] + self.dirty = False def destroy(self): pass @@ -374,18 +374,23 @@ class Context(Object): _state.ucp = ucp self.real.set_clip(_state) + def dump_constant_buffer(self, buffer): + if verbose < 2: + return + + data = buffer.read() + format = '4f' + index = 0 + for offset in range(0, len(data), struct.calcsize(format)): + x, y, z, w = unpack_from(format, data, offset) + sys.stdout.write('\tCONST[%2u] = {%10.4f, %10.4f, %10.4f, %10.4f}\n' % (index, x, y, z, w)) + index += 1 + def set_constant_buffer(self, shader, index, state): if state is not None: self.real.set_constant_buffer(shader, index, state.buffer) - if 1: - data = state.buffer.read() - format = '4f' - index = 0 - for offset in range(0, len(data), struct.calcsize(format)): - x, y, z, w = struct.unpack_from(format, data, offset) - sys.stdout.write('\tCONST[%2u] = {%10.4f, %10.4f, %10.4f, %10.4f}\n' % (index, x, y, z, w)) - index += 1 + self.dump_constant_buffer(state.buffer) def set_framebuffer_state(self, state): _state = gallium.Framebuffer() @@ -436,6 +441,9 @@ class Context(Object): pass def dump_vertices(self, start, count): + if verbose < 2: + return + for index in range(start, start + count): if index >= start + 16: sys.stdout.write('\t...\n') @@ -454,12 +462,15 @@ class Context(Object): }[velem.src_format] data = vbuf.buffer.read() - values = struct.unpack_from(format, data, offset) + values = unpack_from(format, data, offset) sys.stdout.write('\t\t{' + ', '.join(map(str, values)) + '},\n') assert len(values) == velem.nr_components sys.stdout.write('\t},\n') def dump_indices(self, ibuf, isize, start, count): + if verbose < 2: + return + format = { 1: 'B', 2: 'H', @@ -477,7 +488,7 @@ class Context(Object): sys.stdout.write('\t...\n') break offset = i*isize - index, = struct.unpack_from(format, data, offset) + index, = unpack_from(format, data, offset) sys.stdout.write('\t\t%u,\n' % index) minindex = min(minindex, index) maxindex = max(maxindex, index) @@ -489,25 +500,35 @@ class Context(Object): self.dump_vertices(start, count) self.real.draw_arrays(mode, start, count) + + self.dirty = True def draw_elements(self, indexBuffer, indexSize, mode, start, count): - minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count) - self.dump_vertices(minindex, maxindex - minindex) + if verbose >= 2: + minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count) + self.dump_vertices(minindex, maxindex - minindex) self.real.draw_elements(indexBuffer, indexSize, mode, start, count) + + self.dirty = True def draw_range_elements(self, indexBuffer, indexSize, minIndex, maxIndex, mode, start, count): - minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count) - minindex = min(minindex, minIndex) - maxindex = min(maxindex, maxIndex) - self.dump_vertices(minindex, maxindex - minindex) + if verbose >= 2: + minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count) + minindex = min(minindex, minIndex) + maxindex = min(maxindex, maxIndex) + self.dump_vertices(minindex, maxindex - minindex) self.real.draw_range_elements(indexBuffer, indexSize, minIndex, maxIndex, mode, start, count) + + self.dirty = True def flush(self, flags): self.real.flush(flags) - if flags & gallium.PIPE_FLUSH_FRAME: - self._update() + if self.dirty: + if flags & gallium.PIPE_FLUSH_FRAME: + self._update() + self.dirty = False return None def clear(self, surface, value): @@ -553,7 +574,8 @@ class Interpreter(parser.TraceDumper): if (call.klass, call.method) in self.ignore_calls: return - parser.TraceDumper.handle_call(self, call) + if verbose >= 1: + parser.TraceDumper.handle_call(self, call) args = [self.interpret_arg(arg) for name, arg in call.args] diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py index 193479f7d6..d3ccb6c2f4 100644 --- a/src/gallium/state_trackers/python/samples/tri.py +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -31,20 +31,10 @@ from gallium import * def make_image(surface): - pixels = FloatArray(surface.height*surface.width*4) - surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + data = surface.get_tile_rgba8(0, 0, surface.width, surface.height) import Image - outimage = Image.new( - mode='RGB', - size=(surface.width, surface.height), - color=(0,0,0)) - outpixels = outimage.load() - for y in range(0, surface.height): - for x in range(0, surface.width): - offset = (y*surface.width + x)*4 - r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] - outpixels[x, y] = r, g, b + outimage = Image.fromstring('RGBA', (surface.width, surface.height), data, "raw", 'RGBA', 0, 1) return outimage def save_image(filename, surface): diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index 20dd8d269d..366d4eb19a 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -51,11 +51,20 @@ st_device_really_destroy(struct st_device *st_dev) } +static void +st_device_reference(struct st_device **ptr, struct st_device *st_dev) +{ + struct st_device *old_dev = *ptr; + + if (pipe_reference((struct pipe_reference **)ptr, &st_dev->reference)) + st_device_really_destroy(old_dev); +} + + void st_device_destroy(struct st_device *st_dev) { - if(!--st_dev->refcount) - st_device_really_destroy(st_dev); + st_device_reference(&st_dev, NULL); } @@ -72,7 +81,7 @@ st_device_create_from_st_winsys(const struct st_winsys *st_ws) if(!st_dev) return NULL; - st_dev->refcount = 1; + pipe_reference_init(&st_dev->reference, 1); st_dev->st_ws = st_ws; st_dev->real_screen = st_ws->screen_create(); @@ -124,8 +133,7 @@ st_context_destroy(struct st_context *st_ctx) FREE(st_ctx); - if(!--st_dev->refcount) - st_device_really_destroy(st_dev); + st_device_reference(&st_dev, NULL); } } @@ -139,8 +147,7 @@ st_context_create(struct st_device *st_dev) if(!st_ctx) return NULL; - st_ctx->st_dev = st_dev; - ++st_dev->refcount; + st_device_reference(&st_ctx->st_dev, st_dev); st_ctx->real_pipe = st_dev->st_ws->context_create(st_dev->real_screen); if(!st_ctx->real_pipe) { @@ -185,8 +192,7 @@ st_context_create(struct st_device *st_dev) memset(&rasterizer, 0, sizeof(rasterizer)); rasterizer.front_winding = PIPE_WINDING_CW; rasterizer.cull_mode = PIPE_WINDING_NONE; - rasterizer.bypass_clipping = 1; - /*rasterizer.bypass_vs = 1;*/ + rasterizer.bypass_vs_clip_and_viewport = 1; cso_set_rasterizer(st_ctx->cso, &rasterizer); } @@ -225,7 +231,7 @@ st_context_create(struct st_device *st_dev) { struct pipe_screen *screen = st_dev->screen; struct pipe_texture templat; - struct pipe_surface *surface; + struct pipe_transfer *transfer; unsigned i; memset( &templat, 0, sizeof( templat ) ); @@ -241,17 +247,21 @@ st_context_create(struct st_device *st_dev) st_ctx->default_texture = screen->texture_create( screen, &templat ); if(st_ctx->default_texture) { - surface = screen->get_tex_surface( screen, - st_ctx->default_texture, 0, 0, 0, - PIPE_BUFFER_USAGE_CPU_WRITE ); - if(surface) { + transfer = screen->get_tex_transfer(screen, + st_ctx->default_texture, + 0, 0, 0, + PIPE_TRANSFER_WRITE, + 0, 0, + st_ctx->default_texture->width[0], + st_ctx->default_texture->height[0]); + if (transfer) { uint32_t *map; - map = (uint32_t *) pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_WRITE ); + map = (uint32_t *) screen->transfer_map(screen, transfer); if(map) { *map = 0x00000000; - pipe_surface_unmap( surface ); + screen->transfer_unmap(screen, transfer); } - pipe_surface_reference(&surface, NULL); + screen->tex_transfer_destroy(transfer); } } @@ -263,24 +273,19 @@ st_context_create(struct st_device *st_dev) /* vertex shader */ { - struct pipe_shader_state vert_shader; - const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, TGSI_SEMANTIC_GENERIC }; const uint semantic_indexes[] = { 0, 0 }; st_ctx->vs = util_make_vertex_passthrough_shader(st_ctx->pipe, 2, semantic_names, - semantic_indexes, - &vert_shader); + semantic_indexes); cso_set_vertex_shader_handle(st_ctx->cso, st_ctx->vs); } /* fragment shader */ { - struct pipe_shader_state frag_shader; - st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, - &frag_shader); + st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe); cso_set_fragment_shader_handle(st_ctx->cso, st_ctx->fs); } @@ -292,8 +297,7 @@ void st_buffer_destroy(struct st_buffer *st_buf) { if(st_buf) { - struct pipe_screen *screen = st_buf->st_dev->screen; - pipe_buffer_reference(screen, &st_buf->buffer, NULL); + pipe_buffer_reference(&st_buf->buffer, NULL); FREE(st_buf); } } diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h index 7cfe6de9f6..0641aff149 100644 --- a/src/gallium/state_trackers/python/st_device.h +++ b/src/gallium/state_trackers/python/st_device.h @@ -68,13 +68,13 @@ struct st_context { struct st_device { + /* FIXME: we also need to refcount for textures and surfaces... */ + struct pipe_reference reference; + const struct st_winsys *st_ws; struct pipe_screen *real_screen; struct pipe_screen *screen; - - /* FIXME: we also need to refcount for textures and surfaces... */ - unsigned refcount; }; diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c index 7765df3c4a..c2ffe9fce1 100644 --- a/src/gallium/state_trackers/python/st_sample.c +++ b/src/gallium/state_trackers/python/st_sample.c @@ -451,6 +451,7 @@ st_sample_dxt_pixel_block(enum pipe_format format, break; default: assert(0); + return; } i = st_random() % n; @@ -524,26 +525,42 @@ st_sample_pixel_block(enum pipe_format format, void st_sample_surface(struct pipe_surface *surface, float *rgba) { - const struct pipe_format_block *block = &surface->block; - unsigned rgba_stride = surface->width*4; + struct pipe_screen *screen = surface->texture->screen; + uint rgba_stride = surface->width * 4; + struct pipe_transfer *transfer; void *raw; - unsigned x, y; - raw = pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_READ); - if(!raw) + transfer = screen->get_tex_transfer(screen, + surface->texture, + surface->face, + surface->level, + surface->zslice, + PIPE_TRANSFER_READ, + 0, 0, + surface->width, + surface->height); + if (!transfer) return; - for (y = 0; y < surface->nblocksy; ++y) { - for(x = 0; x < surface->nblocksx; ++x) { - st_sample_pixel_block(surface->format, - block, - (uint8_t*)raw + y*surface->stride + x*block->size, - rgba + y*block->height*rgba_stride + x*block->width*4, - rgba_stride, - MIN2(block->width, surface->width - x*block->width), - MIN2(block->height, surface->height - y*block->height)); - } + raw = screen->transfer_map(screen, transfer); + if (raw) { + const struct pipe_format_block *block = &transfer->block; + uint x, y; + + for (y = 0; y < transfer->nblocksy; ++y) { + for (x = 0; x < transfer->nblocksx; ++x) { + st_sample_pixel_block(surface->format, + block, + (uint8_t *) raw + y * transfer->stride + x * block->size, + rgba + y * block->height * rgba_stride + x * block->width * 4, + rgba_stride, + MIN2(block->width, surface->width - x*block->width), + MIN2(block->height, surface->height - y*block->height)); + } + } + + screen->transfer_unmap(screen, transfer); } - pipe_surface_unmap(surface); + screen->tex_transfer_destroy(transfer); } diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index 4d798df99b..426f347d18 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -124,7 +124,7 @@ st_softpipe_buffer_create(struct pipe_winsys *winsys, { struct st_softpipe_buffer *buffer = CALLOC_STRUCT(st_softpipe_buffer); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; @@ -149,7 +149,7 @@ st_softpipe_user_buffer_create(struct pipe_winsys *winsys, if(!buffer) return NULL; - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.size = bytes; buffer->userBuffer = TRUE; buffer->data = ptr; diff --git a/src/gallium/state_trackers/wgl/SConscript b/src/gallium/state_trackers/wgl/SConscript index c72f495735..2141b02d68 100644 --- a/src/gallium/state_trackers/wgl/SConscript +++ b/src/gallium/state_trackers/wgl/SConscript @@ -14,13 +14,10 @@ if env['platform'] in ['windows']: env.Append(CPPDEFINES = [ '_GDI32_', # prevent wgl* being declared __declspec(dllimport) 'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers - '__GL_EXPORTS', - '_GNU_H_WINDOWS32_DEFINES', + 'WIN32_THREADS', # use Win32 thread API ]) sources = [ - 'stw.c', - 'icd/stw_icd.c', 'wgl/stw_wgl.c', diff --git a/src/gallium/state_trackers/wgl/icd/stw_icd.c b/src/gallium/state_trackers/wgl/icd/stw_icd.c index 1aa4b8a6e2..faf7f2f410 100644 --- a/src/gallium/state_trackers/wgl/icd/stw_icd.c +++ b/src/gallium/state_trackers/wgl/icd/stw_icd.c @@ -30,82 +30,17 @@ #include "GL/gl.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "pipe/p_thread.h" #include "shared/stw_public.h" #include "icd/stw_icd.h" -#include "stw.h" +#define DBG 0 -#define DRV_CONTEXT_MAX 32 +static GLCLTPROCTABLE cpt; +static boolean cpt_initialized = FALSE; -struct stw_icd -{ - pipe_mutex mutex; - - GLCLTPROCTABLE cpt; - boolean cpt_initialized; - - struct { - struct stw_context *ctx; - } ctx_array[DRV_CONTEXT_MAX]; -}; - - -static struct stw_icd *stw_icd = NULL; - - -boolean -stw_icd_init( void ) -{ - static struct stw_icd stw_icd_storage; - - assert(!stw_icd); - - stw_icd = &stw_icd_storage; - memset(stw_icd, 0, sizeof *stw_icd); - - pipe_mutex_init( stw_icd->mutex ); - - return TRUE; -} - -void -stw_icd_cleanup(void) -{ - int i; - - if (!stw_icd) - return; - - pipe_mutex_lock( stw_icd->mutex ); - { - /* Ensure all contexts are destroyed */ - for (i = 0; i < DRV_CONTEXT_MAX; i++) - if (stw_icd->ctx_array[i].ctx) - stw_delete_context( stw_icd->ctx_array[i].ctx ); - } - pipe_mutex_unlock( stw_icd->mutex ); - - pipe_mutex_init( stw_icd->mutex ); - stw_icd = NULL; -} - - -static struct stw_context * -lookup_context( struct stw_icd *icd, - DHGLRC dhglrc ) -{ - if (dhglrc == 0 || - dhglrc >= DRV_CONTEXT_MAX) - return NULL; - - if (icd == NULL) - return NULL; - - return icd->ctx_array[dhglrc - 1].ctx; -} BOOL APIENTRY DrvCopyContext( @@ -113,63 +48,16 @@ DrvCopyContext( DHGLRC dhrcDest, UINT fuMask ) { - BOOL ret = FALSE; - - if (!stw_icd) - return FALSE; - - pipe_mutex_lock( stw_icd->mutex ); - { - struct stw_context *src = lookup_context( stw_icd, dhrcSource ); - struct stw_context *dst = lookup_context( stw_icd, dhrcDest ); - - if (src == NULL || dst == NULL) - goto done; - - ret = stw_copy_context( src, dst, fuMask ); - } -done: - pipe_mutex_unlock( stw_icd->mutex ); - - return ret; + return stw_copy_context(dhrcSource, dhrcDest, fuMask); } + DHGLRC APIENTRY DrvCreateLayerContext( HDC hdc, INT iLayerPlane ) { - DHGLRC handle = 0; - - if (!stw_icd) - return handle; - - pipe_mutex_lock( stw_icd->mutex ); - { - int i; - - for (i = 0; i < DRV_CONTEXT_MAX; i++) { - if (stw_icd->ctx_array[i].ctx == NULL) - break; - } - - /* No slot available, fail: - */ - if (i == DRV_CONTEXT_MAX) - goto done; - - stw_icd->ctx_array[i].ctx = stw_create_context( hdc, iLayerPlane ); - if (stw_icd->ctx_array[i].ctx == NULL) - goto done; - - /* success: - */ - handle = (DHGLRC) i + 1; - } -done: - pipe_mutex_unlock( stw_icd->mutex ); - - return handle; + return stw_create_layer_context( hdc, iLayerPlane ); } DHGLRC APIENTRY @@ -183,30 +71,7 @@ BOOL APIENTRY DrvDeleteContext( DHGLRC dhglrc ) { - BOOL ret = FALSE; - - if (!stw_icd) - return ret; - - pipe_mutex_lock( stw_icd->mutex ); - { - struct stw_context *ctx; - - ctx = lookup_context( stw_icd, dhglrc ); - if (ctx == NULL) - goto done; - - if (stw_delete_context( ctx ) == FALSE) - goto done; - - stw_icd->ctx_array[dhglrc - 1].ctx = NULL; - ret = TRUE; - - } -done: - pipe_mutex_unlock( stw_icd->mutex ); - - return ret; + return stw_delete_context( dhglrc ); } BOOL APIENTRY @@ -217,7 +82,8 @@ DrvDescribeLayerPlane( UINT nBytes, LPLAYERPLANEDESCRIPTOR plpd ) { - debug_printf( "%s\n", __FUNCTION__ ); + if (DBG) + debug_printf( "%s\n", __FUNCTION__ ); return FALSE; } @@ -233,8 +99,9 @@ DrvDescribePixelFormat( r = stw_pixelformat_describe( hdc, iPixelFormat, cjpfd, ppfd ); - debug_printf( "%s( %p, %d, %u, %p ) = %d\n", - __FUNCTION__, hdc, iPixelFormat, cjpfd, ppfd, r ); + if (DBG) + debug_printf( "%s( %p, %d, %u, %p ) = %d\n", + __FUNCTION__, hdc, iPixelFormat, cjpfd, ppfd, r ); return r; } @@ -247,7 +114,8 @@ DrvGetLayerPaletteEntries( INT cEntries, COLORREF *pcr ) { - debug_printf( "%s\n", __FUNCTION__ ); + if (DBG) + debug_printf( "%s\n", __FUNCTION__ ); return 0; } @@ -260,7 +128,8 @@ DrvGetProcAddress( r = stw_get_proc_address( lpszProc ); - debug_printf( "%s( \", __FUNCTION__%s\" ) = %p\n", lpszProc, r ); + if (DBG) + debug_printf( "%s( \", __FUNCTION__%s\" ) = %p\n", lpszProc, r ); return r; } @@ -271,7 +140,8 @@ DrvRealizeLayerPalette( INT iLayerPlane, BOOL bRealize ) { - debug_printf( "%s\n", __FUNCTION__ ); + if (DBG) + debug_printf( "%s\n", __FUNCTION__ ); return FALSE; } @@ -280,32 +150,7 @@ BOOL APIENTRY DrvReleaseContext( DHGLRC dhglrc ) { - BOOL ret = FALSE; - - if (!stw_icd) - return ret; - - pipe_mutex_lock( stw_icd->mutex ); - { - struct stw_context *ctx; - - /* XXX: The expectation is that ctx is the same context which is - * current for this thread. We should check that and return False - * if not the case. - */ - ctx = lookup_context( stw_icd, dhglrc ); - if (ctx == NULL) - goto done; - - if (stw_make_current( NULL, NULL ) == FALSE) - goto done; - - ret = TRUE; - } -done: - pipe_mutex_unlock( stw_icd->mutex ); - - return ret; + return stw_release_context(dhglrc); } void APIENTRY @@ -313,7 +158,8 @@ DrvSetCallbackProcs( INT nProcs, PROC *pProcs ) { - debug_printf( "%s( %d, %p )\n", __FUNCTION__, nProcs, pProcs ); + if (DBG) + debug_printf( "%s( %d, %p )\n", __FUNCTION__, nProcs, pProcs ); return; } @@ -671,38 +517,21 @@ DrvSetContext( DHGLRC dhglrc, PFN_SETPROCTABLE pfnSetProcTable ) { - PGLCLTPROCTABLE result = NULL; - - if (!stw_icd) - return result; - - pipe_mutex_lock( stw_icd->mutex ); - { - struct stw_context *ctx; - + if (DBG) debug_printf( "%s( 0x%p, %u, 0x%p )\n", __FUNCTION__, hdc, dhglrc, pfnSetProcTable ); - /* Although WGL allows different dispatch entrypoints per - */ - if (!stw_icd->cpt_initialized) { - init_proc_table( &stw_icd->cpt ); - stw_icd->cpt_initialized = TRUE; - } - - ctx = lookup_context( stw_icd, dhglrc ); - if (ctx == NULL) - goto done; - - if (!stw_make_current( hdc, ctx )) - goto done; - - result = &stw_icd->cpt; + /* Although WGL allows different dispatch entrypoints per + */ + if (!cpt_initialized) { + init_proc_table( &cpt ); + cpt_initialized = TRUE; } -done: - pipe_mutex_unlock( stw_icd->mutex ); - return result; + if (!stw_make_current( hdc, dhglrc )) + return NULL; + + return &cpt; } int APIENTRY @@ -713,7 +542,8 @@ DrvSetLayerPaletteEntries( INT cEntries, CONST COLORREF *pcr ) { - debug_printf( "%s\n", __FUNCTION__ ); + if (DBG) + debug_printf( "%s\n", __FUNCTION__ ); return 0; } @@ -727,7 +557,8 @@ DrvSetPixelFormat( r = stw_pixelformat_set( hdc, iPixelFormat ); - debug_printf( "%s( %p, %d ) = %s\n", __FUNCTION__, hdc, iPixelFormat, r ? "TRUE" : "FALSE" ); + if (DBG) + debug_printf( "%s( %p, %d ) = %s\n", __FUNCTION__, hdc, iPixelFormat, r ? "TRUE" : "FALSE" ); return r; } @@ -737,7 +568,8 @@ DrvShareLists( DHGLRC dhglrc1, DHGLRC dhglrc2 ) { - debug_printf( "%s\n", __FUNCTION__ ); + if (DBG) + debug_printf( "%s\n", __FUNCTION__ ); return FALSE; } @@ -746,7 +578,8 @@ BOOL APIENTRY DrvSwapBuffers( HDC hdc ) { - debug_printf( "%s( %p )\n", __FUNCTION__, hdc ); + if (DBG) + debug_printf( "%s( %p )\n", __FUNCTION__, hdc ); return stw_swap_buffers( hdc ); } @@ -756,7 +589,8 @@ DrvSwapLayerBuffers( HDC hdc, UINT fuPlanes ) { - debug_printf( "%s\n", __FUNCTION__ ); + if (DBG) + debug_printf( "%s\n", __FUNCTION__ ); return FALSE; } @@ -765,7 +599,8 @@ BOOL APIENTRY DrvValidateVersion( ULONG ulVersion ) { - debug_printf( "%s( %u )\n", __FUNCTION__, ulVersion ); + if (DBG) + debug_printf( "%s( %u )\n", __FUNCTION__, ulVersion ); /* TODO: get the expected version from the winsys */ diff --git a/src/gallium/state_trackers/wgl/opengl32.mingw.def b/src/gallium/state_trackers/wgl/opengl32.mingw.def new file mode 100644 index 0000000000..1f03ea3b37 --- /dev/null +++ b/src/gallium/state_trackers/wgl/opengl32.mingw.def @@ -0,0 +1,387 @@ +EXPORTS +; GlmfBeginGlsBlock = GlmfBeginGlsBlock@4 +; GlmfCloseMetaFile = GlmfCloseMetaFile@4 +; GlmfEndGlsBlock = GlmfEndGlsBlock@4 +; GlmfEndPlayback = GlmfEndPlayback@4 +; GlmfInitPlayback = GlmfInitPlayback@12 +; GlmfPlayGlsRecord = GlmfPlayGlsRecord@16 + glAccum = glAccum@8 + glAlphaFunc = glAlphaFunc@8 + glAreTexturesResident = glAreTexturesResident@12 + glArrayElement = glArrayElement@4 + glBegin = glBegin@4 + glBindTexture = glBindTexture@8 + glBitmap = glBitmap@28 + glBlendFunc = glBlendFunc@8 + glCallList = glCallList@4 + glCallLists = glCallLists@12 + glClear = glClear@4 + glClearAccum = glClearAccum@16 + glClearColor = glClearColor@16 + glClearDepth = glClearDepth@8 + glClearIndex = glClearIndex@4 + glClearStencil = glClearStencil@4 + glClipPlane = glClipPlane@8 + glColor3b = glColor3b@12 + glColor3bv = glColor3bv@4 + glColor3d = glColor3d@24 + glColor3dv = glColor3dv@4 + glColor3f = glColor3f@12 + glColor3fv = glColor3fv@4 + glColor3i = glColor3i@12 + glColor3iv = glColor3iv@4 + glColor3s = glColor3s@12 + glColor3sv = glColor3sv@4 + glColor3ub = glColor3ub@12 + glColor3ubv = glColor3ubv@4 + glColor3ui = glColor3ui@12 + glColor3uiv = glColor3uiv@4 + glColor3us = glColor3us@12 + glColor3usv = glColor3usv@4 + glColor4b = glColor4b@16 + glColor4bv = glColor4bv@4 + glColor4d = glColor4d@32 + glColor4dv = glColor4dv@4 + glColor4f = glColor4f@16 + glColor4fv = glColor4fv@4 + glColor4i = glColor4i@16 + glColor4iv = glColor4iv@4 + glColor4s = glColor4s@16 + glColor4sv = glColor4sv@4 + glColor4ub = glColor4ub@16 + glColor4ubv = glColor4ubv@4 + glColor4ui = glColor4ui@16 + glColor4uiv = glColor4uiv@4 + glColor4us = glColor4us@16 + glColor4usv = glColor4usv@4 + glColorMask = glColorMask@16 + glColorMaterial = glColorMaterial@8 + glColorPointer = glColorPointer@16 + glCopyPixels = glCopyPixels@20 + glCopyTexImage1D = glCopyTexImage1D@28 + glCopyTexImage2D = glCopyTexImage2D@32 + glCopyTexSubImage1D = glCopyTexSubImage1D@24 + glCopyTexSubImage2D = glCopyTexSubImage2D@32 + glCullFace = glCullFace@4 +; glDebugEntry = glDebugEntry@8 + glDeleteLists = glDeleteLists@8 + glDeleteTextures = glDeleteTextures@8 + glDepthFunc = glDepthFunc@4 + glDepthMask = glDepthMask@4 + glDepthRange = glDepthRange@16 + glDisable = glDisable@4 + glDisableClientState = glDisableClientState@4 + glDrawArrays = glDrawArrays@12 + glDrawBuffer = glDrawBuffer@4 + glDrawElements = glDrawElements@16 + glDrawPixels = glDrawPixels@20 + glEdgeFlag = glEdgeFlag@4 + glEdgeFlagPointer = glEdgeFlagPointer@8 + glEdgeFlagv = glEdgeFlagv@4 + glEnable = glEnable@4 + glEnableClientState = glEnableClientState@4 + glEnd = glEnd@0 + glEndList = glEndList@0 + glEvalCoord1d = glEvalCoord1d@8 + glEvalCoord1dv = glEvalCoord1dv@4 + glEvalCoord1f = glEvalCoord1f@4 + glEvalCoord1fv = glEvalCoord1fv@4 + glEvalCoord2d = glEvalCoord2d@16 + glEvalCoord2dv = glEvalCoord2dv@4 + glEvalCoord2f = glEvalCoord2f@8 + glEvalCoord2fv = glEvalCoord2fv@4 + glEvalMesh1 = glEvalMesh1@12 + glEvalMesh2 = glEvalMesh2@20 + glEvalPoint1 = glEvalPoint1@4 + glEvalPoint2 = glEvalPoint2@8 + glFeedbackBuffer = glFeedbackBuffer@12 + glFinish = glFinish@0 + glFlush = glFlush@0 + glFogf = glFogf@8 + glFogfv = glFogfv@8 + glFogi = glFogi@8 + glFogiv = glFogiv@8 + glFrontFace = glFrontFace@4 + glFrustum = glFrustum@48 + glGenLists = glGenLists@4 + glGenTextures = glGenTextures@8 + glGetBooleanv = glGetBooleanv@8 + glGetClipPlane = glGetClipPlane@8 + glGetDoublev = glGetDoublev@8 + glGetError = glGetError@0 + glGetFloatv = glGetFloatv@8 + glGetIntegerv = glGetIntegerv@8 + glGetLightfv = glGetLightfv@12 + glGetLightiv = glGetLightiv@12 + glGetMapdv = glGetMapdv@12 + glGetMapfv = glGetMapfv@12 + glGetMapiv = glGetMapiv@12 + glGetMaterialfv = glGetMaterialfv@12 + glGetMaterialiv = glGetMaterialiv@12 + glGetPixelMapfv = glGetPixelMapfv@8 + glGetPixelMapuiv = glGetPixelMapuiv@8 + glGetPixelMapusv = glGetPixelMapusv@8 + glGetPointerv = glGetPointerv@8 + glGetPolygonStipple = glGetPolygonStipple@4 + glGetString = glGetString@4 + glGetTexEnvfv = glGetTexEnvfv@12 + glGetTexEnviv = glGetTexEnviv@12 + glGetTexGendv = glGetTexGendv@12 + glGetTexGenfv = glGetTexGenfv@12 + glGetTexGeniv = glGetTexGeniv@12 + glGetTexImage = glGetTexImage@20 + glGetTexLevelParameterfv = glGetTexLevelParameterfv@16 + glGetTexLevelParameteriv = glGetTexLevelParameteriv@16 + glGetTexParameterfv = glGetTexParameterfv@12 + glGetTexParameteriv = glGetTexParameteriv@12 + glHint = glHint@8 + glIndexMask = glIndexMask@4 + glIndexPointer = glIndexPointer@12 + glIndexd = glIndexd@8 + glIndexdv = glIndexdv@4 + glIndexf = glIndexf@4 + glIndexfv = glIndexfv@4 + glIndexi = glIndexi@4 + glIndexiv = glIndexiv@4 + glIndexs = glIndexs@4 + glIndexsv = glIndexsv@4 + glIndexub = glIndexub@4 + glIndexubv = glIndexubv@4 + glInitNames = glInitNames@0 + glInterleavedArrays = glInterleavedArrays@12 + glIsEnabled = glIsEnabled@4 + glIsList = glIsList@4 + glIsTexture = glIsTexture@4 + glLightModelf = glLightModelf@8 + glLightModelfv = glLightModelfv@8 + glLightModeli = glLightModeli@8 + glLightModeliv = glLightModeliv@8 + glLightf = glLightf@12 + glLightfv = glLightfv@12 + glLighti = glLighti@12 + glLightiv = glLightiv@12 + glLineStipple = glLineStipple@8 + glLineWidth = glLineWidth@4 + glListBase = glListBase@4 + glLoadIdentity = glLoadIdentity@0 + glLoadMatrixd = glLoadMatrixd@4 + glLoadMatrixf = glLoadMatrixf@4 + glLoadName = glLoadName@4 + glLogicOp = glLogicOp@4 + glMap1d = glMap1d@32 + glMap1f = glMap1f@24 + glMap2d = glMap2d@56 + glMap2f = glMap2f@40 + glMapGrid1d = glMapGrid1d@20 + glMapGrid1f = glMapGrid1f@12 + glMapGrid2d = glMapGrid2d@40 + glMapGrid2f = glMapGrid2f@24 + glMaterialf = glMaterialf@12 + glMaterialfv = glMaterialfv@12 + glMateriali = glMateriali@12 + glMaterialiv = glMaterialiv@12 + glMatrixMode = glMatrixMode@4 + glMultMatrixd = glMultMatrixd@4 + glMultMatrixf = glMultMatrixf@4 + glNewList = glNewList@8 + glNormal3b = glNormal3b@12 + glNormal3bv = glNormal3bv@4 + glNormal3d = glNormal3d@24 + glNormal3dv = glNormal3dv@4 + glNormal3f = glNormal3f@12 + glNormal3fv = glNormal3fv@4 + glNormal3i = glNormal3i@12 + glNormal3iv = glNormal3iv@4 + glNormal3s = glNormal3s@12 + glNormal3sv = glNormal3sv@4 + glNormalPointer = glNormalPointer@12 + glOrtho = glOrtho@48 + glPassThrough = glPassThrough@4 + glPixelMapfv = glPixelMapfv@12 + glPixelMapuiv = glPixelMapuiv@12 + glPixelMapusv = glPixelMapusv@12 + glPixelStoref = glPixelStoref@8 + glPixelStorei = glPixelStorei@8 + glPixelTransferf = glPixelTransferf@8 + glPixelTransferi = glPixelTransferi@8 + glPixelZoom = glPixelZoom@8 + glPointSize = glPointSize@4 + glPolygonMode = glPolygonMode@8 + glPolygonOffset = glPolygonOffset@8 + glPolygonStipple = glPolygonStipple@4 + glPopAttrib = glPopAttrib@0 + glPopClientAttrib = glPopClientAttrib@0 + glPopMatrix = glPopMatrix@0 + glPopName = glPopName@0 + glPrioritizeTextures = glPrioritizeTextures@12 + glPushAttrib = glPushAttrib@4 + glPushClientAttrib = glPushClientAttrib@4 + glPushMatrix = glPushMatrix@0 + glPushName = glPushName@4 + glRasterPos2d = glRasterPos2d@16 + glRasterPos2dv = glRasterPos2dv@4 + glRasterPos2f = glRasterPos2f@8 + glRasterPos2fv = glRasterPos2fv@4 + glRasterPos2i = glRasterPos2i@8 + glRasterPos2iv = glRasterPos2iv@4 + glRasterPos2s = glRasterPos2s@8 + glRasterPos2sv = glRasterPos2sv@4 + glRasterPos3d = glRasterPos3d@24 + glRasterPos3dv = glRasterPos3dv@4 + glRasterPos3f = glRasterPos3f@12 + glRasterPos3fv = glRasterPos3fv@4 + glRasterPos3i = glRasterPos3i@12 + glRasterPos3iv = glRasterPos3iv@4 + glRasterPos3s = glRasterPos3s@12 + glRasterPos3sv = glRasterPos3sv@4 + glRasterPos4d = glRasterPos4d@32 + glRasterPos4dv = glRasterPos4dv@4 + glRasterPos4f = glRasterPos4f@16 + glRasterPos4fv = glRasterPos4fv@4 + glRasterPos4i = glRasterPos4i@16 + glRasterPos4iv = glRasterPos4iv@4 + glRasterPos4s = glRasterPos4s@16 + glRasterPos4sv = glRasterPos4sv@4 + glReadBuffer = glReadBuffer@4 + glReadPixels = glReadPixels@28 + glRectd = glRectd@32 + glRectdv = glRectdv@8 + glRectf = glRectf@16 + glRectfv = glRectfv@8 + glRecti = glRecti@16 + glRectiv = glRectiv@8 + glRects = glRects@16 + glRectsv = glRectsv@8 + glRenderMode = glRenderMode@4 + glRotated = glRotated@32 + glRotatef = glRotatef@16 + glScaled = glScaled@24 + glScalef = glScalef@12 + glScissor = glScissor@16 + glSelectBuffer = glSelectBuffer@8 + glShadeModel = glShadeModel@4 + glStencilFunc = glStencilFunc@12 + glStencilMask = glStencilMask@4 + glStencilOp = glStencilOp@12 + glTexCoord1d = glTexCoord1d@8 + glTexCoord1dv = glTexCoord1dv@4 + glTexCoord1f = glTexCoord1f@4 + glTexCoord1fv = glTexCoord1fv@4 + glTexCoord1i = glTexCoord1i@4 + glTexCoord1iv = glTexCoord1iv@4 + glTexCoord1s = glTexCoord1s@4 + glTexCoord1sv = glTexCoord1sv@4 + glTexCoord2d = glTexCoord2d@16 + glTexCoord2dv = glTexCoord2dv@4 + glTexCoord2f = glTexCoord2f@8 + glTexCoord2fv = glTexCoord2fv@4 + glTexCoord2i = glTexCoord2i@8 + glTexCoord2iv = glTexCoord2iv@4 + glTexCoord2s = glTexCoord2s@8 + glTexCoord2sv = glTexCoord2sv@4 + glTexCoord3d = glTexCoord3d@24 + glTexCoord3dv = glTexCoord3dv@4 + glTexCoord3f = glTexCoord3f@12 + glTexCoord3fv = glTexCoord3fv@4 + glTexCoord3i = glTexCoord3i@12 + glTexCoord3iv = glTexCoord3iv@4 + glTexCoord3s = glTexCoord3s@12 + glTexCoord3sv = glTexCoord3sv@4 + glTexCoord4d = glTexCoord4d@32 + glTexCoord4dv = glTexCoord4dv@4 + glTexCoord4f = glTexCoord4f@16 + glTexCoord4fv = glTexCoord4fv@4 + glTexCoord4i = glTexCoord4i@16 + glTexCoord4iv = glTexCoord4iv@4 + glTexCoord4s = glTexCoord4s@16 + glTexCoord4sv = glTexCoord4sv@4 + glTexCoordPointer = glTexCoordPointer@16 + glTexEnvf = glTexEnvf@12 + glTexEnvfv = glTexEnvfv@12 + glTexEnvi = glTexEnvi@12 + glTexEnviv = glTexEnviv@12 + glTexGend = glTexGend@16 + glTexGendv = glTexGendv@12 + glTexGenf = glTexGenf@12 + glTexGenfv = glTexGenfv@12 + glTexGeni = glTexGeni@12 + glTexGeniv = glTexGeniv@12 + glTexImage1D = glTexImage1D@32 + glTexImage2D = glTexImage2D@36 + glTexParameterf = glTexParameterf@12 + glTexParameterfv = glTexParameterfv@12 + glTexParameteri = glTexParameteri@12 + glTexParameteriv = glTexParameteriv@12 + glTexSubImage1D = glTexSubImage1D@28 + glTexSubImage2D = glTexSubImage2D@36 + glTranslated = glTranslated@24 + glTranslatef = glTranslatef@12 + glVertex2d = glVertex2d@16 + glVertex2dv = glVertex2dv@4 + glVertex2f = glVertex2f@8 + glVertex2fv = glVertex2fv@4 + glVertex2i = glVertex2i@8 + glVertex2iv = glVertex2iv@4 + glVertex2s = glVertex2s@8 + glVertex2sv = glVertex2sv@4 + glVertex3d = glVertex3d@24 + glVertex3dv = glVertex3dv@4 + glVertex3f = glVertex3f@12 + glVertex3fv = glVertex3fv@4 + glVertex3i = glVertex3i@12 + glVertex3iv = glVertex3iv@4 + glVertex3s = glVertex3s@12 + glVertex3sv = glVertex3sv@4 + glVertex4d = glVertex4d@32 + glVertex4dv = glVertex4dv@4 + glVertex4f = glVertex4f@16 + glVertex4fv = glVertex4fv@4 + glVertex4i = glVertex4i@16 + glVertex4iv = glVertex4iv@4 + glVertex4s = glVertex4s@16 + glVertex4sv = glVertex4sv@4 + glVertexPointer = glVertexPointer@16 + glViewport = glViewport@16 + wglChoosePixelFormat = wglChoosePixelFormat@8 + wglCopyContext = wglCopyContext@12 + wglCreateContext = wglCreateContext@4 + wglCreateLayerContext = wglCreateLayerContext@8 + wglDeleteContext = wglDeleteContext@4 + wglDescribeLayerPlane = wglDescribeLayerPlane@20 + wglDescribePixelFormat = wglDescribePixelFormat@16 + wglGetCurrentContext = wglGetCurrentContext@0 + wglGetCurrentDC = wglGetCurrentDC@0 +; wglGetDefaultProcAddress = wglGetDefaultProcAddress@4 + wglGetLayerPaletteEntries = wglGetLayerPaletteEntries@20 + wglGetPixelFormat = wglGetPixelFormat@4 + wglGetProcAddress = wglGetProcAddress@4 + wglMakeCurrent = wglMakeCurrent@8 + wglRealizeLayerPalette = wglRealizeLayerPalette@12 + wglSetLayerPaletteEntries = wglSetLayerPaletteEntries@20 + wglSetPixelFormat = wglSetPixelFormat@12 + wglShareLists = wglShareLists@8 + wglSwapBuffers = wglSwapBuffers@4 + wglSwapLayerBuffers = wglSwapLayerBuffers@8 +; wglSwapMultipleBuffers = wglSwapMultipleBuffers@8 + wglUseFontBitmapsA = wglUseFontBitmapsA@16 + wglUseFontBitmapsW = wglUseFontBitmapsW@16 + wglUseFontOutlinesA = wglUseFontOutlinesA@32 + wglUseFontOutlinesW = wglUseFontOutlinesW@32 + DrvCopyContext = DrvCopyContext@12 + DrvCreateContext = DrvCreateContext@4 + DrvCreateLayerContext = DrvCreateLayerContext@8 + DrvDeleteContext = DrvDeleteContext@4 + DrvDescribeLayerPlane = DrvDescribeLayerPlane@20 + DrvDescribePixelFormat = DrvDescribePixelFormat@16 + DrvGetLayerPaletteEntries = DrvGetLayerPaletteEntries@20 + DrvGetProcAddress = DrvGetProcAddress@4 + DrvRealizeLayerPalette = DrvRealizeLayerPalette@12 + DrvReleaseContext = DrvReleaseContext@4 + DrvSetCallbackProcs = DrvSetCallbackProcs@8 + DrvSetContext = DrvSetContext@12 + DrvSetLayerPaletteEntries = DrvSetLayerPaletteEntries@20 + DrvSetPixelFormat = DrvSetPixelFormat@8 + DrvShareLists = DrvShareLists@8 + DrvSwapBuffers = DrvSwapBuffers@4 + DrvSwapLayerBuffers = DrvSwapLayerBuffers@8 + DrvValidateVersion = DrvValidateVersion@4 diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.c b/src/gallium/state_trackers/wgl/shared/stw_context.c index 2abf97b5ad..d77daac39c 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_context.c +++ b/src/gallium/state_trackers/wgl/shared/stw_context.c @@ -40,26 +40,38 @@ #include "stw_public.h" #include "stw_context.h" -static struct stw_context *ctx_head = NULL; - static HDC current_hdc = NULL; -static struct stw_context *current_hrc = NULL; +static UINT_PTR current_hglrc = 0; BOOL stw_copy_context( - struct stw_context *src, - struct stw_context *dst, + UINT_PTR hglrcSrc, + UINT_PTR hglrcDst, UINT mask ) { - (void) src; - (void) dst; - (void) mask; + struct stw_context *src; + struct stw_context *dst; + BOOL ret = FALSE; + + pipe_mutex_lock( stw_dev->mutex ); + + src = stw_lookup_context( hglrcSrc ); + dst = stw_lookup_context( hglrcDst ); + + if (src && dst) { + /* FIXME */ + (void) src; + (void) dst; + (void) mask; + } - return FALSE; + pipe_mutex_unlock( stw_dev->mutex ); + + return ret; } -struct stw_context * -stw_create_context( +UINT_PTR +stw_create_layer_context( HDC hdc, int iLayerPlane ) { @@ -68,19 +80,23 @@ stw_create_context( struct stw_context *ctx = NULL; GLvisual *visual = NULL; struct pipe_context *pipe = NULL; + UINT_PTR hglrc = 0; + if(!stw_dev) + return 0; + if (iLayerPlane != 0) - return NULL; + return 0; pfi = stw_pixelformat_get( hdc ); if (pfi == 0) - return NULL; + return 0; pf = pixelformat_get_info( pfi - 1 ); ctx = CALLOC_STRUCT( stw_context ); if (ctx == NULL) - return NULL; + return 0; ctx->hdc = hdc; ctx->color_bits = GetDeviceCaps( ctx->hdc, BITSPIXEL ); @@ -119,10 +135,26 @@ stw_create_context( ctx->st->ctx->DriverCtx = ctx; - ctx->next = ctx_head; - ctx_head = ctx; + pipe_mutex_lock( stw_dev->mutex ); + { + UINT_PTR i; + + for (i = 0; i < STW_CONTEXT_MAX; i++) { + if (stw_dev->ctx_array[i].ctx == NULL) { + /* success: + */ + stw_dev->ctx_array[i].ctx = ctx; + hglrc = i + 1; + break; + } + } + } + pipe_mutex_unlock( stw_dev->mutex ); - return ctx; + /* Success? + */ + if (hglrc != 0) + return hglrc; fail: if (visual) @@ -132,47 +164,83 @@ fail: pipe->destroy( pipe ); FREE( ctx ); - return NULL; + return 0; } - BOOL stw_delete_context( - struct stw_context *hglrc ) + UINT_PTR hglrc ) { - struct stw_context **link = &ctx_head; - struct stw_context *ctx = ctx_head; + struct stw_context *ctx ; + BOOL ret = FALSE; + + if (!stw_dev) + return FALSE; - while (ctx != NULL) { - if (ctx == hglrc) { - GLcontext *glctx = ctx->st->ctx; - GET_CURRENT_CONTEXT( glcurctx ); - struct stw_framebuffer *fb; + pipe_mutex_lock( stw_dev->mutex ); - /* Unbind current if deleting current context. - */ - if (glcurctx == glctx) - st_make_current( NULL, NULL, NULL ); + ctx = stw_lookup_context(hglrc); + if (ctx) { + GLcontext *glctx = ctx->st->ctx; + GET_CURRENT_CONTEXT( glcurctx ); + struct stw_framebuffer *fb; - fb = framebuffer_from_hdc( ctx->hdc ); - if (fb) - framebuffer_destroy( fb ); + /* Unbind current if deleting current context. + */ + if (glcurctx == glctx) + st_make_current( NULL, NULL, NULL ); - if (WindowFromDC( ctx->hdc ) != NULL) - ReleaseDC( WindowFromDC( ctx->hdc ), ctx->hdc ); + fb = framebuffer_from_hdc( ctx->hdc ); + if (fb) + framebuffer_destroy( fb ); - st_destroy_context( ctx->st ); + if (WindowFromDC( ctx->hdc ) != NULL) + ReleaseDC( WindowFromDC( ctx->hdc ), ctx->hdc ); - *link = ctx->next; - FREE( ctx ); - return TRUE; - } + st_destroy_context( ctx->st ); - link = &ctx->next; - ctx = ctx->next; + FREE( ctx ); + + stw_dev->ctx_array[hglrc - 1].ctx = NULL; + + ret = TRUE; } - return FALSE; + pipe_mutex_unlock( stw_dev->mutex ); + + return ret; +} + +BOOL +stw_release_context( + UINT_PTR hglrc ) +{ + BOOL ret = FALSE; + + if (!stw_dev) + return ret; + + pipe_mutex_lock( stw_dev->mutex ); + { + struct stw_context *ctx; + + /* XXX: The expectation is that ctx is the same context which is + * current for this thread. We should check that and return False + * if not the case. + */ + ctx = stw_lookup_context( hglrc ); + if (ctx == NULL) + goto done; + + if (stw_make_current( NULL, 0 ) == FALSE) + goto done; + + ret = TRUE; + } +done: + pipe_mutex_unlock( stw_dev->mutex ); + + return ret; } /* Find the width and height of the window named by hdc. @@ -193,10 +261,10 @@ get_window_size( HDC hdc, GLuint *width, GLuint *height ) } } -struct stw_context * +UINT_PTR stw_get_current_context( void ) { - return current_hrc; + return current_hglrc; } HDC @@ -208,35 +276,43 @@ stw_get_current_dc( void ) BOOL stw_make_current( HDC hdc, - struct stw_context *hglrc ) + UINT_PTR hglrc ) { - struct stw_context *ctx = ctx_head; + struct stw_context *ctx; GET_CURRENT_CONTEXT( glcurctx ); struct stw_framebuffer *fb; GLuint width = 0; GLuint height = 0; + struct stw_context *curctx; + + if (!stw_dev) + return FALSE; + + pipe_mutex_lock( stw_dev->mutex ); + ctx = stw_lookup_context( hglrc ); + pipe_mutex_unlock( stw_dev->mutex ); + + if (ctx == NULL) + return FALSE; current_hdc = hdc; - current_hrc = hglrc; + current_hglrc = hglrc; - if (hdc == NULL || hglrc == NULL) { - st_make_current( NULL, NULL, NULL ); - return TRUE; + if (glcurctx != NULL) { + curctx = (struct stw_context *) glcurctx->DriverCtx; + + if (curctx != ctx) + st_flush(glcurctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); } - while (ctx != NULL) { - if (ctx == hglrc) - break; - ctx = ctx->next; + if (hdc == NULL || hglrc == 0) { + st_make_current( NULL, NULL, NULL ); + return TRUE; } - if (ctx == NULL) - return FALSE; /* Return if already current. */ if (glcurctx != NULL) { - struct stw_context *curctx = (struct stw_context *) glcurctx->DriverCtx; - if (curctx != NULL && curctx == ctx && ctx->hdc == hdc) return TRUE; } @@ -273,20 +349,3 @@ stw_make_current( return TRUE; } - -struct stw_context * -stw_context_from_hdc( - HDC hdc ) -{ - struct stw_context *ctx = ctx_head; - - while (ctx != NULL) { - if (ctx->hdc == hdc) - return ctx; - ctx = ctx->next; - } - return NULL; -} - - - diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.h b/src/gallium/state_trackers/wgl/shared/stw_context.h index 89a8f900d8..b289615272 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_context.h +++ b/src/gallium/state_trackers/wgl/shared/stw_context.h @@ -37,17 +37,6 @@ struct stw_context struct st_context *st; HDC hdc; DWORD color_bits; - struct stw_context *next; }; -struct stw_context * -stw_context_from_hdc(HDC hdc ); - - - - - - - - #endif /* STW_CONTEXT_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.c b/src/gallium/state_trackers/wgl/shared/stw_device.c index 6873e813ee..0dca856d73 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_device.c +++ b/src/gallium/state_trackers/wgl/shared/stw_device.c @@ -27,14 +27,19 @@ #include <windows.h> -#include "pipe/p_debug.h" +#include "glapi/glthread.h" +#include "util/u_debug.h" #include "pipe/p_screen.h" #include "shared/stw_device.h" #include "shared/stw_winsys.h" #include "shared/stw_pixelformat.h" #include "shared/stw_public.h" -#include "stw.h" + +#ifdef WIN32_THREADS +extern _glthread_Mutex OneTimeLock; +extern void FreeAllTSD(void); +#endif struct stw_device *stw_dev = NULL; @@ -57,23 +62,35 @@ st_flush_frontbuffer(struct pipe_screen *screen, boolean -stw_shared_init(const struct stw_winsys *stw_winsys) +st_init(const struct stw_winsys *stw_winsys) { static struct stw_device stw_dev_storage; + debug_printf("%s\n", __FUNCTION__); + assert(!stw_dev); stw_dev = &stw_dev_storage; memset(stw_dev, 0, sizeof(*stw_dev)); +#ifdef DEBUG + stw_dev->memdbg_no = debug_memory_begin(); +#endif + stw_dev->stw_winsys = stw_winsys; +#ifdef WIN32_THREADS + _glthread_INIT_MUTEX(OneTimeLock); +#endif + stw_dev->screen = stw_winsys->create_screen(); if(!stw_dev->screen) goto error1; stw_dev->screen->flush_frontbuffer = st_flush_frontbuffer; + pipe_mutex_init( stw_dev->mutex ); + pixelformat_init(); return TRUE; @@ -85,7 +102,51 @@ error1: void -stw_shared_cleanup(void) +st_cleanup(void) { + UINT_PTR i; + + debug_printf("%s\n", __FUNCTION__); + + if (!stw_dev) + return; + + pipe_mutex_lock( stw_dev->mutex ); + { + /* Ensure all contexts are destroyed */ + for (i = 0; i < STW_CONTEXT_MAX; i++) + if (stw_dev->ctx_array[i].ctx) + stw_delete_context( i + 1 ); + } + pipe_mutex_unlock( stw_dev->mutex ); + + pipe_mutex_destroy( stw_dev->mutex ); + + stw_dev->screen->destroy(stw_dev->screen); + +#ifdef WIN32_THREADS + _glthread_DESTROY_MUTEX(OneTimeLock); + FreeAllTSD(); +#endif + +#ifdef DEBUG + debug_memory_end(stw_dev->memdbg_no); +#endif + stw_dev = NULL; } + + +struct stw_context * +stw_lookup_context( UINT_PTR dhglrc ) +{ + if (dhglrc == 0 || + dhglrc >= STW_CONTEXT_MAX) + return NULL; + + if (stw_dev == NULL) + return NULL; + + return stw_dev->ctx_array[dhglrc - 1].ctx; +} + diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.h b/src/gallium/state_trackers/wgl/shared/stw_device.h index bc0bce37c6..80da14b84f 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_device.h +++ b/src/gallium/state_trackers/wgl/shared/stw_device.h @@ -25,8 +25,15 @@ * **************************************************************************/ -#ifndef ST_DEVICE_H_ -#define ST_DEVICE_H_ +#ifndef STW_DEVICE_H_ +#define STW_DEVICE_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_thread.h" + + +#define STW_CONTEXT_MAX 32 struct pipe_screen; @@ -35,10 +42,22 @@ struct stw_device { const struct stw_winsys *stw_winsys; struct pipe_screen *screen; + + pipe_mutex mutex; + + struct { + struct stw_context *ctx; + } ctx_array[STW_CONTEXT_MAX]; + +#ifdef DEBUG + unsigned long memdbg_no; +#endif }; +struct stw_context * +stw_lookup_context( UINT_PTR hglrc ); extern struct stw_device *stw_dev; -#endif /* ST_DEVICE_H_ */ +#endif /* STW_DEVICE_H_ */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c index c70b31a488..17c96c411f 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c +++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c @@ -79,6 +79,14 @@ window_proc( return CallWindowProc( fb->WndProc, hWnd, uMsg, wParam, lParam ); } +static INLINE boolean +stw_is_supported_depth_stencil(enum pipe_format format) +{ + struct pipe_screen *screen = stw_dev->screen; + return screen->is_format_supported(screen, format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); +} + /* Create a new framebuffer object which will correspond to the given HDC. */ struct stw_framebuffer * @@ -101,18 +109,43 @@ framebuffer_create( if (visual->depthBits == 0) depthFormat = PIPE_FORMAT_NONE; - else if (visual->depthBits <= 16) + else if (visual->depthBits <= 16 && + stw_is_supported_depth_stencil(PIPE_FORMAT_Z16_UNORM)) depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits <= 24) + else if (visual->depthBits <= 24 && visual->stencilBits != 8 && + stw_is_supported_depth_stencil(PIPE_FORMAT_X8Z24_UNORM)) { + depthFormat = PIPE_FORMAT_X8Z24_UNORM; + } + else if (visual->depthBits <= 24 && visual->stencilBits != 8 && + stw_is_supported_depth_stencil(PIPE_FORMAT_Z24X8_UNORM)) { + depthFormat = PIPE_FORMAT_Z24X8_UNORM; + } + else if (visual->depthBits <= 24 && visual->stencilBits == 8 && + stw_is_supported_depth_stencil(PIPE_FORMAT_S8Z24_UNORM)) { depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else + } + else if (visual->depthBits <= 24 && visual->stencilBits == 8 && + stw_is_supported_depth_stencil(PIPE_FORMAT_Z24S8_UNORM)) { + depthFormat = PIPE_FORMAT_Z24S8_UNORM; + } + else if(stw_is_supported_depth_stencil(PIPE_FORMAT_Z32_UNORM)) { depthFormat = PIPE_FORMAT_Z32_UNORM; + } + else if(stw_is_supported_depth_stencil(PIPE_FORMAT_Z32_FLOAT)) { + depthFormat = PIPE_FORMAT_Z32_FLOAT; + } + else { + assert(0); + depthFormat = PIPE_FORMAT_NONE; + } - if (visual->stencilBits == 8) { - if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) - stencilFormat = depthFormat; - else - stencilFormat = PIPE_FORMAT_S8_UNORM; + if (depthFormat == PIPE_FORMAT_S8Z24_UNORM || + depthFormat == PIPE_FORMAT_Z24S8_UNORM) { + stencilFormat = depthFormat; + } + else if (visual->stencilBits == 8 && + stw_is_supported_depth_stencil(PIPE_FORMAT_S8_UNORM)) { + stencilFormat = PIPE_FORMAT_S8_UNORM; } else { stencilFormat = PIPE_FORMAT_NONE; diff --git a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c index 84b7b287b9..2992a1ac0a 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c +++ b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "stw_pixelformat.h" #include "stw_public.h" @@ -76,6 +76,18 @@ add_standard_pixelformats( pf->flags = flags; pf->color = color24; + pf->alpha = alpha8; + pf->depth = depth16; + pf++; + + pf->flags = flags; + pf->color = color24; + pf->alpha = alpha8; + pf->depth = depth24s8; + pf++; + + pf->flags = flags; + pf->color = color24; pf->alpha = noalpha; pf->depth = depth16; pf++; diff --git a/src/gallium/state_trackers/wgl/shared/stw_public.h b/src/gallium/state_trackers/wgl/shared/stw_public.h index 75b504a50f..39d377c16b 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_public.h +++ b/src/gallium/state_trackers/wgl/shared/stw_public.h @@ -29,31 +29,24 @@ #define STW_PUBLIC_H #include <windows.h> -#include "pipe/p_compiler.h" -struct stw_winsys; -struct stw_context; - -boolean -st_shared_init(const struct stw_winsys *stw_winsys); - -void -st_shared_cleanup(void); - - -BOOL stw_copy_context( struct stw_context *src, - struct stw_context *dst, +BOOL stw_copy_context( UINT_PTR hglrcSrc, + UINT_PTR hglrcDst, UINT mask ); -struct stw_context *stw_create_context( HDC hdc, int iLayerPlane ); +UINT_PTR stw_create_layer_context( HDC hdc, + int iLayerPlane ); + +BOOL stw_delete_context( UINT_PTR hglrc ); -BOOL stw_delete_context( struct stw_context *ctx ); +BOOL +stw_release_context( UINT_PTR dhglrc ); -struct stw_context *stw_get_current_context( void ); +UINT_PTR stw_get_current_context( void ); HDC stw_get_current_dc( void ); -BOOL stw_make_current( HDC hdc, struct stw_context *ctx ); +BOOL stw_make_current( HDC hdc, UINT_PTR hglrc ); BOOL stw_swap_buffers( HDC hdc ); diff --git a/src/gallium/state_trackers/wgl/shared/stw_quirks.c b/src/gallium/state_trackers/wgl/shared/stw_quirks.c index 0961ce3bb0..2f7091a52c 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_quirks.c +++ b/src/gallium/state_trackers/wgl/shared/stw_quirks.c @@ -111,3 +111,6 @@ void gl_dispatch_stub_770(void){} void gl_dispatch_stub_771(void){} void gl_dispatch_stub_772(void){} void gl_dispatch_stub_773(void){} +void gl_dispatch_stub_774(void){} +void gl_dispatch_stub_750(void){} +void gl_dispatch_stub_742(void){} diff --git a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c index f50b79b4e1..e06d2640b4 100644 --- a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c +++ b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c @@ -27,25 +27,9 @@ #include <windows.h> -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "shared/stw_public.h" #include "stw_wgl.h" -#include "stw.h" - -boolean stw_wgl_init( void ) -{ - debug_printf("%s\n", __FUNCTION__); - return TRUE; -} - -void stw_wgl_cleanup( void ) -{ -} - -static INLINE struct stw_context *stw_context( HGLRC hglrc ) -{ - return (struct stw_context *)hglrc; -} WINGDIAPI BOOL APIENTRY @@ -54,8 +38,8 @@ wglCopyContext( HGLRC hglrcDst, UINT mask ) { - return stw_copy_context( stw_context(hglrcSrc), - stw_context(hglrcDst), + return stw_copy_context( (UINT_PTR)hglrcSrc, + (UINT_PTR)hglrcDst, mask ); } @@ -63,7 +47,7 @@ WINGDIAPI HGLRC APIENTRY wglCreateContext( HDC hdc ) { - return (HGLRC) stw_create_context( hdc, 0 ); + return wglCreateLayerContext(hdc, 0); } WINGDIAPI HGLRC APIENTRY @@ -71,21 +55,21 @@ wglCreateLayerContext( HDC hdc, int iLayerPlane ) { - return (HGLRC) stw_create_context( hdc, iLayerPlane ); + return (HGLRC) stw_create_layer_context( hdc, iLayerPlane ); } WINGDIAPI BOOL APIENTRY wglDeleteContext( HGLRC hglrc ) { - return stw_delete_context( stw_context(hglrc) ); + return stw_delete_context( (UINT_PTR)hglrc ); } WINGDIAPI HGLRC APIENTRY wglGetCurrentContext( VOID ) { - return (HGLRC) stw_get_current_context(); + return (HGLRC)stw_get_current_context(); } WINGDIAPI HDC APIENTRY @@ -99,7 +83,7 @@ wglMakeCurrent( HDC hdc, HGLRC hglrc ) { - return stw_make_current( hdc, stw_context(hglrc) ); + return stw_make_current( hdc, (UINT_PTR)hglrc ); } diff --git a/src/gallium/state_trackers/wgl/wgl/stw_wgl.h b/src/gallium/state_trackers/wgl/wgl/stw_wgl.h index b86cc240f2..a98179944a 100644 --- a/src/gallium/state_trackers/wgl/wgl/stw_wgl.h +++ b/src/gallium/state_trackers/wgl/wgl/stw_wgl.h @@ -31,7 +31,7 @@ #include <windows.h> -#include "GL/gl.h" +#include <GL/gl.h> /* diff --git a/src/gallium/state_trackers/xorg/Makefile b/src/gallium/state_trackers/xorg/Makefile new file mode 100644 index 0000000000..a00ea3e2a4 --- /dev/null +++ b/src/gallium/state_trackers/xorg/Makefile @@ -0,0 +1,29 @@ +TARGET = libxorgtracker.a +CFILES = $(wildcard ./*.c) +OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) +GALLIUMDIR = ../.. +TOP = ../../../.. + +include $(TOP)/configs/current + +CFLAGS = -DHAVE_CONFIG_H \ + -g -Wall -Wimplicit-function-declaration -fPIC \ + $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \ + -I$(GALLIUMDIR)/include \ + -I$(GALLIUMDIR)/auxiliary \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main + +############################################# + +.PHONY = all clean + +all: $(TARGET) + +$(TARGET): $(OBJECTS) + ar rcs $(TARGET) $(OBJECTS) + +clean: + rm -rf $(OBJECTS) $(TARGET) diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c new file mode 100644 index 0000000000..7304113a65 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_crtc.c @@ -0,0 +1,314 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + +#include <unistd.h> +#include <string.h> +#include <assert.h> +#include <stdlib.h> +#include <math.h> +#include <stdint.h> + +#include "xorg-server.h" +#include <xf86.h> +#include <xf86i2c.h> +#include <xf86Crtc.h> +#include "xorg_tracker.h" +#include "xf86Modes.h" + +#define DPMS_SERVER +#include <X11/extensions/dpms.h> + +#include "pipe/p_inlines.h" + +struct crtc_private +{ + drmModeCrtcPtr drm_crtc; + + /* hwcursor */ + struct pipe_buffer *cursor_buf; + unsigned cursor_handle; +}; + +static void +crtc_dpms(xf86CrtcPtr crtc, int mode) +{ + //ScrnInfoPtr pScrn = crtc->scrn; + + switch (mode) { + case DPMSModeOn: + case DPMSModeStandby: + case DPMSModeSuspend: + break; + case DPMSModeOff: + break; + } +} + +static Bool +crtc_lock(xf86CrtcPtr crtc) +{ + return FALSE; +} + +static void +crtc_unlock(xf86CrtcPtr crtc) +{ +} + +static void +crtc_prepare(xf86CrtcPtr crtc) +{ +} + +static void +crtc_commit(xf86CrtcPtr crtc) +{ +} + +static Bool +crtc_mode_fixup(xf86CrtcPtr crtc, DisplayModePtr mode, + DisplayModePtr adjusted_mode) +{ + return TRUE; +} + +static void +crtc_mode_set(xf86CrtcPtr crtc, DisplayModePtr mode, + DisplayModePtr adjusted_mode, int x, int y) +{ + xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(crtc->scrn); + modesettingPtr ms = modesettingPTR(crtc->scrn); + xf86OutputPtr output = config->output[config->compat_output]; + drmModeConnectorPtr drm_connector = output->driver_private; + struct crtc_private *crtcp = crtc->driver_private; + drmModeCrtcPtr drm_crtc = crtcp->drm_crtc; + drmModeModeInfo drm_mode; + + drm_mode.clock = mode->Clock; + drm_mode.hdisplay = mode->HDisplay; + drm_mode.hsync_start = mode->HSyncStart; + drm_mode.hsync_end = mode->HSyncEnd; + drm_mode.htotal = mode->HTotal; + drm_mode.vdisplay = mode->VDisplay; + drm_mode.vsync_start = mode->VSyncStart; + drm_mode.vsync_end = mode->VSyncEnd; + drm_mode.vtotal = mode->VTotal; + drm_mode.flags = mode->Flags; + drm_mode.hskew = mode->HSkew; + drm_mode.vscan = mode->VScan; + drm_mode.vrefresh = mode->VRefresh; + if (!mode->name) + xf86SetModeDefaultName(mode); + strncpy(drm_mode.name, mode->name, DRM_DISPLAY_MODE_LEN); + + drmModeSetCrtc(ms->fd, drm_crtc->crtc_id, ms->fb_id, x, y, + &drm_connector->connector_id, 1, &drm_mode); +} + +#if 0 +static void +crtc_load_lut(xf86CrtcPtr crtc) +{ + //ScrnInfoPtr pScrn = crtc->scrn; +} +#endif + +static void +crtc_gamma_set(xf86CrtcPtr crtc, CARD16 * red, CARD16 * green, CARD16 * blue, + int size) +{ +} + +static void * +crtc_shadow_allocate(xf86CrtcPtr crtc, int width, int height) +{ + //ScrnInfoPtr pScrn = crtc->scrn; + + return NULL; +} + +static PixmapPtr +crtc_shadow_create(xf86CrtcPtr crtc, void *data, int width, int height) +{ + //ScrnInfoPtr pScrn = crtc->scrn; + + return NULL; +} + +static void +crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data) +{ + //ScrnInfoPtr pScrn = crtc->scrn; +} + +static void +crtc_destroy(xf86CrtcPtr crtc) +{ + modesettingPtr ms = modesettingPTR(crtc->scrn); + struct crtc_private *crtcp = crtc->driver_private; + + if (crtcp->cursor_buf) + pipe_buffer_reference(&crtcp->cursor_buf, NULL); + + drmModeFreeCrtc(crtcp->drm_crtc); + xfree(crtcp); +} + +static void +crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image) +{ + unsigned char *ptr; + modesettingPtr ms = modesettingPTR(crtc->scrn); + struct crtc_private *crtcp = crtc->driver_private; + + if (!crtcp->cursor_buf) { + crtcp->cursor_buf = pipe_buffer_create(ms->screen, + 0, + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ, + 64*64*4); + drm_api_hooks.handle_from_buffer(ms->screen, + crtcp->cursor_buf, + &crtcp->cursor_handle); + } + + ptr = pipe_buffer_map(ms->screen, crtcp->cursor_buf, PIPE_BUFFER_USAGE_CPU_WRITE); + + if (ptr) + memcpy(ptr, image, 64 * 64 * 4); + + pipe_buffer_unmap(ms->screen, crtcp->cursor_buf); +} + +static void +crtc_set_cursor_position(xf86CrtcPtr crtc, int x, int y) +{ + modesettingPtr ms = modesettingPTR(crtc->scrn); + struct crtc_private *crtcp = crtc->driver_private; + + drmModeMoveCursor(ms->fd, crtcp->drm_crtc->crtc_id, x, y); +} + +static void +crtc_show_cursor(xf86CrtcPtr crtc) +{ + modesettingPtr ms = modesettingPTR(crtc->scrn); + struct crtc_private *crtcp = crtc->driver_private; + + if (crtcp->cursor_buf) + drmModeSetCursor(ms->fd, crtcp->drm_crtc->crtc_id, + crtcp->cursor_handle, 64, 64); +} + +static void +crtc_hide_cursor(xf86CrtcPtr crtc) +{ + modesettingPtr ms = modesettingPTR(crtc->scrn); + struct crtc_private *crtcp = crtc->driver_private; + + drmModeSetCursor(ms->fd, crtcp->drm_crtc->crtc_id, 0, 0, 0); +} + +static const xf86CrtcFuncsRec crtc_funcs = { + .dpms = crtc_dpms, + .save = NULL, + .restore = NULL, + .lock = crtc_lock, + .unlock = crtc_unlock, + .mode_fixup = crtc_mode_fixup, + .prepare = crtc_prepare, + .mode_set = crtc_mode_set, + .commit = crtc_commit, + .gamma_set = crtc_gamma_set, + .shadow_create = crtc_shadow_create, + .shadow_allocate = crtc_shadow_allocate, + .shadow_destroy = crtc_shadow_destroy, + .set_cursor_position = crtc_set_cursor_position, + .show_cursor = crtc_show_cursor, + .hide_cursor = crtc_hide_cursor, + .load_cursor_image = NULL, /* lets convert to argb only */ + .set_cursor_colors = NULL, /* using argb only */ + .load_cursor_argb = crtc_load_cursor_argb, + .destroy = crtc_destroy, +}; + +void +cursor_destroy(xf86CrtcPtr crtc) +{ + modesettingPtr ms = modesettingPTR(crtc->scrn); + struct crtc_private *crtcp = crtc->driver_private; + + if (crtcp->cursor_buf) { + pipe_buffer_reference(&crtcp->cursor_buf, NULL); + } +} + +void +crtc_init(ScrnInfoPtr pScrn) +{ + modesettingPtr ms = modesettingPTR(pScrn); + xf86CrtcPtr crtc; + drmModeResPtr res; + drmModeCrtcPtr drm_crtc = NULL; + struct crtc_private *crtcp; + int c; + + res = drmModeGetResources(ms->fd); + if (res == 0) { + ErrorF("Failed drmModeGetResources %d\n", errno); + return; + } + + for (c = 0; c < res->count_crtcs; c++) { + drm_crtc = drmModeGetCrtc(ms->fd, res->crtcs[c]); + if (!drm_crtc) + continue; + + crtc = xf86CrtcCreate(pScrn, &crtc_funcs); + if (crtc == NULL) + goto out; + + crtcp = xcalloc(1, sizeof(struct crtc_private)); + if (!crtcp) { + xf86CrtcDestroy(crtc); + goto out; + } + + crtcp->drm_crtc = drm_crtc; + + crtc->driver_private = crtcp; + + } + + out: + drmModeFreeResources(res); +} + +/* vim: set sw=4 ts=8 sts=4: */ diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c new file mode 100644 index 0000000000..d04204e1bf --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -0,0 +1,212 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + +#include "xf86.h" +#include "xf86_OSproc.h" + +#include "xorg_tracker.h" + +#include "dri2.h" + +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" + +typedef struct { + PixmapPtr pPixmap; + struct pipe_texture *tex; + struct pipe_buffer *buf; +} *BufferPrivatePtr; + +static DRI2BufferPtr +driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) +{ + struct pipe_texture *depth, *tex; + struct pipe_buffer *buf; + ScreenPtr pScreen = pDraw->pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + BufferPrivatePtr privates; + DRI2BufferPtr buffers; + PixmapPtr pPixmap; + unsigned stride, handle; + int i; + + buffers = xcalloc(count, sizeof *buffers); + if (!buffers) + goto fail_buffers; + + privates = xcalloc(count, sizeof *privates); + if (!privates) + goto fail_privates; + + depth = NULL; + for (i = 0; i < count; i++) { + pPixmap = NULL; + tex = NULL; + buf = NULL; + if (attachments[i] == DRI2BufferFrontLeft) { + if (pDraw->type == DRAWABLE_PIXMAP) + pPixmap = (PixmapPtr) pDraw; + else + pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr) pDraw); + pPixmap->refcnt++; + tex = xorg_exa_get_texture(pPixmap); + } else if (attachments[i] == DRI2BufferStencil) { + pipe_texture_reference(&tex, depth); + } else if (attachments[i] == DRI2BufferDepth) { + struct pipe_texture template; + + memset(&template, 0, sizeof(template)); + template.target = PIPE_TEXTURE_2D; + template.compressed = 0; + template.format = PIPE_FORMAT_S8Z24_UNORM; + pf_get_block(template.format, &template.block); + template.width[0] = pDraw->width; + template.height[0] = pDraw->height; + template.depth[0] = 1; + template.last_level = 0; + template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + tex = ms->screen->texture_create(ms->screen, &template); + } else { + struct pipe_texture template; + memset(&template, 0, sizeof(template)); + template.target = PIPE_TEXTURE_2D; + template.compressed = 0; + template.format = PIPE_FORMAT_A8R8G8B8_UNORM; + pf_get_block(template.format, &template.block); + template.width[0] = pDraw->width; + template.height[0] = pDraw->height; + template.depth[0] = 1; + template.last_level = 0; + template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + tex = ms->screen->texture_create(ms->screen, &template); + } + + drm_api_hooks.buffer_from_texture(tex, &buf, &stride); + drm_api_hooks.global_handle_from_buffer(ms->screen, buf, &handle); + + buffers[i].name = handle; + buffers[i].attachment = attachments[i]; + buffers[i].pitch = stride; + buffers[i].cpp = 4; + buffers[i].driverPrivate = &privates[i]; + buffers[i].flags = 0; /* not tiled */ + privates[i].pPixmap = pPixmap; + privates[i].buf = buf; + privates[i].tex = tex; + } + + return buffers; + +fail_privates: + xfree(buffers); +fail_buffers: + return NULL; +} + +static void +driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count) +{ + ScreenPtr pScreen = pDraw->pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + BufferPrivatePtr private; + int i; + + for (i = 0; i < count; i++) { + private = buffers[i].driverPrivate; + + if (private->pPixmap) + (*pScreen->DestroyPixmap)(private->pPixmap); + + pipe_texture_reference(&private->tex, NULL); + pipe_buffer_reference(&private->buf, NULL); + } + + if (buffers) { + xfree(buffers[0].driverPrivate); + xfree(buffers); + } +} + +static void +driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion, + DRI2BufferPtr pDestBuffer, DRI2BufferPtr pSrcBuffer) +{ + ScreenPtr pScreen = pDraw->pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + BufferPrivatePtr dst_priv = pDestBuffer->driverPrivate; + BufferPrivatePtr src_priv = pSrcBuffer->driverPrivate; + + struct pipe_surface *dst_surf = + ms->screen->get_tex_surface(ms->screen, dst_priv->tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + struct pipe_surface *src_surf = + ms->screen->get_tex_surface(ms->screen, src_priv->tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ); + + ms->ctx->surface_copy(ms->ctx, dst_surf, 0, 0, src_surf, + 0, 0, pDraw->width, pDraw->height); + + pipe_surface_reference(&dst_surf, NULL); + pipe_surface_reference(&src_surf, NULL); +} + +Bool +driScreenInit(ScreenPtr pScreen) +{ + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + DRI2InfoRec dri2info; + + dri2info.version = 1; + dri2info.fd = ms->fd; +#if 0 + dri2info.driverName = pScrn->name; +#else + dri2info.driverName = "i915"; /* FIXME */ +#endif + dri2info.deviceName = "/dev/dri/card0"; /* FIXME */ + + dri2info.CreateBuffers = driCreateBuffers; + dri2info.DestroyBuffers = driDestroyBuffers; + dri2info.CopyRegion = driCopyRegion; + + return DRI2ScreenInit(pScreen, &dri2info); +} + +void +driCloseScreen(ScreenPtr pScreen) +{ + DRI2CloseScreen(pScreen); +} + +/* vim: set sw=4 ts=8 sts=4: */ diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c new file mode 100644 index 0000000000..8a2711e70c --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_driver.c @@ -0,0 +1,695 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + + +#include "xorg-server.h" +#include "xf86.h" +#include "xf86_OSproc.h" +#include "compiler.h" +#include "xf86RAC.h" +#include "xf86PciInfo.h" +#include "xf86Pci.h" +#include "xf86Resources.h" +#include "mipointer.h" +#include "micmap.h" +#include <X11/extensions/randr.h> +#include "fb.h" +#include "edid.h" +#include "xf86i2c.h" +#include "xf86Crtc.h" +#include "miscstruct.h" +#include "dixstruct.h" +#include "xf86xv.h" +#include <X11/extensions/Xv.h> +#ifndef XSERVER_LIBPCIACCESS +#error "libpciaccess needed" +#endif + +#include <pciaccess.h> + +#include "xorg_tracker.h" +#include "xorg_winsys.h" + +static void AdjustFrame(int scrnIndex, int x, int y, int flags); +static Bool CloseScreen(int scrnIndex, ScreenPtr pScreen); +static Bool EnterVT(int scrnIndex, int flags); +static Bool SaveHWState(ScrnInfoPtr pScrn); +static Bool RestoreHWState(ScrnInfoPtr pScrn); + + +static ModeStatus ValidMode(int scrnIndex, DisplayModePtr mode, Bool verbose, + int flags); +static void FreeScreen(int scrnIndex, int flags); +static void LeaveVT(int scrnIndex, int flags); +static Bool SwitchMode(int scrnIndex, DisplayModePtr mode, int flags); +static Bool ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, + char **argv); +static Bool PreInit(ScrnInfoPtr pScrn, int flags); + +typedef enum +{ + OPTION_SW_CURSOR, +} modesettingOpts; + +static const OptionInfoRec Options[] = { + {OPTION_SW_CURSOR, "SWcursor", OPTV_BOOLEAN, {0}, FALSE}, + {-1, NULL, OPTV_NONE, {0}, FALSE} +}; + +/* + * Functions that might be needed + */ + +static const char *exaSymbols[] = { + "exaGetVersion", + "exaDriverInit", + "exaDriverFini", + "exaOffscreenAlloc", + "exaOffscreenFree", + "exaWaitSync", + NULL +}; + +static const char *fbSymbols[] = { + "fbPictureInit", + "fbScreenInit", + NULL +}; + +static const char *ddcSymbols[] = { + "xf86PrintEDID", + "xf86SetDDCproperties", + NULL +}; + +/* + * Exported Xorg driver functions to winsys + */ + +void +xorg_tracker_loader_ref_sym_lists() +{ + LoaderRefSymLists(exaSymbols, fbSymbols, ddcSymbols, NULL); +} + +const OptionInfoRec * +xorg_tracker_available_options(int chipid, int busid) +{ + return Options; +} + +void +xorg_tracker_set_functions(ScrnInfoPtr scrn) +{ + scrn->PreInit = PreInit; + scrn->ScreenInit = ScreenInit; + scrn->SwitchMode = SwitchMode; + scrn->AdjustFrame = AdjustFrame; + scrn->EnterVT = EnterVT; + scrn->LeaveVT = LeaveVT; + scrn->FreeScreen = FreeScreen; + scrn->ValidMode = ValidMode; +} + +/* + * Static Xorg funtctions + */ + +static Bool +GetRec(ScrnInfoPtr pScrn) +{ + if (pScrn->driverPrivate) + return TRUE; + + pScrn->driverPrivate = xnfcalloc(sizeof(modesettingRec), 1); + + return TRUE; +} + +static void +FreeRec(ScrnInfoPtr pScrn) +{ + if (!pScrn) + return; + + if (!pScrn->driverPrivate) + return; + + xfree(pScrn->driverPrivate); + + pScrn->driverPrivate = NULL; +} + +static void +ProbeDDC(ScrnInfoPtr pScrn, int index) +{ + ConfiguredMonitor = NULL; +} + +static Bool +CreateFrontBuffer(ScrnInfoPtr pScrn) +{ + modesettingPtr ms = modesettingPTR(pScrn); + ScreenPtr pScreen = pScrn->pScreen; + PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen); + + ms->noEvict = TRUE; + pScreen->ModifyPixmapHeader(rootPixmap, + pScrn->virtualX, pScrn->virtualY, + pScrn->depth, pScrn->bitsPerPixel, + pScrn->displayWidth * pScrn->bitsPerPixel / 8, + NULL); + ms->noEvict = FALSE; + + drmModeAddFB(ms->fd, + pScrn->virtualX, + pScrn->virtualY, + pScrn->depth, + pScrn->bitsPerPixel, + pScrn->displayWidth * pScrn->bitsPerPixel / 8, + xorg_exa_get_pixmap_handle(rootPixmap), &ms->fb_id); + + pScrn->frameX0 = 0; + pScrn->frameY0 = 0; + AdjustFrame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0); + + return TRUE; +} + +static Bool +crtc_resize(ScrnInfoPtr pScrn, int width, int height) +{ + modesettingPtr ms = modesettingPTR(pScrn); + //ScreenPtr pScreen = pScrn->pScreen; + //PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen); + //Bool fbAccessDisabled; + //CARD8 *fbstart; + + if (width == pScrn->virtualX && height == pScrn->virtualY) + return TRUE; + + ErrorF("RESIZING TO %dx%d\n", width, height); + + pScrn->virtualX = width; + pScrn->virtualY = height; + + /* HW dependent - FIXME */ + pScrn->displayWidth = pScrn->virtualX; + + drmModeRmFB(ms->fd, ms->fb_id); + + /* now create new frontbuffer */ + return CreateFrontBuffer(pScrn); +} + +static const xf86CrtcConfigFuncsRec crtc_config_funcs = { + crtc_resize +}; + +static Bool +PreInit(ScrnInfoPtr pScrn, int flags) +{ + xf86CrtcConfigPtr xf86_config; + modesettingPtr ms; + rgb defaultWeight = { 0, 0, 0 }; + EntityInfoPtr pEnt; + EntPtr msEnt = NULL; + char *BusID; + int max_width, max_height; + + if (pScrn->numEntities != 1) + return FALSE; + + pEnt = xf86GetEntityInfo(pScrn->entityList[0]); + + if (flags & PROBE_DETECT) { + ProbeDDC(pScrn, pEnt->index); + return TRUE; + } + + /* Allocate driverPrivate */ + if (!GetRec(pScrn)) + return FALSE; + + ms = modesettingPTR(pScrn); + ms->SaveGeneration = -1; + ms->pEnt = pEnt; + + pScrn->displayWidth = 640; /* default it */ + + if (ms->pEnt->location.type != BUS_PCI) + return FALSE; + + ms->PciInfo = xf86GetPciInfoForEntity(ms->pEnt->index); + + /* Allocate an entity private if necessary */ + if (xf86IsEntityShared(pScrn->entityList[0])) { + FatalError("Entity"); +#if 0 + msEnt = xf86GetEntityPrivate(pScrn->entityList[0], + modesettingEntityIndex)->ptr; + ms->entityPrivate = msEnt; +#else + (void)msEnt; +#endif + } else + ms->entityPrivate = NULL; + + if (xf86RegisterResources(ms->pEnt->index, NULL, ResNone)) { + return FALSE; + } + + if (xf86IsEntityShared(pScrn->entityList[0])) { + if (xf86IsPrimInitDone(pScrn->entityList[0])) { + /* do something */ + } else { + xf86SetPrimInitDone(pScrn->entityList[0]); + } + } + + BusID = xalloc(64); + sprintf(BusID, "PCI:%d:%d:%d", + ((ms->PciInfo->domain << 8) | ms->PciInfo->bus), + ms->PciInfo->dev, ms->PciInfo->func + ); + + ms->fd = drmOpen(NULL, BusID); + + if (ms->fd < 0) + return FALSE; + + pScrn->racMemFlags = RAC_FB | RAC_COLORMAP; + pScrn->monitor = pScrn->confScreen->monitor; + pScrn->progClock = TRUE; + pScrn->rgbBits = 8; + + if (!xf86SetDepthBpp + (pScrn, 0, 0, 0, + PreferConvert24to32 | SupportConvert24to32 | Support32bppFb)) + return FALSE; + + switch (pScrn->depth) { + case 15: + case 16: + case 24: + break; + default: + xf86DrvMsg(pScrn->scrnIndex, X_ERROR, + "Given depth (%d) is not supported by the driver\n", + pScrn->depth); + return FALSE; + } + xf86PrintDepthBpp(pScrn); + + if (!xf86SetWeight(pScrn, defaultWeight, defaultWeight)) + return FALSE; + if (!xf86SetDefaultVisual(pScrn, -1)) + return FALSE; + + /* Process the options */ + xf86CollectOptions(pScrn, NULL); + if (!(ms->Options = xalloc(sizeof(Options)))) + return FALSE; + memcpy(ms->Options, Options, sizeof(Options)); + xf86ProcessOptions(pScrn->scrnIndex, pScrn->options, ms->Options); + + /* Allocate an xf86CrtcConfig */ + xf86CrtcConfigInit(pScrn, &crtc_config_funcs); + xf86_config = XF86_CRTC_CONFIG_PTR(pScrn); + + max_width = 8192; + max_height = 8192; + xf86CrtcSetSizeRange(pScrn, 320, 200, max_width, max_height); + + if (xf86ReturnOptValBool(ms->Options, OPTION_SW_CURSOR, FALSE)) { + ms->SWCursor = TRUE; + } + + SaveHWState(pScrn); + + crtc_init(pScrn); + output_init(pScrn); + + if (!xf86InitialConfiguration(pScrn, TRUE)) { + xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "No valid modes.\n"); + RestoreHWState(pScrn); + return FALSE; + } + + RestoreHWState(pScrn); + + /* + * If the driver can do gamma correction, it should call xf86SetGamma() here. + */ + { + Gamma zeros = { 0.0, 0.0, 0.0 }; + + if (!xf86SetGamma(pScrn, zeros)) { + return FALSE; + } + } + + if (pScrn->modes == NULL) { + xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "No modes.\n"); + return FALSE; + } + + pScrn->currentMode = pScrn->modes; + + /* Set display resolution */ + xf86SetDpi(pScrn, 0, 0); + + /* Load the required sub modules */ + if (!xf86LoadSubModule(pScrn, "fb")) { + return FALSE; + } + + xf86LoaderReqSymLists(fbSymbols, NULL); + + xf86LoadSubModule(pScrn, "exa"); + +#ifdef DRI2 + xf86LoadSubModule(pScrn, "dri2"); +#endif + + return TRUE; +} + +static Bool +SaveHWState(ScrnInfoPtr pScrn) +{ + /*xf86CrtcConfigPtr xf86_config = XF86_CRTC_CONFIG_PTR(pScrn);*/ + + return TRUE; +} + +static Bool +RestoreHWState(ScrnInfoPtr pScrn) +{ + /*xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn);*/ + + return TRUE; +} + +static Bool +CreateScreenResources(ScreenPtr pScreen) +{ + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + PixmapPtr rootPixmap; + Bool ret; + + ms->noEvict = TRUE; + + pScreen->CreateScreenResources = ms->createScreenResources; + ret = pScreen->CreateScreenResources(pScreen); + pScreen->CreateScreenResources = CreateScreenResources; + + rootPixmap = pScreen->GetScreenPixmap(pScreen); + + if (!pScreen->ModifyPixmapHeader(rootPixmap, -1, -1, -1, -1, -1, NULL)) + FatalError("Couldn't adjust screen pixmap\n"); + + ms->noEvict = FALSE; + + drmModeAddFB(ms->fd, + pScrn->virtualX, + pScrn->virtualY, + pScrn->depth, + pScrn->bitsPerPixel, + pScrn->displayWidth * pScrn->bitsPerPixel / 8, + xorg_exa_get_pixmap_handle(rootPixmap), &ms->fb_id); + + AdjustFrame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0); + + return ret; +} + +static Bool +ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv) +{ + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + VisualPtr visual; + + /* deal with server regeneration */ + if (ms->fd < 0) { + char *BusID; + + BusID = xalloc(64); + sprintf(BusID, "PCI:%d:%d:%d", + ((ms->PciInfo->domain << 8) | ms->PciInfo->bus), + ms->PciInfo->dev, ms->PciInfo->func + ); + + ms->fd = drmOpen(NULL, BusID); + + if (ms->fd < 0) + return FALSE; + } + + if (!ms->screen) { + ms->screen = drm_api_hooks.create_screen(ms->fd, ms->PciInfo->device_id); + + if (!ms->screen) { + FatalError("Could not init pipe_screen\n"); + return FALSE; + } + } + + pScrn->pScreen = pScreen; + + /* HW dependent - FIXME */ + pScrn->displayWidth = pScrn->virtualX; + + miClearVisualTypes(); + + if (!miSetVisualTypes(pScrn->depth, + miGetDefaultVisualMask(pScrn->depth), + pScrn->rgbBits, pScrn->defaultVisual)) + return FALSE; + + if (!miSetPixmapDepths()) + return FALSE; + + pScrn->memPhysBase = 0; + pScrn->fbOffset = 0; + + if (!fbScreenInit(pScreen, NULL, + pScrn->virtualX, pScrn->virtualY, + pScrn->xDpi, pScrn->yDpi, + pScrn->displayWidth, pScrn->bitsPerPixel)) + return FALSE; + + if (pScrn->bitsPerPixel > 8) { + /* Fixup RGB ordering */ + visual = pScreen->visuals + pScreen->numVisuals; + while (--visual >= pScreen->visuals) { + if ((visual->class | DynamicClass) == DirectColor) { + visual->offsetRed = pScrn->offset.red; + visual->offsetGreen = pScrn->offset.green; + visual->offsetBlue = pScrn->offset.blue; + visual->redMask = pScrn->mask.red; + visual->greenMask = pScrn->mask.green; + visual->blueMask = pScrn->mask.blue; + } + } + } + + fbPictureInit(pScreen, NULL, 0); + + ms->createScreenResources = pScreen->CreateScreenResources; + pScreen->CreateScreenResources = CreateScreenResources; + + xf86SetBlackWhitePixels(pScreen); + + ms->exa = xorg_exa_init(pScrn); + + miInitializeBackingStore(pScreen); + xf86SetBackingStore(pScreen); + xf86SetSilkenMouse(pScreen); + miDCInitialize(pScreen, xf86GetPointerScreenFuncs()); + + /* Need to extend HWcursor support to handle mask interleave */ + if (!ms->SWCursor) + xf86_cursors_init(pScreen, 64, 64, + HARDWARE_CURSOR_SOURCE_MASK_INTERLEAVE_64 | + HARDWARE_CURSOR_ARGB); + + /* Must force it before EnterVT, so we are in control of VT and + * later memory should be bound when allocating, e.g rotate_mem */ + pScrn->vtSema = TRUE; + + pScreen->SaveScreen = xf86SaveScreen; + ms->CloseScreen = pScreen->CloseScreen; + pScreen->CloseScreen = CloseScreen; + + if (!xf86CrtcScreenInit(pScreen)) + return FALSE; + + if (!miCreateDefColormap(pScreen)) + return FALSE; + + xf86DPMSInit(pScreen, xf86DPMSSet, 0); + + if (serverGeneration == 1) + xf86ShowUnusedOptions(pScrn->scrnIndex, pScrn->options); + +#if 1 +#ifdef DRI2 + driScreenInit(pScreen); +#endif +#endif + + return EnterVT(scrnIndex, 1); +} + +static void +AdjustFrame(int scrnIndex, int x, int y, int flags) +{ + ScrnInfoPtr pScrn = xf86Screens[scrnIndex]; + xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn); + xf86OutputPtr output = config->output[config->compat_output]; + xf86CrtcPtr crtc = output->crtc; + + if (crtc && crtc->enabled) { + crtc->funcs->mode_set(crtc, pScrn->currentMode, pScrn->currentMode, x, + y); + crtc->x = output->initial_x + x; + crtc->y = output->initial_y + y; + } +} + +static void +FreeScreen(int scrnIndex, int flags) +{ + FreeRec(xf86Screens[scrnIndex]); +} + +/* HACK */ +void +cursor_destroy(xf86CrtcPtr crtc); + +static void +LeaveVT(int scrnIndex, int flags) +{ + ScrnInfoPtr pScrn = xf86Screens[scrnIndex]; + modesettingPtr ms = modesettingPTR(pScrn); + xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn); + int o; + + for (o = 0; o < config->num_crtc; o++) { + xf86CrtcPtr crtc = config->crtc[o]; + + cursor_destroy(crtc); + + if (crtc->rotatedPixmap || crtc->rotatedData) { + crtc->funcs->shadow_destroy(crtc, crtc->rotatedPixmap, + crtc->rotatedData); + crtc->rotatedPixmap = NULL; + crtc->rotatedData = NULL; + } + } + + drmModeRmFB(ms->fd, ms->fb_id); + + RestoreHWState(pScrn); + + pScrn->vtSema = FALSE; +} + +/* + * This gets called when gaining control of the VT, and from ScreenInit(). + */ +static Bool +EnterVT(int scrnIndex, int flags) +{ + ScrnInfoPtr pScrn = xf86Screens[scrnIndex]; + modesettingPtr ms = modesettingPTR(pScrn); + + /* + * Only save state once per server generation since that's what most + * drivers do. Could change this to save state at each VT enter. + */ + if (ms->SaveGeneration != serverGeneration) { + ms->SaveGeneration = serverGeneration; + SaveHWState(pScrn); + } + + if (!flags) /* signals startup as we'll do this in CreateScreenResources */ + CreateFrontBuffer(pScrn); + + if (!xf86SetDesiredModes(pScrn)) + return FALSE; + + return TRUE; +} + +static Bool +SwitchMode(int scrnIndex, DisplayModePtr mode, int flags) +{ + ScrnInfoPtr pScrn = xf86Screens[scrnIndex]; + + return xf86SetSingleMode(pScrn, mode, RR_Rotate_0); +} + +static Bool +CloseScreen(int scrnIndex, ScreenPtr pScreen) +{ + ScrnInfoPtr pScrn = xf86Screens[scrnIndex]; + modesettingPtr ms = modesettingPTR(pScrn); + + if (pScrn->vtSema) { + LeaveVT(scrnIndex, 0); + } +#ifdef DRI2 + driCloseScreen(pScreen); +#endif + + pScreen->CreateScreenResources = ms->createScreenResources; + + if (ms->exa) + xorg_exa_close(pScrn); + + drmClose(ms->fd); + ms->fd = -1; + + pScrn->vtSema = FALSE; + pScreen->CloseScreen = ms->CloseScreen; + return (*pScreen->CloseScreen) (scrnIndex, pScreen); +} + +static ModeStatus +ValidMode(int scrnIndex, DisplayModePtr mode, Bool verbose, int flags) +{ + return MODE_OK; +} + +/* vim: set sw=4 ts=8 sts=4: */ diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c new file mode 100644 index 0000000000..56c8fdccb2 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_exa.c @@ -0,0 +1,528 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + +#include "xf86.h" +#include "xorg_tracker.h" + +#include "pipe/p_format.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" + +struct exa_entity +{ + ExaDriverPtr pExa; + struct pipe_context *ctx; + struct pipe_screen *scrn; +}; + +struct PixmapPriv +{ + int flags; + struct pipe_texture *tex; + unsigned int color; + struct pipe_surface *src_surf; /* for copies */ + + struct pipe_transfer *map_transfer; + unsigned map_count; +}; + +/* + * Helper functions + */ + +static void +exa_get_pipe_format(int depth, enum pipe_format *format, int *bbp) +{ + switch (depth) { + case 32: + *format = PIPE_FORMAT_A8R8G8B8_UNORM; + assert(*bbp == 32); + break; + case 24: + *format = PIPE_FORMAT_X8R8G8B8_UNORM; + assert(*bbp == 32); + break; + case 16: + *format = PIPE_FORMAT_R5G6B5_UNORM; + assert(*bbp == 16); + break; + case 15: + *format = PIPE_FORMAT_A1R5G5B5_UNORM; + assert(*bbp == 16); + break; + case 8: + case 4: + case 1: + *format = PIPE_FORMAT_A8R8G8B8_UNORM; /* bad bad bad */ + break; + default: + assert(0); + break; + } +} + +/* + * Static exported EXA functions + */ + +static void +ExaWaitMarker(ScreenPtr pScreen, int marker) +{ +} + +static int +ExaMarkSync(ScreenPtr pScreen) +{ + return 1; +} + +static Bool +ExaPrepareAccess(PixmapPtr pPix, int index) +{ + ScreenPtr pScreen = pPix->drawable.pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + struct PixmapPriv *priv; + + priv = exaGetPixmapDriverPrivate(pPix); + + if (!priv) + return FALSE; + + if (!priv->tex) + return FALSE; + + if (priv->map_count++ == 0) + { + priv->map_transfer = + exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, + PIPE_TRANSFER_READ_WRITE, + 0, 0, priv->tex->width[0], priv->tex->height[0]); + + pPix->devPrivate.ptr = + exa->scrn->transfer_map(exa->scrn, priv->map_transfer); + pPix->devKind = priv->map_transfer->stride; + } + + return TRUE; +} + +static void +ExaFinishAccess(PixmapPtr pPix, int index) +{ + ScreenPtr pScreen = pPix->drawable.pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + struct PixmapPriv *priv; + priv = exaGetPixmapDriverPrivate(pPix); + + if (!priv) + return; + + if (!priv->map_transfer) + return; + + if (--priv->map_count == 0) { + assert(priv->map_transfer); + exa->scrn->transfer_unmap(exa->scrn, priv->map_transfer); + exa->scrn->tex_transfer_destroy(priv->map_transfer); + priv->map_transfer = NULL; + } +} + +static void +ExaDone(PixmapPtr pPixmap) +{ + ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_entity *exa = ms->exa; + + if (!priv) + return; + + if (priv->src_surf) + exa->scrn->tex_surface_destroy(priv->src_surf); + priv->src_surf = NULL; +} + +static void +ExaDoneComposite(PixmapPtr pPixmap) +{ + +} + +static Bool +ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg) +{ + ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_entity *exa = ms->exa; + + if (1) + return FALSE; + + if (pPixmap->drawable.depth < 15) + return FALSE; + + if (!EXA_PM_IS_SOLID(&pPixmap->drawable, planeMask)) + return FALSE; + + if (!priv || !priv->tex) + return FALSE; + + if (alu != GXcopy) + return FALSE; + + if (!exa->ctx || !exa->ctx->surface_fill) + return FALSE; + + priv->color = fg; + + return TRUE; +} + +static void +ExaSolid(PixmapPtr pPixmap, int x0, int y0, int x1, int y1) +{ + ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct pipe_surface *surf = + exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + exa->ctx->surface_fill(exa->ctx, surf, x0, y0, x1 - x0, y1 - y0, + priv->color); + + exa->scrn->tex_surface_destroy(surf); +} + +static Bool +ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir, + int ydir, int alu, Pixel planeMask) +{ + ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap); + struct PixmapPriv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap); + + if (1) + return FALSE; + + if (alu != GXcopy) + return FALSE; + + if (pSrcPixmap->drawable.depth < 15 || pDstPixmap->drawable.depth < 15) + return FALSE; + + if (!EXA_PM_IS_SOLID(&pSrcPixmap->drawable, planeMask)) + return FALSE; + + if (!priv || !src_priv) + return FALSE; + + if (!priv->tex || !src_priv->tex) + return FALSE; + + if (!exa->ctx || !exa->ctx->surface_copy) + return FALSE; + + priv->src_surf = + exa->scrn->get_tex_surface(exa->scrn, src_priv->tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + return TRUE; +} + +static void +ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY, + int width, int height) +{ + ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap); + struct pipe_surface *surf = + exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + exa->ctx->surface_copy(exa->ctx, surf, dstX, dstY, priv->src_surf, + srcX, srcY, width, height); + exa->scrn->tex_surface_destroy(surf); +} + +static Bool +ExaPrepareComposite(int op, PicturePtr pSrcPicture, + PicturePtr pMaskPicture, PicturePtr pDstPicture, + PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst) +{ + return FALSE; +} + +static void +ExaComposite(PixmapPtr pDst, int srcX, int srcY, int maskX, int maskY, + int dstX, int dstY, int width, int height) +{ +} + +static Bool +ExaCheckComposite(int op, + PicturePtr pSrcPicture, PicturePtr pMaskPicture, + PicturePtr pDstPicture) +{ + return FALSE; +} + +static void * +ExaCreatePixmap(ScreenPtr pScreen, int size, int align) +{ + struct PixmapPriv *priv; + + priv = xcalloc(1, sizeof(struct PixmapPriv)); + if (!priv) + return NULL; + + return priv; +} + +static void +ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv) +{ + struct PixmapPriv *priv = (struct PixmapPriv *)dPriv; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + + if (!priv) + return; + + if (priv->tex) + ms->screen->texture_destroy(priv->tex); + + xfree(priv); +} + +static Bool +ExaPixmapIsOffscreen(PixmapPtr pPixmap) +{ + struct PixmapPriv *priv; + + priv = exaGetPixmapDriverPrivate(pPixmap); + + if (!priv) + return FALSE; + + if (priv->tex) + return TRUE; + + return FALSE; +} + +unsigned +xorg_exa_get_pixmap_handle(PixmapPtr pPixmap) +{ + ScreenPtr pScreen = pPixmap->drawable.pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct PixmapPriv *priv; + struct pipe_buffer *buffer = NULL; + unsigned handle; + unsigned stride; + + if (!ms->exa) { + FatalError("NO MS->EXA\n"); + return 0; + } + + priv = exaGetPixmapDriverPrivate(pPixmap); + + if (!priv) { + FatalError("NO PIXMAP PRIVATE\n"); + return 0; + } + + drm_api_hooks.buffer_from_texture(priv->tex, &buffer, &stride); + drm_api_hooks.handle_from_buffer(ms->screen, buffer, &handle); + pipe_buffer_reference(&buffer, NULL); + return handle; +} + +static Bool +ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, + int depth, int bitsPerPixel, int devKind, + pointer pPixData) +{ + ScreenPtr pScreen = pPixmap->drawable.pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + + if (!priv) + return FALSE; + + if (depth <= 0) + depth = pPixmap->drawable.depth; + + if (bitsPerPixel <= 0) + bitsPerPixel = pPixmap->drawable.bitsPerPixel; + + if (width <= 0) + width = pPixmap->drawable.width; + + if (height <= 0) + height = pPixmap->drawable.height; + + if (width <= 0 || height <= 0 || depth <= 0) + return FALSE; + + miModifyPixmapHeader(pPixmap, width, height, depth, + bitsPerPixel, devKind, NULL); + + /* Deal with screen resize */ + if (priv->tex && (priv->tex->width[0] != width || priv->tex->height[0] != height)) { + pipe_texture_reference(&priv->tex, NULL); + } + + if (!priv->tex) { + struct pipe_texture template; + + memset(&template, 0, sizeof(template)); + template.target = PIPE_TEXTURE_2D; + template.compressed = 0; + exa_get_pipe_format(depth, &template.format, &bitsPerPixel); + pf_get_block(template.format, &template.block); + template.width[0] = width; + template.height[0] = height; + template.depth[0] = 1; + template.last_level = 0; + template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + priv->tex = exa->scrn->texture_create(exa->scrn, &template); + } + + return TRUE; +} + +struct pipe_texture * +xorg_exa_get_texture(PixmapPtr pPixmap) +{ + struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct pipe_texture *tex = NULL; + pipe_texture_reference(&tex, priv->tex); + return tex; +} + +void +xorg_exa_close(ScrnInfoPtr pScrn) +{ + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa = ms->exa; + + if (exa->ctx) + exa->ctx->destroy(exa->ctx); + + exaDriverFini(pScrn->pScreen); + xfree(exa); + ms->exa = NULL; +} + +void * +xorg_exa_init(ScrnInfoPtr pScrn) +{ + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_entity *exa; + ExaDriverPtr pExa; + + exa = xcalloc(1, sizeof(struct exa_entity)); + if (!exa) + return NULL; + + pExa = exaDriverAlloc(); + if (!pExa) { + goto out_err; + } + + memset(pExa, 0, sizeof(*pExa)); + pExa->exa_major = 2; + pExa->exa_minor = 2; + pExa->memoryBase = 0; + pExa->memorySize = 0; + pExa->offScreenBase = 0; + pExa->pixmapOffsetAlign = 0; + pExa->pixmapPitchAlign = 1; + pExa->flags = EXA_OFFSCREEN_PIXMAPS | EXA_HANDLES_PIXMAPS; + pExa->maxX = 8191; /* FIXME */ + pExa->maxY = 8191; /* FIXME */ + pExa->WaitMarker = ExaWaitMarker; + pExa->MarkSync = ExaMarkSync; + pExa->PrepareSolid = ExaPrepareSolid; + pExa->Solid = ExaSolid; + pExa->DoneSolid = ExaDone; + pExa->PrepareCopy = ExaPrepareCopy; + pExa->Copy = ExaCopy; + pExa->DoneCopy = ExaDone; + pExa->CheckComposite = ExaCheckComposite; + pExa->PrepareComposite = ExaPrepareComposite; + pExa->Composite = ExaComposite; + pExa->DoneComposite = ExaDoneComposite; + pExa->PixmapIsOffscreen = ExaPixmapIsOffscreen; + pExa->PrepareAccess = ExaPrepareAccess; + pExa->FinishAccess = ExaFinishAccess; + pExa->CreatePixmap = ExaCreatePixmap; + pExa->DestroyPixmap = ExaDestroyPixmap; + pExa->ModifyPixmapHeader = ExaModifyPixmapHeader; + + if (!exaDriverInit(pScrn->pScreen, pExa)) { + goto out_err; + } + + exa->scrn = ms->screen; + exa->ctx = drm_api_hooks.create_context(exa->scrn); + /* Share context with DRI */ + ms->ctx = exa->ctx; + + return (void *)exa; + + out_err: + xorg_exa_close(pScrn); + + return NULL; +} + +/* vim: set sw=4 ts=8 sts=4: */ diff --git a/src/gallium/state_trackers/xorg/xorg_output.c b/src/gallium/state_trackers/xorg/xorg_output.c new file mode 100644 index 0000000000..950af942f5 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_output.c @@ -0,0 +1,296 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + +#include "xorg-server.h" +#include <xf86.h> +#include <xf86i2c.h> +#include <xf86Crtc.h> +#include <errno.h> +#include <fcntl.h> +#include <unistd.h> +#include <stdio.h> +#include <stdlib.h> +#include <stdint.h> +#include <string.h> +#include <sys/stat.h> +#include <sys/types.h> + +#define DPMS_SERVER +#include <X11/extensions/dpms.h> + +#include "X11/Xatom.h" + +#include "xorg_tracker.h" + +static char *connector_enum_list[] = { + "Unknown", + "VGA", + "DVI-I", + "DVI-D", + "DVI-A", + "Composite", + "SVIDEO", + "LVDS", + "Component", + "9-pin DIN", + "DisplayPort", + "HDMI Type A", + "HDMI Type B", +}; + +static void +dpms(xf86OutputPtr output, int mode) +{ +} + +static void +save(xf86OutputPtr output) +{ +} + +static void +restore(xf86OutputPtr output) +{ +} + +static int +mode_valid(xf86OutputPtr output, DisplayModePtr pMode) +{ + return MODE_OK; +} + +static Bool +mode_fixup(xf86OutputPtr output, DisplayModePtr mode, + DisplayModePtr adjusted_mode) +{ + return TRUE; +} + +static void +prepare(xf86OutputPtr output) +{ + dpms(output, DPMSModeOff); +} + +static void +mode_set(xf86OutputPtr output, DisplayModePtr mode, + DisplayModePtr adjusted_mode) +{ +} + +static void +commit(xf86OutputPtr output) +{ + dpms(output, DPMSModeOn); + + if (output->scrn->pScreen != NULL) + xf86_reload_cursors(output->scrn->pScreen); +} + +static xf86OutputStatus +detect(xf86OutputPtr output) +{ + drmModeConnectorPtr drm_connector = output->driver_private; + + switch (drm_connector->connection) { + case DRM_MODE_CONNECTED: + return XF86OutputStatusConnected; + case DRM_MODE_DISCONNECTED: + return XF86OutputStatusDisconnected; + default: + return XF86OutputStatusUnknown; + } +} + +static DisplayModePtr +get_modes(xf86OutputPtr output) +{ + drmModeConnectorPtr drm_connector = output->driver_private; + drmModeModeInfoPtr drm_mode = NULL; + DisplayModePtr modes = NULL, mode = NULL; + int i; + + for (i = 0; i < drm_connector->count_modes; i++) { + drm_mode = &drm_connector->modes[i]; + if (drm_mode) { + mode = xcalloc(1, sizeof(DisplayModeRec)); + if (!mode) + continue; + mode->type = 0; + mode->Clock = drm_mode->clock; + mode->HDisplay = drm_mode->hdisplay; + mode->HSyncStart = drm_mode->hsync_start; + mode->HSyncEnd = drm_mode->hsync_end; + mode->HTotal = drm_mode->htotal; + mode->VDisplay = drm_mode->vdisplay; + mode->VSyncStart = drm_mode->vsync_start; + mode->VSyncEnd = drm_mode->vsync_end; + mode->VTotal = drm_mode->vtotal; + mode->Flags = drm_mode->flags; + mode->HSkew = drm_mode->hskew; + mode->VScan = drm_mode->vscan; + mode->VRefresh = xf86ModeVRefresh(mode); + mode->Private = (void *)drm_mode; + xf86SetModeDefaultName(mode); + modes = xf86ModesAdd(modes, mode); + xf86PrintModeline(0, mode); + } + } + + return modes; +} + +static void +destroy(xf86OutputPtr output) +{ + drmModeFreeConnector(output->driver_private); +} + +static void +create_resources(xf86OutputPtr output) +{ +#ifdef RANDR_12_INTERFACE +#endif /* RANDR_12_INTERFACE */ +} + +#ifdef RANDR_12_INTERFACE +static Bool +set_property(xf86OutputPtr output, Atom property, RRPropertyValuePtr value) +{ + return TRUE; +} +#endif /* RANDR_12_INTERFACE */ + +#ifdef RANDR_13_INTERFACE +static Bool +get_property(xf86OutputPtr output, Atom property) +{ + return TRUE; +} +#endif /* RANDR_13_INTERFACE */ + +#ifdef RANDR_GET_CRTC_INTERFACE +static xf86CrtcPtr +get_crtc(xf86OutputPtr output) +{ + return NULL; +} +#endif + +static const xf86OutputFuncsRec output_funcs = { + .create_resources = create_resources, + .dpms = dpms, + .save = save, + .restore = restore, + .mode_valid = mode_valid, + .mode_fixup = mode_fixup, + .prepare = prepare, + .mode_set = mode_set, + .commit = commit, + .detect = detect, + .get_modes = get_modes, +#ifdef RANDR_12_INTERFACE + .set_property = set_property, +#endif +#ifdef RANDR_13_INTERFACE + .get_property = get_property, +#endif + .destroy = destroy, +#ifdef RANDR_GET_CRTC_INTERFACE + .get_crtc = get_crtc, +#endif +}; + +void +output_init(ScrnInfoPtr pScrn) +{ + modesettingPtr ms = modesettingPTR(pScrn); + xf86OutputPtr output; + drmModeResPtr res; + drmModeConnectorPtr drm_connector = NULL; + drmModeEncoderPtr drm_encoder = NULL; + char *name; + int c, v, p; + + res = drmModeGetResources(ms->fd); + if (res == 0) { + DRV_ERROR("Failed drmModeGetResources\n"); + return; + } + + for (c = 0; c < res->count_connectors; c++) { + drm_connector = drmModeGetConnector(ms->fd, res->connectors[c]); + if (!drm_connector) + goto out; + +#if 0 + for (p = 0; p < drm_connector->count_props; p++) { + drmModePropertyPtr prop; + + prop = drmModeGetProperty(ms->fd, drm_connector->props[p]); + + name = NULL; + if (prop) { + ErrorF("VALUES %d\n", prop->count_values); + + for (v = 0; v < prop->count_values; v++) + ErrorF("%s %lld\n", prop->name, prop->values[v]); + } + } +#else + (void)p; + (void)v; +#endif + + name = connector_enum_list[drm_connector->connector_type]; + + output = xf86OutputCreate(pScrn, &output_funcs, name); + if (!output) + continue; + + drm_encoder = drmModeGetEncoder(ms->fd, drm_connector->encoders[0]); + if (drm_encoder) { + output->possible_crtcs = drm_encoder->possible_crtcs; + output->possible_clones = drm_encoder->possible_clones; + } else { + output->possible_crtcs = 0; + output->possible_clones = 0; + } + output->driver_private = drm_connector; + output->subpixel_order = SubPixelHorizontalRGB; + output->interlaceAllowed = FALSE; + output->doubleScanAllowed = FALSE; + } + + out: + drmModeFreeResources(res); +} + +/* vim: set sw=4 ts=8 sts=4: */ diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h new file mode 100644 index 0000000000..82c3890dfb --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_tracker.h @@ -0,0 +1,130 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + +#ifndef _XORG_TRACKER_H_ +#define _XORG_TRACKER_H_ + +#include <errno.h> +#include <drm.h> +#include <xf86drm.h> +#include <xf86drmMode.h> +#include "exa.h" + +#include "pipe/p_screen.h" +#include "state_tracker/drm_api.h" + +#define DRV_ERROR(msg) xf86DrvMsg(pScrn->scrnIndex, X_ERROR, msg); + +typedef struct +{ + int lastInstance; + int refCount; + ScrnInfoPtr pScrn_1; + ScrnInfoPtr pScrn_2; +} EntRec, *EntPtr; + +typedef struct _modesettingRec +{ + /* drm */ + int fd; + unsigned fb_id; + + /* X */ + EntPtr entityPrivate; + + int Chipset; + EntityInfoPtr pEnt; + struct pci_device *PciInfo; + + Bool noAccel; + Bool SWCursor; + CloseScreenProcPtr CloseScreen; + + /* Broken-out options. */ + OptionInfoPtr Options; + + unsigned int SaveGeneration; + + CreateScreenResourcesProcPtr createScreenResources; + + /* gallium */ + struct pipe_screen *screen; + struct pipe_context *ctx; + + /* exa */ + void *exa; + Bool noEvict; + +} modesettingRec, *modesettingPtr; + +#define modesettingPTR(p) ((modesettingPtr)((p)->driverPrivate)) + + +/*********************************************************************** + * xorg_exa.c + */ +struct pipe_texture * +xorg_exa_get_texture(PixmapPtr pPixmap); + +unsigned +xorg_exa_get_pixmap_handle(PixmapPtr pPixmap); + +void * +xorg_exa_init(ScrnInfoPtr pScrn); + +void +xorg_exa_close(ScrnInfoPtr pScrn); + + +/*********************************************************************** + * xorg_dri2.c + */ +Bool +driScreenInit(ScreenPtr pScreen); + +void +driCloseScreen(ScreenPtr pScreen); + + +/*********************************************************************** + * xorg_crtc.c + */ +void +crtc_init(ScrnInfoPtr pScrn); + + +/*********************************************************************** + * xorg_output.c + */ +void +output_init(ScrnInfoPtr pScrn); + + +#endif /* _XORG_TRACKER_H_ */ diff --git a/src/gallium/state_trackers/wgl/stw.h b/src/gallium/state_trackers/xorg/xorg_winsys.h index 450af4ccb6..d523080e90 100644 --- a/src/gallium/state_trackers/wgl/stw.h +++ b/src/gallium/state_trackers/xorg/xorg_winsys.h @@ -1,5 +1,4 @@ -/************************************************************************** - * +/* * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * @@ -23,31 +22,30 @@ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * - **************************************************************************/ - -#ifndef STW_H -#define STW_H - -#include "pipe/p_compiler.h" - -struct stw_winsys; - -/* Public interface: + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * */ -boolean stw_init( const struct stw_winsys *stw_winsys ); -void stw_cleanup( void ); +/* + * File with all the junk needed to personalize the a xorg driver. + */ +#ifndef _XORG_WINSYS_H_ +#define _XORG_WINSYS_H_ -/* Internal functions - */ -boolean stw_shared_init( const struct stw_winsys *stw_winsys ); -boolean stw_icd_init( void ); -boolean stw_wgl_init( void ); +#include "xorg-server.h" +#include "xf86.h" +#include "xf86Resources.h" +#include "pciaccess.h" -void stw_shared_cleanup( void ); -void stw_icd_cleanup( void ); -void stw_wgl_cleanup( void ); +#ifndef XSERVER_LIBPCIACCESS +#error "libpciaccess needed" +#endif +void xorg_tracker_set_functions(ScrnInfoPtr scrn); +const OptionInfoRec * xorg_tracker_available_options(int chipid, int busid); +void xorg_tracker_loader_ref_sym_lists(void); -#endif /* STW_H */ +#endif diff --git a/src/gallium/winsys/Makefile b/src/gallium/winsys/Makefile index 2360a6a94a..bce5b3f9e0 100644 --- a/src/gallium/winsys/Makefile +++ b/src/gallium/winsys/Makefile @@ -1,24 +1,12 @@ +# src/gallium/winsys/Makefile TOP = ../../.. include $(TOP)/configs/current - SUBDIRS = $(GALLIUM_WINSYS_DIRS) - -default: subdirs - - -subdirs: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - - -clean: - rm -f `find . -name \*.[oa]` - - -# Dummy install target -install: diff --git a/src/gallium/winsys/drm/Makefile b/src/gallium/winsys/drm/Makefile index a84fcd8418..fee0191643 100644 --- a/src/gallium/winsys/drm/Makefile +++ b/src/gallium/winsys/drm/Makefile @@ -1,38 +1,12 @@ -# src/gallium/winsys/drm/Makefile - +# src/gallium/winsys/Makefile TOP = ../../../.. - include $(TOP)/configs/current SUBDIRS = $(GALLIUM_WINSYS_DRM_DIRS) -default: $(TOP)/$(LIB_DIR)/gallium subdirs - - -$(TOP)/$(LIB_DIR)/gallium: - -mkdir -p $(TOP)/$(LIB_DIR)/gallium - - -subdirs: - @for dir in $(SUBDIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ - fi \ - done - - -install: - @for dir in $(SUBDIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE) install) || exit 1 ; \ - fi \ - done - - -clean: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE) clean) ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - -rm -f common/*.o diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template index cccf8abdfd..9f92cb4207 100644 --- a/src/gallium/winsys/drm/Makefile.template +++ b/src/gallium/winsys/drm/Makefile.template @@ -79,25 +79,24 @@ SHARED_INCLUDES = \ ##### TARGETS ##### -default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME_EGL) - +default: depend symlinks $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME) $(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template - $(TOP)/bin/mklib -noprefix -o $@ \ + $(MKLIB) -noprefix -o $@ \ $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) $(DRIVER_EXTRAS) $(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS) - $(TOP)/bin/mklib -o $(LIBNAME_EGL) \ + $(MKLIB) -o $(LIBNAME_EGL) \ -linker "$(CC)" \ -noprefix \ $(OBJECTS) $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) $(PIPE_DRIVERS) $(WINOBJ) $(DRI_LIB_DEPS) \ --whole-archive $(LIBS) $(GALLIUM_AUXILIARIES) --no-whole-archive $(DRIVER_EXTRAS) -$(TOP)/$(LIB_DIR)/gallium/$(LIBNAME): $(LIBNAME) - $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR)/gallium +$(TOP)/$(LIB_DIR)/gallium: + mkdir -p $@ -$(TOP)/$(LIB_DIR)/gallium/$(LIBNAME_EGL): $(LIBNAME_EGL) - $(INSTALL) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/gallium +$(TOP)/$(LIB_DIR)/gallium/$(LIBNAME): $(LIBNAME) $(TOP)/$(LIB_DIR)/gallium + $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR)/gallium depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) rm -f depend @@ -118,8 +117,8 @@ clean: install: $(LIBNAME) - $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) - $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + $(INSTALL) -d $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR) + $(INSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR) include depend diff --git a/src/gallium/winsys/drm/intel/Makefile b/src/gallium/winsys/drm/intel/Makefile index eede9fc866..d8feef6824 100644 --- a/src/gallium/winsys/drm/intel/Makefile +++ b/src/gallium/winsys/drm/intel/Makefile @@ -1,25 +1,12 @@ +# src/gallium/winsys/drm/intel/Makefile TOP = ../../../../.. include $(TOP)/configs/current +SUBDIRS = gem $(GALLIUM_STATE_TRACKERS_DIRS) -SUBDIRS = gem egl - - -default: subdirs - - -subdirs: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - - -clean: - rm -f `find . -name \*.[oa]` - rm -f `find . -name depend` - - -# Dummy install target -install: diff --git a/src/gallium/winsys/drm/intel/common/Makefile b/src/gallium/winsys/drm/intel/common/Makefile deleted file mode 100644 index bf1a7d691f..0000000000 --- a/src/gallium/winsys/drm/intel/common/Makefile +++ /dev/null @@ -1,23 +0,0 @@ -TOP = ../../../../../.. -include $(TOP)/configs/current - -LIBNAME = inteldrm - -C_SOURCES = \ - intel_be_batchbuffer.c \ - intel_be_context.c \ - intel_be_device.c \ - ws_dri_bufmgr.c \ - ws_dri_drmpool.c \ - ws_dri_fencemgr.c \ - ws_dri_mallocpool.c \ - ws_dri_slabpool.c - - -include ./Makefile.template - -DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \ - && pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") -symlinks: - diff --git a/src/gallium/winsys/drm/intel/common/Makefile.template b/src/gallium/winsys/drm/intel/common/Makefile.template deleted file mode 100644 index 02ed363a43..0000000000 --- a/src/gallium/winsys/drm/intel/common/Makefile.template +++ /dev/null @@ -1,64 +0,0 @@ -# -*-makefile-*- - - -# We still have a dependency on the "dri" buffer manager. Most likely -# the interface can be reused in non-dri environments, and also as a -# frontend to simpler memory managers. -# -COMMON_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/include \ - $(DRIVER_INCLUDES) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks $(LIBNAME) - - -$(LIBNAME): $(OBJECTS) Makefile Makefile.template - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean:: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -include depend diff --git a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c deleted file mode 100644 index bc13a5761e..0000000000 --- a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c +++ /dev/null @@ -1,429 +0,0 @@ - -#include "intel_be_batchbuffer.h" -#include "intel_be_context.h" -#include "intel_be_device.h" -#include <errno.h> - -#include "xf86drm.h" - -static void -intel_realloc_relocs(struct intel_be_batchbuffer *batch, int num_relocs) -{ - unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; - - size *= sizeof(uint32_t); - batch->reloc = realloc(batch->reloc, size); - batch->reloc_size = num_relocs; -} - - -void -intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) -{ - /* - * Get a new, free batchbuffer. - */ - drmBO *bo; - struct drm_bo_info_req *req; - - driBOUnrefUserList(batch->list); - driBOResetList(batch->list); - - /* base.size is the size available to the i915simple driver */ - batch->base.size = batch->device->max_batch_size - BATCH_RESERVED; - batch->base.actual_size = batch->device->max_batch_size; - driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0); - - /* - * Add the batchbuffer to the validate list. - */ - - driBOAddListItem(batch->list, batch->buffer, - DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, - &batch->dest_location, &batch->node); - - req = &batch->node->bo_arg.d.req.bo_req; - - /* - * Set up information needed for us to make relocations - * relative to the underlying drm buffer objects. - */ - - driReadLockKernelBO(); - bo = driBOKernel(batch->buffer); - req->presumed_offset = (uint64_t) bo->offset; - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - batch->drmBOVirtual = (uint8_t *) bo->virtual; - driReadUnlockKernelBO(); - - /* - * Adjust the relocation buffer size. - */ - - if (batch->reloc_size > INTEL_MAX_RELOCS || - batch->reloc == NULL) - intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); - - assert(batch->reloc != NULL); - batch->reloc[0] = 0; /* No relocs yet. */ - batch->reloc[1] = 1; /* Reloc type 1 */ - batch->reloc[2] = 0; /* Only a single relocation list. */ - batch->reloc[3] = 0; /* Only a single relocation list. */ - - batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->poolOffset = driBOPoolOffset(batch->buffer); - batch->base.ptr = batch->base.map; - batch->dirty_state = ~0; - batch->nr_relocs = 0; - batch->flags = 0; - batch->id = 0;//batch->intel->intelScreen->batch_id++; -} - -/*====================================================================== - * Public functions - */ -struct intel_be_batchbuffer * -intel_be_batchbuffer_alloc(struct intel_be_context *intel) -{ - struct intel_be_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel = intel; - batch->device = intel->device; - - driGenBuffers(intel->device->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - batch->list = driBOCreateList(20); - batch->reloc = NULL; - intel_be_batchbuffer_reset(batch); - return batch; -} - -void -intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE, FALSE); - driFenceUnReference(&batch->last_fence); - } - if (batch->base.map) { - driBOUnmap(batch->buffer); - batch->base.map = NULL; - } - driBOUnReference(batch->buffer); - driBOFreeList(batch->list); - if (batch->reloc) - free(batch->reloc); - batch->buffer = NULL; - free(batch); -} - -void -intel_be_offset_relocation(struct intel_be_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask) -{ - int itemLoc; - struct _drmBONode *node; - uint32_t *reloc; - struct drm_bo_info_req *req; - - driBOAddListItem(batch->list, driBO, val_flags, val_mask, - &itemLoc, &node); - req = &node->bo_arg.d.req.bo_req; - - if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { - - /* - * Stop other threads from tampering with the underlying - * drmBO while we're reading its offset. - */ - - driReadLockKernelBO(); - req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; - driReadUnlockKernelBO(); - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - } - - pre_add += driBOPoolOffset(driBO); - - if (batch->nr_relocs == batch->reloc_size) - intel_realloc_relocs(batch, batch->reloc_size * 2); - - reloc = batch->reloc + - (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); - - reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual); - i915_batchbuffer_dword(&batch->base, req->presumed_offset + pre_add); - reloc[1] = pre_add; - reloc[2] = itemLoc; - reloc[3] = batch->dest_location; - batch->nr_relocs++; -} - -static void -i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) -{ - buf->handle = rep->handle; - buf->flags = rep->flags; - buf->size = rep->size; - buf->offset = rep->offset; - buf->mapHandle = rep->arg_handle; - buf->proposedFlags = rep->proposed_flags; - buf->start = rep->buffer_start; - buf->fenceFlags = rep->fence_flags; - buf->replyFlags = rep->rep_flags; - buf->pageAlignment = rep->page_alignment; -} - -static int -i915_execbuf(struct intel_be_batchbuffer *batch, - unsigned int used, - boolean ignore_cliprects, - drmBOList *list, - struct drm_i915_execbuffer *ea) -{ -// struct intel_be_context *intel = batch->intel; - drmBONode *node; - drmMMListHead *l; - struct drm_i915_op_arg *arg, *first; - struct drm_bo_op_req *req; - struct drm_bo_info_rep *rep; - uint64_t *prevNext = NULL; - drmBO *buf; - int ret = 0; - uint32_t count = 0; - - first = NULL; - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - - arg = &node->bo_arg; - req = &arg->d.req; - - if (!first) - first = arg; - - if (prevNext) - *prevNext = (unsigned long)arg; - - prevNext = &arg->next; - req->bo_req.handle = node->buf->handle; - req->op = drm_bo_validate; - req->bo_req.flags = node->arg0; - req->bo_req.mask = node->arg1; - req->bo_req.hint |= 0; - count++; - } - - memset(ea, 0, sizeof(*ea)); - ea->num_buffers = count; - ea->batch.start = batch->poolOffset; - ea->batch.used = used; -#if 0 /* ZZZ JB: no cliprects used */ - ea->batch.cliprects = intel->pClipRects; - ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | - (((GLuint) intel->drawY) << 16)); -#else - ea->batch.cliprects = NULL; - ea->batch.num_cliprects = 0; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0; -#endif - ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; - ea->ops_list = (unsigned long) first; - first->reloc_ptr = (unsigned long) batch->reloc; - batch->reloc[0] = batch->nr_relocs; - - //return -EFAULT; - do { - ret = drmCommandWriteRead(batch->device->fd, DRM_I915_EXECBUFFER, ea, - sizeof(*ea)); - } while (ret == -EAGAIN); - - if (ret != 0) - return ret; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - arg = &node->bo_arg; - rep = &arg->d.rep.bo_info; - - if (!arg->handled) { - return -EFAULT; - } - if (arg->d.rep.ret) - return arg->d.rep.ret; - - buf = node->buf; - i915_drm_copy_reply(rep, buf); - } - return 0; -} - -/* TODO: Push this whole function into bufmgr. - */ -static struct _DriFenceObject * -do_flush_locked(struct intel_be_batchbuffer *batch, - unsigned int used, - boolean ignore_cliprects, boolean allow_unlock) -{ - struct intel_be_context *intel = batch->intel; - struct _DriFenceObject *fo; - drmFence fence; - drmBOList *boList; - struct drm_i915_execbuffer ea; - int ret = 0; - - driBOValidateUserList(batch->list); - boList = driGetdrmBOList(batch->list); - -#if 0 /* ZZZ JB Allways run */ - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { -#else - if (1) { -#endif - ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); - } else { - driPutdrmBOList(batch->list); - fo = NULL; - goto out; - } - driPutdrmBOList(batch->list); - if (ret) - abort(); - - if (ea.fence_arg.error != 0) { - - /* - * The hardware has been idled by the kernel. - * Don't fence the driBOs. - */ - - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); -#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ - _mesa_printf("fence error\n"); -#endif - batch->last_fence = NULL; - fo = NULL; - goto out; - } - - fence.handle = ea.fence_arg.handle; - fence.fence_class = ea.fence_arg.fence_class; - fence.type = ea.fence_arg.type; - fence.flags = ea.fence_arg.flags; - fence.signaled = ea.fence_arg.signaled; - - fo = driBOFenceUserList(batch->device->fenceMgr, batch->list, - "SuperFence", &fence); - - if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); - /* - * FIXME: Context last fence?? - */ - batch->last_fence = fo; - driFenceReference(fo); - } - out: -#if 0 /* ZZZ JB: fix this */ - intel->vtbl.lost_hardware(intel); -#else - (void)intel; -#endif - return fo; -} - - -struct _DriFenceObject * -intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch) -{ - struct intel_be_context *intel = batch->intel; - unsigned int used = batch->base.ptr - batch->base.map; - boolean was_locked = batch->intel->hardware_locked(intel); - struct _DriFenceObject *fence; - - if (used == 0) { - driFenceReference(batch->last_fence); - return batch->last_fence; - } - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ -#if 0 /* ZZZ JB: what should we do here? */ - if (used & 4) { - ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->base.ptr)[1] = 0; - ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } -#else - if (used & 4) { - ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->base.ptr)[1] = 0; - ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 8; - } -#endif - driBOUnmap(batch->buffer); - batch->base.ptr = NULL; - batch->base.map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ - if (!was_locked) - intel->hardware_lock(intel); - - fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - FALSE); - - if (!was_locked) - intel->hardware_unlock(intel); - - /* Reset the buffer: - */ - intel_be_batchbuffer_reset(batch); - return fence; -} - -void -intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_be_batchbuffer_flush(batch); - driFenceFinish(fence, driFenceType(fence), FALSE); - driFenceUnReference(&fence); -} - -#if 0 -void -intel_be_batchbuffer_data(struct intel_be_batchbuffer *batch, - const void *data, unsigned int bytes, unsigned int flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->base.ptr, data, bytes); - batch->base.ptr += bytes; -} -#endif diff --git a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h deleted file mode 100644 index f150e3a674..0000000000 --- a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h +++ /dev/null @@ -1,69 +0,0 @@ - -#ifndef INTEL_BE_BATCHBUFFER_H -#define INTEL_BE_BATCHBUFFER_H - -#include "i915simple/i915_batch.h" - -#include "ws_dri_bufmgr.h" - -#define BATCH_RESERVED 16 - -#define INTEL_DEFAULT_RELOCS 100 -#define INTEL_MAX_RELOCS 400 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct intel_be_context; -struct intel_be_device; - -struct intel_be_batchbuffer -{ - struct i915_batchbuffer base; - - struct intel_be_context *intel; - struct intel_be_device *device; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - uint32_t flags; - - struct _DriBufferList *list; - size_t list_count; - - uint32_t *reloc; - size_t reloc_size; - size_t nr_relocs; - - uint32_t dirty_state; - uint32_t id; - - uint32_t poolOffset; - uint8_t *drmBOVirtual; - struct _drmBONode *node; /* Validation list node for this buffer */ - int dest_location; /* Validation list sequence for this buffer */ -}; - -struct intel_be_batchbuffer * -intel_be_batchbuffer_alloc(struct intel_be_context *intel); - -void -intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch); - -void -intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch); - -struct _DriFenceObject * -intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch); - -void -intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch); - -void -intel_be_offset_relocation(struct intel_be_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask); - -#endif diff --git a/src/gallium/winsys/drm/intel/common/intel_be_context.c b/src/gallium/winsys/drm/intel/common/intel_be_context.c deleted file mode 100644 index 1af39674f4..0000000000 --- a/src/gallium/winsys/drm/intel/common/intel_be_context.c +++ /dev/null @@ -1,107 +0,0 @@ - -/* - * Authors: Jakob Bornecrantz <jakob-at-tungstengraphics.com> - */ - -#include "ws_dri_fencemgr.h" -#include "intel_be_device.h" -#include "intel_be_context.h" -#include "intel_be_batchbuffer.h" - -static INLINE struct intel_be_context * -intel_be_context(struct i915_winsys *sws) -{ - return (struct intel_be_context *)sws; -} - -/* Simple batchbuffer interface: - */ - -static struct i915_batchbuffer* -intel_i915_batch_get(struct i915_winsys *sws) -{ - struct intel_be_context *intel = intel_be_context(sws); - return &intel->batch->base; -} - -static void intel_i915_batch_reloc(struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta) -{ - struct intel_be_context *intel = intel_be_context(sws); - - unsigned flags = DRM_BO_FLAG_MEM_TT; - unsigned mask = DRM_BO_MASK_MEM; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - flags |= DRM_BO_FLAG_WRITE; - mask |= DRM_BO_FLAG_WRITE; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - flags |= DRM_BO_FLAG_READ; - mask |= DRM_BO_FLAG_READ; - } - - intel_be_offset_relocation(intel->batch, - delta, - dri_bo(buf), - flags, - mask); -} - -static void intel_i915_batch_flush(struct i915_winsys *sws, - struct pipe_fence_handle **fence) -{ - struct intel_be_context *intel = intel_be_context(sws); - - union { - struct _DriFenceObject *dri; - struct pipe_fence_handle *pipe; - } fu; - - if (fence) - assert(!*fence); - - fu.dri = intel_be_batchbuffer_flush(intel->batch); - - if (!fu.dri) { - assert(0); - *fence = NULL; - return; - } - - if (fu.dri) { - if (fence) - *fence = fu.pipe; - else - driFenceUnReference(&fu.dri); - } - -} - -boolean -intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device) -{ - assert(intel); - assert(device); - - intel->device = device; - - /* TODO move framebuffer createion to the driver */ - - intel->base.batch_get = intel_i915_batch_get; - intel->base.batch_reloc = intel_i915_batch_reloc; - intel->base.batch_flush = intel_i915_batch_flush; - - intel->batch = intel_be_batchbuffer_alloc(intel); - - return true; -} - -void -intel_be_destroy_context(struct intel_be_context *intel) -{ - intel_be_batchbuffer_free(intel->batch); -} diff --git a/src/gallium/winsys/drm/intel/common/intel_be_context.h b/src/gallium/winsys/drm/intel/common/intel_be_context.h deleted file mode 100644 index d5cbc93594..0000000000 --- a/src/gallium/winsys/drm/intel/common/intel_be_context.h +++ /dev/null @@ -1,40 +0,0 @@ -/* These need to be diffrent from the intel winsys */ -#ifndef INTEL_BE_CONTEXT_H -#define INTEL_BE_CONTEXT_H - -#include "i915simple/i915_winsys.h" - -struct intel_be_context -{ - /** Interface to i915simple driver */ - struct i915_winsys base; - - struct intel_be_device *device; - struct intel_be_batchbuffer *batch; - - /* - * Hardware lock functions. - * - * Needs to be filled in by the winsys. - */ - void (*hardware_lock)(struct intel_be_context *context); - void (*hardware_unlock)(struct intel_be_context *context); - boolean (*hardware_locked)(struct intel_be_context *context); -}; - -/** - * Intialize a allocated intel_be_context struct. - * - * Remember to set the hardware_* functions. - */ -boolean -intel_be_init_context(struct intel_be_context *intel, - struct intel_be_device *device); - -/** - * Destroy a intel_be_context. - * Does not free the struct that is up to the winsys. - */ -void -intel_be_destroy_context(struct intel_be_context *intel); -#endif diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.c b/src/gallium/winsys/drm/intel/common/intel_be_device.c deleted file mode 100644 index 85ab1a2684..0000000000 --- a/src/gallium/winsys/drm/intel/common/intel_be_device.c +++ /dev/null @@ -1,296 +0,0 @@ - - -/* - * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com> - * Jakob Bornecrantz <jakob-at-tungstengraphics-dot-com> - */ - -#include "intel_be_device.h" -#include "ws_dri_bufmgr.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" - -#include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" -#include "pipe/p_inlines.h" -#include "util/u_memory.h" - -#include "i915simple/i915_screen.h" - -/* Turn a pipe winsys into an intel/pipe winsys: - */ -static INLINE struct intel_be_device * -intel_be_device( struct pipe_winsys *winsys ) -{ - return (struct intel_be_device *)winsys; -} - - -/* - * Buffer functions. - * - * Most callbacks map direcly onto dri_bufmgr operations: - */ - -static void *intel_be_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags ) -{ - unsigned drm_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - drm_flags |= DRM_BO_FLAG_WRITE; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - drm_flags |= DRM_BO_FLAG_READ; - - return driBOMap( dri_bo(buf), drm_flags, 0 ); -} - -static void intel_be_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - -static void -intel_be_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnReference( dri_bo(buf) ); - FREE(buf); -} - -static struct pipe_buffer * -intel_be_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size ) -{ - struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); - struct intel_be_device *iws = intel_be_device(winsys); - unsigned flags = 0; - struct _DriBufferPool *pool; - - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - - if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { - flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; - pool = iws->mallocPool; - } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { - /* For vertex buffers */ - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - pool = iws->vertexPool; - } else { - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - pool = iws->regionPool; - } - - if (usage & PIPE_BUFFER_USAGE_GPU_READ) - flags |= DRM_BO_FLAG_READ; - - if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) - flags |= DRM_BO_FLAG_WRITE; - - /* drm complains if we don't set any read/write flags. - */ - if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) - flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - - buffer->pool = pool; - driGenBuffers( buffer->pool, - "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - - driBOData( buffer->driBO, size, NULL, buffer->pool, 0 ); - - return &buffer->base; -} - - -static struct pipe_buffer * -intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); - struct intel_be_device *iws = intel_be_device(winsys); - - driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes ); - - buffer->base.refcount = 1; - - return &buffer->base; -} - -struct pipe_buffer * -intel_be_buffer_from_handle(struct intel_be_device *device, - const char* name, unsigned handle) -{ - struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf)); - struct pipe_buffer *buffer; - - if (!be_buf) - goto err; - - memset(be_buf, 0, sizeof(*be_buf)); - - driGenBuffers(device->staticPool, name, 1, &be_buf->driBO, 0, 0, 0); - driBOSetReferenced(be_buf->driBO, handle); - - if (0) /** XXX TODO check error */ - goto err_bo; - - buffer = &be_buf->base; - buffer->refcount = 1; - buffer->alignment = 0; - buffer->usage = 0; - buffer->size = driBOSize(be_buf->driBO); - - return buffer; -err_bo: - free(be_buf); -err: - return NULL; -} - - -static struct pipe_buffer * -intel_i915_surface_buffer_create(struct pipe_winsys *winsys, - unsigned width, unsigned height, - enum pipe_format format, - unsigned usage, - unsigned *stride) -{ - const unsigned alignment = 64; - struct pipe_format_block block; - unsigned nblocksx, nblocksy; - - pf_get_block(format, &block); - nblocksx = pf_get_nblocksx(&block, width); - nblocksy = pf_get_nblocksy(&block, height); - *stride = round_up(nblocksx * block.size, alignment); - - return winsys->buffer_create(winsys, alignment, - usage, - *stride * nblocksy); -} - - -/* - * Fence functions - */ - -static void -intel_be_fence_reference( struct pipe_winsys *sws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *fence ) -{ - if (*ptr) - driFenceUnReference((struct _DriFenceObject **)ptr); - - if (fence) - *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); -} - -static int -intel_be_fence_signalled( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceSignaled((struct _DriFenceObject *)fence, flag); -} - -static int -intel_be_fence_finish( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); -} - - -/* - * Misc functions - */ - -boolean -intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) -{ - dev->fd = fd; - dev->max_batch_size = 16 * 4096; - dev->max_vertex_size = 128 * 4096; - - dev->base.buffer_create = intel_be_buffer_create; - dev->base.user_buffer_create = intel_be_user_buffer_create; - dev->base.buffer_map = intel_be_buffer_map; - dev->base.buffer_unmap = intel_be_buffer_unmap; - dev->base.buffer_destroy = intel_be_buffer_destroy; - dev->base.surface_buffer_create = intel_i915_surface_buffer_create; - dev->base.fence_reference = intel_be_fence_reference; - dev->base.fence_signalled = intel_be_fence_signalled; - dev->base.fence_finish = intel_be_fence_finish; - -#if 0 /* Set by the winsys */ - dev->base.flush_frontbuffer = intel_flush_frontbuffer; - dev->base.get_name = intel_get_name; -#endif - - dev->fMan = driInitFreeSlabManager(10, 10); - dev->fenceMgr = driFenceMgrTTMInit(dev->fd); - - dev->mallocPool = driMallocPoolInit(); - dev->staticPool = driDRMPoolInit(dev->fd); - /* Sizes: 64 128 256 512 1024 2048 4096 8192 16384 32768 */ - dev->regionPool = driSlabPoolInit(dev->fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - 64, - 10, 120, 4096 * 64, 0, - dev->fMan); - - dev->vertexPool = driSlabPoolInit(dev->fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - dev->max_vertex_size, - 1, 120, dev->max_vertex_size * 4, 0, - dev->fMan); - - dev->batchPool = driSlabPoolInit(dev->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - dev->max_batch_size, - 1, 40, dev->max_batch_size * 16, 0, - dev->fMan); - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - dev->screen = i915_create_screen(&dev->base, id); - - return true; -} - -void -intel_be_destroy_device(struct intel_be_device *dev) -{ - driPoolTakeDown(dev->mallocPool); - driPoolTakeDown(dev->staticPool); - driPoolTakeDown(dev->regionPool); - driPoolTakeDown(dev->vertexPool); - driPoolTakeDown(dev->batchPool); - - /** TODO takedown fenceMgr and fMan */ -} diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.h b/src/gallium/winsys/drm/intel/common/intel_be_device.h deleted file mode 100644 index 534d638b6a..0000000000 --- a/src/gallium/winsys/drm/intel/common/intel_be_device.h +++ /dev/null @@ -1,72 +0,0 @@ -#ifndef INTEL_DRM_DEVICE_H -#define INTEL_DRM_DEVICE_H - -#include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_context.h" - -/* - * Device - */ - -struct intel_be_device -{ - struct pipe_winsys base; - - /** - * Hw level screen - */ - struct pipe_screen *screen; - - int fd; /**< Drm file discriptor */ - - size_t max_batch_size; - size_t max_vertex_size; - - struct _DriFenceMgr *fenceMgr; - - struct _DriBufferPool *batchPool; - struct _DriBufferPool *regionPool; - struct _DriBufferPool *mallocPool; - struct _DriBufferPool *vertexPool; - struct _DriBufferPool *staticPool; - struct _DriFreeSlabManager *fMan; -}; - -boolean -intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); - -void -intel_be_destroy_device(struct intel_be_device *dev); - -/* - * Buffer - */ - -struct intel_be_buffer { - struct pipe_buffer base; - struct _DriBufferPool *pool; - struct _DriBufferObject *driBO; -}; - -/** - * Create a be buffer from a drm bo handle - * - * Takes a reference - */ -struct pipe_buffer * -intel_be_buffer_from_handle(struct intel_be_device *device, - const char* name, unsigned handle); - -static INLINE struct intel_be_buffer * -intel_be_buffer(struct pipe_buffer *buf) -{ - return (struct intel_be_buffer *)buf; -} - -static INLINE struct _DriBufferObject * -dri_bo(struct pipe_buffer *buf) -{ - return intel_be_buffer(buf)->driBO; -} - -#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c deleted file mode 100644 index 517a97b3ee..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c +++ /dev/null @@ -1,949 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com> - * Keith Whitwell <keithw-at-tungstengraphics-dot-com> - */ - -#include <xf86drm.h> -#include <stdlib.h> -#include <stdio.h> -#include "pipe/p_thread.h" -#include "errno.h" -#include "ws_dri_bufmgr.h" -#include "string.h" -#include "pipe/p_debug.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" - - -/* - * This lock is here to protect drmBO structs changing underneath us during a - * validate list call, since validatelist cannot take individiual locks for - * each drmBO. Validatelist takes this lock in write mode. Any access to an - * individual drmBO should take this lock in read mode, since in that case, the - * driBufferObject mutex will protect the access. Locking order is - * driBufferObject mutex - > this rw lock. - */ - -pipe_static_mutex(bmMutex); -pipe_static_condvar(bmCond); - -static int kernelReaders = 0; -static int num_buffers = 0; -static int num_user_buffers = 0; - -static drmBO *drmBOListBuf(void *iterator) -{ - drmBONode *node; - drmMMListHead *l = (drmMMListHead *) iterator; - node = DRMLISTENTRY(drmBONode, l, head); - return node->buf; -} - -static void *drmBOListIterator(drmBOList *list) -{ - void *ret = list->list.next; - - if (ret == &list->list) - return NULL; - return ret; -} - -static void *drmBOListNext(drmBOList *list, void *iterator) -{ - void *ret; - - drmMMListHead *l = (drmMMListHead *) iterator; - ret = l->next; - if (ret == &list->list) - return NULL; - return ret; -} - -static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, - uint64_t arg0, - uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } - else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADD(&node->head, &list->list); - list->numOnList++; - return node; -} - -static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, - uint64_t mask, int *newItem) -{ - drmBONode *node, *cur; - drmMMListHead *l; - - *newItem = 0; - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - } - if (!cur) { - cur = drmAddListItem(list, buf, flags, mask); - if (!cur) { - return -ENOMEM; - } - *newItem = 1; - cur->arg0 = flags; - cur->arg1 = mask; - } - else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - return 0; -} - -static void drmBOFreeList(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->list.next; - while(l != &list->list) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->list.next; - list->numCurrent--; - list->numOnList--; - } - - l = list->free.next; - while(l != &list->free) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->free.next; - list->numCurrent--; - } -} - -static int drmAdjustListNodes(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - int ret = 0; - - while(list->numCurrent < list->numTarget) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - ret = -ENOMEM; - break; - } - list->numCurrent++; - DRMLISTADD(&node->head, &list->free); - } - - while(list->numCurrent > list->numTarget) { - l = list->free.next; - if (l == &list->free) - break; - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - list->numCurrent--; - } - return ret; -} - -static int drmBOCreateList(int numTarget, drmBOList *list) -{ - DRMINITLISTHEAD(&list->list); - DRMINITLISTHEAD(&list->free); - list->numTarget = numTarget; - list->numCurrent = 0; - list->numOnList = 0; - return drmAdjustListNodes(list); -} - -static int drmBOResetList(drmBOList *list) -{ - drmMMListHead *l; - int ret; - - ret = drmAdjustListNodes(list); - if (ret) - return ret; - - l = list->list.next; - while (l != &list->list) { - DRMLISTDEL(l); - DRMLISTADD(l, &list->free); - list->numOnList--; - l = list->list.next; - } - return drmAdjustListNodes(list); -} - -void driWriteLockKernelBO(void) -{ - pipe_mutex_lock(bmMutex); - while(kernelReaders != 0) - pipe_condvar_wait(bmCond, bmMutex); -} - -void driWriteUnlockKernelBO(void) -{ - pipe_mutex_unlock(bmMutex); -} - -void driReadLockKernelBO(void) -{ - pipe_mutex_lock(bmMutex); - kernelReaders++; - pipe_mutex_unlock(bmMutex); -} - -void driReadUnlockKernelBO(void) -{ - pipe_mutex_lock(bmMutex); - if (--kernelReaders == 0) - pipe_condvar_broadcast(bmCond); - pipe_mutex_unlock(bmMutex); -} - - - - -/* - * TODO: Introduce fence pools in the same way as - * buffer object pools. - */ - -typedef struct _DriBufferObject -{ - DriBufferPool *pool; - pipe_mutex mutex; - int refCount; - const char *name; - uint64_t flags; - unsigned hint; - unsigned alignment; - unsigned createdByReference; - void *private; - /* user-space buffer: */ - unsigned userBuffer; - void *userData; - unsigned userSize; -} DriBufferObject; - -typedef struct _DriBufferList { - drmBOList drmBuffers; /* List of kernel buffers needing validation */ - drmBOList driBuffers; /* List of user-space buffers needing validation */ -} DriBufferList; - - -void -bmError(int val, const char *file, const char *function, int line) -{ - printf("Fatal video memory manager error \"%s\".\n" - "Check kernel logs or set the LIBGL_DEBUG\n" - "environment variable to \"verbose\" for more info.\n" - "Detected in file %s, line %d, function %s.