diff options
22 files changed, 2541 insertions, 386 deletions
diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index da7eded21f..67d78bdbbd 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -26,8 +26,11 @@ C_SOURCES = \ draw_pt_emit.c \ draw_pt_fetch.c \ draw_pt_fetch_emit.c \ + draw_pt_fetch_shade_emit.c \ + draw_pt_middle_fse.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_post_vs.c \ + draw_pt_util.c \ draw_pt_varray.c \ draw_pt_vcache.c \ draw_vertex.c \ diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 46afb0f41f..1d26706dee 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -212,6 +212,71 @@ void draw_pipeline_run( struct draw_context *draw, draw->pipeline.vertex_count = 0; } +#define QUAD(i0,i1,i2,i3) \ + do_triangle( draw, \ + ( DRAW_PIPE_RESET_STIPPLE | \ + DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_2 ), \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i1), \ + verts + stride * (i3)); \ + do_triangle( draw, \ + ( DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_1 ), \ + verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i2), \ + verts + stride * (i3)) + +#define TRIANGLE(flags,i0,i1,i2) \ + do_triangle( draw, \ + flags, /* flags */ \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i1), \ + verts + stride * (i2)) + +#define LINE(flags,i0,i1) \ + do_line( draw, \ + flags, \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i+1)) + +#define POINT(i0) \ + do_point( draw, \ + verts + stride * i0 ) + +#define FUNC pipe_run_linear +#define ARGS \ + struct draw_context *draw, \ + unsigned prim, \ + struct vertex_header *vertices, \ + unsigned stride + +#define LOCAL_VARS \ + char *verts = (char *)vertices; \ + boolean flatfirst = (draw->rasterizer->flatshade && \ + draw->rasterizer->flatshade_first); \ + unsigned i, flags + +#define FLUSH + +#include "draw_pt_decompose.h" + +void draw_pipeline_run_linear( struct draw_context *draw, + unsigned prim, + struct vertex_header *vertices, + unsigned count, + unsigned stride ) +{ + char *verts = (char *)vertices; + draw->pipeline.verts = verts; + draw->pipeline.vertex_stride = stride; + draw->pipeline.vertex_count = count; + + pipe_run_linear(draw, prim, vertices, stride, count); + + draw->pipeline.verts = NULL; + draw->pipeline.vertex_count = 0; +} void draw_pipeline_flush( struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 86a7d1c730..67b9a9503d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -427,9 +427,6 @@ static void vbuf_destroy( struct draw_stage *stage ) if(vbuf->indices) align_free( vbuf->indices ); - if(vbuf->translate) - vbuf->translate->release( vbuf->translate ); - if (vbuf->render) vbuf->render->destroy( vbuf->render ); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index cee58bbf73..86b901a3c8 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -124,12 +124,14 @@ struct draw_context struct { struct { struct draw_pt_middle_end *fetch_emit; + struct draw_pt_middle_end *fetch_shade_emit; struct draw_pt_middle_end *general; } middle; struct { struct draw_pt_front_end *vcache; struct draw_pt_front_end *varray; + struct draw_pt_front_end *fetch_shade_emit; /* temp hack */ } front; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; @@ -154,6 +156,7 @@ struct draw_context const void *constants; } user; + boolean test_fse; } pt; struct { @@ -247,6 +250,12 @@ void draw_pipeline_run( struct draw_context *draw, const ushort *elts, unsigned count ); +void draw_pipeline_run_linear( struct draw_context *draw, + unsigned prim, + struct vertex_header *vertices, + unsigned count, + unsigned stride ); + void draw_pipeline_flush( struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index c9c5d18313..91e35db819 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -64,7 +64,7 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_PIPELINE; } - if (!draw->bypass_clipping) { + if (!draw->bypass_clipping && !draw->pt.test_fse) { opt |= PT_CLIPTEST; } @@ -72,17 +72,24 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_SHADE; } - if (opt) - middle = draw->pt.middle.general; - else + + if (opt == 0) middle = draw->pt.middle.fetch_emit; + else if (opt == PT_SHADE && draw->pt.test_fse) + middle = draw->pt.middle.fetch_shade_emit; + else + middle = draw->pt.middle.general; /* Pick the right frontend */ - if (draw->pt.user.elts || - count >= 256) { + if (draw->pt.user.elts || (opt & PT_PIPELINE)) { frontend = draw->pt.front.vcache; +#if 0 + } else if (opt == PT_SHADE && draw->pt.test_fse) { + /* should be a middle end.. */ + frontend = draw->pt.front.fetch_shade_emit; +#endif } else { frontend = draw->pt.front.varray; } @@ -102,6 +109,8 @@ draw_pt_arrays(struct draw_context *draw, boolean draw_pt_init( struct draw_context *draw ) { + draw->pt.test_fse = GETENV("DRAW_FSE") != NULL; + draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) return FALSE; @@ -114,6 +123,17 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_emit) return FALSE; + if (draw->pt.test_fse) { + draw->pt.middle.fetch_shade_emit = draw_pt_middle_fse( draw ); + if (!draw->pt.middle.fetch_shade_emit) + return FALSE; + + draw->pt.front.fetch_shade_emit = draw_pt_fetch_shade_emit( draw ); + if (!draw->pt.front.fetch_shade_emit) + return FALSE; + } + + draw->pt.middle.general = draw_pt_fetch_pipeline_or_emit( draw ); if (!draw->pt.middle.general) return FALSE; @@ -134,6 +154,16 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.middle.fetch_emit = NULL; } + if (draw->pt.middle.fetch_shade_emit) { + draw->pt.middle.fetch_shade_emit->destroy( draw->pt.middle.fetch_shade_emit ); + draw->pt.middle.fetch_shade_emit = NULL; + } + + if (draw->pt.front.fetch_shade_emit) { + draw->pt.front.fetch_shade_emit->destroy( draw->pt.front.fetch_shade_emit ); + draw->pt.front.fetch_shade_emit = NULL; + } + if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); draw->pt.front.vcache = NULL; @@ -147,19 +177,6 @@ void draw_pt_destroy( struct draw_context *draw ) -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; - /** * Draw vertex arrays @@ -172,9 +189,10 @@ void draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { - if (reduced_prim[prim] != draw->reduced_prim) { + unsigned reduced_prim = draw_pt_reduced_prim(prim); + if (reduced_prim != draw->reduced_prim) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->reduced_prim = reduced_prim[prim]; + draw->reduced_prim = reduced_prim; } /* drawing done here: */ diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 2dec376cee..cdae46b8d2 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -92,6 +92,10 @@ struct draw_pt_middle_end { const ushort *draw_elts, unsigned draw_count ); + void (*run_linear)(struct draw_pt_middle_end *, + unsigned start, + unsigned count); + void (*finish)( struct draw_pt_middle_end * ); void (*destroy)( struct draw_pt_middle_end * ); }; @@ -117,6 +121,8 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw); +struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ); + /* Middle-ends: * * Currently one general-purpose case which can do all possibilities, @@ -128,6 +134,7 @@ struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw); * vertex_elements. */ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); +struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); @@ -152,6 +159,13 @@ void draw_pt_emit( struct pt_emit *emit, const ushort *elts, unsigned count ); +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 ); struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); @@ -170,6 +184,11 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, unsigned count, char *verts ); +void draw_pt_fetch_run_linear( struct pt_fetch *fetch, + unsigned start, + unsigned count, + char *verts ); + void draw_pt_fetch_destroy( struct pt_fetch *fetch ); struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ); @@ -194,4 +213,11 @@ struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw ); void draw_pt_post_vs_destroy( struct pt_post_vs *pvs ); +/******************************************************************************* + * Utils: + */ +void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr); +unsigned draw_pt_reduced_prim(unsigned prim); + + #endif diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h new file mode 100644 index 0000000000..dccfde99dd --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h @@ -0,0 +1,153 @@ + + +static void FUNC( ARGS, + unsigned count ) +{ + LOCAL_VARS; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + POINT( (i + 0) ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + LINE( DRAW_PIPE_RESET_STIPPLE, + (i + 0), + (i + 1)); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + flags = DRAW_PIPE_RESET_STIPPLE; + + for (i = 1; i < count; i++, flags = 0) { + LINE( flags, + (i - 1), + (i )); + } + + LINE( flags, + (i - 1), + (0 )); + } + break; + + case PIPE_PRIM_LINE_STRIP: + flags = DRAW_PIPE_RESET_STIPPLE; + for (i = 1; i < count; i++, flags = 0) { + LINE( flags, + (i - 1), + (i )); + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0), + (i + 1), + (i + 2 )); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0), + (i + 1 + (i&1)), + (i + 2 - (i&1))); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0 + (i&1)), + (i + 1 - (i&1)), + (i + 2 )); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 1), + (i + 2), + (0 )); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (0), + (i + 1), + (i + 2 )); + } + } + } + break; + + + case PIPE_PRIM_QUADS: + for (i = 0; i+3 < count; i += 4) { + QUAD( (i + 0), + (i + 1), + (i + 2), + (i + 3)); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + QUAD( (i + 2), + (i + 0), + (i + 1), + (i + 3)); + } + break; + + case PIPE_PRIM_POLYGON: + { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + TRIANGLE( flags, + (i + 1), + (i + 2), + (0)); + } + } + break; + + default: + assert(0); + break; + } + + FLUSH; +} + + +#undef TRIANGLE +#undef QUAD +#undef POINT +#undef LINE +#undef FUNC diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index ce3a153f64..22a83ec78f 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -178,6 +178,53 @@ 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) +{ + struct draw_context *draw = emit->draw; + struct translate *translate = emit->translate; + struct vbuf_render *render = draw->render; + void *hw_verts; + +#if 0 + debug_printf("Linear emit\n"); +#endif + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = render->allocate_vertices(render, + (ushort)translate->key.output_stride, + (ushort)count); + if (!hw_verts) { + assert(0); + return; + } + + translate->set_buffer(translate, 0, + vertex_data, stride); + + translate->set_buffer(translate, 1, + &draw->rasterizer->point_size, + 0); + + translate->run(translate, + 0, + vertex_count, + hw_verts); + + render->draw_arrays(render, start, count); + + render->release_vertices(render, + hw_verts, + translate->key.output_stride, + vertex_count); +} + struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) { struct pt_emit *emit = CALLOC_STRUCT(pt_emit); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 100117a9ae..07f4c99164 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -166,6 +166,42 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, } +void draw_pt_fetch_run_linear( struct pt_fetch *fetch, + unsigned start, + unsigned count, + char *verts ) +{ + struct draw_context *draw = fetch->draw; + struct translate *translate = fetch->translate; + unsigned i; + + for (i = 0; i < draw->pt.nr_vertex_buffers; i++) { + translate->set_buffer(translate, + i, + ((char *)draw->pt.user.vbuffer[i] + + draw->pt.vertex_buffer[i].buffer_offset), + draw->pt.vertex_buffer[i].pitch ); + } + + translate->run( translate, + start, + count, + verts ); + + /* Edgeflags are hard to fit into a translate program, populate + * them separately if required. In the setup above they are + * defaulted to one, so only need this if there is reason to change + * that default: + */ + if (fetch->need_edgeflags) { + for (i = 0; i < count; i++) { + struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size); + vh->edgeflag = draw_pt_get_edgeflag( draw, start + i ); + } + } +} + + struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) { struct pt_fetch *fetch = CALLOC_STRUCT(pt_fetch); @@ -184,7 +220,8 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) void draw_pt_fetch_destroy( struct pt_fetch *fetch ) { - translate_cache_destroy(fetch->cache); + if (fetch->cache) + translate_cache_destroy(fetch->cache); FREE(fetch); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index b7b970a297..a1d041a74f 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -258,6 +258,59 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, } +static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count ) +{ + struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; + struct draw_context *draw = feme->draw; + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)feme->translate->key.output_stride, + (ushort)count ); + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices and emit HW verts. + */ + feme->translate->run( feme->translate, + start, + count, + hw_verts ); + + if (0) { + unsigned i; + for (i = 0; i < count; i++) { + debug_printf("\n\nvertex %d:\n", i); + draw_dump_emitted_vertex( feme->vinfo, + (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i ); + } + } + + /* XXX: Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw_arrays( draw->render, + 0, /*start*/ + count ); + + /* Done -- that was easy, wasn't it: + */ + draw->render->release_vertices( draw->render, + hw_verts, + feme->translate->key.output_stride, + count ); + +} + + static void fetch_emit_finish( struct draw_pt_middle_end *middle ) { @@ -268,7 +321,8 @@ static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; - translate_cache_destroy(feme->cache); + if (feme->cache) + translate_cache_destroy(feme->cache); FREE(middle); } @@ -286,10 +340,11 @@ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ) return