\n", - strerror(-val), file, line, function); -#ifndef NDEBUG - abort(); -#else - abort(); -#endif -} - -extern drmBO * -driBOKernel(struct _DriBufferObject *buf) -{ - drmBO *ret; - - driReadLockKernelBO(); - pipe_mutex_lock(buf->mutex); - assert(buf->private != NULL); - ret = buf->pool->kernel(buf->pool, buf->private); - if (!ret) - BM_CKFATAL(-EINVAL); - pipe_mutex_unlock(buf->mutex); - driReadUnlockKernelBO(); - - return ret; -} - -void -driBOWaitIdle(struct _DriBufferObject *buf, int lazy) -{ - - /* - * This function may block. Is it sane to keep the mutex held during - * that time?? - */ - - pipe_mutex_lock(buf->mutex); - BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); - pipe_mutex_unlock(buf->mutex); -} - -void * -driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) -{ - void *virtual; - int retval; - - if (buf->userBuffer) { - return buf->userData; - } - - pipe_mutex_lock(buf->mutex); - assert(buf->private != NULL); - retval = buf->pool->map(buf->pool, buf->private, flags, hint, - &buf->mutex, &virtual); - pipe_mutex_unlock(buf->mutex); - - return retval == 0 ? virtual : NULL; -} - -void -driBOUnmap(struct _DriBufferObject *buf) -{ - if (buf->userBuffer) - return; - - assert(buf->private != NULL); - pipe_mutex_lock(buf->mutex); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - pipe_mutex_unlock(buf->mutex); -} - -unsigned long -driBOOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - pipe_mutex_lock(buf->mutex); - ret = buf->pool->offset(buf->pool, buf->private); - pipe_mutex_unlock(buf->mutex); - return ret; -} - -unsigned long -driBOPoolOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - pipe_mutex_lock(buf->mutex); - ret = buf->pool->poolOffset(buf->pool, buf->private); - pipe_mutex_unlock(buf->mutex); - return ret; -} - -uint64_t -driBOFlags(struct _DriBufferObject *buf) -{ - uint64_t ret; - - assert(buf->private != NULL); - - driReadLockKernelBO(); - pipe_mutex_lock(buf->mutex); - ret = buf->pool->flags(buf->pool, buf->private); - pipe_mutex_unlock(buf->mutex); - driReadUnlockKernelBO(); - return ret; -} - -struct _DriBufferObject * -driBOReference(struct _DriBufferObject *buf) -{ - pipe_mutex_lock(buf->mutex); - if (++buf->refCount == 1) { - pipe_mutex_unlock(buf->mutex); - BM_CKFATAL(-EINVAL); - } - pipe_mutex_unlock(buf->mutex); - return buf; -} - -void -driBOUnReference(struct _DriBufferObject *buf) -{ - int tmp; - - if (!buf) - return; - - pipe_mutex_lock(buf->mutex); - tmp = --buf->refCount; - if (!tmp) { - pipe_mutex_unlock(buf->mutex); - if (buf->private) { - if (buf->createdByReference) - buf->pool->unreference(buf->pool, buf->private); - else - buf->pool->destroy(buf->pool, buf->private); - } - if (buf->userBuffer) - num_user_buffers--; - else - num_buffers--; - free(buf); - } else - pipe_mutex_unlock(buf->mutex); - -} - - -int -driBOData(struct _DriBufferObject *buf, - unsigned size, const void *data, - DriBufferPool *newPool, - uint64_t flags) -{ - void *virtual = NULL; - int newBuffer; - int retval = 0; - struct _DriBufferPool *pool; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - pipe_mutex_lock(buf->mutex); - pool = buf->pool; - - if (pool == NULL && newPool != NULL) { - buf->pool = newPool; - pool = newPool; - } - if (newPool == NULL) - newPool = pool; - - if (!pool->create) { - assert((size_t)"driBOData called on invalid buffer\n" & 0); - BM_CKFATAL(-EINVAL); - } - - newBuffer = (!buf->private || pool != newPool || - pool->size(pool, buf->private) < size); - - if (!flags) - flags = buf->flags; - - if (newBuffer) { - - if (buf->createdByReference) { - assert((size_t)"driBOData requiring resizing called on shared buffer.\n" & 0); - BM_CKFATAL(-EINVAL); - } - - if (buf->private) - buf->pool->destroy(buf->pool, buf->private); - - pool = newPool; - buf->pool = newPool; - buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (!buf->private) - retval = -ENOMEM; - - if (retval == 0) - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); - } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { - /* - * Buffer is busy. need to create a new one. - */ - - void *newBuf; - - newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (newBuf) { - buf->pool->destroy(buf->pool, buf->private); - buf->private = newBuf; - } - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } else { - uint64_t flag_diff = flags ^ buf->flags; - - /* - * We might need to change buffer flags. - */ - - if (flag_diff){ - assert(pool->setStatus != NULL); - BM_CKFATAL(pool->unmap(pool, buf->private)); - BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, - buf->flags)); - if (!data) - goto out; - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } - } - - if (retval == 0) { - if (data) - memcpy(virtual, data, size); - - BM_CKFATAL(pool->unmap(pool, buf->private)); - } - - out: - pipe_mutex_unlock(buf->mutex); - - return retval; -} - -void -driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, const void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - pipe_mutex_lock(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, - &virtual)); - memcpy((unsigned char *) virtual + offset, data, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - pipe_mutex_unlock(buf->mutex); -} - -void -driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - pipe_mutex_lock(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); - memcpy(data, (unsigned char *) virtual + offset, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - pipe_mutex_unlock(buf->mutex); -} - -void -driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle) -{ - pipe_mutex_lock(buf->mutex); - if (buf->private != NULL) { - assert((size_t)"Invalid buffer for setReferenced\n" & 0); - BM_CKFATAL(-EINVAL); - - } - if (buf->pool->reference == NULL) { - assert((size_t)"Invalid buffer pool for setReferenced\n" & 0); - BM_CKFATAL(-EINVAL); - } - buf->private = buf->pool->reference(buf->pool, handle); - if (!buf->private) { - assert((size_t)"Invalid buffer pool for setStatic\n" & 0); - BM_CKFATAL(-ENOMEM); - } - buf->createdByReference = TRUE; - buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; - pipe_mutex_unlock(buf->mutex); -} - -int -driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint) -{ - struct _DriBufferObject *buf; - int i; - - flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - - ++num_buffers; - - assert(pool); - - for (i = 0; i < n; ++i) { - buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); - if (!buf) - return -ENOMEM; - - pipe_mutex_init(buf->mutex); - pipe_mutex_lock(buf->mutex); - buf->refCount = 1; - buf->flags = flags; - buf->hint = hint; - buf->name = name; - buf->alignment = alignment; - buf->pool = pool; - buf->createdByReference = 0; - pipe_mutex_unlock(buf->mutex); - buffers[i] = buf; - } - return 0; -} - -void -driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject **buffers, - void *ptr, unsigned bytes) -{ - const unsigned alignment = 1, flags = 0, hint = 0; - - --num_buffers; /* JB: is inced in GenBuffes */ - driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); - ++num_user_buffers; - - (*buffers)->userBuffer = 1; - (*buffers)->userData = ptr; - (*buffers)->userSize = bytes; -} - -void -driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) -{ - int i; - - for (i = 0; i < n; ++i) { - driBOUnReference(buffers[i]); - } -} - - -void -driInitBufMgr(int fd) -{ - ; -} - -/* - * Note that lists are per-context and don't need mutex protection. - */ - -struct _DriBufferList * -driBOCreateList(int target) -{ - struct _DriBufferList *list = calloc(sizeof(*list), 1); - - BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); - BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); - return list; -} - -int -driBOResetList(struct _DriBufferList * list) -{ - int ret; - ret = drmBOResetList(&list->drmBuffers); - if (ret) - return ret; - ret = drmBOResetList(&list->driBuffers); - return ret; -} - -void -driBOFreeList(struct _DriBufferList * list) -{ - drmBOFreeList(&list->drmBuffers); - drmBOFreeList(&list->driBuffers); - free(list); -} - - -/* - * Copied from libdrm, because it is needed by driAddValidateItem. - */ - -static drmBONode * -driAddListItem(drmBOList * list, drmBO * item, - uint64_t arg0, uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - memset(&node->bo_arg, 0, sizeof(node->bo_arg)); - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADDTAIL(&node->head, &list->list); - list->numOnList++; - return node; -} - -/* - * Slightly modified version compared to the libdrm version. - * This one returns the list index of the buffer put on the list. - */ - -static int -driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, - uint64_t mask, int *itemLoc, - struct _drmBONode **pnode) -{ - drmBONode *node, *cur; - drmMMListHead *l; - int count = 0; - - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - count++; - } - if (!cur) { - cur = driAddListItem(list, buf, flags, mask); - if (!cur) - return -ENOMEM; - - cur->arg0 = flags; - cur->arg1 = mask; - } else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - *itemLoc = count; - *pnode = cur; - return 0; -} - - -void -driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node) -{ - int newItem; - - pipe_mutex_lock(buf->mutex); - BM_CKFATAL(driAddValidateItem(&list->drmBuffers, - buf->pool->kernel(buf->pool, buf->private), - flags, mask, itemLoc, node)); - BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, - flags, mask, &newItem)); - if (newItem) - buf->refCount++; - - pipe_mutex_unlock(buf->mutex); -} - -drmBOList *driGetdrmBOList(struct _DriBufferList *list) -{ - driWriteLockKernelBO(); - return &list->drmBuffers; -} - -void driPutdrmBOList(struct _DriBufferList *list) -{ - driWriteUnlockKernelBO(); -} - - -void -driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) -{ - pipe_mutex_lock(buf->mutex); - if (buf->pool->fence) - BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); - pipe_mutex_unlock(buf->mutex); - -} - -void -driBOUnrefUserList(struct _DriBufferList *list) -{ - struct _DriBufferObject *buf; - void *curBuf; - - curBuf = drmBOListIterator(&list->driBuffers); - while (curBuf) { - buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - -struct _DriFenceObject * -driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, const char *name, - drmFence *kFence) -{ - struct _DriFenceObject *fence; - struct _DriBufferObject *buf; - void *curBuf; - - fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, - kFence, sizeof(*kFence)); - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space fencing callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - driBOFence(buf, fence); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } - - driBOResetList(list); - return fence; -} - -void -driBOValidateUserList(struct _DriBufferList * list) -{ - void *curBuf; - struct _DriBufferObject *buf; - - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space validation callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - pipe_mutex_lock(buf->mutex); - if (buf->pool->validate) - BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); - pipe_mutex_unlock(buf->mutex); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - - -void -driPoolTakeDown(struct _DriBufferPool *pool) -{ - pool->takeDown(pool); - -} - -unsigned long -driBOSize(struct _DriBufferObject *buf) -{ - unsigned long size; - - pipe_mutex_lock(buf->mutex); - size = buf->pool->size(buf->pool, buf->private); - pipe_mutex_unlock(buf->mutex); - - return size; - -} - -drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) -{ - return &list->drmBuffers; -} - -drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) -{ - return &list->driBuffers; -} - diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h deleted file mode 100644 index e6c0cff0a0..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h +++ /dev/null @@ -1,138 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com> - * Keith Whitwell <keithw-at-tungstengraphics-dot-com> - */ - -#ifndef _PSB_BUFMGR_H_ -#define _PSB_BUFMGR_H_ -#include <xf86mm.h> -#include "i915_drm.h" -#include "ws_dri_fencemgr.h" - -typedef struct _drmBONode -{ - drmMMListHead head; - drmBO *buf; - struct drm_i915_op_arg bo_arg; - uint64_t arg0; - uint64_t arg1; -} drmBONode; - -typedef struct _drmBOList { - unsigned numTarget; - unsigned numCurrent; - unsigned numOnList; - drmMMListHead list; - drmMMListHead free; -} drmBOList; - - -struct _DriFenceObject; -struct _DriBufferObject; -struct _DriBufferPool; -struct _DriBufferList; - -/* - * Return a pointer to the libdrm buffer object this DriBufferObject - * uses. - */ - -extern drmBO *driBOKernel(struct _DriBufferObject *buf); -extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, - unsigned hint); -extern void driBOUnmap(struct _DriBufferObject *buf); -extern unsigned long driBOOffset(struct _DriBufferObject *buf); -extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); - -extern uint64_t driBOFlags(struct _DriBufferObject *buf); -extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); -extern void driBOUnReference(struct _DriBufferObject *buf); - -extern int driBOData(struct _DriBufferObject *r_buf, - unsigned size, const void *data, - struct _DriBufferPool *pool, uint64_t flags); - -extern void driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - const void *data); -extern void driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - void *data); -extern int driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint); -extern void driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject *buffers[], - void *ptr, unsigned bytes); -extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); -extern void driInitBufMgr(int fd); -extern struct _DriBufferList *driBOCreateList(int target); -extern int driBOResetList(struct _DriBufferList * list); -extern void driBOAddListItem(struct _DriBufferList * list, - struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node); - -extern void driBOValidateList(int fd, struct _DriBufferList * list); -extern void driBOFreeList(struct _DriBufferList * list); -extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, - const char *name, - drmFence *kFence); -extern void driBOUnrefUserList(struct _DriBufferList *list); -extern void driBOValidateUserList(struct _DriBufferList * list); -extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); -extern void driPutdrmBOList(struct _DriBufferList *list); - -extern void driBOFence(struct _DriBufferObject *buf, - struct _DriFenceObject *fence); - -extern void driPoolTakeDown(struct _DriBufferPool *pool); -extern void driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle); -unsigned long driBOSize(struct _DriBufferObject *buf); -extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); -extern void driPoolTakeDown(struct _DriBufferPool *pool); - -extern void driReadLockKernelBO(void); -extern void driReadUnlockKernelBO(void); -extern void driWriteLockKernelBO(void); -extern void driWriteUnlockKernelBO(void); - -/* - * For debugging purposes. - */ - -extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); -extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); -#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h deleted file mode 100644 index ad3b6f3931..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com> - */ - -#ifndef _PSB_BUFPOOL_H_ -#define _PSB_BUFPOOL_H_ - -#include <xf86drm.h> -#include "pipe/p_thread.h" -struct _DriFenceObject; - -typedef struct _DriBufferPool -{ - int fd; - int (*map) (struct _DriBufferPool * pool, void *private, - unsigned flags, int hint, pipe_mutex *mutex, - void **virtual); - int (*unmap) (struct _DriBufferPool * pool, void *private); - int (*destroy) (struct _DriBufferPool * pool, void *private); - unsigned long (*offset) (struct _DriBufferPool * pool, void *private); - unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); - uint64_t (*flags) (struct _DriBufferPool * pool, void *private); - unsigned long (*size) (struct _DriBufferPool * pool, void *private); - void *(*create) (struct _DriBufferPool * pool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment); - void *(*reference) (struct _DriBufferPool * pool, unsigned handle); - int (*unreference) (struct _DriBufferPool * pool, void *private); - int (*fence) (struct _DriBufferPool * pool, void *private, - struct _DriFenceObject * fence); - drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); - int (*validate) (struct _DriBufferPool * pool, void *private, pipe_mutex *mutex); - int (*waitIdle) (struct _DriBufferPool *pool, void *private, pipe_mutex *mutex, - int lazy); - int (*setStatus) (struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags); - void (*takeDown) (struct _DriBufferPool * pool); - void *data; -} DriBufferPool; - -extern void bmError(int val, const char *file, const char *function, - int line); -#define BM_CKFATAL(val) \ - do{ \ - int tstVal = (val); \ - if (tstVal) \ - bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ - } while(0); - - -/* - * Builtin pools. - */ - -/* - * Kernel buffer objects. Size in multiples of page size. Page size aligned. - */ - -extern struct _DriBufferPool *driDRMPoolInit(int fd); -extern struct _DriBufferPool *driMallocPoolInit(void); - -struct _DriFreeSlabManager; -extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan); -extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); -extern struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); - - -#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c deleted file mode 100644 index 54618b1c82..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c +++ /dev/null @@ -1,268 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com> - */ - -#include <xf86drm.h> -#include <stdlib.h> -#include <unistd.h> -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" -#include "assert.h" - -/* - * Buffer pool implementation using DRM buffer objects as DRI buffer objects. - */ - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - unsigned pageSize = getpagesize(); - - if (!buf) - return NULL; - - if ((alignment > pageSize) && (alignment % pageSize)) { - free(buf); - return NULL; - } - - ret = drmBOCreate(pool->fd, size, alignment / pageSize, - NULL, - flags, hint, buf); - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static void * -pool_reference(struct _DriBufferPool *pool, unsigned handle) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - - if (!buf) - return NULL; - - ret = drmBOReference(pool->fd, handle, buf); - - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unreference(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, pipe_mutex *mutex, void **virtual) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOMap(pool->fd, buf, flags, hint, virtual); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOUnmap(pool->fd, buf); - driReadUnlockKernelBO(); - - return ret; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long offset; - - driReadLockKernelBO(); - assert(buf->flags & DRM_BO_FLAG_NO_MOVE); - offset = buf->offset; - driReadUnlockKernelBO(); - - return buf->offset; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - uint64_t flags; - - driReadLockKernelBO(); - flags = buf->flags; - driReadUnlockKernelBO(); - - return flags; -} - - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long size; - - driReadLockKernelBO(); - size = buf->size; - driReadUnlockKernelBO(); - - return buf->size; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - /* - * Noop. The kernel handles all fencing. - */ - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - return (drmBO *) private; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, pipe_mutex *mutex, - int lazy) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); - driReadUnlockKernelBO(); - - return ret; -} - - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - -/*static int -pool_setStatus(struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags) -{ - drmBO *buf = (drmBO *) private; - uint64_t new_flags = old_flags ^ flag_diff; - int ret; - - driReadLockKernelBO(); - ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, - 0, 0, 0); - driReadUnlockKernelBO(); - return ret; -}*/ - -struct _DriBufferPool * -driDRMPoolInit(int fd) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - - if (!pool) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - pool->reference = &pool_reference; - pool->unreference = &pool_unreference; - pool->data = NULL; - return pool; -} diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c deleted file mode 100644 index 831c75d30c..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c +++ /dev/null @@ -1,377 +0,0 @@ -#include "ws_dri_fencemgr.h" -#include "pipe/p_thread.h" -#include <xf86mm.h> -#include <string.h> -#include <unistd.h> - -/* - * Note: Locking order is - * _DriFenceObject::mutex - * _DriFenceMgr::mutex - */ - -struct _DriFenceMgr { - /* - * Constant members. Need no mutex protection. - */ - struct _DriFenceMgrCreateInfo info; - void *private; - - /* - * These members are protected by this->mutex - */ - pipe_mutex mutex; - int refCount; - drmMMListHead *heads; - int num_fences; -}; - -struct _DriFenceObject { - - /* - * These members are constant and need no mutex protection. - */ - struct _DriFenceMgr *mgr; - uint32_t fence_class; - uint32_t fence_type; - - /* - * These members are protected by mgr->mutex. - */ - drmMMListHead head; - int refCount; - - /* - * These members are protected by this->mutex. - */ - pipe_mutex mutex; - uint32_t signaled_type; - void *private; -}; - -uint32_t -driFenceType(struct _DriFenceObject *fence) -{ - return fence->fence_type; -} - -struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) -{ - struct _DriFenceMgr *tmp; - uint32_t i; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - pipe_mutex_init(tmp->mutex); - pipe_mutex_lock(tmp->mutex); - tmp->refCount = 1; - tmp->info = *info; - tmp->num_fences = 0; - tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); - if (!tmp->heads) - goto out_err; - - for (i=0; i<tmp->info.num_classes; ++i) { - DRMINITLISTHEAD(&tmp->heads[i]); - } - pipe_mutex_unlock(tmp->mutex); - return tmp; - - out_err: - if (tmp) - free(tmp); - return NULL; -} - -static void -driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) -{ - struct _DriFenceMgr *mgr = *pMgr; - - *pMgr = NULL; - if (--mgr->refCount == 0) - free(mgr); - else - pipe_mutex_unlock(mgr->mutex); -} - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr) -{ - pipe_mutex_lock((*pMgr)->mutex); - driFenceMgrUnrefUnlock(pMgr); -} - -static void -driFenceUnReferenceLocked(struct _DriFenceObject **pFence) -{ - struct _DriFenceObject *fence = *pFence; - struct _DriFenceMgr *mgr = fence->mgr; - - *pFence = NULL; - if (--fence->refCount == 0) { - DRMLISTDELINIT(&fence->head); - if (fence->private) - mgr->info.unreference(mgr, &fence->private); - --mgr->num_fences; - fence->mgr = NULL; - --mgr->refCount; - free(fence); - - } -} - - -static void -driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, - drmMMListHead *list, - uint32_t fence_class, - uint32_t fence_type) -{ - struct _DriFenceObject *entry; - drmMMListHead *prev; - - while(list != &mgr->heads[fence_class]) { - entry = DRMLISTENTRY(struct _DriFenceObject, list, head); - - /* - * Up refcount so that entry doesn't disappear from under us - * when we unlock-relock mgr to get the correct locking order. - */ - - ++entry->refCount; - pipe_mutex_unlock(mgr->mutex); - pipe_mutex_lock(entry->mutex); - pipe_mutex_lock(mgr->mutex); - - prev = list->prev; - - - - if (list->prev == list) { - - /* - * Somebody else removed the entry from the list. - */ - - pipe_mutex_unlock(entry->mutex); - driFenceUnReferenceLocked(&entry); - return; - } - - entry->signaled_type |= (fence_type & entry->fence_type); - if (entry->signaled_type == entry->fence_type) { - DRMLISTDELINIT(list); - mgr->info.unreference(mgr, &entry->private); - } - pipe_mutex_unlock(entry->mutex); - driFenceUnReferenceLocked(&entry); - list = prev; - } -} - - -int -driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint) -{ - struct _DriFenceMgr *mgr = fence->mgr; - int ret = 0; - - pipe_mutex_lock(fence->mutex); - - if ((fence->signaled_type & fence_type) == fence_type) - goto out0; - - ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); - if (ret) - goto out0; - - pipe_mutex_lock(mgr->mutex); - pipe_mutex_unlock(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - fence_type); - pipe_mutex_unlock(mgr->mutex); - return 0; - - out0: - pipe_mutex_unlock(fence->mutex); - return ret; -} - -uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) -{ - uint32_t ret; - - pipe_mutex_lock(fence->mutex); - ret = fence->signaled_type; - pipe_mutex_unlock(fence->mutex); - - return ret; -} - -int -driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, - uint32_t *signaled) -{ - int ret = 0; - struct _DriFenceMgr *mgr; - - pipe_mutex_lock(fence->mutex); - mgr = fence->mgr; - *signaled = fence->signaled_type; - if ((fence->signaled_type & flush_type) == flush_type) - goto out0; - - ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); - if (ret) { - *signaled = fence->signaled_type; - goto out0; - } - - if ((fence->signaled_type | *signaled) == fence->signaled_type) - goto out0; - - pipe_mutex_lock(mgr->mutex); - pipe_mutex_unlock(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - *signaled); - - pipe_mutex_unlock(mgr->mutex); - return 0; - out0: - pipe_mutex_unlock(fence->mutex); - return ret; -} - -struct _DriFenceObject * -driFenceReference(struct _DriFenceObject *fence) -{ - pipe_mutex_lock(fence->mgr->mutex); - ++fence->refCount; - pipe_mutex_unlock(fence->mgr->mutex); - return fence; -} - -void -driFenceUnReference(struct _DriFenceObject **pFence) -{ - struct _DriFenceMgr *mgr; - - if (*pFence == NULL) - return; - - mgr = (*pFence)->mgr; - pipe_mutex_lock(mgr->mutex); - ++mgr->refCount; - driFenceUnReferenceLocked(pFence); - driFenceMgrUnrefUnlock(&mgr); -} - -struct _DriFenceObject -*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, - uint32_t fence_type, void *private, size_t private_size) -{ - struct _DriFenceObject *fence; - size_t fence_size = sizeof(*fence); - - if (private_size) - fence_size = ((fence_size + 15) & ~15); - - fence = calloc(1, fence_size + private_size); - - if (!fence) { - int ret = mgr->info.finish(mgr, private, fence_type, 0); - - if (ret) - usleep(10000000); - - return NULL; - } - - pipe_mutex_init(fence->mutex); - pipe_mutex_lock(fence->mutex); - pipe_mutex_lock(mgr->mutex); - fence->refCount = 1; - DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); - fence->mgr = mgr; - ++mgr->refCount; - ++mgr->num_fences; - pipe_mutex_unlock(mgr->mutex); - fence->fence_class = fence_class; - fence->fence_type = fence_type; - fence->signaled_type = 0; - fence->private = private; - if (private_size) { - fence->private = (void *)(((uint8_t *) fence) + fence_size); - memcpy(fence->private, private, private_size); - } - - pipe_mutex_unlock(fence->mutex); - return fence; -} - - -static int -tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type) -{ - long fd = (long) mgr->private; - int dummy; - drmFence *fence = (drmFence *) private; - int ret; - - *signaled_type = 0; - ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); - if (ret) - return ret; - - *signaled_type = fence->signaled; - - return 0; -} - -static int -tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, - int lazy_hint) -{ - long fd = (long) mgr->private; - unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; - - return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); -} - -static int -tUnref(struct _DriFenceMgr *mgr, void **private) -{ - long fd = (long) mgr->private; - drmFence *fence = (drmFence *) *private; - *private = NULL; - - return drmFenceUnreference(fd, fence); -} - -struct _DriFenceMgr *driFenceMgrTTMInit(int fd) -{ - struct _DriFenceMgrCreateInfo info; - struct _DriFenceMgr *mgr; - - info.flags = DRI_FENCE_CLASS_ORDERED; - info.num_classes = 4; - info.signaled = tSignaled; - info.finish = tFinish; - info.unreference = tUnref; - - mgr = driFenceMgrCreate(&info); - if (mgr == NULL) - return NULL; - - mgr->private = (void *) (long) fd; - return mgr; -} - diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h deleted file mode 100644 index 4ea58dfe18..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h +++ /dev/null @@ -1,115 +0,0 @@ -#ifndef DRI_FENCEMGR_H -#define DRI_FENCEMGR_H - -#include <stdint.h> -#include <stdlib.h> - -struct _DriFenceObject; -struct _DriFenceMgr; - -/* - * Do a quick check to see if the fence manager has registered the fence - * object as signaled. Note that this function may return a false negative - * answer. - */ -extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); - -/* - * Check if the fence object is signaled. This function can be substantially - * more expensive to call than the above function, but will not return a false - * negative answer. The argument "flush_type" sets the types that the - * underlying mechanism must make sure will eventually signal. - */ -extern int driFenceSignaledType(struct _DriFenceObject *fence, - uint32_t flush_type, uint32_t *signaled); - -/* - * Convenience functions. - */ - -static inline int driFenceSignaled(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types; - int ret = driFenceSignaledType(fence, flush_type, &signaled_types); - if (ret) - return 0; - return ((signaled_types & flush_type) == flush_type); -} - -static inline int driFenceSignaledCached(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types = - driFenceSignaledTypeCached(fence); - - return ((signaled_types & flush_type) == flush_type); -} - -/* - * Reference a fence object. - */ -extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); - -/* - * Unreference a fence object. The fence object pointer will be reset to NULL. - */ - -extern void driFenceUnReference(struct _DriFenceObject **pFence); - - -/* - * Wait for a fence to signal the indicated fence_type. - * If "lazy_hint" is true, it indicates that the wait may sleep to avoid - * busy-wait polling. - */ -extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint); - -/* - * Create a DriFenceObject for manager "mgr". - * - * "private" is a pointer that should be used for the callbacks in - * struct _DriFenceMgrCreateInfo. - * - * if private_size is nonzero, then the info stored at *private, with size - * private size will be copied and the fence manager will instead use a - * pointer to the copied data for the callbacks in - * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by - * "private" may be destroyed after the call to driFenceCreate. - */ -extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, - uint32_t fence_class, - uint32_t fence_type, - void *private, - size_t private_size); - -extern uint32_t driFenceType(struct _DriFenceObject *fence); - -/* - * Fence creations are ordered. If a fence signals a fence_type, - * it is safe to assume that all fences of the same class that was - * created before that fence has signaled the same type. - */ - -#define DRI_FENCE_CLASS_ORDERED (1 << 0) - -struct _DriFenceMgrCreateInfo { - uint32_t flags; - uint32_t num_classes; - int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type); - int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); - int (*unreference) (struct _DriFenceMgr *mgr, void **private); -}; - -extern struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr); - -extern struct _DriFenceMgr * -driFenceMgrTTMInit(int fd); - -#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c deleted file mode 100644 index 60924eac9e..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c +++ /dev/null @@ -1,161 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com> - */ - -#include <xf86drm.h> -#include <stdlib.h> -#include <errno.h> -#include "pipe/p_debug.h" -#include "pipe/p_thread.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - unsigned long *private = malloc(size + 2*sizeof(unsigned long)); - if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) - abort(); - - *private = size; - return (void *)private; -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - free(private); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, - pipe_mutex *mutex, int lazy) -{ - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, pipe_mutex *mutex, void **virtual) -{ - *virtual = (void *)((unsigned long *)private + 2); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); - return 0UL; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - return *(unsigned long *) private; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - abort(); - return 0UL; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - abort(); - return NULL; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - - -struct _DriBufferPool * -driMallocPoolInit(void) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = NULL; - pool->fd = -1; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c deleted file mode 100644 index 391cea50a7..0000000000 --- a/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c +++ /dev/null @@ -1,968 +0,0 @@ -/************************************************************************** - * - * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellstrom <thomas-at-tungstengraphics-dot-com> - */ - -#include <stdint.h> -#include <sys/time.h> -#include <errno.h> -#include <unistd.h> -#include <assert.h> -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" -#include "ws_dri_bufmgr.h" -#include "pipe/p_thread.h" - -#define DRI_SLABPOOL_ALLOC_RETRIES 100 - -struct _DriSlab; - -struct _DriSlabBuffer { - int isSlabBuffer; - drmBO *bo; - struct _DriFenceObject *fence; - struct _DriSlab *parent; - drmMMListHead head; - uint32_t mapCount; - uint32_t start; - uint32_t fenceType; - int unFenced; - pipe_condvar event; -}; - -struct _DriKernelBO { - int fd; - drmBO bo; - drmMMListHead timeoutHead; - drmMMListHead head; - struct timeval timeFreed; - uint32_t pageAlignment; - void *virtual; -}; - -struct _DriSlab{ - drmMMListHead head; - drmMMListHead freeBuffers; - uint32_t numBuffers; - uint32_t numFree; - struct _DriSlabBuffer *buffers; - struct _DriSlabSizeHeader *header; - struct _DriKernelBO *kbo; -}; - - -struct _DriSlabSizeHeader { - drmMMListHead slabs; - drmMMListHead freeSlabs; - drmMMListHead delayedBuffers; - uint32_t numDelayed; - struct _DriSlabPool *slabPool; - uint32_t bufSize; - pipe_mutex mutex; -}; - -struct _DriFreeSlabManager { - struct timeval slabTimeout; - struct timeval checkInterval; - struct timeval nextCheck; - drmMMListHead timeoutList; - drmMMListHead unCached; - drmMMListHead cached; - pipe_mutex mutex; -}; - - -struct _DriSlabPool { - - /* - * The data of this structure remains constant after - * initialization and thus needs no mutex protection. - */ - - struct _DriFreeSlabManager *fMan; - uint64_t proposedFlags; - uint64_t validMask; - uint32_t *bucketSizes; - uint32_t numBuckets; - uint32_t pageSize; - int fd; - int pageAlignment; - int maxSlabSize; - int desiredNumBuffers; - struct _DriSlabSizeHeader *headers; -}; - -/* - * FIXME: Perhaps arrange timeout slabs in size buckets for fast - * retreival?? - */ - - -static inline int -driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) -{ - return ((arg1->tv_sec > arg2->tv_sec) || - ((arg1->tv_sec == arg2->tv_sec) && - (arg1->tv_usec > arg2->tv_usec))); -} - -static inline void -driTimeAdd(struct timeval *arg, struct timeval *add) -{ - unsigned int sec; - - arg->tv_sec += add->tv_sec; - arg->tv_usec += add->tv_usec; - sec = arg->tv_usec / 1000000; - arg->tv_sec += sec; - arg->tv_usec -= sec*1000000; -} - -static void -driFreeKernelBO(struct _DriKernelBO *kbo) -{ - if (!kbo) - return; - - (void) drmBOUnreference(kbo->fd, &kbo->bo); - free(kbo); -} - - -static void -driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, - struct timeval *time) -{ - drmMMListHead *list, *next; - struct _DriKernelBO *kbo; - - if (!driTimeAfterEq(time, &fMan->nextCheck)) - return; - - for (list = fMan->timeoutList.next, next = list->next; - list != &fMan->timeoutList; - list = next, next = list->next) { - - kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); - - if (!driTimeAfterEq(time, &kbo->timeFreed)) - break; - - DRMLISTDELINIT(&kbo->timeoutHead); - DRMLISTDELINIT(&kbo->head); - driFreeKernelBO(kbo); - } - - fMan->nextCheck = *time; - driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); -} - - -/* - * Add a _DriKernelBO to the free slab manager. - * This means that it is available for reuse, but if it's not - * reused in a while, it will be freed. - */ - -static void -driSetKernelBOFree(struct _DriFreeSlabManager *fMan, - struct _DriKernelBO *kbo) -{ - struct timeval time; - - pipe_mutex_lock(fMan->mutex); - gettimeofday(&time, NULL); - driTimeAdd(&time, &fMan->slabTimeout); - - kbo->timeFreed = time; - - if (kbo->bo.flags & DRM_BO_FLAG_CACHED) - DRMLISTADD(&kbo->head, &fMan->cached); - else - DRMLISTADD(&kbo->head, &fMan->unCached); - - DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); - driFreeTimeoutKBOsLocked(fMan, &time); - - pipe_mutex_unlock(fMan->mutex); -} - -/* - * Get a _DriKernelBO for us to use as storage for a slab. - * - */ - -static struct _DriKernelBO * -driAllocKernelBO(struct _DriSlabSizeHeader *header) - -{ - struct _DriSlabPool *slabPool = header->slabPool; - struct _DriFreeSlabManager *fMan = slabPool->fMan; - drmMMListHead *list, *next, *head; - uint32_t size = header->bufSize * slabPool->desiredNumBuffers; - struct _DriKernelBO *kbo; - struct _DriKernelBO *kboTmp; - int ret; - - /* - * FIXME: We should perhaps allow some variation in slabsize in order - * to efficiently reuse slabs. - */ - - size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; - size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); - pipe_mutex_lock(fMan->mutex); - - kbo = NULL; - - retry: - head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? - &fMan->cached : &fMan->unCached; - - for (list = head->next, next = list->next; - list != head; - list = next, next = list->next) { - - kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); - - if ((kboTmp->bo.size == size) && - (slabPool->pageAlignment == 0 || - (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { - - if (!kbo) - kbo = kboTmp; - - if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) - break; - - } - } - - if (kbo) { - DRMLISTDELINIT(&kbo->head); - DRMLISTDELINIT(&kbo->timeoutHead); - } - - pipe_mutex_unlock(fMan->mutex); - - if (kbo) { - uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; - - ret = 0; - if (new_mask) { - ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, - new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); - } - if (ret == 0) - return kbo; - - driFreeKernelBO(kbo); - kbo = NULL; - goto retry; - } - - kbo = calloc(1, sizeof(struct _DriKernelBO)); - if (!kbo) - return NULL; - - kbo->fd = slabPool->fd; - DRMINITLISTHEAD(&kbo->head); - DRMINITLISTHEAD(&kbo->timeoutHead); - ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, - slabPool->proposedFlags, - DRM_BO_HINT_DONT_FENCE, &kbo->bo); - if (ret) - goto out_err0; - - ret = drmBOMap(kbo->fd, &kbo->bo, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &kbo->virtual); - - if (ret) - goto out_err1; - - ret = drmBOUnmap(kbo->fd, &kbo->bo); - if (ret) - goto out_err1; - - return kbo; - - out_err1: - drmBOUnreference(kbo->fd, &kbo->bo); - out_err0: - free(kbo); - return NULL; -} - - -static int -driAllocSlab(struct _DriSlabSizeHeader *header) -{ - struct _DriSlab *slab; - struct _DriSlabBuffer *buf; - uint32_t numBuffers; - int ret; - int i; - - slab = calloc(1, sizeof(*slab)); - if (!slab) - return -ENOMEM; - - slab->kbo = driAllocKernelBO(header); - if (!slab->kbo) { - ret = -ENOMEM; - goto out_err0; - } - - numBuffers = slab->kbo->bo.size / header->bufSize; - - slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); - if (!slab->buffers) { - ret = -ENOMEM; - goto out_err1; - } - - DRMINITLISTHEAD(&slab->head); - DRMINITLISTHEAD(&slab->freeBuffers); - slab->numBuffers = numBuffers; - slab->numFree = 0; - slab->header = header; - - buf = slab->buffers; - for (i=0; i < numBuffers; ++i) { - buf->parent = slab; - buf->start = i* header->bufSize; - buf->mapCount = 0; - buf->isSlabBuffer = 1; - pipe_condvar_init(buf->event); - DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); - slab->numFree++; - buf++; - } - - DRMLISTADDTAIL(&slab->head, &header->slabs); - - return 0; - - out_err1: - driSetKernelBOFree(header->slabPool->fMan, slab->kbo); - free(slab->buffers); - out_err0: - free(slab); - return ret; -} - -/* - * Delete a buffer from the slab header delayed list and put - * it on the slab free list. - */ - -static void -driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) -{ - struct _DriSlab *slab = buf->parent; - struct _DriSlabSizeHeader *header = slab->header; - drmMMListHead *list = &buf->head; - - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &slab->freeBuffers); - slab->numFree++; - - if (slab->head.next == &slab->head) - DRMLISTADDTAIL(&slab->head, &header->slabs); - - if (slab->numFree == slab->numBuffers) { - list = &slab->head; - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &header->freeSlabs); - } - - if (header->slabs.next == &header->slabs || - slab->numFree != slab->numBuffers) { - - drmMMListHead *next; - struct _DriFreeSlabManager *fMan = header->slabPool->fMan; - - for (list = header->freeSlabs.next, next = list->next; - list != &header->freeSlabs; - list = next, next = list->next) { - - slab = DRMLISTENTRY(struct _DriSlab, list, head); - - DRMLISTDELINIT(list); - driSetKernelBOFree(fMan, slab->kbo); - free(slab->buffers); - free(slab); - } - } -} - -static void -driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) -{ - drmMMListHead *list, *prev, *first; - struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - int firstWasSignaled = 1; - int signaled; - int i; - int ret; - - /* - * Rerun the freeing test if the youngest tested buffer - * was signaled, since there might be more idle buffers - * in the delay list. - */ - - while (firstWasSignaled) { - firstWasSignaled = 0; - signaled = 0; - first = header->delayedBuffers.next; - - /* Only examine the oldest 1/3 of delayed buffers: - */ - if (header->numDelayed > 3) { - for (i = 0; i < header->numDelayed; i += 3) { - first = first->next; - } - } - - for (list = first, prev = list->prev; - list != &header->delayedBuffers; - list = prev, prev = list->prev) { - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - slab = buf->parent; - - if (!signaled) { - if (wait) { - ret = driFenceFinish(buf->fence, buf->fenceType, 0); - if (ret) - break; - signaled = 1; - wait = 0; - } else { - signaled = driFenceSignaled(buf->fence, buf->fenceType); - } - if (signaled) { - if (list == first) - firstWasSignaled = 1; - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } - } -} - - -static struct _DriSlabBuffer * -driSlabAllocBuffer(struct _DriSlabSizeHeader *header) -{ - static struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - drmMMListHead *list; - int count = DRI_SLABPOOL_ALLOC_RETRIES; - - pipe_mutex_lock(header->mutex); - while(header->slabs.next == &header->slabs && count > 0) { - driSlabCheckFreeLocked(header, 0); - if (header->slabs.next != &header->slabs) - break; - - pipe_mutex_unlock(header->mutex); - if (count != DRI_SLABPOOL_ALLOC_RETRIES) - usleep(1); - pipe_mutex_lock(header->mutex); - (void) driAllocSlab(header); - count--; - } - - list = header->slabs.next; - if (list == &header->slabs) { - pipe_mutex_unlock(header->mutex); - return NULL; - } - slab = DRMLISTENTRY(struct _DriSlab, list, head); - if (--slab->numFree == 0) - DRMLISTDELINIT(list); - - list = slab->freeBuffers.next; - DRMLISTDELINIT(list); - - pipe_mutex_unlock(header->mutex); - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - return buf; -} - -static void * -pool_create(struct _DriBufferPool *driPool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment) -{ - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - struct _DriSlabSizeHeader *header; - struct _DriSlabBuffer *buf; - void *dummy; - int i; - int ret; - - /* - * FIXME: Check for compatibility. - */ - - header = pool->headers; - for (i=0; i<pool->numBuckets; ++i) { - if (header->bufSize >= size) - break; - header++; - } - - if (i < pool->numBuckets) - return driSlabAllocBuffer(header); - - - /* - * Fall back to allocate a buffer object directly from DRM. - * and wrap it in a driBO structure. - */ - - - buf = calloc(1, sizeof(*buf)); - - if (!buf) - return NULL; - - buf->bo = calloc(1, sizeof(*buf->bo)); - if (!buf->bo) - goto out_err0; - - if (alignment) { - if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) - goto out_err1; - if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) - goto out_err1; - } - - ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, - flags, hint, buf->bo); - if (ret) - goto out_err1; - - ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &dummy); - if (ret) - goto out_err2; - - ret = drmBOUnmap(pool->fd, buf->bo); - if (ret) - goto out_err2; - - return buf; - out_err2: - drmBOUnreference(pool->fd, buf->bo); - out_err1: - free(buf->bo); - out_err0: - free(buf); - return NULL; -} - -static int -pool_destroy(struct _DriBufferPool *driPool, void *private) -{ - struct _DriSlabBuffer *buf = - (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - int ret; - - ret = drmBOUnreference(pool->fd, buf->bo); - free(buf->bo); - free(buf); - return ret; - } - - slab = buf->parent; - header = slab->header; - - pipe_mutex_lock(header->mutex); - buf->unFenced = 0; - buf->mapCount = 0; - - if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { - DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); - header->numDelayed++; - } else { - if (buf->fence) - driFenceUnReference(&buf->fence); - driSlabFreeBufferLocked(buf); - } - - pipe_mutex_unlock(header->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *driPool, void *private, - pipe_mutex *mutex, int lazy) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - while(buf->unFenced) - pipe_condvar_wait(buf->event, *mutex); - - if (!buf->fence) - return 0; - - driFenceFinish(buf->fence, buf->fenceType, lazy); - driFenceUnReference(&buf->fence); - - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, pipe_mutex *mutex, void **virtual) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - int busy; - - if (buf->isSlabBuffer) - busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); - else - busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); - - - if (busy) { - if (hint & DRM_BO_HINT_DONT_BLOCK) - return -EBUSY; - else { - (void) pool_waitIdle(pool, private, mutex, 0); - } - } - - ++buf->mapCount; - *virtual = (buf->isSlabBuffer) ? - (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : - (void *) buf->bo->virtual; - - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - --buf->mapCount; - if (buf->mapCount == 0 && buf->isSlabBuffer) - pipe_condvar_broadcast(buf->event); - - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return buf->bo->offset; - } - - slab = buf->parent; - header = slab->header; - - (void) header; - assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return slab->kbo->bo.offset + buf->start; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return buf->start; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return buf->bo->flags; - - return buf->parent->kbo->bo.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - if (!buf->isSlabBuffer) - return buf->bo->size; - - return buf->parent->header->bufSize; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - drmBO *bo; - - if (buf->fence) - driFenceUnReference(&buf->fence); - - buf->fence = driFenceReference(fence); - bo = (buf->isSlabBuffer) ? - &buf->parent->kbo->bo: - buf->bo; - buf->fenceType = bo->fenceFlags; - - buf->unFenced = 0; - pipe_condvar_broadcast(buf->event); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private, - pipe_mutex *mutex) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return 0; - - while(buf->mapCount != 0) - pipe_condvar_wait(buf->event, *mutex); - - buf->unFenced = 1; - return 0; -} - - -struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) -{ - struct _DriFreeSlabManager *tmp; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - pipe_mutex_init(tmp->mutex); - pipe_mutex_lock(tmp->mutex); - tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; - tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; - tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; - - tmp->checkInterval.tv_usec = checkIntervalMsec*1000; - tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; - tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; - - gettimeofday(&tmp->nextCheck, NULL); - driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); - DRMINITLISTHEAD(&tmp->timeoutList); - DRMINITLISTHEAD(&tmp->unCached); - DRMINITLISTHEAD(&tmp->cached); - pipe_mutex_unlock(tmp->mutex); - - return tmp; -} - -void -driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) -{ - struct timeval time; - - time = fMan->nextCheck; - driTimeAdd(&time, &fMan->checkInterval); - - pipe_mutex_lock(fMan->mutex); - driFreeTimeoutKBOsLocked(fMan, &time); - pipe_mutex_unlock(fMan->mutex); - - assert(fMan->timeoutList.next == &fMan->timeoutList); - assert(fMan->unCached.next == &fMan->unCached); - assert(fMan->cached.next == &fMan->cached); - - free(fMan); -} - -static void -driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, - struct _DriSlabSizeHeader *header) -{ - pipe_mutex_init(header->mutex); - pipe_mutex_lock(header->mutex); - - DRMINITLISTHEAD(&header->slabs); - DRMINITLISTHEAD(&header->freeSlabs); - DRMINITLISTHEAD(&header->delayedBuffers); - - header->numDelayed = 0; - header->slabPool = pool; - header->bufSize = size; - - pipe_mutex_unlock(header->mutex); -} - -static void -driFinishSizeHeader(struct _DriSlabSizeHeader *header) -{ - drmMMListHead *list, *next; - struct _DriSlabBuffer *buf; - - pipe_mutex_lock(header->mutex); - for (list = header->delayedBuffers.next, next = list->next; - list != &header->delayedBuffers; - list = next, next = list->next) { - - buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); - if (buf->fence) { - (void) driFenceFinish(buf->fence, buf->fenceType, 0); - driFenceUnReference(&buf->fence); - } - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - pipe_mutex_unlock(header->mutex); -} - -static void -pool_takedown(struct _DriBufferPool *driPool) -{ - struct _DriSlabPool *pool = driPool->data; - int i; - - for (i=0; i<pool->numBuckets; ++i) { - driFinishSizeHeader(&pool->headers[i]); - } - - free(pool->headers); - free(pool->bucketSizes); - free(pool); - free(driPool); -} - -struct _DriBufferPool * -driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan) -{ - struct _DriBufferPool *driPool; - struct _DriSlabPool *pool; - uint32_t i; - - driPool = calloc(1, sizeof(*driPool)); - if (!driPool) - return NULL; - - pool = calloc(1, sizeof(*pool)); - if (!pool) - goto out_err0; - - pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); - if (!pool->bucketSizes) - goto out_err1; - - pool->headers = calloc(numSizes, sizeof(*pool->headers)); - if (!pool->headers) - goto out_err2; - - pool->fMan = fMan; - pool->proposedFlags = flags; - pool->validMask = validMask; - pool->numBuckets = numSizes; - pool->pageSize = getpagesize(); - pool->fd = fd; - pool->pageAlignment = pageAlignment; - pool->maxSlabSize = maxSlabSize; - pool->desiredNumBuffers = desiredNumBuffers; - - for (i=0; i<pool->numBuckets; ++i) { - pool->bucketSizes[i] = (smallestSize << i); - driInitSizeHeader(pool, pool->bucketSizes[i], - &pool->headers[i]); - } - - driPool->data = (void *) pool; - driPool->map = &pool_map; - driPool->unmap = &pool_unmap; - driPool->destroy = &pool_destroy; - driPool->offset = &pool_offset; - driPool->poolOffset = &pool_poolOffset; - driPool->flags = &pool_flags; - driPool->size = &pool_size; - driPool->create = &pool_create; - driPool->fence = &pool_fence; - driPool->kernel = &pool_kernel; - driPool->validate = &pool_validate; - driPool->waitIdle = &pool_waitIdle; - driPool->takeDown = &pool_takedown; - - return driPool; - - out_err2: - free(pool->bucketSizes); - out_err1: - free(pool); - out_err0: - free(driPool); - - return NULL; -} diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile deleted file mode 100644 index 2046441a22..0000000000 --- a/src/gallium/winsys/drm/intel/dri/Makefile +++ /dev/null @@ -1,33 +0,0 @@ -TOP = ../../../../../.. -include $(TOP)/configs/current - -LIBNAME = i915_dri.so -LIBNAME_EGL = egl_i915_dri.so - -PIPE_DRIVERS = \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - ../common/libinteldrm.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a - - -DRIVER_SOURCES = \ - intel_winsys_softpipe.c \ - intel_swapbuffers.c \ - intel_context.c \ - intel_lock.c \ - intel_screen.c - -C_SOURCES = \ - $(COMMON_GALLIUM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I../common $(shell pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -include ../../Makefile.template - -#intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c - -symlinks: diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript deleted file mode 100644 index 6a4f50afcc..0000000000 --- a/src/gallium/winsys/drm/intel/dri/SConscript +++ /dev/null @@ -1,41 +0,0 @@ -Import('*') - -if 'mesa' in env['statetrackers']: - - env = drienv.Clone() - - env.Append(CPPPATH = [ - '../intel', - 'server' - ]) - - #MINIGLX_SOURCES = server/intel_dri.c - - DRIVER_SOURCES = [ - 'intel_winsys_pipe.c', - 'intel_winsys_softpipe.c', - 'intel_winsys_i915.c', - 'intel_batchbuffer.c', - 'intel_swapbuffers.c', - 'intel_context.c', - 'intel_lock.c', - 'intel_screen.c', - 'intel_batchpool.c', - ] - - sources = \ - COMMON_GALLIUM_SOURCES + \ - COMMON_BM_SOURCES + \ - DRIVER_SOURCES - - drivers = [ - softpipe, - i915simple - ] - - # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions - env.SharedLibrary( - target ='i915tex_dri.so', - source = sources, - LIBS = drivers + mesa + auxiliaries + env['LIBS'], - ) diff --git a/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h b/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h deleted file mode 100644 index 3e95326168..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h +++ /dev/null @@ -1,24 +0,0 @@ -#ifndef INTEL_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "intel_be_batchbuffer.h" - -/* - * Need to redefine the BATCH defines - */ - -#undef BEGIN_BATCH -#define BEGIN_BATCH(dwords, relocs) \ - (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) - -#undef OUT_BATCH -#define OUT_BATCH(d) \ - i915_batchbuffer_dword(&intel->base.batch->base, d) - -#undef OUT_RELOC -#define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ -} while (0) - -#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_context.c b/src/gallium/winsys/drm/intel/dri/intel_context.c deleted file mode 100644 index 97ef731aaa..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_context.c +++ /dev/null @@ -1,337 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i830_dri.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" -#include "intel_winsys_softpipe.h" - -#include "i915simple/i915_screen.h" - -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_context.h" - -#include "utils.h" - - -#ifdef DEBUG -int __intel_debug = 0; -#endif - - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, - {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, - {NULL, NULL} -}; - - - -#ifdef DEBUG -static const struct dri_debug_control debug_control[] = { - {"ioctl", DEBUG_IOCTL}, - {"bat", DEBUG_BATCH}, - {"lock", DEBUG_LOCK}, - {"swap", DEBUG_SWAP}, - {NULL, 0} -}; -#endif - - - -static void -intel_lock_hardware(struct intel_be_context *context) -{ - struct intel_context *intel = (struct intel_context *)context; - LOCK_HARDWARE(intel); -} - -static void -intel_unlock_hardware(struct intel_be_context *context) -{ - struct intel_context *intel = (struct intel_context *)context; - UNLOCK_HARDWARE(intel); -} - -static boolean -intel_locked_hardware(struct intel_be_context *context) -{ - struct intel_context *intel = (struct intel_context *)context; - return intel->locked ? TRUE : FALSE; -} - -GLboolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct intel_context *intel = CALLOC_STRUCT(intel_context); - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *saPriv = intelScreen->sarea; - int fthrottle_mode; - GLboolean havePools; - struct pipe_context *pipe; - struct st_context *st_share = NULL; - - if (sharedContextPrivate) { - st_share = ((struct intel_context *) sharedContextPrivate)->st; - } - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - - /* - * memory pools - */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - // ZZZ JB should be per screen and not be done per context - havePools = intelCreatePools(sPriv); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - if (!havePools) - return GL_FALSE; - - - /* Dri stuff */ - intel->hHWContext = driContextPriv->hHWContext; - intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; - intel->irqsEmitted = 0; - - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - -#ifdef DEBUG - __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); -#endif - intel->base.hardware_lock = intel_lock_hardware; - intel->base.hardware_unlock = intel_unlock_hardware; - intel->base.hardware_locked = intel_locked_hardware; - - intel_be_init_context(&intel->base, &intelScreen->base); - - /* - * Pipe-related setup - */ - if (getenv("INTEL_SP")) { - /* use softpipe driver instead of hw */ - pipe = intel_create_softpipe( intel, &intelScreen->base.base ); - } - else { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - pipe = i915_create_context(intelScreen->base.screen, - &intelScreen->base.base, - &intel->base.base); - break; - default: - fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); - - pipe = intel_create_softpipe( intel, &intelScreen->base.base ); - break; - } - } - - pipe->priv = intel; - - intel->st = st_create_context(pipe, visual, st_share); - - driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE ); - - return GL_TRUE; -} - - -void -intelDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - - assert(intel); /* should never be null */ - if (intel) { - st_finish(intel->st); - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - if (intel->first_swap_fence) { - driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = NULL; - } - - if (intel->intelScreen->dummyContext == intel) - intel->intelScreen->dummyContext = NULL; - - st_destroy_context(intel->st); - intel_be_destroy_context(&intel->base); - free(intel); - } -} - - -GLboolean -intelUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL); - /* XXX make_current(NULL)? */ - return GL_TRUE; -} - - -GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - if (driContextPriv) { - struct intel_context *intel = intel_context(driContextPriv); - struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); - struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); - - assert(draw_fb->stfb); - assert(read_fb->stfb); - - /* This is for situations in which we need a rendering context but - * there may not be any currently bound. - */ - intel->intelScreen->dummyContext = intel; - - st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); - - if ((intel->driDrawable != driDrawPriv) || - (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelUpdateWindowSize(driDrawPriv); - intel->lastStamp = driDrawPriv->lastStamp; - } - - /* The size of the draw buffer will have been updated above. - * If the readbuffer is a different window, check/update its size now. - */ - if (driReadPriv != driDrawPriv) { - intelUpdateWindowSize(driReadPriv); - } - - } - else { - st_make_current(NULL, NULL, NULL); - } - - return GL_TRUE; -} diff --git a/src/gallium/winsys/drm/intel/dri/intel_context.h b/src/gallium/winsys/drm/intel/dri/intel_context.h deleted file mode 100644 index 5d22a422af..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_context.h +++ /dev/null @@ -1,164 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_CONTEXT_H -#define INTEL_CONTEXT_H - -#include <stdint.h> -#include "drm.h" - -#include "pipe/p_debug.h" - -#include "intel_screen.h" -#include "i915_drm.h" - -#include "intel_be_context.h" - - -struct pipe_context; -struct intel_context; -struct _DriBufferObject; -struct st_context; - - -#define INTEL_MAX_FIXUP 64 - -/** - * Intel rendering context, contains a state tracker and intel-specific info. - */ -struct intel_context -{ - struct intel_be_context base; - struct st_context *st; - - struct _DriFenceObject *last_swap_fence; - struct _DriFenceObject *first_swap_fence; - -// struct intel_batchbuffer *batch; - - boolean locked; - char *prevLockFile; - int prevLockLine; - - uint irqsEmitted; - drm_i915_irq_wait_t iw; - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - __DRIdrawablePrivate *driDrawable; - __DRIscreenPrivate *driScreen; - struct intel_screen *intelScreen; - drmI830Sarea *sarea; - - uint lastStamp; - - /** - * Configuration cache - */ - driOptionCache optionCache; -}; - - - -/** - * Intel framebuffer. - */ -struct intel_framebuffer -{ - struct st_framebuffer *stfb; - - /* other fields TBD */ - int other; -}; - - - - -/* These are functions now: - */ -void LOCK_HARDWARE( struct intel_context *intel ); -void UNLOCK_HARDWARE( struct intel_context *intel ); - -extern char *__progname; - - - -/* ================================================================ - * Debugging: - */ -#ifdef DEBUG -extern int __intel_debug; - -#define DEBUG_SWAP 0x1 -#define DEBUG_LOCK 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BATCH 0x8 - -#define DBG(flag, ...) do { \ - if (__intel_debug & (DEBUG_##flag)) \ - printf(__VA_ARGS__); \ -} while(0) - -#else -#define DBG(flag, ...) -#endif - - - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -/** Cast wrapper */ -static INLINE struct intel_context * -intel_context(__DRIcontextPrivate *driContextPriv) -{ - return (struct intel_context *) driContextPriv->driverPrivate; -} - - -/** Cast wrapper */ -static INLINE struct intel_framebuffer * -intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) -{ - return (struct intel_framebuffer *) driDrawPriv->driverPrivate; -} - - -#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_lock.c b/src/gallium/winsys/drm/intel/dri/intel_lock.c deleted file mode 100644 index ad1c202429..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_lock.c +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "main/glheader.h" -#include "pipe/p_thread.h" -#include <GL/internal/glcore.h> -#include "state_tracker/st_public.h" -#include "intel_context.h" -#include "i830_dri.h" - - - -pipe_static_mutex( lockMutex ); - - -static void -intelContendedLock(struct intel_context *intel, uint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - DBG(LOCK, "%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { - - intelUpdateScreenRotation(sPriv, sarea); - } -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret = 0; - - pipe_mutex_lock(lockMutex); - assert(!intel->locked); - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - DBG(LOCK, "%s - locked\n", __progname); - - intel->locked = 1; -} - - -/* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - assert(intel->locked); - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - pipe_mutex_unlock(lockMutex); - - DBG(LOCK, "%s - unlocked\n", __progname); -} diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.c b/src/gallium/winsys/drm/intel/dri/intel_screen.c deleted file mode 100644 index ed75368982..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_screen.c +++ /dev/null @@ -1,703 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "utils.h" -#include "vblank.h" -#include "xmlpool.h" - -#include "intel_context.h" -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_swapbuffers.h" - -#include "i830_dri.h" -#include "ws_dri_bufpool.h" - -#include "pipe/p_context.h" -#include "pipe/p_screen.h" -#include "pipe/p_inlines.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_cb_fbo.h" - -static void -intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle); - -static void -intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle) -{ - struct pipe_screen *screen = intelScreen->base.screen; - struct pipe_texture *texture; - struct pipe_texture templat; - struct pipe_surface *surface; - struct pipe_buffer *buffer; - unsigned pitch; - - assert(intelScreen->front.cpp == 4); - - buffer = intel_be_buffer_from_handle(&intelScreen->base, - "front", handle); - - if (!buffer) - return; - - intelScreen->front.buffer = dri_bo(buffer); - - memset(&templat, 0, sizeof(templat)); - templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; - templat.target = PIPE_TEXTURE_2D; - templat.last_level = 0; - templat.depth[0] = 1; - templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; - templat.width[0] = intelScreen->front.width; - templat.height[0] = intelScreen->front.height; - pf_get_block(templat.format, &templat.block); - pitch = intelScreen->front.pitch; - - texture = screen->texture_blanket(screen, - &templat, - &pitch, - buffer); - - /* Unref the buffer we don't need it anyways */ - pipe_buffer_reference(screen, &buffer, NULL); - - surface = screen->get_tex_surface(screen, - texture, - 0, - 0, - 0, - PIPE_BUFFER_USAGE_GPU_WRITE); - - intelScreen->front.texture = texture; - intelScreen->front.surface = surface; -} - -PUBLIC const char __driConfigOptions[] = - DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY -// DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END DRI_CONF_END; - -const uint __driNConfigOptions = 3; - -#ifdef USE_NEW_INTERFACE -static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE */ - -extern const struct dri_extension card_extensions[]; - -static GLboolean -intel_get_param(__DRIscreenPrivate *psp, int param, int *value) -{ - int ret; - struct drm_i915_getparam gp; - - gp.param = param; - gp.value = value; - - ret = drmCommandWriteRead(psp->fd, DRM_I915_GETPARAM, &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "drm_i915_getparam: %d\n", ret); - return GL_FALSE; - } - - return GL_TRUE; -} - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, int texname, - unsigned long long offset, int depth, uint pitch) -{ - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->pt) - st->pipe->texture_release(intel->st->pipe, &stObj->pt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -#if 0 -static void -intelHandleDrawableConfig(__DRIdrawablePrivate *dPriv, - __DRIcontextPrivate *pcp, - __DRIDrawableConfigEvent *event) -{ - (void) dPriv; - (void) pcp; - (void) event; -} -#endif - -#if 0 -static void -intelHandleBufferAttach(__DRIdrawablePrivate *dPriv, - __DRIcontextPrivate *pcp, - __DRIBufferAttachEvent *ba) -{ - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - - switch (ba->buffer.attachment) { - case DRI_DRAWABLE_BUFFER_FRONT_LEFT: - intelScreen->front.width = dPriv->w; - intelScreen->front.height = dPriv->h; - intelScreen->front.cpp = ba->buffer.cpp; - intelScreen->front.pitch = ba->buffer.pitch; - driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); - driBOSetReferenced(intelScreen->front.buffer, ba->buffer.handle); - break; - - case DRI_DRAWABLE_BUFFER_BACK_LEFT: - case DRI_DRAWABLE_BUFFER_DEPTH: - case DRI_DRAWABLE_BUFFER_STENCIL: - case DRI_DRAWABLE_BUFFER_ACCUM: - /* anything ?? */ - break; - - default: - fprintf(stderr, "unhandled buffer attach event, attachment type %d\n", - ba->buffer.attachment); - return; - } -} -#endif - -static const __DRItexOffsetExtension intelTexOffsetExtension = { - { __DRI_TEX_OFFSET }, - intelSetTexOffset, -}; - -#if 0 -static const __DRItexBufferExtension intelTexBufferExtension = { - { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION }, - intelSetTexBuffer, -}; -#endif - -static const __DRIextension *intelScreenExtensions[] = { - &driReadDrawableExtension, - &driCopySubBufferExtension.base, - &driSwapControlExtension.base, - &driFrameTrackingExtension.base, - &driMediaStreamCounterExtension.base, - &intelTexOffsetExtension.base, -// &intelTexBufferExtension.base, - NULL -}; - - -static void -intelPrintDRIInfo(struct intel_screen * intelScreen, - __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) -{ - fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->front.size, intelScreen->front.offset, - intelScreen->front.pitch); - fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); -} - - -#if 0 -static void -intelPrintSAREA(const drmI830Sarea * sarea) -{ - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, - sarea->height); - fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); - fprintf(stderr, - "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->front_offset, sarea->front_size, - (unsigned) sarea->front_handle); - fprintf(stderr, - "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->back_offset, sarea->back_size, - (unsigned) sarea->back_handle); - fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->depth_offset, sarea->depth_size, - (unsigned) sarea->depth_handle); - fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); - fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); - fprintf(stderr, - "SAREA: rotated offset: 0x%08x size: 0x%x\n", - sarea->rotated_offset, sarea->rotated_size); - fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); -} -#endif - - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->front.map) { - drmUnmap(intelScreen->front.map, intelScreen->front.size); - intelScreen->front.map = NULL; - } - - if (intelScreen->front.buffer) - driDeleteBuffers(1, &intelScreen->front.buffer); - - intelScreen->front.width = sarea->width; - intelScreen->front.height = sarea->height; - intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; - intelScreen->front.size = sarea->front_size; - intelScreen->front.handle = sarea->front_handle; - - assert( sarea->front_size >= - intelScreen->front.pitch * intelScreen->front.height ); - -#if 0 /* JB not important */ - if (!sarea->front_handle) - return; - - if (drmMap(sPriv->fd, - sarea->front_handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - fprintf(stderr, "drmMap(frontbuffer) failed!\n"); - return; - } -#endif - -#if 0 /* JB */ - if (intelScreen->staticPool) { - driGenBuffers(intelScreen->staticPool, "static region", 1, - &intelScreen->front.buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - - driBOSetStatic(intelScreen->front.buffer, - intelScreen->front.offset, - intelScreen->front.pitch * intelScreen->front.height, - intelScreen->front.map, 0); - } -#else - if (intelScreen->base.staticPool) { - if (intelScreen->front.buffer) { - driBOUnReference(intelScreen->front.buffer); - pipe_surface_reference(&intelScreen->front.surface, NULL); - pipe_texture_reference(&intelScreen->front.texture, NULL); - } - intelCreateSurface(intelScreen, &intelScreen->base.base, sarea->front_bo_handle); - } -#endif -} - - -boolean -intelCreatePools(__DRIscreenPrivate * sPriv) -{ - //unsigned batchPoolSize = 1024*1024; - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->havePools) - return GL_TRUE; - - intelScreen->havePools = GL_TRUE; - - if (intelScreen->sarea) - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - return GL_TRUE; -} - -static const char * -intel_get_name( struct pipe_winsys *winsys ) -{ - return "Intel/DRI/ttm"; -} - -/* - * The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelDisplaySurface(dPriv, surf, NULL); -} - -static boolean -intelInitDriver(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr, - "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) - return GL_FALSE; - - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - sPriv->private = (void *) intelScreen; - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); - - intelScreen->deviceID = gDRIPriv->deviceID; - - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drm_version.minor; - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - - sPriv->extensions = intelScreenExtensions; - - intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; - intelScreen->base.base.get_name = intel_get_name; - intel_be_init_device(&intelScreen->base, sPriv->fd, intelScreen->deviceID); - - return GL_TRUE; -} - - -static void -intelDestroyScreen(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - intel_be_destroy_device(&intelScreen->base); - /* intelUnmapScreenRegions(intelScreen); */ - - FREE(intelScreen); - sPriv->private = NULL; -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static boolean -intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * visual, boolean isPixmap) -{ - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); - - if (!intelfb) - return GL_FALSE; - - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits == 8) - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - else - stencilFormat = PIPE_FORMAT_NONE; - - intelfb->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - driDrawPriv->w, - driDrawPriv->h, - (void*) intelfb); - if (!intelfb->stfb) { - free(intelfb); - return GL_FALSE; - } - - driDrawPriv->driverPrivate = (void *) intelfb; - return GL_TRUE; - } -} - -static void -intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); - assert(intelfb->stfb); - st_unreference_framebuffer(intelfb->stfb); - free(intelfb); -} - - -/** - * Get information about previous buffer swaps. - */ -static int -intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) -{ - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) - || (sInfo == NULL)) { - return -1; - } - - return 0; -} - -static __DRIconfig ** -intelFillInModes(__DRIscreenPrivate *psp, - unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) -{ - __DRIconfig **configs; - __GLcontextModes *m; - unsigned num_modes; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - int i; - - /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't - * support pageflipping at all. - */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - uint8_t depth_bits_array[3]; - uint8_t stencil_bits_array[3]; - uint8_t msaa_samples_array[1]; - - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - msaa_samples_array[0] = 0; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - num_modes = depth_buffer_factor * back_buffer_factor * 4; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - configs = driCreateConfigs(fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, msaa_samples_array, 1); - if (configs == NULL) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - - /* Mark the visual as slow if there are "fake" stencil bits. - */ - for (i = 0; configs[i]; i++) { - m = &configs[i]->modes; - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - - return configs; -} - -/** - * This is the driver specific part of the createNewScreen entry point. - * - * \todo maybe fold this into intelInitDriver - * - * \return the __GLcontextModes supported by this driver - */ -static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp) -{ -#ifdef I915 - static const __DRIversion ddx_expected = { 1, 5, 0 }; -#else - static const __DRIversion ddx_expected = { 1, 6, 0 }; -#endif - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 5, 0 }; - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - - if (!driCheckDriDdxDrmVersions2("i915", - &psp->dri_version, &dri_expected, - &psp->ddx_version, &ddx_expected, - &psp->drm_version, &drm_expected)) { - return NULL; - } - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - //intelInitExtensions(NULL, GL_TRUE); - - if (!intelInitDriver(psp)) - return NULL; - - psp->extensions = intelScreenExtensions; - - return (const __DRIconfig **) - intelFillInModes(psp, dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); -} - -/** - * This is the driver specific part of the createNewScreen entry point. - * - * \return the __GLcontextModes supported by this driver - */ -static const -__DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp) -{ - struct intel_screen *intelScreen; - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - //intelInitExtensions(NULL, GL_TRUE); - - /* Allocate the private area */ - intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) { - fprintf(stderr, "\nERROR! Allocating private area failed\n"); - return GL_FALSE; - } - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - psp->private = (void *) intelScreen; - - intelScreen->drmMinor = psp->drm_version.minor; - - /* Determine chipset ID? */ - if (!intel_get_param(psp, I915_PARAM_CHIPSET_ID, - &intelScreen->deviceID)) - return GL_FALSE; - - psp->extensions = intelScreenExtensions; - - intel_be_init_device(&intelScreen->base, psp->fd, intelScreen->deviceID); - intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; - intelScreen->base.base.get_name = intel_get_name; - - return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1), - intelFillInModes(psp, 32, 24, 8, 1)); -} - -const struct __DriverAPIRec driDriverAPI = { - .InitScreen = intelInitScreen, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetDrawableMSC = driDrawableGetMSC32, - .WaitForMSC = driWaitForMSC32, - .CopySubBuffer = intelCopySubBuffer, - - //.InitScreen2 = intelInitScreen2, - //.HandleDrawableConfig = intelHandleDrawableConfig, - //.HandleBufferAttach = intelHandleBufferAttach, -}; diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.h b/src/gallium/winsys/drm/intel/dri/intel_screen.h deleted file mode 100644 index 0bb43a915c..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_screen.h +++ /dev/null @@ -1,122 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SCREEN_H_ -#define _INTEL_SCREEN_H_ - -#include "dri_util.h" -#include "i830_common.h" -#include "xmlconfig.h" -#include "ws_dri_bufpool.h" - -#include "pipe/p_compiler.h" - -#include "intel_be_device.h" - -struct intel_screen -{ - struct intel_be_device base; - - struct { - drm_handle_t handle; - - /* We create a static dri buffer for the frontbuffer. - */ - struct _DriBufferObject *buffer; - struct pipe_surface *surface; - struct pipe_texture *texture; - - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ - int width; - int height; - int size; - int cpp; /* for front and back buffers */ - } front; - - int deviceID; - int drmMinor; - - drmI830Sarea *sarea; - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; - - boolean havePools; - - /** - * Temporary(?) context to use for SwapBuffers or other situations in - * which we need a rendering context, but none is currently bound. - */ - struct intel_context *dummyContext; - - /* - * New stuff form the i915tex integration - */ - unsigned batch_id; - - - struct pipe_winsys *winsys; -}; - - - -/** cast wrapper */ -static INLINE struct intel_screen * -intel_screen(__DRIscreenPrivate *sPriv) -{ - return (struct intel_screen *) sPriv->private; -} - - -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); - - -extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); - -extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); - -extern boolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - - -extern boolean -intelCreatePools(__DRIscreenPrivate *sPriv); - -extern boolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - - -#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c deleted file mode 100644 index 34ad7eebe1..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c +++ /dev/null @@ -1,260 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_swapbuffers.h" - -#include "intel_reg.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" - -#include "ws_dri_bufmgr.h" -#include "intel_batchbuffer.h" - -/** - * Display a colorbuffer surface in an X window. - * Used for SwapBuffers and flushing front buffer rendering. - * - * \param dPriv the window/drawable to display into - * \param surf the surface to display - * \param rect optional subrect of surface to display (may be NULL). - */ -void -intelDisplaySurface(__DRIdrawablePrivate *dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t *rect) -{ - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_context *intel = intelScreen->dummyContext; - - DBG(SWAP, "%s\n", __FUNCTION__); - - if (!intel) { - /* XXX this is where some kind of extra/meta context could be useful */ - return; - } - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - * current context (which is what intelScreenContext should return) might - * not get a contended lock and thus cliprects not updated (tests/manywin) - */ - if (intel_context(dPriv->driContextPriv) != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - const int srcWidth = surf->width; - const int srcHeight = surf->height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; - const int cpp = intelScreen->front.cpp; - const int srcpitch = surf->stride / cpp; - int BR13, CMD; - int i; - - ASSERT(surf->buffer); - - DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, surf->stride); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) { - /* invalid cliprect, skip it */ - continue; - } - - box = *pbox; - - if (rect) { - /* intersect cliprect with user-provided src rect */ - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > srcWidth) - box.x2 = srcWidth + box.x1; - if (box.y2 - box.y1 > srcHeight) - box.y2 = srcHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - assert(box.x1 < box.x2); - assert(box.y1 < box.y2); - - /* XXX this could be done with pipe->surface_copy() */ - /* XXX should have its own batch buffer */ - if (!BEGIN_BATCH(8, 2)) { - /* - * Since we share this batch buffer with a context - * we can't flush it since that risks a GPU lockup - */ - assert(0); - continue; - } - - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(surf->buffer), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - } - - if (intel->first_swap_fence) - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); - } - - UNLOCK_HARDWARE(intel); - - if (intel->lastStamp != dPriv->lastStamp) { - intelUpdateWindowSize(dPriv); - intel->lastStamp = dPriv->lastStamp; - } -} - - - -/** - * This will be called whenever the currently bound window is moved/resized. - */ -void -intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); - assert(intelfb->stfb); - st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); -} - - - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, NULL); - st_notify_swapbuffers_complete(intel_fb->stfb); - } -} - - -/** - * Called via glXCopySubBufferMESA() to copy a subrect of the back - * buffer to the front buffer/screen. - */ -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - drm_clip_rect_t rect; - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, &rect); - } -} diff --git a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h deleted file mode 100644 index 46c9bab3af..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_SWAPBUFFERS_H -#define INTEL_SWAPBUFFERS_H - - -struct pipe_surface; - - -extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t * rect); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h); - -extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); - - -#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c deleted file mode 100644 index 20920a2052..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c +++ /dev/null @@ -1,82 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com> - */ - -#include "intel_context.h" -#include "intel_winsys_softpipe.h" -#include "pipe/p_defines.h" -#include "pipe/p_format.h" -#include "util/u_memory.h" -#include "softpipe/sp_winsys.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - -/** - * Return list of surface formats supported by this driver. - */ -static boolean -intel_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - switch(format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - } -} - - -/** - * Create rendering context which uses software rendering. - */ -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - struct pipe_screen *screen = softpipe_create_screen(winsys); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - */ - isws->sws.is_format_supported = intel_is_format_supported; - isws->intel = intel; - - /* Create the softpipe context: - */ - return softpipe_create( screen, winsys, &isws->sws ); -} diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h deleted file mode 100644 index 5fa14cb749..0000000000 --- a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h +++ /dev/null @@ -1,39 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_SOFTPIPE_H -#define INTEL_SOFTPIPE_H - -struct pipe_winsys; -struct pipe_context; -struct intel_context; - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ); - -#endif diff --git a/src/gallium/winsys/drm/intel/dri/server/i830_common.h b/src/gallium/winsys/drm/intel/dri/server/i830_common.h deleted file mode 100644 index 3452ddb3c9..0000000000 --- a/src/gallium/winsys/drm/intel/dri/server/i830_common.h +++ /dev/null @@ -1,255 +0,0 @@ -/************************************************************************** - -Copyright 2001 VA Linux Systems Inc., Fremont, California. -Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the "Software"), -to deal in the Software without restriction, including without limitation -on the rights to use, copy, modify, merge, publish, distribute, sub -license, and/or sell copies of the Software, and to permit persons to whom -the Software is furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice (including the next -paragraph) shall be included in all copies or substantial portions of the -Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL -ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, -DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR -OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE -USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - - -#ifndef _I830_COMMON_H_ -#define _I830_COMMON_H_ - - -#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ -#define I830_LOG_MIN_TEX_REGION_SIZE 14 - - -/* Driver specific DRM command indices - * NOTE: these are not OS specific, but they are driver specific - */ -#define DRM_I830_INIT 0x00 -#define DRM_I830_FLUSH 0x01 -#define DRM_I830_FLIP 0x02 -#define DRM_I830_BATCHBUFFER 0x03 -#define DRM_I830_IRQ_EMIT 0x04 -#define DRM_I830_IRQ_WAIT 0x05 -#define DRM_I830_GETPARAM 0x06 -#define DRM_I830_SETPARAM 0x07 -#define DRM_I830_ALLOC 0x08 -#define DRM_I830_FREE 0x09 -#define DRM_I830_INIT_HEAP 0x0a -#define DRM_I830_CMDBUFFER 0x0b -#define DRM_I830_DESTROY_HEAP 0x0c -#define DRM_I830_SET_VBLANK_PIPE 0x0d -#define DRM_I830_GET_VBLANK_PIPE 0x0e -#define DRM_I830_MMIO 0x10 - -typedef struct { - enum { - I830_INIT_DMA = 0x01, - I830_CLEANUP_DMA = 0x02, - I830_RESUME_DMA = 0x03 - } func; - unsigned int mmio_offset; - int sarea_priv_offset; - unsigned int ring_start; - unsigned int ring_end; - unsigned int ring_size; - unsigned int front_offset; - unsigned int back_offset; - unsigned int depth_offset; - unsigned int w; - unsigned int h; - unsigned int pitch; - unsigned int pitch_bits; - unsigned int back_pitch; - unsigned int depth_pitch; - unsigned int cpp; - unsigned int chipset; -} drmI830Init; - -typedef struct { - drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; - int last_upload; /* last time texture was uploaded */ - int last_enqueue; /* last time a buffer was enqueued */ - int last_dispatch; /* age of the most recently dispatched buffer */ - int ctxOwner; /* last context to upload state */ - /** Last context that used the buffer manager. */ - int texAge; - int pf_enabled; /* is pageflipping allowed? */ - int pf_active; - int pf_current_page; /* which buffer is being displayed? */ - int perf_boxes; /* performance boxes to be displayed */ - int width, height; /* screen size in pixels */ - - drm_handle_t front_handle; - int front_offset; - int front_size; - - drm_handle_t back_handle; - int back_offset; - int back_size; - - drm_handle_t depth_handle; - int depth_offset; - int depth_size; - - drm_handle_t tex_handle; - int tex_offset; - int tex_size; - int log_tex_granularity; - int pitch; - int rotation; /* 0, 90, 180 or 270 */ - int rotated_offset; - int rotated_size; - int rotated_pitch; - int virtualX, virtualY; - - unsigned int front_tiled; - unsigned int back_tiled; - unsigned int depth_tiled; - unsigned int rotated_tiled; - unsigned int rotated2_tiled; - - int planeA_x; - int planeA_y; - int planeA_w; - int planeA_h; - int planeB_x; - int planeB_y; - int planeB_w; - int planeB_h; - - /* Triple buffering */ - drm_handle_t third_handle; - int third_offset; - int third_size; - unsigned int third_tiled; - - /* buffer object handles for the static buffers. May change - * over the lifetime of the client, though it doesn't in our current - * implementation. - */ - unsigned int front_bo_handle; - unsigned int back_bo_handle; - unsigned int third_bo_handle; - unsigned int depth_bo_handle; -} drmI830Sarea; - -/* Flags for perf_boxes - */ -#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ -#define I830_BOX_FLIP 0x2 /* populated by kernel */ -#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ -#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ -#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ - - -typedef struct { - int start; /* agp offset */ - int used; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830BatchBuffer; - -typedef struct { - char *buf; /* agp offset */ - int sz; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830CmdBuffer; - -typedef struct { - int *irq_seq; -} drmI830IrqEmit; - -typedef struct { - int irq_seq; -} drmI830IrqWait; - -typedef struct { - int param; - int *value; -} drmI830GetParam; - -#define I830_PARAM_IRQ_ACTIVE 1 -#define I830_PARAM_ALLOW_BATCHBUFFER 2 - -typedef struct { - int param; - int value; -} drmI830SetParam; - -#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 -#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 -#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 - - -/* A memory manager for regions of shared memory: - */ -#define I830_MEM_REGION_AGP 1 - -typedef struct { - int region; - int alignment; - int size; - int *region_offset; /* offset from start of fb or agp */ -} drmI830MemAlloc; - -typedef struct { - int region; - int region_offset; -} drmI830MemFree; - -typedef struct { - int region; - int size; - int start; -} drmI830MemInitHeap; - -typedef struct { - int region; -} drmI830MemDestroyHeap; - -#define DRM_I830_VBLANK_PIPE_A 1 -#define DRM_I830_VBLANK_PIPE_B 2 - -typedef struct { - int pipe; -} drmI830VBlankPipe; - -#define MMIO_READ 0 -#define MMIO_WRITE 1 - -#define MMIO_REGS_IA_PRIMATIVES_COUNT 0 -#define MMIO_REGS_IA_VERTICES_COUNT 1 -#define MMIO_REGS_VS_INVOCATION_COUNT 2 -#define MMIO_REGS_GS_PRIMITIVES_COUNT 3 -#define MMIO_REGS_GS_INVOCATION_COUNT 4 -#define MMIO_REGS_CL_PRIMITIVES_COUNT 5 -#define MMIO_REGS_CL_INVOCATION_COUNT 6 -#define MMIO_REGS_PS_INVOCATION_COUNT 7 -#define MMIO_REGS_PS_DEPTH_COUNT 8 - -typedef struct { - unsigned int read_write:1; - unsigned int reg:31; - void __user *data; -} drmI830MMIO; - -#endif /* _I830_DRM_H_ */ diff --git a/src/gallium/winsys/drm/intel/dri/server/i830_dri.h b/src/gallium/winsys/drm/intel/dri/server/i830_dri.h deleted file mode 100644 index 0d514b6c38..0000000000 --- a/src/gallium/winsys/drm/intel/dri/server/i830_dri.h +++ /dev/null @@ -1,62 +0,0 @@ - -#ifndef _I830_DRI_H -#define _I830_DRI_H - -#include "xf86drm.h" -#include "i830_common.h" - -#define I830_MAX_DRAWABLES 256 - -#define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 7 -#define I830_PATCHLEVEL 2 - -#define I830_REG_SIZE 0x80000 - -typedef struct _I830DRIRec { - drm_handle_t regs; - drmSize regsSize; - - drmSize unused1; /* backbufferSize */ - drm_handle_t unused2; /* backbuffer */ - - drmSize unused3; /* depthbufferSize */ - drm_handle_t unused4; /* depthbuffer */ - - drmSize unused5; /* rotatedSize */ - drm_handle_t unused6; /* rotatedbuffer */ - - drm_handle_t unused7; /* textures */ - int unused8; /* textureSize */ - - drm_handle_t unused9; /* agp_buffers */ - drmSize unused10; /* agp_buf_size */ - - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - - int unused12; /* logTextureGranularity */ - int unused13; /* textureOffset */ - - int irq; - int sarea_priv_offset; -} I830DRIRec, *I830DRIPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830ConfigPrivRec, *I830ConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830DRIContextRec, *I830DRIContextPtr; - - -#endif diff --git a/src/gallium/winsys/drm/intel/dri2/Makefile b/src/gallium/winsys/drm/intel/dri2/Makefile new file mode 100644 index 0000000000..1a02109274 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri2/Makefile @@ -0,0 +1,22 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = i915_dri.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \ + $(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a + + +DRIVER_SOURCES = + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +include ../../Makefile.template + +DRI_LIB_DEPS += -ldrm_intel + +symlinks: diff --git a/src/gallium/winsys/drm/intel/egl/Makefile b/src/gallium/winsys/drm/intel/egl/Makefile index 7147d89e0d..c5217ad2d6 100644 --- a/src/gallium/winsys/drm/intel/egl/Makefile +++ b/src/gallium/winsys/drm/intel/egl/Makefile @@ -1,19 +1,16 @@ TOP = ../../../../../.. +GALLIUMDIR = ../../../.. include $(TOP)/configs/current LIBNAME = EGL_i915.so PIPE_DRIVERS = \ - ../gem/libinteldrm.a \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ $(TOP)/src/gallium/state_trackers/egl/libegldrm.a \ + $(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a - -DRIVER_SOURCES = \ - intel_context.c \ - intel_device.c \ - intel_api.c +DRIVER_SOURCES = C_SOURCES = \ $(COMMON_GALLIUM_SOURCES) \ diff --git a/src/gallium/winsys/drm/intel/egl/intel_api.c b/src/gallium/winsys/drm/intel/egl/intel_api.c deleted file mode 100644 index 5dc4a7b052..0000000000 --- a/src/gallium/winsys/drm/intel/egl/intel_api.c +++ /dev/null @@ -1,10 +0,0 @@ - -#include "intel_api.h" - -struct drm_api drm_api_hocks = -{ - .create_screen = intel_create_screen, - .create_context = intel_create_context, - .buffer_from_handle = intel_be_buffer_from_handle, - .handle_from_buffer = intel_be_handle_from_buffer, -}; diff --git a/src/gallium/winsys/drm/intel/egl/intel_api.h b/src/gallium/winsys/drm/intel/egl/intel_api.h deleted file mode 100644 index 8ec165ab01..0000000000 --- a/src/gallium/winsys/drm/intel/egl/intel_api.h +++ /dev/null @@ -1,14 +0,0 @@ - -#ifndef _INTEL_API_H_ -#define _INTEL_API_H_ - -#include "pipe/p_compiler.h" - -#include "state_tracker/drm_api.h" - -#include "intel_be_device.h" - -struct pipe_screen *intel_create_screen(int drmFD, int pciID); -struct pipe_context *intel_create_context(struct pipe_screen *screen); - -#endif diff --git a/src/gallium/winsys/drm/intel/egl/intel_context.c b/src/gallium/winsys/drm/intel/egl/intel_context.c deleted file mode 100644 index 57e5ff7bc1..0000000000 --- a/src/gallium/winsys/drm/intel/egl/intel_context.c +++ /dev/null @@ -1,83 +0,0 @@ - -#include "i915simple/i915_screen.h" - -#include "intel_be_device.h" -#include "intel_be_context.h" - -#include "pipe/p_defines.h" -#include "pipe/p_context.h" - -#include "intel_api.h" - -struct intel_context -{ - struct intel_be_context base; - - /* stuff */ -}; - -/* - * Hardware lock functions. - * Doesn't do anything in EGL - */ - -static void -intel_lock_hardware(struct intel_be_context *context) -{ - (void)context; -} - -static void -intel_unlock_hardware(struct intel_be_context *context) -{ - (void)context; -} - -static boolean -intel_locked_hardware(struct intel_be_context *context) -{ - (void)context; - return FALSE; -} - - -/* - * Misc functions. - */ -static void -intel_destroy_be_context(struct i915_winsys *winsys) -{ - struct intel_context *intel = (struct intel_context *)winsys; - - intel_be_destroy_context(&intel->base); - free(intel); -} - -struct pipe_context * -intel_create_context(struct pipe_screen *screen) -{ - struct intel_context *intel; - struct pipe_context *pipe; - struct intel_be_device *device = (struct intel_be_device *)screen->winsys; - - intel = (struct intel_context *)malloc(sizeof(*intel)); - memset(intel, 0, sizeof(*intel)); - - intel->base.hardware_lock = intel_lock_hardware; - intel->base.hardware_unlock = intel_unlock_hardware; - intel->base.hardware_locked = intel_locked_hardware; - - intel_be_init_context(&intel->base, device); - - intel->base.base.destroy = intel_destroy_be_context; - -#if 0 - pipe = intel_create_softpipe(intel, screen->winsys); -#else - pipe = i915_create_context(screen, &device->base, &intel->base.base); -#endif - - pipe->priv = intel; - - return pipe; -} diff --git a/src/gallium/winsys/drm/intel/egl/intel_device.c b/src/gallium/winsys/drm/intel/egl/intel_device.c deleted file mode 100644 index 6b281402d5..0000000000 --- a/src/gallium/winsys/drm/intel/egl/intel_device.c +++ /dev/null @@ -1,48 +0,0 @@ - -#include <stdio.h> -#include "pipe/p_defines.h" -#include "intel_be_device.h" -#include "i915simple/i915_screen.h" - -#include "intel_api.h" - -struct intel_device -{ - struct intel_be_device base; - - int deviceID; -}; - -static void -intel_destroy_winsys(struct pipe_winsys *winsys) -{ - struct intel_device *dev = (struct intel_device *)winsys; - - intel_be_destroy_device(&dev->base); - - free(dev); -} - -struct pipe_screen * -intel_create_screen(int drmFD, int deviceID) -{ - struct intel_device *dev; - struct pipe_screen *screen; - - /* Allocate the private area */ - dev = malloc(sizeof(*dev)); - if (!dev) - return NULL; - memset(dev, 0, sizeof(*dev)); - - dev->deviceID = deviceID; - - intel_be_init_device(&dev->base, drmFD, deviceID); - - /* we need to hock our own destroy function in here */ - dev->base.base.destroy = intel_destroy_winsys; - - screen = i915_create_screen(&dev->base.base, deviceID); - - return screen; -} diff --git a/src/gallium/winsys/drm/intel/gem/Makefile b/src/gallium/winsys/drm/intel/gem/Makefile index b25fc258f4..7ab1a2a771 100644 --- a/src/gallium/winsys/drm/intel/gem/Makefile +++ b/src/gallium/winsys/drm/intel/gem/Makefile @@ -6,13 +6,11 @@ LIBNAME = inteldrm C_SOURCES = \ intel_be_batchbuffer.c \ intel_be_context.c \ - intel_be_device.c + intel_be_device.c \ + intel_be_api.c +LIBRARY_INCLUDES = $(shell pkg-config libdrm --cflags-only-I) -include ./Makefile.template - -DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \ - && pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") -symlinks: +LIBRARY_DEFINES = $(shell pkg-config libdrm --cflags-only-other) +include ../../../../Makefile.template diff --git a/src/gallium/winsys/drm/intel/gem/Makefile.template b/src/gallium/winsys/drm/intel/gem/Makefile.template deleted file mode 100644 index b60e978894..0000000000 --- a/src/gallium/winsys/drm/intel/gem/Makefile.template +++ /dev/null @@ -1,64 +0,0 @@ -# -*-makefile-*- - - -# We still have a dependency on the "dri" buffer manager. Most likely -# the interface can be reused in non-dri environments, and also as a -# frontend to simpler memory managers. -# -COMMON_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/include \ - $(DRIVER_INCLUDES) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks lib$(LIBNAME).a - - -lib$(LIBNAME).a: $(OBJECTS) Makefile Makefile.template - $(TOP)/bin/mklib -o $(LIBNAME) -static $(OBJECTS) $(DRIVER_LIBS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean:: - -rm -f *.o */*.o *~ *.so *.a *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -include depend diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.c b/src/gallium/winsys/drm/intel/gem/intel_be_api.c new file mode 100644 index 0000000000..f4ef7c2d88 --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.c @@ -0,0 +1,15 @@ + +#include "intel_be_api.h" +#include "i915simple/i915_winsys.h" + +struct drm_api drm_api_hooks = +{ + /* intel_be_context.c */ + .create_context = intel_be_create_context, + /* intel_be_device.c */ + .create_screen = intel_be_create_screen, + .buffer_from_texture = i915_get_texture_buffer, + .buffer_from_handle = intel_be_buffer_from_handle, + .handle_from_buffer = intel_be_handle_from_buffer, + .global_handle_from_buffer = intel_be_global_handle_from_buffer, +}; diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h new file mode 100644 index 0000000000..73e458d4ba --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.h @@ -0,0 +1,14 @@ + +#ifndef _INTEL_BE_API_H_ +#define _INTEL_BE_API_H_ + +#include "pipe/p_compiler.h" + +#include "state_tracker/drm_api.h" + +#include "intel_be_device.h" + +struct pipe_screen *intel_be_create_screen(int drmFD, int pciID); +struct pipe_context *intel_be_create_context(struct pipe_screen *screen); + +#endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c index d9556e1f38..d5e63c3bae 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c @@ -21,7 +21,7 @@ intel_be_batchbuffer_alloc(struct intel_be_context *intel) batch->base.size = 0; batch->base.actual_size = intel->device->max_batch_size; batch->base.relocs = 0; - batch->base.max_relocs = INTEL_DEFAULT_RELOCS; + batch->base.max_relocs = 500;/*INTEL_DEFAULT_RELOCS;*/ batch->base.map = malloc(batch->base.actual_size); memset(batch->base.map, 0, batch->base.actual_size); @@ -47,7 +47,6 @@ intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) batch->base.size = batch->base.actual_size - BATCH_RESERVED; batch->base.relocs = 0; - batch->base.max_relocs = INTEL_DEFAULT_RELOCS; batch->bo = drm_intel_bo_alloc(dev->pools.gem, "gallium3d_batch_buffer", @@ -115,10 +114,10 @@ intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch, if (fence) { if (*fence) - intel_be_fence_unreference(*fence); + intel_be_fence_reference(fence, NULL); (*fence) = CALLOC_STRUCT(intel_be_fence); - (*fence)->refcount = 1; + pipe_reference_init(&(*fence)->reference, 1); (*fence)->bo = NULL; } } diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c index 95e761d78d..bb6f1b916c 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c @@ -1,10 +1,14 @@ +#include "pipe/p_screen.h" + #include "intel_be_device.h" #include "intel_be_context.h" #include "intel_be_batchbuffer.h" #include "i915_drm.h" +#include "intel_be_api.h" + static struct i915_batchbuffer * intel_be_batch_get(struct i915_winsys *sws) { @@ -57,6 +61,21 @@ intel_be_batch_flush(struct i915_winsys *sws, intel_be_batchbuffer_flush(intel->batch, f); } + +/* + * Misc functions. + */ + +static void +intel_be_destroy_context(struct i915_winsys *winsys) +{ + struct intel_be_context *intel = intel_be_context(winsys); + + intel_be_batchbuffer_free(intel->batch); + + free(intel); +} + boolean intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device) { @@ -68,13 +87,32 @@ intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *de intel->base.batch_reloc = intel_be_batch_reloc; intel->base.batch_flush = intel_be_batch_flush; + intel->base.destroy = intel_be_destroy_context; + intel->batch = intel_be_batchbuffer_alloc(intel); return true; } -void -intel_be_destroy_context(struct intel_be_context *intel) +struct pipe_context * +intel_be_create_context(struct pipe_screen *screen) { - intel_be_batchbuffer_free(intel->batch); + struct intel_be_context *intel; + struct pipe_context *pipe; + struct intel_be_device *device = intel_be_device(screen->winsys); + + intel = (struct intel_be_context *)malloc(sizeof(*intel)); + memset(intel, 0, sizeof(*intel)); + + intel_be_init_context(intel, device); + +#if 0 + pipe = intel_create_softpipe(intel, screen->winsys); +#else + pipe = i915_create_context(screen, &device->base, &intel->base); +#endif + + pipe->priv = intel; + + return pipe; } diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.h b/src/gallium/winsys/drm/intel/gem/intel_be_context.h index 9cee1a4e52..5a369669c0 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.h @@ -11,15 +11,6 @@ struct intel_be_context struct intel_be_device *device; struct intel_be_batchbuffer *batch; - - /* - * Hardware lock functions. - * - * Needs to be filled in by the winsys. - */ - void (*hardware_lock)(struct intel_be_context *context); - void (*hardware_unlock)(struct intel_be_context *context); - boolean (*hardware_locked)(struct intel_be_context *context); }; static INLINE struct intel_be_context * @@ -37,12 +28,4 @@ boolean intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device); -/** - * Destroy a intel_be_context. - * - * Does not free the struct that is up to the winsys. - */ -void -intel_be_destroy_context(struct intel_be_context *intel); - #endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c index 82c1cb2f32..0f6300323b 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.c @@ -9,17 +9,9 @@ #include "intel_be_fence.h" -#include "i915simple/i915_screen.h" +#include "i915simple/i915_winsys.h" - -/** - * Turn a pipe winsys into an intel/pipe winsys: - */ -static INLINE struct intel_be_device * -intel_be_device(struct pipe_winsys *winsys) -{ - return (struct intel_be_device *)winsys; -} +#include "intel_be_api.h" /* * Buffer @@ -34,6 +26,12 @@ intel_be_buffer_map(struct pipe_winsys *winsys, int write = 0; int ret; + if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { + /* Remove this when drm_intel_bo_map supports DONTBLOCK + */ + return NULL; + } + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) write = 1; @@ -53,8 +51,7 @@ intel_be_buffer_unmap(struct pipe_winsys *winsys, } static void -intel_be_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) +intel_be_buffer_destroy(struct pipe_buffer *buf) { drm_intel_bo_unreference(intel_bo(buf)); free(buf); @@ -74,10 +71,12 @@ intel_be_buffer_create(struct pipe_winsys *winsys, if (!buffer) return NULL; - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; + buffer->flinked = FALSE; + buffer->flink = 0; if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { /* Local buffer */ @@ -115,7 +114,7 @@ intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned byte if (!buffer) return NULL; - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.alignment = 0; buffer->base.usage = 0; buffer->base.size = bytes; @@ -141,10 +140,10 @@ err: } struct pipe_buffer * -intel_be_buffer_from_handle(struct pipe_winsys *winsys, +intel_be_buffer_from_handle(struct pipe_screen *screen, const char* name, unsigned handle) { - struct intel_be_device *dev = intel_be_device(winsys); + struct intel_be_device *dev = intel_be_device(screen->winsys); struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer); if (!buffer) @@ -155,7 +154,8 @@ intel_be_buffer_from_handle(struct pipe_winsys *winsys, if (!buffer->bo) goto err; - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); + buffer->base.screen = screen; buffer->base.alignment = buffer->bo->align; buffer->base.usage = PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE | @@ -170,14 +170,39 @@ err: return NULL; } -unsigned -intel_be_handle_from_buffer(struct pipe_winsys *winsys, - struct pipe_buffer *buf) +boolean +intel_be_handle_from_buffer(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle) { - drm_intel_bo *bo = intel_bo(buf); - return bo->handle; + drm_intel_bo *bo; + + if (!buffer) + return FALSE; + + *handle = intel_bo(buffer)->handle; + return TRUE; } +boolean +intel_be_global_handle_from_buffer(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle) +{ + struct intel_be_buffer *buf = intel_be_buffer(buffer); + + if (!buffer) + return FALSE; + + if (!buf->flinked) { + if (drm_intel_bo_flink(intel_bo(buffer), &buf->flink)) + return FALSE; + buf->flinked = TRUE; + } + + *handle = buf->flink; + return TRUE; +} /* * Fence */ @@ -190,15 +215,7 @@ intel_be_fence_refunref(struct pipe_winsys *sws, struct intel_be_fence **p = (struct intel_be_fence **)ptr; struct intel_be_fence *f = (struct intel_be_fence *)fence; - assert(p); - - if (f) - intel_be_fence_reference(f); - - if (*p) - intel_be_fence_unreference(*p); - - *p = f; + intel_be_fence_reference(p, f); } static int @@ -233,10 +250,21 @@ intel_be_fence_finish(struct pipe_winsys *sws, * Misc functions */ +static void +intel_be_destroy_winsys(struct pipe_winsys *winsys) +{ + struct intel_be_device *dev = intel_be_device(winsys); + + drm_intel_bufmgr_destroy(dev->pools.gem); + + free(dev); +} + boolean intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) { dev->fd = fd; + dev->id = id; dev->max_batch_size = 16 * 4096; dev->max_vertex_size = 128 * 4096; @@ -253,13 +281,28 @@ intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) dev->base.fence_signalled = intel_be_fence_signalled; dev->base.fence_finish = intel_be_fence_finish; + dev->base.destroy = intel_be_destroy_winsys; + dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size); return true; } -void -intel_be_destroy_device(struct intel_be_device *dev) +struct pipe_screen * +intel_be_create_screen(int drmFD, int deviceID) { - drm_intel_bufmgr_destroy(dev->pools.gem); + struct intel_be_device *dev; + struct pipe_screen *screen; + + /* Allocate the private area */ + dev = malloc(sizeof(*dev)); + if (!dev) + return NULL; + memset(dev, 0, sizeof(*dev)); + + intel_be_init_device(dev, drmFD, deviceID); + + screen = i915_create_screen(&dev->base, deviceID); + + return screen; } diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h index f06890163c..47d2176cb4 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.h @@ -18,6 +18,8 @@ struct intel_be_device int fd; /**< Drm file discriptor */ + unsigned id; + size_t max_batch_size; size_t max_vertex_size; @@ -26,12 +28,15 @@ struct intel_be_device } pools; }; +static INLINE struct intel_be_device * +intel_be_device(struct pipe_winsys *winsys) +{ + return (struct intel_be_device *)winsys; +} + boolean intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); -void -intel_be_destroy_device(struct intel_be_device *dev); - /* * Buffer */ @@ -39,6 +44,8 @@ intel_be_destroy_device(struct intel_be_device *dev); struct intel_be_buffer { struct pipe_buffer base; drm_intel_bo *bo; + boolean flinked; + unsigned flink; }; /** @@ -47,7 +54,7 @@ struct intel_be_buffer { * Takes a reference. */ struct pipe_buffer * -intel_be_buffer_from_handle(struct pipe_winsys *winsys, +intel_be_buffer_from_handle(struct pipe_screen *screen, const char* name, unsigned handle); /** @@ -55,9 +62,20 @@ intel_be_buffer_from_handle(struct pipe_winsys *winsys, * * If buffer is destroyed handle may become invalid. */ -unsigned -intel_be_handle_from_buffer(struct pipe_winsys *winsys, - struct pipe_buffer *buffer); +boolean +intel_be_handle_from_buffer(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle); + +/** + * Gets the global handle from a buffer. + * + * If buffer is destroyed handle may become invalid. + */ +boolean +intel_be_global_handle_from_buffer(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle); static INLINE struct intel_be_buffer * intel_be_buffer(struct pipe_buffer *buf) diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h b/src/gallium/winsys/drm/intel/gem/intel_be_fence.h index 0fe18f66f8..a8abb01a9e 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_fence.h @@ -15,23 +15,19 @@ */ struct intel_be_fence { - uint32_t refcount; + struct pipe_reference reference; drm_intel_bo *bo; }; static INLINE void -intel_be_fence_reference(struct intel_be_fence *f) +intel_be_fence_reference(struct intel_be_fence **ptr, struct intel_be_fence *f) { - f->refcount++; -} + struct intel_be_fence *old_fence = *ptr; -static INLINE void -intel_be_fence_unreference(struct intel_be_fence *f) -{ - if (!--f->refcount) { - if (f->bo) - drm_intel_bo_unreference(f->bo); - free(f); + if (pipe_reference((struct pipe_reference**)ptr, &f->reference)) { + if (old_fence->bo) + drm_intel_bo_unreference(old_fence->bo); + free(old_fence); } } diff --git a/src/gallium/winsys/drm/intel/xorg/Makefile b/src/gallium/winsys/drm/intel/xorg/Makefile new file mode 100644 index 0000000000..a45ca570db --- /dev/null +++ b/src/gallium/winsys/drm/intel/xorg/Makefile @@ -0,0 +1,43 @@ +TARGET = modesetting_drv.so +CFILES = $(wildcard ./*.c) +OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) +GALLIUMDIR = ../../../.. +TOP = ../../../../../.. + +include ${TOP}/configs/current + +CFLAGS = -DHAVE_CONFIG_H \ + -g -Wall -Wimplicit-function-declaration -fPIC \ + $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \ + -I../gem \ + -I${GALLIUMDIR}/include \ + -I${GALLIUMDIR}/drivers \ + -I${GALLIUMDIR}/auxiliary \ + -I${TOP}/src/mesa \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main + +LIBS = \ + $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \ + $(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ + $(GALLIUM_AUXILIARIES) + +############################################# + + + +all default: $(TARGET) + +$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a + $(TOP)/bin/mklib -noprefix -o $@ \ + $(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel + +clean: + rm -rf $(OBJECTS) $(TARGET) + +install: + $(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + $(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + +.PHONY = all clean install diff --git a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c new file mode 100644 index 0000000000..aea39247e5 --- /dev/null +++ b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c @@ -0,0 +1,156 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * + */ + +#include "../../../../state_trackers/xorg/xorg_winsys.h" + +static void intel_xorg_identify(int flags); +static Bool intel_xorg_pci_probe(DriverPtr driver, + int entity_num, + struct pci_device *device, + intptr_t match_data); + +static const struct pci_id_match intel_xorg_device_match[] = { + {0x8086, 0x2592, 0xffff, 0xffff, 0, 0, 0}, + {0x8086, 0x27A2, 0xffff, 0xffff, 0, 0, 0}, + {0, 0, 0}, +}; + +static SymTabRec intel_xorg_chipsets[] = { + {0x2592, "Intel Graphics Device"}, + {0x27A2, "Intel Graphics Device"}, + {-1, NULL} +}; + +static PciChipsets intel_xorg_pci_devices[] = { + {0x2592, 0x2592, RES_SHARED_VGA}, + {0x27A2, 0x27A2, RES_SHARED_VGA}, + {-1, -1, RES_UNDEFINED} +}; + +static XF86ModuleVersionInfo intel_xorg_version = { + "modesetting", + MODULEVENDORSTRING, + MODINFOSTRING1, + MODINFOSTRING2, + XORG_VERSION_CURRENT, + 0, 1, 0, /* major, minor, patch */ + ABI_CLASS_VIDEODRV, + ABI_VIDEODRV_VERSION, + MOD_CLASS_VIDEODRV, + {0, 0, 0, 0} +}; + +/* + * Xorg driver exported structures + */ + +_X_EXPORT DriverRec modesetting = { + 1, + "modesetting", + intel_xorg_identify, + NULL, + xorg_tracker_available_options, + NULL, + 0, + NULL, + intel_xorg_device_match, + intel_xorg_pci_probe +}; + +static MODULESETUPPROTO(intel_xorg_setup); + +_X_EXPORT XF86ModuleData modesettingModuleData = { + &intel_xorg_version, + intel_xorg_setup, + NULL +}; + +/* + * Xorg driver functions + */ + +static pointer +intel_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin) +{ + static Bool setupDone = 0; + + /* This module should be loaded only once, but check to be sure. + */ + if (!setupDone) { + setupDone = 1; + xf86AddDriver(&modesetting, module, HaveDriverFuncs); + + /* + * Tell the loader about symbols from other modules that this module + * might refer to. + */ + xorg_tracker_loader_ref_sym_lists(); + + /* + * The return value must be non-NULL on success even though there + * is no TearDownProc. + */ + return (pointer) 1; + } else { + if (errmaj) + *errmaj = LDR_ONCEONLY; + return NULL; + } +} + +static void +intel_xorg_identify(int flags) +{ + xf86PrintChipsets("modesetting", "Driver for Modesetting Kernel Drivers", + intel_xorg_chipsets); +} + +static Bool +intel_xorg_pci_probe(DriverPtr driver, + int entity_num, struct pci_device *device, intptr_t match_data) +{ + ScrnInfoPtr scrn = NULL; + EntityInfoPtr entity; + + scrn = xf86ConfigPciEntity(scrn, 0, entity_num, intel_xorg_pci_devices, + NULL, NULL, NULL, NULL, NULL); + if (scrn != NULL) { + scrn->driverVersion = 1; + scrn->driverName = "modesetting"; + scrn->name = "modesetting"; + scrn->Probe = NULL; + + entity = xf86GetEntityInfo(entity_num); + + /* Use all the functions from the xorg tracker */ + xorg_tracker_set_functions(scrn); + } + return scrn != NULL; +} diff --git a/src/gallium/winsys/drm/nouveau/Makefile b/src/gallium/winsys/drm/nouveau/Makefile index b5735329ec..f8c8135854 100644 --- a/src/gallium/winsys/drm/nouveau/Makefile +++ b/src/gallium/winsys/drm/nouveau/Makefile @@ -1,25 +1,12 @@ +# src/gallium/winsys/drm/nouveau/Makefile TOP = ../../../../.. include $(TOP)/configs/current +SUBDIRS = drm dri dri2 -SUBDIRS = common dri - - -default: subdirs - - -subdirs: +default install clean: @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ + (cd $$dir && $(MAKE) $@) || exit 1; \ fi \ done - - -clean: - rm -f `find . -name \*.[oa]` - rm -f `find . -name depend` - - -# Dummy install target -install: diff --git a/src/gallium/winsys/drm/nouveau/common/Makefile b/src/gallium/winsys/drm/nouveau/common/Makefile deleted file mode 100644 index c6dd6dd7f9..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/Makefile +++ /dev/null @@ -1,22 +0,0 @@ -TOP = ../../../../../.. -include $(TOP)/configs/current - -LIBNAME = nouveaudrm - -C_SOURCES = \ - nouveau_context.c \ - nouveau_lock.c \ - nouveau_screen.c \ - nouveau_winsys.c \ - nouveau_winsys_pipe.c \ - nouveau_winsys_softpipe.c - -include ./Makefile.template - -DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \ - && pkg-config libdrm --atleast-version=2.3.1 \ - && pkg-config libdrm_nouveau --exact-version=0.5 \ - && pkg-config libdrm_nouveau --cflags \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") -symlinks: - diff --git a/src/gallium/winsys/drm/nouveau/common/Makefile.template b/src/gallium/winsys/drm/nouveau/common/Makefile.template deleted file mode 100644 index f0d098bb1c..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/Makefile.template +++ /dev/null @@ -1,59 +0,0 @@ -# -*-makefile-*- - -COMMON_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/include \ - $(DRIVER_INCLUDES) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks lib$(LIBNAME).a - - -lib$(LIBNAME).a: $(OBJECTS) Makefile Makefile.template - $(TOP)/bin/mklib -o $(LIBNAME) -static $(OBJECTS) $(DRIVER_LIBS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean:: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -include depend diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_context.c b/src/gallium/winsys/drm/nouveau/common/nouveau_context.c deleted file mode 100644 index d6ae0827cd..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_context.c +++ /dev/null @@ -1,206 +0,0 @@ -#include <pipe/p_defines.h> -#include <pipe/p_context.h> -#include <pipe/p_screen.h> -#include <util/u_memory.h> -#include "nouveau_context.h" -#include "nouveau_dri.h" -#include "nouveau_local.h" -#include "nouveau_screen.h" -#include "nouveau_winsys_pipe.h" - -static void -nouveau_channel_context_destroy(struct nouveau_channel_context *nvc) -{ - nouveau_channel_free(&nvc->channel); - - FREE(nvc); -} - -static struct nouveau_channel_context * -nouveau_channel_context_create(struct nouveau_device *dev) -{ - struct nouveau_channel_context *nvc; - int ret; - - nvc = CALLOC_STRUCT(nouveau_channel_context); - if (!nvc) - return NULL; - - if ((ret = nouveau_channel_alloc(dev, 0x8003d001, 0x8003d002, - &nvc->channel))) { - NOUVEAU_ERR("Error creating GPU channel: %d\n", ret); - nouveau_channel_context_destroy(nvc); - return NULL; - } - - nvc->next_handle = 0x77000000; - return nvc; -} - -int -nouveau_context_init(struct nouveau_screen *nv_screen, - drm_context_t hHWContext, drmLock *sarea_lock, - struct nouveau_context *nv_share, - struct nouveau_context *nv) -{ - struct pipe_context *pipe = NULL; - struct nouveau_channel_context *nvc = NULL; - struct nouveau_device *dev = nv_screen->device; - int i; - - switch (dev->chipset & 0xf0) { - case 0x00: - /* NV04 */ - case 0x10: - case 0x20: - /* NV10 */ - case 0x30: - /* NV30 */ - case 0x40: - case 0x60: - /* NV40 */ - case 0x50: - case 0x80: - case 0x90: - /* G80 */ - break; - default: - NOUVEAU_ERR("Unsupported chipset: NV%02x\n", dev->chipset); - return 1; - } - - nv->nv_screen = nv_screen; - - { - struct nouveau_device_priv *nvdev = nouveau_device(dev); - - nvdev->ctx = hHWContext; - nvdev->lock = sarea_lock; - } - - /* Attempt to share a single channel between multiple contexts from - * a single process. - */ - nvc = nv_screen->nvc; - if (!nvc && nv_share) - nvc = nv_share->nvc; - - /*XXX: temporary - disable multi-context/single-channel on pre-NV4x */ - switch (dev->chipset & 0xf0) { - case 0x40: - case 0x60: - /* NV40 class */ - case 0x50: - case 0x80: - case 0x90: - /* G80 class */ - break; - default: - nvc = NULL; - break; - } - - if (!nvc) { - nvc = nouveau_channel_context_create(dev); - if (!nvc) { - NOUVEAU_ERR("Failed initialising GPU context\n"); - return 1; - } - nv_screen->nvc = nvc; - } - - nvc->refcount++; - nv->nvc = nvc; - - /* Find a free slot for a pipe context, allocate a new one if needed */ - nv->pctx_id = -1; - for (i = 0; i < nvc->nr_pctx; i++) { - if (nvc->pctx[i] == NULL) { - nv->pctx_id = i; - break; - } - } - - if (nv->pctx_id < 0) { - nv->pctx_id = nvc->nr_pctx++; - nvc->pctx = - realloc(nvc->pctx, - sizeof(struct pipe_context *) * nvc->nr_pctx); - } - - /* Create pipe */ - if (!getenv("NOUVEAU_FORCE_SOFTPIPE")) { - struct pipe_screen *pscreen; - - pipe = nouveau_pipe_create(nv); - if (!pipe) - NOUVEAU_ERR("Couldn't create hw pipe\n"); - pscreen = nvc->pscreen; - - nv->cap.hw_vertex_buffer = - pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF); - nv->cap.hw_index_buffer = - pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF); - } - - if (!pipe) { - NOUVEAU_MSG("Using softpipe\n"); - pipe = nouveau_create_softpipe(nv); - if (!pipe) { - NOUVEAU_ERR("Error creating pipe, bailing\n"); - return 1; - } - } - - { - struct pipe_texture *fb_tex; - struct pipe_surface *fb_surf; - struct nouveau_pipe_buffer *fb_buf; - enum pipe_format format; - - fb_buf = calloc(1, sizeof(struct nouveau_pipe_buffer)); - fb_buf->base.refcount = 1; - fb_buf->base.usage = PIPE_BUFFER_USAGE_PIXEL; - - nouveau_bo_fake(dev, nv_screen->front_offset, NOUVEAU_BO_VRAM, - nv_screen->front_pitch*nv_screen->front_height, - NULL, &fb_buf->bo); - - if (nv_screen->front_cpp == 4) - format = PIPE_FORMAT_A8R8G8B8_UNORM; - else - format = PIPE_FORMAT_R5G6B5_UNORM; - - fb_surf = nouveau_surface_buffer_ref(nv, &fb_buf->base, format, - nv_screen->front_pitch / - nv_screen->front_cpp, - nv_screen->front_height, - nv_screen->front_pitch, - &fb_tex); - - nv->frontbuffer = fb_surf; - nv->frontbuffer_texture = fb_tex; - } - - pipe->priv = nv; - return 0; -} - -void -nouveau_context_cleanup(struct nouveau_context *nv) -{ - struct nouveau_channel_context *nvc = nv->nvc; - - assert(nv); - - if (nv->pctx_id >= 0) { - nvc->pctx[nv->pctx_id] = NULL; - if (--nvc->refcount <= 0) { - nouveau_channel_context_destroy(nvc); - nv->nv_screen->nvc = NULL; - } - } - - /* XXX: Who cleans up the pipe? */ -} - diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_context.h b/src/gallium/winsys/drm/nouveau/common/nouveau_context.h deleted file mode 100644 index 02d2745680..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_context.h +++ /dev/null @@ -1,59 +0,0 @@ -#ifndef __NOUVEAU_CONTEXT_H__ -#define __NOUVEAU_CONTEXT_H__ - -#include "nouveau/nouveau_winsys.h" -#include "nouveau_drmif.h" -#include "nouveau_device.h" -#include "nouveau_channel.h" -#include "nouveau_pushbuf.h" -#include "nouveau_bo.h" -#include "nouveau_grobj.h" -#include "nouveau_notifier.h" -#include "nouveau_class.h" -#include "nouveau_local.h" - -struct nouveau_channel_context { - struct pipe_screen *pscreen; - int refcount; - - unsigned cur_pctx; - unsigned nr_pctx; - struct pipe_context **pctx; - - struct nouveau_channel *channel; - unsigned next_handle; -}; - -struct nouveau_context { - int locked; - struct nouveau_screen *nv_screen; - struct pipe_surface *frontbuffer; - struct pipe_texture *frontbuffer_texture; - - struct { - int hw_vertex_buffer; - int hw_index_buffer; - } cap; - - /* Hardware context */ - struct nouveau_channel_context *nvc; - int pctx_id; -}; - -extern int nouveau_context_init(struct nouveau_screen *nv_screen, - drm_context_t hHWContext, drmLock *sarea_lock, - struct nouveau_context *nv_share, - struct nouveau_context *nv); -extern void nouveau_context_cleanup(struct nouveau_context *nv); - -extern void LOCK_HARDWARE(struct nouveau_context *); -extern void UNLOCK_HARDWARE(struct nouveau_context *); - -extern uint32_t *nouveau_pipe_dma_beginp(struct nouveau_grobj *, int, int); -extern void nouveau_pipe_dma_kickoff(struct nouveau_channel *); - -/* Must be provided by clients of common code */ -extern void -nouveau_contended_lock(struct nouveau_context *nv); - -#endif diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_local.h b/src/gallium/winsys/drm/nouveau/common/nouveau_local.h deleted file mode 100644 index 11175bce7a..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_local.h +++ /dev/null @@ -1,19 +0,0 @@ -#ifndef __NOUVEAU_LOCAL_H__ -#define __NOUVEAU_LOCAL_H__ - -#include "pipe/p_compiler.h" -#include "nouveau_winsys_pipe.h" -#include <stdio.h> - -/* Debug output */ -#define NOUVEAU_MSG(fmt, args...) do { \ - fprintf(stdout, "nouveau: "fmt, ##args); \ - fflush(stdout); \ -} while(0) - -#define NOUVEAU_ERR(fmt, args...) do { \ - fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); \ - fflush(stderr); \ -} while(0) - -#endif diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_screen.c b/src/gallium/winsys/drm/nouveau/common/nouveau_screen.c deleted file mode 100644 index 422fbf0207..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_screen.c +++ /dev/null @@ -1,31 +0,0 @@ -#include <util/u_memory.h> -#include "nouveau_dri.h" -#include "nouveau_local.h" -#include "nouveau_screen.h" - -int -nouveau_screen_init(struct nouveau_dri *nv_dri, int dev_fd, - struct nouveau_screen *nv_screen) -{ - int ret; - - ret = nouveau_device_open_existing(&nv_screen->device, 0, - dev_fd, 0); - if (ret) { - NOUVEAU_ERR("Failed opening nouveau device: %d\n", ret); - return 1; - } - - nv_screen->front_offset = nv_dri->front_offset; - nv_screen->front_pitch = nv_dri->front_pitch * (nv_dri->bpp / 8); - nv_screen->front_cpp = nv_dri->bpp / 8; - nv_screen->front_height = nv_dri->height; - - return 0; -} - -void -nouveau_screen_cleanup(struct nouveau_screen *nv_screen) -{ - nouveau_device_close(&nv_screen->device); -} diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_screen.h b/src/gallium/winsys/drm/nouveau/common/nouveau_screen.h deleted file mode 100644 index 3e68e219d8..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_screen.h +++ /dev/null @@ -1,27 +0,0 @@ -#ifndef __NOUVEAU_SCREEN_H__ -#define __NOUVEAU_SCREEN_H__ - -#include <stdint.h> - -struct nouveau_device; -struct nouveau_dri; - -struct nouveau_screen { - struct nouveau_device *device; - - uint32_t front_offset; - uint32_t front_pitch; - uint32_t front_cpp; - uint32_t front_height; - - void *nvc; -}; - -int -nouveau_screen_init(struct nouveau_dri *nv_dri, int dev_fd, - struct nouveau_screen *nv_screen); - -void -nouveau_screen_cleanup(struct nouveau_screen *nv_screen); - -#endif diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.c b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.c deleted file mode 100644 index 881df98556..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.c +++ /dev/null @@ -1,242 +0,0 @@ -#include "pipe/internal/p_winsys_screen.h" -#include <pipe/p_defines.h> -#include <pipe/p_inlines.h> -#include <util/u_memory.h> -#include "nouveau_context.h" -#include "nouveau_local.h" -#include "nouveau_screen.h" -#include "nouveau_winsys_pipe.h" - -static const char * -nouveau_get_name(struct pipe_winsys *pws) -{ - return "Nouveau/DRI"; -} - -static uint32_t -nouveau_flags_from_usage(struct nouveau_context *nv, unsigned usage) -{ - struct nouveau_device *dev = nv->nv_screen->device; - uint32_t flags = NOUVEAU_BO_LOCAL; - - if (usage & PIPE_BUFFER_USAGE_PIXEL) { - if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE) - flags |= NOUVEAU_BO_GART; - if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE)) - flags |= NOUVEAU_BO_VRAM; - - switch (dev->chipset & 0xf0) { - case 0x50: - case 0x80: - case 0x90: - flags |= NOUVEAU_BO_TILED; - if (usage & NOUVEAU_BUFFER_USAGE_ZETA) - flags |= NOUVEAU_BO_ZTILE; - break; - default: - break; - } - } - - if (usage & PIPE_BUFFER_USAGE_VERTEX) { - if (nv->cap.hw_vertex_buffer) - flags |= NOUVEAU_BO_GART; - } - - if (usage & PIPE_BUFFER_USAGE_INDEX) { - if (nv->cap.hw_index_buffer) - flags |= NOUVEAU_BO_GART; - } - - return flags; -} - -static struct pipe_buffer * -nouveau_pipe_bo_create(struct pipe_winsys *pws, unsigned alignment, - unsigned usage, unsigned size) -{ - struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws; - struct nouveau_context *nv = nvpws->nv; - struct nouveau_device *dev = nv->nv_screen->device; - struct nouveau_pipe_buffer *nvbuf; - uint32_t flags; - - nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer); - if (!nvbuf) - return NULL; - nvbuf->base.refcount = 1; - nvbuf->base.alignment = alignment; - nvbuf->base.usage = usage; - nvbuf->base.size = size; - - flags = nouveau_flags_from_usage(nv, usage); - - if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) { - FREE(nvbuf); - return NULL; - } - - return &nvbuf->base; -} - -static struct pipe_buffer * -nouveau_pipe_bo_user_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) -{ - struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws; - struct nouveau_device *dev = nvpws->nv->nv_screen->device; - struct nouveau_pipe_buffer *nvbuf; - - nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer); - if (!nvbuf) - return NULL; - nvbuf->base.refcount = 1; - nvbuf->base.size = bytes; - - if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) { - FREE(nvbuf); - return NULL; - } - - return &nvbuf->base; -} - -static void -nouveau_pipe_bo_del(struct pipe_winsys *ws, struct pipe_buffer *buf) -{ - struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); - - nouveau_bo_ref(NULL, &nvbuf->bo); - FREE(nvbuf); -} - -static void * -nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf, - unsigned flags) -{ - struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); - uint32_t map_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - map_flags |= NOUVEAU_BO_RD; - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - map_flags |= NOUVEAU_BO_WR; - -#if 0 - if (flags & PIPE_BUFFER_USAGE_DISCARD && - !(flags & PIPE_BUFFER_USAGE_CPU_READ) && - nouveau_bo_busy(nvbuf->bo, map_flags)) { - struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws; - struct nouveau_context *nv = nvpws->nv; - struct nouveau_device *dev = nv->nv_screen->device; - struct nouveau_bo *rename; - uint32_t flags = nouveau_flags_from_usage(nv, buf->usage); - - if (!nouveau_bo_new(dev, flags, buf->alignment, buf->size, &rename)) { - nouveau_bo_ref(NULL, &nvbuf->bo); - nvbuf->bo = rename; - } - } -#endif - - if (nouveau_bo_map(nvbuf->bo, map_flags)) - return NULL; - return nvbuf->bo->map; -} - -static void -nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) -{ - struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); - - nouveau_bo_unmap(nvbuf->bo); -} - -static void -nouveau_pipe_fence_reference(struct pipe_winsys *ws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *pfence) -{ - *ptr = pfence; -} - -static int -nouveau_pipe_fence_signalled(struct pipe_winsys *ws, - struct pipe_fence_handle *pfence, unsigned flag) -{ - return 0; -} - -static int -nouveau_pipe_fence_finish(struct pipe_winsys *ws, - struct pipe_fence_handle *pfence, unsigned flag) -{ - return 0; -} - -struct pipe_surface * -nouveau_surface_buffer_ref(struct nouveau_context *nv, struct pipe_buffer *pb, - enum pipe_format format, int w, int h, - unsigned pitch, struct pipe_texture **ppt) -{ - struct pipe_screen *pscreen = nv->nvc->pscreen; - struct pipe_texture tmpl, *pt; - struct pipe_surface *ps; - - memset(&tmpl, 0, sizeof(tmpl)); - tmpl.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET | - NOUVEAU_TEXTURE_USAGE_LINEAR; - tmpl.target = PIPE_TEXTURE_2D; - tmpl.width[0] = w; - tmpl.height[0] = h; - tmpl.depth[0] = 1; - tmpl.format = format; - pf_get_block(tmpl.format, &tmpl.block); - tmpl.nblocksx[0] = pf_get_nblocksx(&tmpl.block, w); - tmpl.nblocksy[0] = pf_get_nblocksy(&tmpl.block, h); - - pt = pscreen->texture_blanket(pscreen, &tmpl, &pitch, pb); - if (!pt) - return NULL; - - ps = pscreen->get_tex_surface(pscreen, pt, 0, 0, 0, - PIPE_BUFFER_USAGE_GPU_WRITE); - - *ppt = pt; - return ps; -} - -static void -nouveau_destroy(struct pipe_winsys *pws) -{ - FREE(pws); -} - -struct pipe_winsys * -nouveau_create_pipe_winsys(struct nouveau_context *nv) -{ - struct nouveau_pipe_winsys *nvpws; - struct pipe_winsys *pws; - - nvpws = CALLOC_STRUCT(nouveau_pipe_winsys); - if (!nvpws) - return NULL; - nvpws->nv = nv; - pws = &nvpws->pws; - - pws->flush_frontbuffer = nouveau_flush_frontbuffer; - - pws->buffer_create = nouveau_pipe_bo_create; - pws->buffer_destroy = nouveau_pipe_bo_del; - pws->user_buffer_create = nouveau_pipe_bo_user_create; - pws->buffer_map = nouveau_pipe_bo_map; - pws->buffer_unmap = nouveau_pipe_bo_unmap; - - pws->fence_reference = nouveau_pipe_fence_reference; - pws->fence_signalled = nouveau_pipe_fence_signalled; - pws->fence_finish = nouveau_pipe_fence_finish; - - pws->get_name = nouveau_get_name; - pws->destroy = nouveau_destroy; - - return &nvpws->pws; -} diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.h deleted file mode 100644 index 1eb8043478..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.h +++ /dev/null @@ -1,44 +0,0 @@ -#ifndef NOUVEAU_PIPE_WINSYS_H -#define NOUVEAU_PIPE_WINSYS_H - -#include "pipe/p_context.h" -#include "pipe/internal/p_winsys_screen.h" -#include "nouveau_context.h" - -struct nouveau_pipe_buffer { - struct pipe_buffer base; - struct nouveau_bo *bo; -}; - -static INLINE struct nouveau_pipe_buffer * -nouveau_pipe_buffer(struct pipe_buffer *buf) -{ - return (struct nouveau_pipe_buffer *)buf; -} - -struct nouveau_pipe_winsys { - struct pipe_winsys pws; - - struct nouveau_context *nv; -}; - -extern struct pipe_winsys * -nouveau_create_pipe_winsys(struct nouveau_context *nv); - -struct pipe_context * -nouveau_create_softpipe(struct nouveau_context *nv); - -struct pipe_context * -nouveau_pipe_create(struct nouveau_context *nv); - -/* Must be provided by clients of common code */ -extern void -nouveau_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf, - void *context_private); - -struct pipe_surface * -nouveau_surface_buffer_ref(struct nouveau_context *nv, struct pipe_buffer *pb, - enum pipe_format format, int w, int h, - unsigned pitch, struct pipe_texture **ppt); - -#endif diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_softpipe.c b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_softpipe.c deleted file mode 100644 index 396e4f2a2e..0000000000 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_softpipe.c +++ /dev/null @@ -1,101 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com> - */ - -#include "pipe/internal/p_winsys_screen.h" -#include <pipe/p_screen.h> -#include <pipe/p_defines.h> -#include <pipe/p_format.h> -#include <softpipe/sp_winsys.h> -#include <util/u_memory.h> -#include "nouveau_context.h" -#include "nouveau_winsys_pipe.h" - -struct nouveau_softpipe_winsys { - struct softpipe_winsys sws; - struct nouveau_context *nv; -}; - -/** - * Return list of surface formats supported by this driver. - */ -static boolean -nouveau_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - switch (format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_Z24S8_UNORM: - return TRUE; - default: - break; - }; - - return FALSE; -} - -struct pipe_context * -nouveau_create_softpipe(struct nouveau_context *nv) -{ - struct nouveau_softpipe_winsys *nvsws; - struct pipe_screen *pscreen; - struct pipe_winsys *ws; - struct pipe_context *pipe; - - ws = nouveau_create_pipe_winsys(nv); - if (!ws) - return NULL; - pscreen = softpipe_create_screen(ws); - if (!pscreen) { - ws->destroy(ws); - return NULL; - } - nvsws = CALLOC_STRUCT(nouveau_softpipe_winsys); - if (!nvsws) { - ws->destroy(ws); - pscreen->destroy(pscreen); - return NULL; - } - - nvsws->sws.is_format_supported = nouveau_is_format_supported; - nvsws->nv = nv; - - pipe = softpipe_create(pscreen, ws, &nvsws->sws); - if (!pipe) { - ws->destroy(ws); - pscreen->destroy(pscreen); - FREE(nvsws); - return NULL; - } - - return pipe; -} - diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile index 3f3553b61d..f7db6201fe 100644 --- a/src/gallium/winsys/drm/nouveau/dri/Makefile +++ b/src/gallium/winsys/drm/nouveau/dri/Makefile @@ -6,19 +6,19 @@ LIBNAME = nouveau_dri.so MINIGLX_SOURCES = PIPE_DRIVERS = \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \ $(TOP)/src/gallium/drivers/nv04/libnv04.a \ $(TOP)/src/gallium/drivers/nv10/libnv10.a \ $(TOP)/src/gallium/drivers/nv20/libnv20.a \ $(TOP)/src/gallium/drivers/nv30/libnv30.a \ $(TOP)/src/gallium/drivers/nv40/libnv40.a \ $(TOP)/src/gallium/drivers/nv50/libnv50.a - + DRIVER_SOURCES = \ - nouveau_context_dri.c \ - nouveau_screen_dri.c \ + nouveau_context.c \ + nouveau_screen.c \ nouveau_swapbuffers.c \ - ../common/libnouveaudrm.a + nouveau_lock.c C_SOURCES = \ $(COMMON_GALLIUM_SOURCES) \ diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c index aacfe984d1..deb6ffcff1 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.c +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c @@ -5,23 +5,15 @@ #include <state_tracker/st_public.h> #include <state_tracker/st_context.h> +#include <state_tracker/drm_api.h> #include <pipe/p_defines.h> #include <pipe/p_context.h> #include <pipe/p_screen.h> -#include "../common/nouveau_winsys_pipe.h" -#include "../common/nouveau_dri.h" -#include "../common/nouveau_local.h" -#include "nouveau_context_dri.h" -#include "nouveau_screen_dri.h" +#include "nouveau_context.h" +#include "nouveau_screen.h" -#ifdef DEBUG -static const struct dri_debug_control debug_control[] = { - { "bo", DEBUG_BO }, - { NULL, 0 } -}; -int __nouveau_debug = 0; -#endif +#include "nouveau_drmif.h" GLboolean nouveau_context_create(const __GLcontextModes *glVis, @@ -29,34 +21,38 @@ nouveau_context_create(const __GLcontextModes *glVis, void *sharedContextPrivate) { __DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv; - struct nouveau_screen_dri *nv_screen = driScrnPriv->private; - struct nouveau_context_dri *nv = CALLOC_STRUCT(nouveau_context_dri); - struct st_context *st_share = NULL; - struct nouveau_context_dri *nv_share = NULL; + struct nouveau_screen *nv_screen = driScrnPriv->private; + struct nouveau_context *nv; struct pipe_context *pipe; + struct st_context *st_share = NULL; - if (sharedContextPrivate) { - st_share = ((struct nouveau_context_dri *)sharedContextPrivate)->st; - nv_share = st_share->pipe->priv; + if (sharedContextPrivate) + st_share = ((struct nouveau_context *)sharedContextPrivate)->st; + + nv = CALLOC_STRUCT(nouveau_context); + if (!nv) + return GL_FALSE; + + { + struct nouveau_device_priv *nvdev = + nouveau_device(nv_screen->device); + + nvdev->ctx = driContextPriv->hHWContext; + nvdev->lock = (drmLock *)&driScrnPriv->pSAREA->lock; } - if (nouveau_context_init(&nv_screen->base, driContextPriv->hHWContext, - (drmLock *)&driScrnPriv->pSAREA->lock, - &nv_share->base, &nv->base)) { + pipe = drm_api_hooks.create_context(nv_screen->pscreen); + if (!pipe) { + FREE(nv); return GL_FALSE; } + pipe->priv = nv; - pipe = nv->base.nvc->pctx[nv->base.pctx_id]; - driContextPriv->driverPrivate = (void *)nv; - //nv->nv_screen = nv_screen; + driContextPriv->driverPrivate = nv; nv->dri_screen = driScrnPriv; driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache, nv->dri_screen->myNum, "nouveau"); -#ifdef DEBUG - __nouveau_debug = driParseDebugString(getenv("NOUVEAU_DEBUG"), - debug_control); -#endif nv->st = st_create_context(pipe, glVis, st_share); return GL_TRUE; @@ -65,15 +61,13 @@ nouveau_context_create(const __GLcontextModes *glVis, void nouveau_context_destroy(__DRIcontextPrivate *driContextPriv) { - struct nouveau_context_dri *nv = driContextPriv->driverPrivate; + struct nouveau_context *nv = driContextPriv->driverPrivate; assert(nv); st_finish(nv->st); st_destroy_context(nv->st); - nouveau_context_cleanup(&nv->base); - FREE(nv); } @@ -82,7 +76,7 @@ nouveau_context_bind(__DRIcontextPrivate *driContextPriv, __DRIdrawablePrivate *driDrawPriv, __DRIdrawablePrivate *driReadPriv) { - struct nouveau_context_dri *nv; + struct nouveau_context *nv; struct nouveau_framebuffer *draw, *read; if (!driContextPriv) { @@ -115,7 +109,7 @@ nouveau_context_bind(__DRIcontextPrivate *driContextPriv, GLboolean nouveau_context_unbind(__DRIcontextPrivate *driContextPriv) { - struct nouveau_context_dri *nv = driContextPriv->driverPrivate; + struct nouveau_context *nv = driContextPriv->driverPrivate; (void)nv; st_flush(nv->st, 0, NULL); diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h index 64cf326411..2779b092e6 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.h +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h @@ -3,15 +3,16 @@ #include <dri_util.h> #include <xmlconfig.h> -#include <nouveau/nouveau_winsys.h> -#include "../common/nouveau_context.h" + +#include "nouveau/nouveau_winsys.h" + +#define NOUVEAU_ERR(fmt, args...) debug_printf("%s: "fmt, __func__, ##args) struct nouveau_framebuffer { struct st_framebuffer *stfb; }; -struct nouveau_context_dri { - struct nouveau_context base; +struct nouveau_context { struct st_context *st; /* DRI stuff */ @@ -21,6 +22,7 @@ struct nouveau_context_dri { driOptionCache dri_option_cache; drm_context_t drm_context; drmLock drm_lock; + int locked; }; extern GLboolean nouveau_context_create(const __GLcontextModes *, @@ -31,6 +33,10 @@ extern GLboolean nouveau_context_bind(__DRIcontextPrivate *, __DRIdrawablePrivate *read); extern GLboolean nouveau_context_unbind(__DRIcontextPrivate *); +extern void nouveau_contended_lock(struct nouveau_context *nv); +extern void LOCK_HARDWARE(struct nouveau_context *nv); +extern void UNLOCK_HARDWARE(struct nouveau_context *nv); + #ifdef DEBUG extern int __nouveau_debug; diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_dri.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h index 1207c2d609..1207c2d609 100644 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_dri.h +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_lock.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c index e8cf051ed9..92f5bd09c9 100644 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_lock.c +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c @@ -28,6 +28,7 @@ #include <pipe/p_thread.h> #include "nouveau_context.h" #include "nouveau_screen.h" +#include "nouveau_drmif.h" pipe_static_mutex(lockMutex); @@ -36,7 +37,7 @@ pipe_static_mutex(lockMutex); void LOCK_HARDWARE(struct nouveau_context *nv) { - struct nouveau_screen *nv_screen = nv->nv_screen; + struct nouveau_screen *nv_screen = nv->dri_screen->private; struct nouveau_device *dev = nv_screen->device; struct nouveau_device_priv *nvdev = nouveau_device(dev); char __ret=0; @@ -59,7 +60,7 @@ LOCK_HARDWARE(struct nouveau_context *nv) void UNLOCK_HARDWARE(struct nouveau_context *nv) { - struct nouveau_screen *nv_screen = nv->nv_screen; + struct nouveau_screen *nv_screen = nv->dri_screen->private; struct nouveau_device *dev = nv_screen->device; struct nouveau_device_priv *nvdev = nouveau_device(dev); diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c index 964a9028aa..c4cbbc2124 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.c +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c @@ -5,12 +5,15 @@ #include <pipe/p_context.h> #include <state_tracker/st_public.h> #include <state_tracker/st_cb_fbo.h> -#include <nouveau_drm.h> -#include "../common/nouveau_dri.h" -#include "../common/nouveau_local.h" -#include "nouveau_context_dri.h" -#include "nouveau_screen_dri.h" +#include <state_tracker/drm_api.h> + +#include "nouveau_context.h" +#include "nouveau_screen.h" #include "nouveau_swapbuffers.h" +#include "nouveau_dri.h" + +#include "nouveau_drm.h" +#include "nouveau_drmif.h" #if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12 #error nouveau_drm.h version does not match expected version @@ -178,11 +181,58 @@ nouveau_fill_in_modes(__DRIscreenPrivate *psp, return configs; } +static struct pipe_surface * +dri_surface_from_handle(struct pipe_screen *screen, + unsigned handle, + enum pipe_format format, + unsigned width, + unsigned height, + unsigned pitch) +{ + struct pipe_surface *surface = NULL; + struct pipe_texture *texture = NULL; + struct pipe_texture templat; + struct pipe_buffer *buf = NULL; + + buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle); + if (!buf) + return NULL; + + memset(&templat, 0, sizeof(templat)); + templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; + templat.target = PIPE_TEXTURE_2D; + templat.last_level = 0; + templat.depth[0] = 1; + templat.format = format; + templat.width[0] = width; + templat.height[0] = height; + pf_get_block(templat.format, &templat.block); + + texture = screen->texture_blanket(screen, + &templat, + &pitch, + buf); + + /* we don't need the buffer from this point on */ + pipe_buffer_reference(&buf, NULL); + + if (!texture) + return NULL; + + surface = screen->get_tex_surface(screen, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* we don't need the texture from this point on */ + pipe_texture_reference(&texture, NULL); + return surface; +} + static const __DRIconfig ** nouveau_screen_create(__DRIscreenPrivate *psp) { struct nouveau_dri *nv_dri = psp->pDevPriv; - struct nouveau_screen_dri *nv_screen; + struct nouveau_screen *nv_screen; static const __DRIversion ddx_expected = { 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL }; static const __DRIversion dri_expected = { 4, 0, 0 }; @@ -210,17 +260,38 @@ nouveau_screen_create(__DRIscreenPrivate *psp) return NULL; } - nv_screen = CALLOC_STRUCT(nouveau_screen_dri); + nv_screen = CALLOC_STRUCT(nouveau_screen); if (!nv_screen) return NULL; - driParseOptionInfo(&nv_screen->option_cache, - __driConfigOptions, __driNConfigOptions); + nouveau_device_open_existing(&nv_screen->device, 0, psp->fd, 0); - if (nouveau_screen_init(nv_dri, psp->fd, &nv_screen->base)) { + nv_screen->pscreen = drm_api_hooks.create_screen(psp->fd, 0); + if (!nv_screen->pscreen) { FREE(nv_screen); return NULL; } + nv_screen->pscreen->flush_frontbuffer = nouveau_flush_frontbuffer; + + { + enum pipe_format format; + + if (nv_dri->bpp == 16) + format = PIPE_FORMAT_R5G6B5_UNORM; + else + format = PIPE_FORMAT_A8R8G8B8_UNORM; + + nv_screen->fb = dri_surface_from_handle(nv_screen->pscreen, + nv_dri->front_offset, + format, + nv_dri->width, + nv_dri->height, + nv_dri->front_pitch * + nv_dri->bpp / 8); + } + + driParseOptionInfo(&nv_screen->option_cache, + __driConfigOptions, __driNConfigOptions); nv_screen->driScrnPriv = psp; psp->private = (void *)nv_screen; @@ -234,10 +305,9 @@ nouveau_screen_create(__DRIscreenPrivate *psp) static void nouveau_screen_destroy(__DRIscreenPrivate *driScrnPriv) { - struct nouveau_screen_dri *nv_screen = driScrnPriv->private; + struct nouveau_screen *nv_screen = driScrnPriv->private; driScrnPriv->private = NULL; - nouveau_screen_cleanup(&nv_screen->base); FREE(nv_screen); } diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h index 1498087819..ac078f3c63 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.h +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h @@ -1,13 +1,16 @@ #ifndef __NOUVEAU_SCREEN_DRI_H__ #define __NOUVEAU_SCREEN_DRI_H__ -#include "../common/nouveau_screen.h" #include "xmlconfig.h" -struct nouveau_screen_dri { - struct nouveau_screen base; +struct nouveau_screen { __DRIscreenPrivate *driScrnPriv; driOptionCache option_cache; + + struct nouveau_device *device; + + struct pipe_screen *pscreen; + struct pipe_surface *fb; }; #endif diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c index 58cb6f7265..9c841a0b2d 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c @@ -7,23 +7,25 @@ #include <state_tracker/st_context.h> #include <state_tracker/st_cb_fbo.h> -#include "../common/nouveau_local.h" -#include "nouveau_context_dri.h" -#include "nouveau_screen_dri.h" +#include "nouveau_context.h" +#include "nouveau_screen.h" #include "nouveau_swapbuffers.h" +#include "nouveau_pushbuf.h" + void nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf, const drm_clip_rect_t *rect) { - struct nouveau_context_dri *nv = dPriv->driContextPriv->driverPrivate; - struct pipe_context *pipe = nv->base.nvc->pctx[nv->base.pctx_id]; + struct nouveau_context *nv = dPriv->driContextPriv->driverPrivate; + struct nouveau_screen *nv_screen = nv->dri_screen->private; + struct pipe_context *pipe = nv->st->pipe; drm_clip_rect_t *pbox; int nbox, i; - LOCK_HARDWARE(&nv->base); + LOCK_HARDWARE(nv); if (!dPriv->numClipRects) { - UNLOCK_HARDWARE(&nv->base); + UNLOCK_HARDWARE(nv); return; } pbox = dPriv->pClipRects; @@ -39,12 +41,12 @@ nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf, w = pbox->x2 - pbox->x1; h = pbox->y2 - pbox->y1; - pipe->surface_copy(pipe, FALSE, nv->base.frontbuffer, - dx, dy, surf, sx, sy, w, h); + pipe->surface_copy(pipe, nv_screen->fb, dx, dy, surf, + sx, sy, w, h); } - FIRE_RING(nv->base.nvc->channel); - UNLOCK_HARDWARE(&nv->base); + pipe->flush(pipe, 0, NULL); + UNLOCK_HARDWARE(nv); if (nv->last_stamp != dPriv->lastStamp) { struct nouveau_framebuffer *nvfb = dPriv->driverPrivate; @@ -86,19 +88,19 @@ nouveau_swap_buffers(__DRIdrawablePrivate *dPriv) } void -nouveau_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf, +nouveau_flush_frontbuffer(struct pipe_screen *pscreen, struct pipe_surface *ps, void *context_private) { - struct nouveau_context_dri *nv = context_private; + struct nouveau_context *nv = context_private; __DRIdrawablePrivate *dPriv = nv->dri_drawable; - nouveau_copy_buffer(dPriv, surf, NULL); + nouveau_copy_buffer(dPriv, ps, NULL); } void nouveau_contended_lock(struct nouveau_context *nv) { - struct nouveau_context_dri *nv_sub = (struct nouveau_context_dri*)nv; + struct nouveau_context *nv_sub = (struct nouveau_context*)nv; __DRIdrawablePrivate *dPriv = nv_sub->dri_drawable; __DRIscreenPrivate *sPriv = nv_sub->dri_screen; diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h index 825d3da6da..4ca9cc2283 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h +++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h @@ -1,10 +1,11 @@ #ifndef __NOUVEAU_SWAPBUFFERS_H__ #define __NOUVEAU_SWAPBUFFERS_H__ -extern void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *, - const drm_clip_rect_t *); -extern void nouveau_copy_sub_buffer(__DRIdrawablePrivate *, - int x, int y, int w, int h); -extern void nouveau_swap_buffers(__DRIdrawablePrivate *); +void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *, + const drm_clip_rect_t *); +void nouveau_copy_sub_buffer(__DRIdrawablePrivate *, int x, int y, int w, int h); +void nouveau_swap_buffers(__DRIdrawablePrivate *); +void nouveau_flush_frontbuffer(struct pipe_screen *, struct pipe_surface *, + void *context_private); #endif diff --git a/src/gallium/winsys/drm/nouveau/dri2/Makefile b/src/gallium/winsys/drm/nouveau/dri2/Makefile new file mode 100644 index 0000000000..728870d2e1 --- /dev/null +++ b/src/gallium/winsys/drm/nouveau/dri2/Makefile @@ -0,0 +1,26 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = nouveau_dri2.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \ + $(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \ + $(TOP)/src/gallium/drivers/nv04/libnv04.a \ + $(TOP)/src/gallium/drivers/nv10/libnv10.a \ + $(TOP)/src/gallium/drivers/nv20/libnv20.a \ + $(TOP)/src/gallium/drivers/nv30/libnv30.a \ + $(TOP)/src/gallium/drivers/nv40/libnv40.a \ + $(TOP)/src/gallium/drivers/nv50/libnv50.a + +DRIVER_SOURCES = + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +include ../../Makefile.template + +DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs) + +symlinks: diff --git a/src/gallium/winsys/drm/nouveau/drm/Makefile b/src/gallium/winsys/drm/nouveau/drm/Makefile new file mode 100644 index 0000000000..2da78d8690 --- /dev/null +++ b/src/gallium/winsys/drm/nouveau/drm/Makefile @@ -0,0 +1,13 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = nouveaudrm + +C_SOURCES = nouveau_drm_api.c \ + nouveau_winsys_pipe.c \ + nouveau_winsys.c + +LIBRARY_INCLUDES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-I) +LIBRARY_DEFINES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-other) + +include ../../../../Makefile.template diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c new file mode 100644 index 0000000000..c0127e803f --- /dev/null +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -0,0 +1,194 @@ +#include "util/u_memory.h" + +#include "nouveau_drm_api.h" +#include "nouveau_winsys_pipe.h" + +#include "nouveau_drmif.h" +#include "nouveau_channel.h" +#include "nouveau_bo.h" + +static struct pipe_screen * +nouveau_drm_create_screen(int fd, int pciid) +{ + struct pipe_winsys *ws; + struct nouveau_winsys *nvws; + struct nouveau_device *dev = NULL; + struct pipe_screen *(*init)(struct pipe_winsys *, + struct nouveau_winsys *); + int ret; + + ret = nouveau_device_open_existing(&dev, 0, fd, 0); + if (ret) + return NULL; + + switch (dev->chipset & 0xf0) { + case 0x00: + init = nv04_screen_create; + break; + case 0x10: + init = nv10_screen_create; + break; + case 0x20: + init = nv20_screen_create; + break; + case 0x30: + init = nv30_screen_create; + break; + case 0x40: + case 0x60: + init = nv40_screen_create; + break; + case 0x80: + case 0x90: + case 0xa0: + init = nv50_screen_create; + break; + default: + debug_printf("%s: unknown chipset nv%02x\n", __func__, + dev->chipset); + return NULL; + } + + ws = nouveau_pipe_winsys_new(dev); + if (!ws) { + nouveau_device_close(&dev); + return NULL; + } + + nvws = nouveau_winsys_new(ws); + if (!nvws) { + ws->destroy(ws); + return NULL; + } + + nouveau_pipe_winsys(ws)->pscreen = init(ws, nvws); + if (!nouveau_pipe_winsys(ws)->pscreen) { + ws->destroy(ws); + return NULL; + } + + return nouveau_pipe_winsys(ws)->pscreen; +} + +static struct pipe_context * +nouveau_drm_create_context(struct pipe_screen *pscreen) +{ + struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen); + struct pipe_context *(*init)(struct pipe_screen *, unsigned); + unsigned chipset = nvpws->channel->device->chipset; + int i; + + switch (chipset & 0xf0) { + case 0x00: + init = nv04_create; + break; + case 0x10: + init = nv10_create; + break; + case 0x20: + init = nv20_create; + break; + case 0x30: + init = nv30_create; + break; + case 0x40: + case 0x60: + init = nv40_create; + break; + case 0x80: + case 0x90: + case 0xa0: + init = nv50_create; + break; + default: + debug_printf("%s: unknown chipset nv%02x\n", __func__, chipset); + return NULL; + } + + /* Find a free slot for a pipe context, allocate a new one if needed */ + for (i = 0; i < nvpws->nr_pctx; i++) { + if (nvpws->pctx[i] == NULL) + break; + } + + if (i == nvpws->nr_pctx) { + nvpws->nr_pctx++; + nvpws->pctx = realloc(nvpws->pctx, + sizeof(*nvpws->pctx) * nvpws->nr_pctx); + } + + nvpws->pctx[i] = init(pscreen, i); + return nvpws->pctx[i]; +} + +static boolean +nouveau_drm_pb_from_pt(struct pipe_texture *pt, struct pipe_buffer **ppb, + unsigned *stride) +{ + return false; +} + +static struct pipe_buffer * +nouveau_drm_pb_from_handle(struct pipe_screen *pscreen, const char *name, + unsigned handle) +{ + struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen); + struct nouveau_device *dev = nvpws->channel->device; + struct nouveau_pipe_buffer *nvpb; + int ret; + + nvpb = CALLOC_STRUCT(nouveau_pipe_buffer); + if (!nvpb) + return NULL; + + ret = nouveau_bo_handle_ref(dev, handle, &nvpb->bo); + if (ret) { + debug_printf("%s: ref name 0x%08x failed with %d\n", + __func__, handle, ret); + FREE(nvpb); + return NULL; + } + + pipe_reference_init(&nvpb->base.reference, 1); + nvpb->base.screen = pscreen; + nvpb->base.alignment = 0; + nvpb->base.usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE | + PIPE_BUFFER_USAGE_CPU_READ_WRITE; + nvpb->base.size = nvpb->bo->size; + return &nvpb->base; +} + +static boolean +nouveau_drm_handle_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, + unsigned *handle) +{ + struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb); + + if (!nvpb) + return FALSE; + + *handle = nvpb->bo->handle; + return TRUE; +} + +static boolean +nouveau_drm_name_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, + unsigned *handle) +{ + struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb); + + if (!nvpb) + return FALSE; + + return nouveau_bo_handle_get(nvpb->bo, handle) == 0; +} + +struct drm_api drm_api_hooks = { + .create_screen = nouveau_drm_create_screen, + .create_context = nouveau_drm_create_context, + .buffer_from_texture = nouveau_drm_pb_from_pt, + .buffer_from_handle = nouveau_drm_pb_from_handle, + .handle_from_buffer = nouveau_drm_handle_from_pb, + .global_handle_from_buffer = nouveau_drm_name_from_pb, +}; + diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h new file mode 100644 index 0000000000..2782c83c0e --- /dev/null +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h @@ -0,0 +1,5 @@ +#ifndef __NOUVEAU_DRM_API_H__ +#define __NOUVEAU_DRM_API_H__ +#include "state_tracker/drm_api.h" + +#endif diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c index b6199f8e6d..e3175fd775 100644 --- a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c @@ -1,18 +1,14 @@ #include "util/u_memory.h" -#include "nouveau_context.h" -#include "nouveau_screen.h" #include "nouveau_winsys_pipe.h" -#include "nouveau/nouveau_winsys.h" - static int nouveau_pipe_notifier_alloc(struct nouveau_winsys *nvws, int count, struct nouveau_notifier **notify) { - struct nouveau_context *nv = nvws->nv; + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws); - return nouveau_notifier_alloc(nv->nvc->channel, nv->nvc->next_handle++, + return nouveau_notifier_alloc(nvpws->channel, nvpws->next_handle++, count, notify); } @@ -20,12 +16,11 @@ static int nouveau_pipe_grobj_alloc(struct nouveau_winsys *nvws, int grclass, struct nouveau_grobj **grobj) { - struct nouveau_context *nv = nvws->nv; - struct nouveau_channel *chan = nv->nvc->channel; + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws); + struct nouveau_channel *chan = nvpws->channel; int ret; - ret = nouveau_grobj_alloc(chan, nv->nvc->next_handle++, - grclass, grobj); + ret = nouveau_grobj_alloc(chan, nvpws->next_handle++, grclass, grobj); if (ret) return ret; @@ -62,55 +57,18 @@ nouveau_pipe_get_bo(struct pipe_buffer *pb) return nouveau_pipe_buffer(pb)->bo; } -struct pipe_context * -nouveau_pipe_create(struct nouveau_context *nv) +struct nouveau_winsys * +nouveau_winsys_new(struct pipe_winsys *ws) { - struct nouveau_channel_context *nvc = nv->nvc; - struct nouveau_winsys *nvws = CALLOC_STRUCT(nouveau_winsys); - struct pipe_screen *(*hws_create)(struct pipe_winsys *, - struct nouveau_winsys *); - struct pipe_context *(*hw_create)(struct pipe_screen *, unsigned); - struct pipe_winsys *ws; - unsigned chipset = nv->nv_screen->device->chipset; + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); + struct nouveau_winsys *nvws; + nvws = CALLOC_STRUCT(nouveau_winsys); if (!nvws) return NULL; - switch (chipset & 0xf0) { - case 0x00: - hws_create = nv04_screen_create; - hw_create = nv04_create; - break; - case 0x10: - hws_create = nv10_screen_create; - hw_create = nv10_create; - break; - case 0x20: - hws_create = nv20_screen_create; - hw_create = nv20_create; - break; - case 0x30: - hws_create = nv30_screen_create; - hw_create = nv30_create; - break; - case 0x40: - case 0x60: - hws_create = nv40_screen_create; - hw_create = nv40_create; - break; - case 0x50: - case 0x80: - case 0x90: - hws_create = nv50_screen_create; - hw_create = nv50_create; - break; - default: - NOUVEAU_ERR("Unknown chipset NV%02x\n", chipset); - return NULL; - } - - nvws->nv = nv; - nvws->channel = nv->nvc->channel; + nvws->ws = ws; + nvws->channel = nvpws->channel; nvws->res_init = nouveau_resource_init; nvws->res_alloc = nouveau_resource_alloc; @@ -131,11 +89,6 @@ nouveau_pipe_create(struct nouveau_context *nv) nvws->get_bo = nouveau_pipe_get_bo; - ws = nouveau_create_pipe_winsys(nv); - - if (!nvc->pscreen) - nvc->pscreen = hws_create(ws, nvws); - nvc->pctx[nv->pctx_id] = hw_create(nvc->pscreen, nv->pctx_id); - return nvc->pctx[nv->pctx_id]; + return nvws; } diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c new file mode 100644 index 0000000000..9e03a9f5db --- /dev/null +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c @@ -0,0 +1,204 @@ +#include "pipe/internal/p_winsys_screen.h" +#include <pipe/p_defines.h> +#include <pipe/p_inlines.h> +#include <util/u_memory.h> + +#include "nouveau_winsys_pipe.h" + +#include "nouveau_drmif.h" +#include "nouveau_bo.h" + +static const char * +nouveau_get_name(struct pipe_winsys *pws) +{ + return "Nouveau/DRI"; +} + +static uint32_t +nouveau_flags_from_usage(struct pipe_winsys *ws, unsigned usage) +{ + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); + struct pipe_screen *pscreen = nvpws->pscreen; + uint32_t flags = NOUVEAU_BO_LOCAL; + + if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER) + flags |= NOUVEAU_BO_GART; + + if (usage & PIPE_BUFFER_USAGE_PIXEL) { + if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE) + flags |= NOUVEAU_BO_GART; + if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE)) + flags |= NOUVEAU_BO_VRAM; + + switch (nvpws->channel->device->chipset & 0xf0) { + case 0x50: + case 0x80: + case 0x90: + flags |= NOUVEAU_BO_TILED; + if (usage & NOUVEAU_BUFFER_USAGE_ZETA) + flags |= NOUVEAU_BO_ZTILE; + break; + default: + break; + } + } + + if (usage & PIPE_BUFFER_USAGE_VERTEX) { + if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF)) + flags |= NOUVEAU_BO_GART; + } + + if (usage & PIPE_BUFFER_USAGE_INDEX) { + if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF)) + flags |= NOUVEAU_BO_GART; + } + + return flags; +} + +static struct pipe_buffer * +nouveau_pipe_bo_create(struct pipe_winsys *ws, unsigned alignment, + unsigned usage, unsigned size) +{ + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); + struct nouveau_device *dev = nvpws->channel->device; + struct nouveau_pipe_buffer *nvbuf; + uint32_t flags; + + nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer); + if (!nvbuf) + return NULL; + pipe_reference_init(&nvbuf->base.reference, 1); + nvbuf->base.alignment = alignment; + nvbuf->base.usage = usage; + nvbuf->base.size = size; + + flags = nouveau_flags_from_usage(ws, usage); + if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) { + FREE(nvbuf); + return NULL; + } + + return &nvbuf->base; +} + +static struct pipe_buffer * +nouveau_pipe_bo_user_create(struct pipe_winsys *ws, void *ptr, unsigned bytes) +{ + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); + struct nouveau_device *dev = nvpws->channel->device; + struct nouveau_pipe_buffer *nvbuf; + + nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer); + if (!nvbuf) + return NULL; + pipe_reference_init(&nvbuf->base.reference, 1); + nvbuf->base.size = bytes; + + if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) { + FREE(nvbuf); + return NULL; + } + + return &nvbuf->base; +} + +static void +nouveau_pipe_bo_del(struct pipe_buffer *buf) +{ + struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); + + nouveau_bo_ref(NULL, &nvbuf->bo); + FREE(nvbuf); +} + +static void * +nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf, + unsigned flags) +{ + struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); + uint32_t map_flags = 0; + + if (flags & PIPE_BUFFER_USAGE_CPU_READ) + map_flags |= NOUVEAU_BO_RD; + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) + map_flags |= NOUVEAU_BO_WR; + + if (nouveau_bo_map(nvbuf->bo, map_flags)) + return NULL; + return nvbuf->bo->map; +} + +static void +nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) +{ + struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); + + nouveau_bo_unmap(nvbuf->bo); +} + +static void +nouveau_pipe_fence_reference(struct pipe_winsys *ws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *pfence) +{ + *ptr = pfence; +} + +static int +nouveau_pipe_fence_signalled(struct pipe_winsys *ws, + struct pipe_fence_handle *pfence, unsigned flag) +{ + return 0; +} + +static int +nouveau_pipe_fence_finish(struct pipe_winsys *ws, + struct pipe_fence_handle *pfence, unsigned flag) +{ + return 0; +} + +static void +nouveau_destroy(struct pipe_winsys *ws) +{ + struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); + + nouveau_device_close(&nvpws->channel->device); + FREE(nvpws); +} + +struct pipe_winsys * +nouveau_pipe_winsys_new(struct nouveau_device *dev) +{ + struct nouveau_pipe_winsys *nvpws; + int ret; + + nvpws = CALLOC_STRUCT(nouveau_pipe_winsys); + if (!nvpws) + return NULL; + + ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202, + &nvpws->channel); + if (ret) { + debug_printf("%s: error opening GPU channel: %d\n", + __func__, ret); + FREE(nvpws); + return NULL; + } + nvpws->next_handle = 0x77000000; + + nvpws->base.buffer_create = nouveau_pipe_bo_create; + nvpws->base.buffer_destroy = nouveau_pipe_bo_del; + nvpws->base.user_buffer_create = nouveau_pipe_bo_user_create; + nvpws->base.buffer_map = nouveau_pipe_bo_map; + nvpws->base.buffer_unmap = nouveau_pipe_bo_unmap; + + nvpws->base.fence_reference = nouveau_pipe_fence_reference; + nvpws->base.fence_signalled = nouveau_pipe_fence_signalled; + nvpws->base.fence_finish = nouveau_pipe_fence_finish; + + nvpws->base.get_name = nouveau_get_name; + nvpws->base.destroy = nouveau_destroy; + return &nvpws->base; +} diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h new file mode 100644 index 0000000000..10e1e269e8 --- /dev/null +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h @@ -0,0 +1,52 @@ +#ifndef NOUVEAU_PIPE_WINSYS_H +#define NOUVEAU_PIPE_WINSYS_H + +#include "pipe/internal/p_winsys_screen.h" +#include "pipe/p_context.h" + +#include "nouveau/nouveau_winsys.h" + +#include "nouveau_device.h" + +struct nouveau_pipe_buffer { + struct pipe_buffer base; + struct nouveau_bo *bo; +}; + +static INLINE struct nouveau_pipe_buffer * +nouveau_pipe_buffer(struct pipe_buffer *buf) +{ + return (struct nouveau_pipe_buffer *)buf; +} + +struct nouveau_pipe_winsys { + struct pipe_winsys base; + + struct pipe_screen *pscreen; + + struct nouveau_channel *channel; + uint32_t next_handle; + + unsigned nr_pctx; + struct pipe_context **pctx; +}; + +static INLINE struct nouveau_pipe_winsys * +nouveau_pipe_winsys(struct pipe_winsys *ws) +{ + return (struct nouveau_pipe_winsys *)ws; +} + +static INLINE struct nouveau_pipe_winsys * +nouveau_screen(struct pipe_screen *pscreen) +{ + return nouveau_pipe_winsys(pscreen->winsys); +} + +struct pipe_winsys * +nouveau_pipe_winsys_new(struct nouveau_device *); + +struct nouveau_winsys * +nouveau_winsys_new(struct pipe_winsys *ws); + +#endif diff --git a/src/gallium/winsys/drm/radeon/Makefile b/src/gallium/winsys/drm/radeon/Makefile index dca1e3233a..bacdf3de28 100644 --- a/src/gallium/winsys/drm/radeon/Makefile +++ b/src/gallium/winsys/drm/radeon/Makefile @@ -1,32 +1,12 @@ - +# src/gallium/winsys/drm/radeon/Makefile TOP = ../../../../.. include $(TOP)/configs/current -LIBNAME = radeon_dri.so - -MINIGLX_SOURCES = - -PIPE_DRIVERS = \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/r300/libr300.a - -DRIVER_SOURCES = \ - radeon_buffer.c \ - radeon_context.c \ - radeon_r300.c \ - radeon_screen.c \ - radeon_winsys_softpipe.c - -C_SOURCES = \ - $(COMMON_GALLIUM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I../../../drivers/r300 - -include ../Makefile.template - -DRI_LIB_DEPS += -ldrm_radeon +SUBDIRS = core $(GALLIUM_STATE_TRACKERS_DIRS) -symlinks: +default install clean: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) $@) || exit 1; \ + fi \ + done diff --git a/src/gallium/winsys/drm/radeon/core/Makefile b/src/gallium/winsys/drm/radeon/core/Makefile new file mode 100644 index 0000000000..42a6f4abc2 --- /dev/null +++ b/src/gallium/winsys/drm/radeon/core/Makefile @@ -0,0 +1,18 @@ + +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = radeonwinsys + +C_SOURCES = \ + radeon_buffer.c \ + radeon_drm.c \ + radeon_r300.c \ + radeon_winsys_softpipe.c + +LIBRARY_INCLUDES = -I$(TOP)/src/gallium/drivers/r300 \ + $(shell pkg-config libdrm --cflags-only-I) + +include ../../../../Makefile.template + +symlinks: diff --git a/src/gallium/winsys/drm/radeon/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 259a505c0a..9dca510c81 100644 --- a/src/gallium/winsys/drm/radeon/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -1,5 +1,6 @@ /* * Copyright © 2008 Jérôme Glisse + * 2009 Corbin Simpson * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining @@ -26,37 +27,31 @@ /* * Authors: * Jérôme Glisse <glisse@freedesktop.org> + * Corbin Simpson <MostAwesomeDude@gmail.com> */ -#include <stdio.h> -#include "dri_util.h" -#include "state_tracker/st_public.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" + #include "radeon_buffer.h" -#include "radeon_screen.h" -#include "radeon_context.h" -#include "radeon_bo.h" -#include "radeon_drm.h" static const char *radeon_get_name(struct pipe_winsys *ws) { - return "RADEON/DRI2"; + return "Radeon/GEM+KMS"; } static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws, - unsigned alignment, - unsigned usage, - unsigned size) + unsigned alignment, + unsigned usage, + unsigned size) { - struct radeon_pipe_winsys *radeon_ws = (struct radeon_pipe_winsys *)ws; + struct radeon_winsys *radeon_ws = (struct radeon_winsys *)ws; struct radeon_pipe_buffer *radeon_buffer; uint32_t domain; - radeon_buffer = calloc(1, sizeof(*radeon_buffer)); + radeon_buffer = CALLOC_STRUCT(radeon_pipe_buffer); if (radeon_buffer == NULL) { return NULL; } - radeon_buffer->base.refcount = 1; + + pipe_reference_init(&radeon_buffer->base.reference, 1); radeon_buffer->base.alignment = alignment; radeon_buffer->base.usage = usage; radeon_buffer->base.size = size; @@ -69,21 +64,21 @@ static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws, if (usage & PIPE_BUFFER_USAGE_VERTEX) { domain |= RADEON_GEM_DOMAIN_GTT; } - if (usage & PIPE_BUFFER_USAGE_INDEX) { domain |= RADEON_GEM_DOMAIN_GTT; } - radeon_buffer->bo = radeon_bo_open(radeon_ws->radeon_screen->bom, 0, - size, alignment, domain, 0); + + radeon_buffer->bo = radeon_bo_open(radeon_ws->bom, 0, size, alignment, + domain, 0); if (radeon_buffer->bo == NULL) { - free(radeon_buffer); + FREE(radeon_buffer); } return &radeon_buffer->base; } static struct pipe_buffer *radeon_buffer_user_create(struct pipe_winsys *ws, - void *ptr, - unsigned bytes) + void *ptr, + unsigned bytes) { struct radeon_pipe_buffer *radeon_buffer; @@ -97,7 +92,7 @@ static struct pipe_buffer *radeon_buffer_user_create(struct pipe_winsys *ws, return &radeon_buffer->base; } -static void radeon_buffer_del(struct pipe_winsys *ws, struct pipe_buffer *buffer) +static void radeon_buffer_del(struct pipe_buffer *buffer) { struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer; @@ -106,15 +101,20 @@ static void radeon_buffer_del(struct pipe_winsys *ws, struct pipe_buffer *buffer } static void *radeon_buffer_map(struct pipe_winsys *ws, - struct pipe_buffer *buffer, - unsigned flags) + struct pipe_buffer *buffer, + unsigned flags) { struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer; int write = 0; + if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { + /* XXX Remove this when radeon_bo_map supports DONTBLOCK */ + return NULL; + } if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) { write = 1; } + if (radeon_bo_map(radeon_buffer->bo, write)) return NULL; return radeon_buffer->bo->ptr; @@ -128,59 +128,62 @@ static void radeon_buffer_unmap(struct pipe_winsys *ws, struct pipe_buffer *buff } static void radeon_fence_reference(struct pipe_winsys *ws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *pfence) + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *pfence) { } static int radeon_fence_signalled(struct pipe_winsys *ws, - struct pipe_fence_handle *pfence, - unsigned flag) + struct pipe_fence_handle *pfence, + unsigned flag) { return 1; } static int radeon_fence_finish(struct pipe_winsys *ws, - struct pipe_fence_handle *pfence, - unsigned flag) + struct pipe_fence_handle *pfence, + unsigned flag) { return 0; } static void radeon_flush_frontbuffer(struct pipe_winsys *pipe_winsys, - struct pipe_surface *pipe_surface, - void *context_private) + struct pipe_surface *pipe_surface, + void *context_private) { /* TODO: call dri2CopyRegion */ } -struct pipe_winsys *radeon_pipe_winsys(struct radeon_screen *radeon_screen) +struct radeon_winsys* radeon_pipe_winsys(int fd) { - struct radeon_pipe_winsys *radeon_ws; + struct radeon_winsys* radeon_ws; + struct radeon_bo_manager* bom; - radeon_ws = calloc(1, sizeof(struct radeon_pipe_winsys)); + radeon_ws = CALLOC_STRUCT(radeon_winsys); if (radeon_ws == NULL) { return NULL; } - radeon_ws->radeon_screen = radeon_screen; - radeon_ws->winsys.flush_frontbuffer = radeon_flush_frontbuffer; + bom = radeon_bo_manager_gem_ctor(fd); + radeon_ws->bom = bom; - radeon_ws->winsys.buffer_create = radeon_buffer_create; - radeon_ws->winsys.buffer_destroy = radeon_buffer_del; - radeon_ws->winsys.user_buffer_create = radeon_buffer_user_create; - radeon_ws->winsys.buffer_map = radeon_buffer_map; - radeon_ws->winsys.buffer_unmap = radeon_buffer_unmap; + radeon_ws->base.flush_frontbuffer = radeon_flush_frontbuffer; - radeon_ws->winsys.fence_reference = radeon_fence_reference; - radeon_ws->winsys.fence_signalled = radeon_fence_signalled; - radeon_ws->winsys.fence_finish = radeon_fence_finish; + radeon_ws->base.buffer_create = radeon_buffer_create; + radeon_ws->base.buffer_destroy = radeon_buffer_del; + radeon_ws->base.user_buffer_create = radeon_buffer_user_create; + radeon_ws->base.buffer_map = radeon_buffer_map; + radeon_ws->base.buffer_unmap = radeon_buffer_unmap; - radeon_ws->winsys.get_name = radeon_get_name; + radeon_ws->base.fence_reference = radeon_fence_reference; + radeon_ws->base.fence_signalled = radeon_fence_signalled; + radeon_ws->base.fence_finish = radeon_fence_finish; - return &radeon_ws->winsys; -} + radeon_ws->base.get_name = radeon_get_name; + return radeon_ws; +} +#if 0 static struct pipe_buffer *radeon_buffer_from_handle(struct radeon_screen *radeon_screen, uint32_t handle) { @@ -196,7 +199,7 @@ static struct pipe_buffer *radeon_buffer_from_handle(struct radeon_screen *radeo radeon_bo_unref(bo); return NULL; } - radeon_buffer->base.refcount = 1; + pipe_reference_init(&radeon_buffer->base.reference, 1); radeon_buffer->base.usage = PIPE_BUFFER_USAGE_PIXEL; radeon_buffer->bo = bo; return &radeon_buffer->base; @@ -231,9 +234,10 @@ struct pipe_surface *radeon_surface_from_handle(struct radeon_context *radeon_co pt = pipe_screen->texture_blanket(pipe_screen, &tmpl, &pitch, pb); if (pt == NULL) { - pipe_buffer_reference(pipe_screen, &pb, NULL); + pipe_buffer_reference(&pb, NULL); } ps = pipe_screen->get_tex_surface(pipe_screen, pt, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE); return ps; } +#endif diff --git a/src/gallium/winsys/drm/radeon/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h index c626c20229..40ad0fc8d1 100644 --- a/src/gallium/winsys/drm/radeon/radeon_buffer.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h @@ -30,22 +30,35 @@ #ifndef RADEON_BUFFER_H #define RADEON_BUFFER_H +#include <stdio.h> + #include "pipe/internal/p_winsys_screen.h" -#include "radeon_screen.h" -#include "radeon_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +//#include "state_tracker/st_public.h" + +#include "util/u_memory.h" + #include "radeon_bo.h" +#include "radeon_drm.h" + struct radeon_pipe_buffer { struct pipe_buffer base; struct radeon_bo *bo; }; -struct radeon_pipe_winsys { - struct pipe_winsys winsys; - struct radeon_screen *radeon_screen; +struct radeon_winsys { + /* Parent class. */ + struct pipe_winsys base; + + /* Radeon BO manager. + * This corresponds to void* radeon_winsys in r300_winsys. */ + struct radeon_bo_manager* bom; }; -struct pipe_winsys *radeon_pipe_winsys(struct radeon_screen *radeon_screen); +struct radeon_winsys* radeon_pipe_winsys(int fb); struct pipe_surface *radeon_surface_from_handle(struct radeon_context *radeon_context, uint32_t handle, enum pipe_format format, diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c new file mode 100644 index 0000000000..3446654e28 --- /dev/null +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c @@ -0,0 +1,122 @@ +/* + * Copyright © 2009 Corbin Simpson + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES + * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS + * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + */ +/* + * Authors: + * Corbin Simpson <MostAwesomeDude@gmail.com> + */ + +#include "radeon_drm.h" + +/* Create a pipe_screen. */ +struct pipe_screen* radeon_create_screen(int drmFB, int pciID) +{ + struct radeon_winsys* winsys = radeon_pipe_winsys(drmFB); + + if (getenv("RADEON_SOFTPIPE")) { + return softpipe_create_screen((struct pipe_winsys*)winsys); + } else { + struct r300_winsys* r300 = radeon_create_r300_winsys(drmFB, winsys); + FREE(winsys); + return r300_create_screen(r300); + } +} + +/* Create a pipe_context. */ +struct pipe_context* radeon_create_context(struct pipe_screen* screen) +{ + if (getenv("RADEON_SOFTPIPE")) { + return radeon_create_softpipe(screen->winsys); + } else { + return r300_create_context(screen, screen->winsys); + } +} + +boolean radeon_buffer_from_texture(struct pipe_texture* texture, + struct pipe_buffer** buffer, + unsigned* stride) +{ + return FALSE; +} + +/* Create a buffer from a handle. */ +/* XXX what's up with name? */ +struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen, + const char* name, + unsigned handle) +{ + struct radeon_bo_manager* bom = + ((struct radeon_winsys*)screen->winsys)->bom; + struct radeon_pipe_buffer* radeon_buffer; + struct radeon_bo* bo = NULL; + + bo = radeon_bo_open(bom, handle, 0, 0, 0, 0); + if (bo == NULL) { + return NULL; + } + + radeon_buffer = CALLOC_STRUCT(radeon_pipe_buffer); + if (radeon_buffer == NULL) { + radeon_bo_unref(bo); + return NULL; + } + + pipe_reference_init(&radeon_buffer->base.reference, 1); + radeon_buffer->base.screen = screen; + radeon_buffer->base.usage = PIPE_BUFFER_USAGE_PIXEL; + radeon_buffer->bo = bo; + return &radeon_buffer->base; +} + +boolean radeon_handle_from_buffer(struct pipe_screen* screen, + struct pipe_buffer* buffer, + unsigned* handle) +{ + struct radeon_pipe_buffer* radeon_buffer = + (struct radeon_pipe_buffer*)buffer; + *handle = radeon_buffer->bo->handle; + return TRUE; +} + +boolean radeon_global_handle_from_buffer(struct pipe_screen* screen, + struct pipe_buffer* buffer, + unsigned* handle) +{ + /* XXX WTF is the difference here? global? */ + struct radeon_pipe_buffer* radeon_buffer = + (struct radeon_pipe_buffer*)buffer; + *handle = radeon_buffer->bo->handle; + return TRUE; +} + +struct drm_api drm_api_hooks = { + .create_screen = radeon_create_screen, + .create_context = radeon_create_context, + /* XXX fix this */ + .buffer_from_texture = r300_get_texture_buffer, + .buffer_from_handle = radeon_buffer_from_handle, + .handle_from_buffer = radeon_handle_from_buffer, + .global_handle_from_buffer = radeon_global_handle_from_buffer, +}; diff --git a/src/gallium/winsys/drm/radeon/radeon_context.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h index d7222b4469..ca2d98ed1a 100644 --- a/src/gallium/winsys/drm/radeon/radeon_context.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h @@ -1,5 +1,5 @@ /* - * Copyright © 2008 Jérôme Glisse + * Copyright © 2009 Corbin Simpson * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining @@ -25,46 +25,39 @@ */ /* * Authors: - * Jérôme Glisse <glisse@freedesktop.org> + * Corbin Simpson <MostAwesomeDude@gmail.com> */ -#ifndef RADEON_CONTEXT_H -#define RADEON_CONTEXT_H +#ifndef RADEON_DRM_H +#define RADEON_DRM_H -#include "dri_util.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "radeon_screen.h" +#include "pipe/p_screen.h" +#include "util/u_memory.h" + +#include "state_tracker/drm_api.h" + +#include "radeon_buffer.h" #include "radeon_r300.h" +#include "radeon_winsys_softpipe.h" + +struct pipe_screen* radeon_create_screen(int drmFB, int pciID); + +struct pipe_context* radeon_create_context(struct pipe_screen* screen); + +boolean radeon_buffer_from_texture(struct pipe_texture* texture, + struct pipe_buffer** buffer, + unsigned* stride); -struct radeon_framebuffer { - struct st_framebuffer *st_framebuffer; - unsigned attachments; -}; +struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen, + const char* name, + unsigned handle); -struct radeon_context { - /* st */ - struct st_context *st_context; - /* pipe */ - struct pipe_screen *pipe_screen; - struct pipe_winsys *pipe_winsys; - /* DRI */ - __DRIscreenPrivate *dri_screen; - __DRIdrawablePrivate *dri_drawable; - __DRIdrawablePrivate *dri_readable; - /* DRM */ - int drm_fd; - /* RADEON */ - struct radeon_screen *radeon_screen; -}; +boolean radeon_handle_from_buffer(struct pipe_screen* screen, + struct pipe_buffer* buffer, + unsigned* handle); -GLboolean radeon_context_create(const __GLcontextModes*, - __DRIcontextPrivate*, - void*); -void radeon_context_destroy(__DRIcontextPrivate*); -GLboolean radeon_context_bind(__DRIcontextPrivate*, - __DRIdrawablePrivate*, - __DRIdrawablePrivate*); -GLboolean radeon_context_unbind(__DRIcontextPrivate*); +boolean radeon_global_handle_from_buffer(struct pipe_screen* screen, + struct pipe_buffer* buffer, + unsigned* handle); #endif diff --git a/src/gallium/winsys/drm/radeon/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index 8fe2375e34..c7b6813014 100644 --- a/src/gallium/winsys/drm/radeon/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -75,9 +75,10 @@ static void do_ioctls(struct r300_winsys* winsys, int fd) } -struct r300_winsys* radeon_create_r300_winsys(int fd) +struct r300_winsys* +radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys) { - struct r300_winsys* winsys = calloc(1, sizeof(struct r300_winsys)); + struct r300_winsys* winsys = CALLOC_STRUCT(r300_winsys); do_ioctls(winsys, fd); @@ -92,5 +93,7 @@ struct r300_winsys* radeon_create_r300_winsys(int fd) winsys->end_cs = radeon_cs_end; winsys->flush_cs = radeon_r300_flush_cs; + memcpy(winsys, old_winsys, sizeof(struct radeon_winsys)); + return winsys; } diff --git a/src/gallium/winsys/drm/radeon/radeon_r300.h b/src/gallium/winsys/drm/radeon/core/radeon_r300.h index 8ed95a3a9b..5c373cd084 100644 --- a/src/gallium/winsys/drm/radeon/radeon_r300.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.h @@ -31,4 +31,5 @@ #include "radeon_buffer.h" -struct r300_winsys* radeon_create_r300_winsys(int fd); +struct r300_winsys* +radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys); diff --git a/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.c b/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.c index 8402e1fa5a..226e16674c 100644 --- a/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.c @@ -28,12 +28,7 @@ /* * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com> */ -#include <stdio.h> -#include "imports.h" -#include "pipe/p_defines.h" -#include "pipe/p_format.h" -#include "softpipe/sp_winsys.h" -#include "radeon_context.h" + #include "radeon_winsys_softpipe.h" struct radeon_softpipe_winsys { @@ -57,21 +52,20 @@ static boolean radeon_is_format_supported(struct softpipe_winsys *sws, uint form return FALSE; } -struct pipe_context *radeon_create_softpipe(struct radeon_context *radeon_context) +struct pipe_context *radeon_create_softpipe(struct pipe_winsys* winsys) { - struct radeon_softpipe_winsys *radeon_sp_ws; + struct softpipe_winsys *sp_winsys; struct pipe_screen *pipe_screen; - pipe_screen = softpipe_create_screen(radeon_context->pipe_winsys); + pipe_screen = softpipe_create_screen(winsys); - radeon_sp_ws = CALLOC_STRUCT(radeon_softpipe_winsys); - if (radeon_sp_ws == NULL) { + sp_winsys = CALLOC_STRUCT(softpipe_winsys); + if (sp_winsys == NULL) { return NULL; } - radeon_context->pipe_screen = pipe_screen; - radeon_sp_ws->radeon_context = radeon_context; - radeon_sp_ws->sp_winsys.is_format_supported = radeon_is_format_supported; + + sp_winsys->is_format_supported = radeon_is_format_supported; return softpipe_create(pipe_screen, - radeon_context->pipe_winsys, - &radeon_sp_ws->sp_winsys); + winsys, + sp_winsys); } diff --git a/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.h b/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.h index 519eab769c..04740e41a5 100644 --- a/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.h @@ -30,8 +30,15 @@ #ifndef RADEON_WINSYS_SOFTPIPE_H #define RADEON_WINSYS_SOFTPIPE_H -#include "radeon_context.h" +#include <stdio.h> -struct pipe_context *radeon_create_softpipe(struct radeon_context *radeon_context); +#include "pipe/p_defines.h" +#include "pipe/p_format.h" + +#include "softpipe/sp_winsys.h" + +#include "util/u_memory.h" + +struct pipe_context *radeon_create_softpipe(struct pipe_winsys* winsys); #endif diff --git a/src/gallium/winsys/drm/radeon/dri2/Makefile b/src/gallium/winsys/drm/radeon/dri2/Makefile new file mode 100644 index 0000000000..f471c44349 --- /dev/null +++ b/src/gallium/winsys/drm/radeon/dri2/Makefile @@ -0,0 +1,25 @@ + +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = radeon_dri.so + +MINIGLX_SOURCES = + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \ + $(TOP)/src/gallium/winsys/drm/radeon/core/libradeonwinsys.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/r300/libr300.a + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +DRI_LIB_DEPS += -ldrm_radeon + +symlinks: diff --git a/src/gallium/winsys/drm/radeon/egl/Makefile b/src/gallium/winsys/drm/radeon/egl/Makefile new file mode 100644 index 0000000000..d989b3aa93 --- /dev/null +++ b/src/gallium/winsys/drm/radeon/egl/Makefile @@ -0,0 +1,25 @@ +TOP = ../../../../../.. +GALLIUMDIR = ../../../.. +include $(TOP)/configs/current + +LIBNAME = EGL_r300.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/egl/libegldrm.a \ + $(GALLIUMDIR)/winsys/drm/radeon/core/libradeonwinsys.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/r300/libr300.a + +DRIVER_SOURCES = + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +DRIVER_EXTRAS = -ldrm_radeon + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: diff --git a/src/gallium/winsys/drm/radeon/radeon_context.c b/src/gallium/winsys/drm/radeon/radeon_context.c deleted file mode 100644 index a9d1577634..0000000000 --- a/src/gallium/winsys/drm/radeon/radeon_context.c +++ /dev/null @@ -1,306 +0,0 @@ -/* - * Copyright © 2008 Jérôme Glisse - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES - * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS - * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - */ -/* - * Authors: - * Jérôme Glisse <glisse@freedesktop.org> - */ -#include <stdio.h> -#include "dri_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "radeon_screen.h" -#include "radeon_context.h" -#include "radeon_buffer.h" -#include "radeon_winsys_softpipe.h" - -#define need_GL_ARB_fragment_program -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_shader_objects -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_vertex_shader -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_compiled_vertex_array -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_VERSION_2_0 -#define need_GL_VERSION_2_1 -#include "extension_helper.h" - -/** - * Extension strings exported by the radeon driver. - */ -const struct dri_extension radeon_card_extensions[] = { - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_shader_objects", GL_ARB_shader_objects_functions}, - {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions}, - {"GL_ARB_shading_language_120", GL_VERSION_2_1_functions}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, - {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {NULL, NULL} -}; - -static void radeon_update_renderbuffers(__DRIcontext *dri_context, - __DRIdrawable *dri_drawable) -{ - struct radeon_framebuffer *radeon_fb; - struct radeon_context *radeon_context; - unsigned attachments[10]; - __DRIbuffer *buffers; - __DRIscreen *screen; - int i, count; - - radeon_context = dri_context->driverPrivate; - screen = dri_drawable->driScreenPriv; - radeon_fb = dri_drawable->driverPrivate; - for (count = 0, i = 0; count < 6; count++) { - if (radeon_fb->attachments & (1 << count)) { - attachments[i++] = count; - } - } - - buffers = (*screen->dri2.loader->getBuffers)(dri_drawable, - &dri_drawable->w, - &dri_drawable->h, - attachments, - i, - &count, - dri_drawable->loaderPrivate); - if (buffers == NULL) { - return; - } - - /* set one cliprect to cover the whole dri_drawable */ - dri_drawable->x = 0; - dri_drawable->y = 0; - dri_drawable->backX = 0; - dri_drawable->backY = 0; - dri_drawable->numClipRects = 1; - dri_drawable->pClipRects[0].x1 = 0; - dri_drawable->pClipRects[0].y1 = 0; - dri_drawable->pClipRects[0].x2 = dri_drawable->w; - dri_drawable->pClipRects[0].y2 = dri_drawable->h; - dri_drawable->numBackClipRects = 1; - dri_drawable->pBackClipRects[0].x1 = 0; - dri_drawable->pBackClipRects[0].y1 = 0; - dri_drawable->pBackClipRects[0].x2 = dri_drawable->w; - dri_drawable->pBackClipRects[0].y2 = dri_drawable->h; - - for (i = 0; i < count; i++) { - struct pipe_surface *ps; - enum pipe_format format = 0; - int index = 0; - - switch (buffers[i].attachment) { - case __DRI_BUFFER_FRONT_LEFT: - index = ST_SURFACE_FRONT_LEFT; - switch (buffers[i].cpp) { - case 4: - format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; - case 2: - format = PIPE_FORMAT_R5G6B5_UNORM; - break; - default: - /* FIXME: error */ - return; - } - break; - case __DRI_BUFFER_BACK_LEFT: - index = ST_SURFACE_BACK_LEFT; - switch (buffers[i].cpp) { - case 4: - format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; - case 2: - format = PIPE_FORMAT_R5G6B5_UNORM; - break; - default: - /* FIXME: error */ - return; - } - break; - case __DRI_BUFFER_STENCIL: - case __DRI_BUFFER_DEPTH: - index = ST_SURFACE_DEPTH; - switch (buffers[i].cpp) { - case 4: - format = PIPE_FORMAT_Z24S8_UNORM; - break; - case 2: - format = PIPE_FORMAT_Z16_UNORM; - break; - default: - /* FIXME: error */ - return; - } - break; - case __DRI_BUFFER_ACCUM: - default: - fprintf(stderr, - "unhandled buffer attach event, attacment type %d\n", - buffers[i].attachment); - return; - } - - ps = radeon_surface_from_handle(radeon_context, - buffers[i].name, - format, - dri_drawable->w, - dri_drawable->h, - buffers[i].pitch); - assert(ps); - st_set_framebuffer_surface(radeon_fb->st_framebuffer, index, ps); - } - st_resize_framebuffer(radeon_fb->st_framebuffer, - dri_drawable->w, - dri_drawable->h); -} - -GLboolean radeon_context_create(const __GLcontextModes *visual, - __DRIcontextPrivate *dri_context, - void *shared_context) -{ - __DRIscreenPrivate *dri_screen; - struct radeon_context *radeon_context; - struct radeon_screen *radeon_screen; - struct pipe_context *pipe; - struct st_context *shared_st_context = NULL; - - dri_context->driverPrivate = NULL; - radeon_context = calloc(1, sizeof(struct radeon_context)); - if (radeon_context == NULL) { - return GL_FALSE; - } - - if (shared_context) { - shared_st_context = ((struct radeon_context*)shared_context)->st_context; - } - - dri_screen = dri_context->driScreenPriv; - radeon_screen = dri_screen->private; - radeon_context->dri_screen = dri_screen; - radeon_context->radeon_screen = radeon_screen; - radeon_context->drm_fd = dri_screen->fd; - - radeon_context->pipe_winsys = radeon_pipe_winsys(radeon_screen); - if (radeon_context->pipe_winsys == NULL) { - free(radeon_context); - return GL_FALSE; - } - - if (!getenv("RADEON_SOFTPIPE")) { - fprintf(stderr, "Creating r300 context...\n"); - pipe = - r300_create_context(NULL, - radeon_context->pipe_winsys, - radeon_create_r300_winsys(radeon_context->drm_fd)); - radeon_context->pipe_screen = pipe->screen; - } else { - pipe = radeon_create_softpipe(radeon_context); - } - radeon_context->st_context = st_create_context(pipe, visual, - shared_st_context); - driInitExtensions(radeon_context->st_context->ctx, - radeon_card_extensions, GL_TRUE); - dri_context->driverPrivate = radeon_context; - return GL_TRUE; -} - -void radeon_context_destroy(__DRIcontextPrivate *dri_context) -{ - struct radeon_context *radeon_context; - - radeon_context = dri_context->driverPrivate; - st_finish(radeon_context->st_context); - st_destroy_context(radeon_context->st_context); - free(radeon_context); -} - -GLboolean radeon_context_bind(__DRIcontextPrivate *dri_context, - __DRIdrawablePrivate *dri_drawable, - __DRIdrawablePrivate *dri_readable) -{ - struct radeon_framebuffer *drawable; - struct radeon_framebuffer *readable; - struct radeon_context *radeon_context; - - if (dri_context == NULL) { - st_make_current(NULL, NULL, NULL); - return GL_TRUE; - } - - radeon_context = dri_context->driverPrivate; - drawable = dri_drawable->driverPrivate; - readable = dri_readable->driverPrivate; - st_make_current(radeon_context->st_context, - drawable->st_framebuffer, - readable->st_framebuffer); - - radeon_update_renderbuffers(dri_context, dri_drawable); - if (dri_drawable != dri_readable) { - radeon_update_renderbuffers(dri_context, dri_readable); - } - return GL_TRUE; -} - -GLboolean radeon_context_unbind(__DRIcontextPrivate *dri_context) -{ - struct radeon_context *radeon_context; - - radeon_context = dri_context->driverPrivate; - st_flush(radeon_context->st_context, PIPE_FLUSH_RENDER_CACHE, NULL); - return GL_TRUE; -} diff --git a/src/gallium/winsys/drm/radeon/radeon_screen.c b/src/gallium/winsys/drm/radeon/radeon_screen.c deleted file mode 100644 index e31caff0bf..0000000000 --- a/src/gallium/winsys/drm/radeon/radeon_screen.c +++ /dev/null @@ -1,288 +0,0 @@ -/* - * Copyright © 2008 Jérôme Glisse - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES - * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS - * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - */ -/* - * Authors: - * Jérôme Glisse <glisse@freedesktop.org> - */ -#include <stdio.h> -#include "pipe/p_screen.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_context.h" -#include "pipe/p_state.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "utils.h" -#include "xf86drm.h" -#include "drm.h" -#include "dri_util.h" -#include "radeon_screen.h" -#include "radeon_context.h" -#include "radeon_buffer.h" -#include "radeon_bo.h" -#include "radeon_bo_gem.h" -#include "radeon_drm.h" - -extern const struct dri_extension radeon_card_extensions[]; - -static const __DRIextension *radeon_screen_extensions[] = { - &driReadDrawableExtension, - &driCopySubBufferExtension.base, - &driSwapControlExtension.base, - &driFrameTrackingExtension.base, - &driMediaStreamCounterExtension.base, - NULL -}; - -static __DRIconfig **radeon_fill_in_modes(unsigned pixel_bits, - unsigned depth_bits, - GLboolean have_back_buffer) -{ - __DRIconfig **configs; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - unsigned num_modes; - GLenum fb_format; - GLenum fb_type; - uint8_t depth_bits_array[3]; - uint8_t stencil_bits_array[3]; - uint8_t msaa_samples_array[1]; - /* TODO: pageflipping ? */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) { - stencil_bits_array[2] = 8; - num_modes = 3; - } - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - depth_buffer_factor = (depth_bits == 24) ? 3 : 2; - - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - msaa_samples_array[0] = 0; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - configs = (__DRIconfig **)driCreateConfigs(fb_format, - fb_type, - depth_bits_array, - stencil_bits_array, - depth_buffer_factor, - back_buffer_modes, - back_buffer_factor, - msaa_samples_array, - 1); - if (configs == NULL) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", - __FILE__, __LINE__); - return NULL; - } - return configs; -} - -static void radeon_screen_destroy(__DRIscreenPrivate *dri_screen) -{ - struct radeon_screen *radeon_screen = (struct radeon_screen*)dri_screen->private; - - radeon_bo_manager_gem_dtor(radeon_screen->bom); - dri_screen = NULL; - free(radeon_screen); -} - -static const __DRIconfig **radeon_screen_init(__DRIscreenPrivate *dri_screen) -{ - struct radeon_screen *radeon_screen; - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, radeon_card_extensions, GL_FALSE); - - radeon_screen = calloc(1, sizeof(struct radeon_screen)); - if (radeon_screen == NULL) { - fprintf(stderr, "\nERROR! Allocating private area failed\n"); - return NULL; - } - dri_screen->private = (void*)radeon_screen; - dri_screen->extensions = radeon_screen_extensions; - radeon_screen->dri_screen = dri_screen; - - radeon_screen->bom = radeon_bo_manager_gem_ctor(dri_screen->fd); - if (radeon_screen->bom == NULL) { - radeon_screen_destroy(dri_screen); - return NULL; - } - - return driConcatConfigs(radeon_fill_in_modes(16, 16, 1), - radeon_fill_in_modes(32, 24, 1)); -} - -static boolean radeon_buffer_create(__DRIscreenPrivate *dri_screen, - __DRIdrawablePrivate *dri_drawable, - const __GLcontextModes *visual, - boolean is_pixmap) -{ - if (is_pixmap) { - /* TODO: implement ? */ - return GL_FALSE; - } else { - enum pipe_format color_format, depth_format, stencil_format; - struct radeon_framebuffer *radeon_fb; - - radeon_fb = calloc(1, sizeof(struct radeon_framebuffer)); - if (radeon_fb == NULL) { - return GL_FALSE; - } - - switch (visual->redBits) { - case 5: - color_format = PIPE_FORMAT_R5G6B5_UNORM; - break; - default: - color_format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; - } - - switch (visual->depthBits) { - case 24: - depth_format = PIPE_FORMAT_S8Z24_UNORM; - break; - case 16: - depth_format = PIPE_FORMAT_Z16_UNORM; - break; - default: - depth_format = PIPE_FORMAT_NONE; - break; - } - - switch (visual->stencilBits) { - case 8: - /* force depth format */ - depth_format = PIPE_FORMAT_S8Z24_UNORM; - stencil_format = PIPE_FORMAT_S8Z24_UNORM; - break; - default: - stencil_format = PIPE_FORMAT_NONE; - break; - } - - radeon_fb->st_framebuffer = st_create_framebuffer(visual, - color_format, - depth_format, - stencil_format, - dri_drawable->w, - dri_drawable->h, - (void*)radeon_fb); - if (radeon_fb->st_framebuffer == NULL) { - free(radeon_fb); - return GL_FALSE; - } - dri_drawable->driverPrivate = (void *) radeon_fb; - - radeon_fb->attachments = (1 << __DRI_BUFFER_FRONT_LEFT); - if (visual->doubleBufferMode) { - radeon_fb->attachments |= (1 << __DRI_BUFFER_BACK_LEFT); - } - if (visual->depthBits || visual->stencilBits) { - radeon_fb->attachments |= (1 << __DRI_BUFFER_DEPTH); - } - - return GL_TRUE; - } -} - -static void radeon_buffer_destroy(__DRIdrawablePrivate * dri_drawable) -{ - struct radeon_framebuffer *radeon_fb; - - radeon_fb = dri_drawable->driverPrivate; - assert(radeon_fb->st_framebuffer); - st_unreference_framebuffer(radeon_fb->st_framebuffer); - free(radeon_fb); -} - -static void radeon_swap_buffers(__DRIdrawablePrivate *dri_drawable) -{ - struct radeon_framebuffer *radeon_fb; - struct pipe_surface *back_surf = NULL; - - radeon_fb = dri_drawable->driverPrivate; - assert(radeon_fb); - assert(radeon_fb->st_framebuffer); - - st_get_framebuffer_surface(radeon_fb->st_framebuffer, - ST_SURFACE_BACK_LEFT, - &back_surf); - if (back_surf) { - st_notify_swapbuffers(radeon_fb->st_framebuffer); - /* TODO: do we want to do anythings ? */ - st_notify_swapbuffers_complete(radeon_fb->st_framebuffer); - } -} - -/** - * Called via glXCopySubBufferMESA() to copy a subrect of the back - * buffer to the front buffer/screen. - */ -static void radeon_copy_sub_buffer(__DRIdrawablePrivate *dri_drawable, - int x, int y, int w, int h) -{ - /* TODO: ... */ -} - -const struct __DriverAPIRec driDriverAPI = { - .InitScreen = NULL, - .DestroyScreen = radeon_screen_destroy, - .CreateContext = radeon_context_create, - .DestroyContext = radeon_context_destroy, - .CreateBuffer = radeon_buffer_create, - .DestroyBuffer = radeon_buffer_destroy, - .SwapBuffers = radeon_swap_buffers, - .MakeCurrent = radeon_context_bind, - .UnbindContext = radeon_context_unbind, - .CopySubBuffer = radeon_copy_sub_buffer, - .InitScreen2 = radeon_screen_init, -}; diff --git a/src/gallium/winsys/drm/radeon/radeon_screen.h b/src/gallium/winsys/drm/radeon/radeon_screen.h deleted file mode 100644 index 01b7fa6531..0000000000 --- a/src/gallium/winsys/drm/radeon/radeon_screen.h +++ /dev/null @@ -1,41 +0,0 @@ -/* - * Copyright © 2008 Jérôme Glisse - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES - * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS - * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - */ -/* - * Authors: - * Jérôme Glisse <glisse@freedesktop.org> - */ -#ifndef RADEON_SCREEN_H -#define RADEON_SCREEN_H - -#include "dri_util.h" -#include "radeon_bo.h" - -struct radeon_screen { - __DRIscreenPrivate *dri_screen; - struct radeon_bo_manager *bom; -}; - -#endif diff --git a/src/gallium/winsys/drm/radeon/xorg/Makefile b/src/gallium/winsys/drm/radeon/xorg/Makefile new file mode 100644 index 0000000000..6ffd4a3a54 --- /dev/null +++ b/src/gallium/winsys/drm/radeon/xorg/Makefile @@ -0,0 +1,42 @@ +TARGET = modesetting_drv.so +CFILES = $(wildcard ./*.c) +OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) +GALLIUMDIR = ../../../.. +TOP = ../../../../../.. + +include ${TOP}/configs/current + +CFLAGS = -DHAVE_CONFIG_H \ + -g -Wall -Wimplicit-function-declaration -fPIC \ + $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \ + -I${GALLIUMDIR}/include \ + -I${GALLIUMDIR}/drivers \ + -I${GALLIUMDIR}/auxiliary \ + -I${TOP}/src/mesa \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main + +LIBS = \ + $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \ + $(GALLIUMDIR)/winsys/drm/radeon/core/libradeonwinsys.a \ + $(TOP)/src/gallium/drivers/r300/libr300.a \ + $(GALLIUM_AUXILIARIES) + +############################################# + + + +all default: $(TARGET) + +$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a + $(TOP)/bin/mklib -noprefix -o $@ \ + $(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_radeon + +clean: + rm -rf $(OBJECTS) $(TARGET) + +install: + $(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + $(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + +.PHONY = all clean install diff --git a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c new file mode 100644 index 0000000000..6f77fbe5de --- /dev/null +++ b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c @@ -0,0 +1,157 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <alanh@tungstengraphics.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + * Author: Corbin Simpson <MostAwesomedude@gmail.com> + * + */ + +#include "../../../../state_trackers/xorg/xorg_winsys.h" + +static void radeon_xorg_identify(int flags); +static Bool radeon_xorg_pci_probe(DriverPtr driver, + int entity_num, + struct pci_device *device, + intptr_t match_data); + +static const struct pci_id_match radeon_xorg_device_match[] = { + {0x1002, 0x5E4D, 0xffff, 0xffff, 0, 0, 0}, + {0x1002, 0x7249, 0xffff, 0xffff, 0, 0, 0}, + {0, 0, 0}, +}; + +static SymTabRec radeon_xorg_chipsets[] = { + {0x5E4D, "Radeon RV410 PCIE (X700)"}, + {0x7249, "Radeon R580 PCIE (X1900 XT)"}, + {-1, NULL} +}; + +static PciChipsets radeon_xorg_pci_devices[] = { + {0x5E4D, 0x5E4D, RES_SHARED_VGA}, + {0x7249, 0x7249, RES_SHARED_VGA}, + {-1, -1, RES_UNDEFINED} +}; + +static XF86ModuleVersionInfo radeon_xorg_version = { + "modesetting", + MODULEVENDORSTRING, + MODINFOSTRING1, + MODINFOSTRING2, + XORG_VERSION_CURRENT, + 0, 1, 0, /* major, minor, patch */ + ABI_CLASS_VIDEODRV, + ABI_VIDEODRV_VERSION, + MOD_CLASS_VIDEODRV, + {0, 0, 0, 0} +}; + +/* + * Xorg driver exported structures + */ + +_X_EXPORT DriverRec modesetting = { + 1, + "modesetting", + radeon_xorg_identify, + NULL, + xorg_tracker_available_options, + NULL, + 0, + NULL, + radeon_xorg_device_match, + radeon_xorg_pci_probe +}; + +static MODULESETUPPROTO(radeon_xorg_setup); + +_X_EXPORT XF86ModuleData modesettingModuleData = { + &radeon_xorg_version, + radeon_xorg_setup, + NULL +}; + +/* + * Xorg driver functions + */ + +static pointer +radeon_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin) +{ + static Bool setupDone = 0; + + /* This module should be loaded only once, but check to be sure. + */ + if (!setupDone) { + setupDone = 1; + xf86AddDriver(&modesetting, module, HaveDriverFuncs); + + /* + * Tell the loader about symbols from other modules that this module + * might refer to. + */ + xorg_tracker_loader_ref_sym_lists(); + + /* + * The return value must be non-NULL on success even though there + * is no TearDownProc. + */ + return (pointer) 1; + } else { + if (errmaj) + *errmaj = LDR_ONCEONLY; + return NULL; + } +} + +static void +radeon_xorg_identify(int flags) +{ + xf86PrintChipsets("modesetting", "Driver for Modesetting Kernel Drivers", + radeon_xorg_chipsets); +} + +static Bool +radeon_xorg_pci_probe(DriverPtr driver, + int entity_num, struct pci_device *device, intptr_t match_data) +{ + ScrnInfoPtr scrn = NULL; + EntityInfoPtr entity; + + scrn = xf86ConfigPciEntity(scrn, 0, entity_num, radeon_xorg_pci_devices, + NULL, NULL, NULL, NULL, NULL); + if (scrn != NULL) { + scrn->driverVersion = 1; + scrn->driverName = "modesetting"; + scrn->name = "modesetting"; + scrn->Probe = NULL; + + entity = xf86GetEntityInfo(entity_num); + + /* Use all the functions from the xorg tracker */ + xorg_tracker_set_functions(scrn); + } + return scrn != NULL; +} diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index c6b0e3d8c5..9ceb67d2ac 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -277,6 +277,7 @@ display_surface(struct pipe_winsys *pws, struct pipe_surface *psurf, struct xlib_egl_surface *xsurf) { + struct softpipe_texture *spt = softpipe_texture(psurf->texture); XImage *ximage; void *data; @@ -293,13 +294,13 @@ display_surface(struct pipe_winsys *pws, assert(ximage->format); assert(ximage->bitmap_unit); - data = pws->buffer_map(pws, softpipe_texture(psurf->texture)->buffer, 0); + data = pws->buffer_map(pws, spt->buffer, 0); /* update XImage's fields */ ximage->data = data; ximage->width = psurf->width; ximage->height = psurf->height; - ximage->bytes_per_line = psurf->stride; + ximage->bytes_per_line = spt->stride[psurf->level]; XPutImage(xsurf->Dpy, xsurf->Win, xsurf->Gc, ximage, 0, 0, 0, 0, psurf->width, psurf->height); @@ -309,7 +310,7 @@ display_surface(struct pipe_winsys *pws, ximage->data = NULL; XDestroyImage(ximage); - pws->buffer_unmap(pws, softpipe_texture(psurf->texture)->buffer); + pws->buffer_unmap(pws, spt->buffer); } diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index 739bfa1c1a..aa1bfa8e88 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -99,7 +99,7 @@ buffer_create(struct pipe_winsys *pws, if (!buffer) return NULL; - buffer->Base.refcount = 1; + pipe_reference_init(&buffer->Base.reference, 1); buffer->Base.alignment = alignment; buffer->Base.usage = usage; buffer->Base.size = size; @@ -121,7 +121,7 @@ user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) if (!buffer) return NULL; - buffer->Base.refcount = 1; + pipe_reference_init(&buffer->Base.reference, 1); buffer->Base.size = bytes; buffer->UserBuffer = TRUE; buffer->Data = ptr; @@ -148,7 +148,7 @@ buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) static void -buffer_destroy(struct pipe_winsys *pws, struct pipe_buffer *buf) +buffer_destroy(struct pipe_buffer *buf) { struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf); diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c index 864be37871..77e46a2054 100644 --- a/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c +++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c @@ -31,7 +31,7 @@ nouveau_copy_buffer(dri_drawable_t *dri_drawable, struct pipe_surface *surf, w = pbox->x2 - pbox->x1; h = pbox->y2 - pbox->y1; - pipe->surface_copy(pipe, FALSE, nv->base.frontbuffer, + pipe->surface_copy(pipe, nv->base.frontbuffer, dx, dy, surf, sx, sy, w, h); } diff --git a/src/gallium/winsys/g3dvl/xsp_winsys.c b/src/gallium/winsys/g3dvl/xsp_winsys.c index 40d683234f..5b9fdb5c1f 100644 --- a/src/gallium/winsys/g3dvl/xsp_winsys.c +++ b/src/gallium/winsys/g3dvl/xsp_winsys.c @@ -1,10 +1,12 @@ #include "vl_winsys.h" #include <X11/Xutil.h> -#include <pipe/p_winsys.h> +#include <pipe/internal/p_winsys_screen.h> #include <pipe/p_state.h> #include <pipe/p_inlines.h> #include <util/u_memory.h> +#include <util/u_math.h> #include <softpipe/sp_winsys.h> +#include <softpipe/sp_texture.h> /* pipe_winsys implementation */ @@ -37,7 +39,7 @@ static struct pipe_buffer* xsp_buffer_create(struct pipe_winsys *pws, unsigned a assert(pws); buffer = calloc(1, sizeof(struct xsp_buffer)); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; @@ -53,7 +55,7 @@ static struct pipe_buffer* xsp_user_buffer_create(struct pipe_winsys *pws, void assert(pws); buffer = calloc(1, sizeof(struct xsp_buffer)); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.size = size; buffer->is_user_buffer = TRUE; buffer->data = data; @@ -96,12 +98,6 @@ static void xsp_buffer_destroy(struct pipe_winsys *pws, struct pipe_buffer *buff free(xsp_buf); } -/* Borrowed from Mesa's xm_winsys */ -static unsigned int round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - static struct pipe_buffer* xsp_surface_buffer_create ( struct pipe_winsys *pws, @@ -119,11 +115,11 @@ static struct pipe_buffer* xsp_surface_buffer_create pf_get_block(format, &block); nblocksx = pf_get_nblocksx(&block, width); nblocksy = pf_get_nblocksy(&block, height); - *stride = round_up(nblocksx * block.size, ALIGNMENT); + *stride = align(nblocksx * block.size, ALIGNMENT); - return winsys->buffer_create(winsys, ALIGNMENT, - usage, - *stride * nblocksy); + return pws->buffer_create(pws, ALIGNMENT, + usage, + *stride * nblocksy); } static void xsp_fence_reference(struct pipe_winsys *pws, struct pipe_fence_handle **ptr, struct pipe_fence_handle *fence) @@ -167,7 +163,7 @@ static void xsp_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface * xsp_winsys->fbimage.width = surface->width; xsp_winsys->fbimage.height = surface->height; xsp_winsys->fbimage.bytes_per_line = surface->width * (xsp_winsys->fbimage.bits_per_pixel >> 3); - xsp_winsys->fbimage.data = pipe_surface_map(surface, 0); + xsp_winsys->fbimage.data = ((struct xsp_buffer *)softpipe_texture(surface->texture)->buffer)->data + surface->offset; XPutImage ( @@ -183,7 +179,6 @@ static void xsp_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface * surface->height ); XFlush(xsp_context->display); - pipe_surface_unmap(surface); } static const char* xsp_get_name(struct pipe_winsys *pws) diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript index f9c1a34668..72b5df8ca2 100644 --- a/src/gallium/winsys/gdi/SConscript +++ b/src/gallium/winsys/gdi/SConscript @@ -11,12 +11,6 @@ if env['platform'] == 'windows': '#src/gallium/state_trackers/wgl', ]) - env.Append(CPPDEFINES = [ - '__GL_EXPORTS', - 'BUILD_GL32', - '_GNU_H_WINDOWS32_DEFINES', - ]) - env.Append(LIBS = [ 'gdi32', 'user32', @@ -24,9 +18,13 @@ if env['platform'] == 'windows': ]) sources = [ - '#src/gallium/state_trackers/wgl/opengl32.def', 'gdi_softpipe_winsys.c', ] + + if env['toolchain'] == 'crossmingw': + sources += ['#src/gallium/state_trackers/wgl/opengl32.mingw.def'] + else: + sources += ['#src/gallium/state_trackers/wgl/opengl32.def'] drivers = [ softpipe, diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c index 4924dbf26a..440666d835 100644 --- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c @@ -87,8 +87,7 @@ gdi_softpipe_buffer_unmap(struct pipe_winsys *winsys, static void -gdi_softpipe_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) +gdi_softpipe_buffer_destroy(struct pipe_buffer *buf) { struct gdi_softpipe_buffer *oldBuf = gdi_softpipe_buffer(buf); @@ -118,7 +117,7 @@ gdi_softpipe_buffer_create(struct pipe_winsys *winsys, { struct gdi_softpipe_buffer *buffer = CALLOC_STRUCT(gdi_softpipe_buffer); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; @@ -143,7 +142,7 @@ gdi_softpipe_user_buffer_create(struct pipe_winsys *winsys, if(!buffer) return NULL; - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.size = bytes; buffer->userBuffer = TRUE; buffer->data = ptr; @@ -273,14 +272,17 @@ gdi_softpipe_flush_frontbuffer(struct pipe_screen *screen, struct pipe_surface *surface, HDC hDC) { + struct softpipe_texture *texture; struct gdi_softpipe_buffer *buffer; BITMAPINFO bmi; - buffer = gdi_softpipe_buffer(softpipe_texture(surface->texture)->buffer); + texture = softpipe_texture(surface->texture); + + buffer = gdi_softpipe_buffer(texture->buffer); memset(&bmi, 0, sizeof(BITMAPINFO)); bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER); - bmi.bmiHeader.biWidth = surface->stride / pf_get_size(surface->format); + bmi.bmiHeader.biWidth = texture->stride[surface->level] / pf_get_size(surface->format); bmi.bmiHeader.biHeight= -(long)surface->height; bmi.bmiHeader.biPlanes = 1; bmi.bmiHeader.biBitCount = pf_get_bits(surface->format); diff --git a/src/gallium/winsys/xlib/xlib_brw_aub.c b/src/gallium/winsys/xlib/xlib_brw_aub.c index 2956e1b960..b6bd849ef2 100644 --- a/src/gallium/winsys/xlib/xlib_brw_aub.c +++ b/src/gallium/winsys/xlib/xlib_brw_aub.c @@ -34,8 +34,9 @@ #include "xlib_brw_aub.h" #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "pipe/p_debug.h" +#include "util/u_debug.h" #include "util/u_memory.h" +#include "softpipe/sp_texture.h" struct brw_aubfile { @@ -322,10 +323,10 @@ void brw_aub_dump_bmp( struct brw_aubfile *aubfile, struct aub_dump_bmp db; unsigned format; - assert(surface->block.width == 1); - assert(surface->block.height == 1); + assert(surface->texture->block.width == 1); + assert(surface->texture->block.height == 1); - if (surface->block.size == 4) + if (surface->texture->block.size == 4) format = 0x7; else format = 0x3; @@ -334,8 +335,9 @@ void brw_aub_dump_bmp( struct brw_aubfile *aubfile, db.xmin = 0; db.ymin = 0; db.format = format; - db.bpp = surface->block.size * 8; - db.pitch = surface->stride/surface->block.size; + db.bpp = surface->texture->block.size * 8; + db.pitch = softpipe_texture(surface->texture)->stride[surface->level] / + surface->texture->block.size; db.xsize = surface->width; db.ysize = surface->height; db.addr = gtt_offset; diff --git a/src/gallium/winsys/xlib/xlib_brw_screen.c b/src/gallium/winsys/xlib/xlib_brw_screen.c index 8e1bfab2f5..fe8dfff767 100644 --- a/src/gallium/winsys/xlib/xlib_brw_screen.c +++ b/src/gallium/winsys/xlib/xlib_brw_screen.c @@ -37,7 +37,6 @@ //#include "state_trackers/xlib/xmesaP.h" #include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "i965simple/brw_winsys.h" @@ -57,10 +56,10 @@ buffer_from_surface(struct pipe_surface *surface) } struct aub_buffer { + struct pipe_reference reference; char *data; unsigned offset; unsigned size; - unsigned refcount; unsigned map_count; boolean dump_on_unmap; }; @@ -144,8 +143,7 @@ static void aub_buffer_unmap(struct pipe_winsys *winsys, static void -aub_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) +aub_buffer_destroy(struct pipe_buffer *buf) { free(buf); } @@ -189,7 +187,7 @@ aub_buffer_create(struct pipe_winsys *winsys, struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); - sbo->refcount = 1; + pipe_reference_init(&sbo->reference, 1); /* Could reuse buffers that are not referenced in current * batchbuffer. Can't do that atm, so always reallocate: diff --git a/src/gallium/winsys/xlib/xlib_cell.c b/src/gallium/winsys/xlib/xlib_cell.c index c87564f4dc..13e609f58f 100644 --- a/src/gallium/winsys/xlib/xlib_cell.c +++ b/src/gallium/winsys/xlib/xlib_cell.c @@ -107,22 +107,21 @@ xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) } static void -xm_buffer_destroy(struct pipe_winsys *pws, +xm_buffer_destroy(/*struct pipe_winsys *pws,*/ struct pipe_buffer *buf) { struct xm_buffer *oldBuf = xm_buffer(buf); - if (oldBuf->data) { - { + if (oldBuf) { + if (oldBuf->data) { if (!oldBuf->userBuffer) { align_free(oldBuf->data); } - } - oldBuf->data = NULL; + oldBuf->data = NULL; + } + free(oldBuf); } - - free(oldBuf); } @@ -222,7 +221,8 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, * This function copies that XImage to the actual X Window. */ XMesaContext xmctx = (XMesaContext) context_private; - xlib_cell_display_surface(xmctx->xm_buffer, surf); + if (xmctx) + xlib_cell_display_surface(xmctx->xm_buffer, surf); } @@ -242,7 +242,7 @@ xm_buffer_create(struct pipe_winsys *pws, { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; @@ -266,7 +266,7 @@ static struct pipe_buffer * xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.size = bytes; buffer->userBuffer = TRUE; buffer->data = ptr; @@ -370,7 +370,7 @@ xlib_create_cell_winsys( void ) static struct pipe_screen * -xlib_create_cell_screen( struct pipe_winsys *pws ) +xlib_create_cell_screen( void ) { struct pipe_winsys *winsys; struct pipe_screen *screen; diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c index 586e1dfca5..762ebd9847 100644 --- a/src/gallium/winsys/xlib/xlib_softpipe.c +++ b/src/gallium/winsys/xlib/xlib_softpipe.c @@ -61,8 +61,10 @@ struct xm_buffer void *mapped; XImage *tempImage; +#ifdef USE_XSHM int shm; XShmSegmentInfo shminfo; +#endif }; @@ -73,7 +75,9 @@ struct xmesa_pipe_winsys { struct pipe_winsys base; /* struct xmesa_visual *xm_visual; */ +#ifdef USE_XSHM int shm; +#endif }; @@ -89,7 +93,13 @@ xm_buffer( struct pipe_buffer *buf ) /** * X Shared Memory Image extension code */ +#ifdef USE_XSHM #define XSHM_ENABLED(b) ((b)->shm) +#else +#define XSHM_ENABLED(b) 0 +#endif + +#ifdef USE_XSHM static volatile int mesaXErrorFlag = 0; @@ -170,8 +180,12 @@ alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, (void) XSetErrorHandler(old_handler); return; } + + b->shm = 1; } +#endif /* USE_XSHM */ + /* Most callbacks map direcly onto dri_bufmgr operations: @@ -193,12 +207,12 @@ xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) } static void -xm_buffer_destroy(struct pipe_winsys *pws, - struct pipe_buffer *buf) +xm_buffer_destroy(struct pipe_buffer *buf) { struct xm_buffer *oldBuf = xm_buffer(buf); if (oldBuf->data) { +#ifdef USE_XSHM if (oldBuf->shminfo.shmid >= 0) { shmdt(oldBuf->shminfo.shmaddr); shmctl(oldBuf->shminfo.shmid, IPC_RMID, 0); @@ -207,6 +221,7 @@ xm_buffer_destroy(struct pipe_winsys *pws, oldBuf->shminfo.shmaddr = (char *) -1; } else +#endif { if (!oldBuf->userBuffer) { align_free(oldBuf->data); @@ -220,7 +235,6 @@ xm_buffer_destroy(struct pipe_winsys *pws, } - /** * Display/copy the image in the surface into the X window specified * by the XMesaBuffer. @@ -230,8 +244,8 @@ xlib_softpipe_display_surface(struct xmesa_buffer *b, struct pipe_surface *surf) { XImage *ximage; - struct xm_buffer *xm_buf = xm_buffer( - softpipe_texture(surf->texture)->buffer); + struct softpipe_texture *spt = softpipe_texture(surf->texture); + struct xm_buffer *xm_buf = xm_buffer(spt->buffer); static boolean no_swap = 0; static boolean firsttime = 1; @@ -243,20 +257,26 @@ xlib_softpipe_display_surface(struct xmesa_buffer *b, if (no_swap) return; +#ifdef USE_XSHM if (XSHM_ENABLED(xm_buf) && (xm_buf->tempImage == NULL)) { - assert(surf->block.width == 1); - assert(surf->block.height == 1); - alloc_shm_ximage(xm_buf, b, surf->stride/surf->block.size, surf->height); + assert(surf->texture->block.width == 1); + assert(surf->texture->block.height == 1); + alloc_shm_ximage(xm_buf, b, spt->stride[surf->level] / + surf->texture->block.size, surf->height); } +#endif ximage = (XSHM_ENABLED(xm_buf)) ? xm_buf->tempImage : b->tempImage; ximage->data = xm_buf->data; /* display image in Window */ +#ifdef USE_XSHM if (XSHM_ENABLED(xm_buf)) { XShmPutImage(b->xm_visual->display, b->drawable, b->gc, ximage, 0, 0, 0, 0, surf->width, surf->height, False); - } else { + } else +#endif + { /* check that the XImage has been previously initialized */ assert(ximage->format); assert(ximage->bitmap_unit); @@ -264,7 +284,7 @@ xlib_softpipe_display_surface(struct xmesa_buffer *b, /* update XImage's fields */ ximage->width = surf->width; ximage->height = surf->height; - ximage->bytes_per_line = surf->stride; + ximage->bytes_per_line = spt->stride[surf->level]; XPutImage(b->xm_visual->display, b->drawable, b->gc, ximage, 0, 0, 0, 0, surf->width, surf->height); @@ -301,12 +321,9 @@ xm_buffer_create(struct pipe_winsys *pws, unsigned size) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); +#ifdef USE_XSHM struct xmesa_pipe_winsys *xpws = (struct xmesa_pipe_winsys *) pws; - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; buffer->shminfo.shmid = -1; buffer->shminfo.shmaddr = (char *) -1; @@ -315,12 +332,17 @@ xm_buffer_create(struct pipe_winsys *pws, if (alloc_shm(buffer, size)) { buffer->data = buffer->shminfo.shmaddr; + buffer->shm = 1; } } +#endif - if (buffer->data == NULL) { - buffer->shm = 0; + pipe_reference_init(&buffer->base.reference, 1); + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + if (buffer->data == NULL) { /* align to 16-byte multiple for Cell */ buffer->data = align_malloc(size, max(alignment, 16)); } @@ -336,11 +358,13 @@ static struct pipe_buffer * xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->base.refcount = 1; + pipe_reference_init(&buffer->base.reference, 1); buffer->base.size = bytes; buffer->userBuffer = TRUE; buffer->data = ptr; +#ifdef USE_XSHM buffer->shm = 0; +#endif return &buffer->base; } |