NULL; } - fetch_emit->base.prepare = fetch_emit_prepare; - fetch_emit->base.run = fetch_emit_run; - fetch_emit->base.finish = fetch_emit_finish; - fetch_emit->base.destroy = fetch_emit_destroy; + fetch_emit->base.prepare = fetch_emit_prepare; + fetch_emit->base.run = fetch_emit_run; + fetch_emit->base.run_linear = fetch_emit_run_linear; + fetch_emit->base.finish = fetch_emit_finish; + fetch_emit->base.destroy = fetch_emit_destroy; fetch_emit->draw = draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c new file mode 100644 index 0000000000..f756d3e0bb --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -0,0 +1,677 @@ +/************************************************************************** + * + * 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> + */ + + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "draw/draw_vs.h" + +#include "translate/translate.h" + +struct fetch_shade_emit; + +struct fse_shader { + struct translate_key key; + + void (*run_linear)( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ); +}; + +/* Prototype fetch, shade, emit-hw-verts all in one go. + */ +struct fetch_shade_emit { + struct draw_pt_front_end base; + + struct draw_context *draw; + + struct translate_key key; + + /* Temporaries: + */ + const float *constants; + unsigned pitch[PIPE_MAX_ATTRIBS]; + const ubyte *src[PIPE_MAX_ATTRIBS]; + unsigned prim; + + /* Points to one of the three hardwired example shaders, below: + */ + struct fse_shader *active; + + /* Temporary: A list of hard-wired shaders. Of course the plan + * would be to generate these for a given (vertex-shader, + * translate-key) pair... + */ + struct fse_shader shader[10]; + int nr_shaders; +}; + + + +/* Not quite passthrough yet -- we're still running the 'shader' here, + * inlined into the vertex fetch function. + */ +static void fetch_xyz_rgb_st( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = fse->constants; + const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; + const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; + const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; + const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; + + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *st = fse->src[2] + start * fse->pitch[2]; + + float *out = (float *)buffer; + + + assert(fse->pitch[1] == 0); + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m0 * ix + m4 * iy + m8 * iz + m12; + out[1] = m1 * ix + m5 * iy + m9 * iz + m13; + out[2] = m2 * ix + m6 * iy + m10 * iz + m14; + out[3] = m3 * ix + m7 * iy + m11 * iz + m15; + xyz += fse->pitch[0]; + } + + { + out[4] = 1.0f; + out[5] = 1.0f; + out[6] = 1.0f; + out[7] = 1.0f; + } + + { + const float *in = (const float *)st; st += fse->pitch[2]; + out[8] = in[0]; + out[9] = in[1]; + out[10] = 0.0f; + out[11] = 1.0f; + } + + out += 12; + } +} + + + +static void fetch_xyz_rgb( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = (const float *)fse->constants; + const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; + const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; + const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; + const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; + + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + + float *out = (float *)buffer; + +// debug_printf("rgb %f %f %f\n", rgb[0], rgb[1], rgb[2]); + + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m0 * ix + m4 * iy + m8 * iz + m12; + out[1] = m1 * ix + m5 * iy + m9 * iz + m13; + out[2] = m2 * ix + m6 * iy + m10 * iz + m14; + out[3] = m3 * ix + m7 * iy + m11 * iz + m15; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + out += 8; + } +} + + + + +static void fetch_xyz_rgb_psiz( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = (const float *)fse->constants; + const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; + const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; + const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; + const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; + + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const float *rgb = (const float *)(fse->src[1] + start * fse->pitch[1]); + const float psiz = 1.0; + + float *out = (float *)buffer; + + + assert(fse->pitch[1] == 0); + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m0 * ix + m4 * iy + m8 * iz + m12; + out[1] = m1 * ix + m5 * iy + m9 * iz + m13; + out[2] = m2 * ix + m6 * iy + m10 * iz + m14; + out[3] = m3 * ix + m7 * iy + m11 * iz + m15; + xyz += fse->pitch[0]; + } + + { + out[4] = rgb[0]; + out[5] = rgb[1]; + out[6] = rgb[2]; + out[7] = 1.0f; + } + + { + out[8] = psiz; + } + + out += 9; + } +} + + + + +static boolean set_prim( struct fetch_shade_emit *fse, + unsigned prim, + unsigned count ) +{ + struct draw_context *draw = fse->draw; + + fse->prim = prim; + + switch (prim) { + case PIPE_PRIM_LINE_LOOP: + if (count > 1024) + return FALSE; + draw->render->set_primitive( draw->render, PIPE_PRIM_LINE_STRIP ); + break; + + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + if (count > 1024) + return FALSE; + draw->render->set_primitive( draw->render, prim ); + break; + + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + draw->render->set_primitive( draw->render, PIPE_PRIM_TRIANGLES ); + break; + + default: + draw->render->set_primitive( draw->render, prim ); + break; + } + + return TRUE; +} + + + + + + +static void fse_prepare( struct draw_pt_front_end *fe, + unsigned prim, + struct draw_pt_middle_end *unused, + unsigned opt ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)fe; + struct draw_context *draw = fse->draw; + unsigned num_vs_inputs = draw->vertex_shader->info.num_inputs; + unsigned num_vs_outputs = draw->vertex_shader->info.num_outputs; + const struct vertex_info *vinfo; + unsigned i; + boolean need_psize = 0; + + + if (draw->pt.user.elts) { + assert(0); + return ; + } + + if (!set_prim(fse, prim, /*count*/1022 )) { + assert(0); + return ; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + + + fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ + num_vs_inputs); /* inputs - fetch from api format */ + + fse->key.output_stride = vinfo->size * 4; + memset(fse->key.element, 0, + fse->key.nr_elements * sizeof(fse->key.element[0])); + + for (i = 0; i < num_vs_inputs; i++) { + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = src->src_format; + + /* Consider ignoring these at this point, ie make generated + * programs independent of this state: + */ + fse->key.element[i].input_buffer = 0; //src->vertex_buffer_index; + fse->key.element[i].input_offset = 0; //src->src_offset; + } + + + { + unsigned dst_offset = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned output_format = PIPE_FORMAT_NONE; + unsigned vs_output = vinfo->src_index[i]; + + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + need_psize = 1; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + vs_output = num_vs_outputs + 1; + + break; + default: + assert(0); + break; + } + + /* The elements in the key correspond to vertex shader output + * numbers, not to positions in the hw vertex description -- + * that's handled by the output_offset field. + */ + fse->key.element[vs_output].output_format = output_format; + fse->key.element[vs_output].output_offset = dst_offset; + + dst_offset += emit_sz; + assert(fse->key.output_stride >= dst_offset); + } + } + + /* To make psize work, really need to tell the vertex shader to + * copy that value from input->output. For 'translate' this was + * implicit for all elements. + */ +#if 0 + if (need_psize) { + unsigned input = num_vs_inputs + 1; + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; + fse->key.element[i].input_buffer = 0; //nr_buffers + 1; + fse->key.element[i].input_offset = 0; + + fse->key.nr_elements += 1; + + } +#endif + + fse->constants = draw->pt.user.constants; + + /* Would normally look up a vertex shader and peruse its list of + * varients somehow. We omitted that step and put all the + * hardcoded "shaders" into an array. We're just making the + * assumption that this happens to be a matching shader... ie + * you're running isosurf, aren't you? + */ + fse->active = NULL; + for (i = 0; i < fse->nr_shaders; i++) { + if (translate_key_compare( &fse->key, &fse->shader[i].key) == 0) + fse->active = &fse->shader[i]; + } + + if (!fse->active) { + assert(0); + return ; + } + + /* Now set buffer pointers: + */ + for (i = 0; i < num_vs_inputs; i++) { + unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; + + fse->src[i] = ((const ubyte *) draw->pt.user.vbuffer[buf] + + draw->pt.vertex_buffer[buf].buffer_offset + + draw->pt.vertex_element[i].src_offset); + + fse->pitch[i] = draw->pt.vertex_buffer[buf].pitch; + + } + + + //return TRUE; +} + + + + + + +#define INDEX(i) (start + (i)) +static void fse_render_linear( struct vbuf_render *render, + unsigned prim, + unsigned start, + unsigned length ) +{ + ushort *tmp = NULL; + unsigned i, j; + + switch (prim) { + case PIPE_PRIM_LINE_LOOP: + tmp = MALLOC( sizeof(ushort) * (length + 1) ); + + for (i = 0; i < length; i++) + tmp[i] = INDEX(i); + tmp[length] = 0; + + render->draw( render, + tmp, + length+1 ); + break; + + + case PIPE_PRIM_QUAD_STRIP: + tmp = MALLOC( sizeof(ushort) * (length / 2 * 6) ); + + for (j = i = 0; i + 3 < length; i += 2, j += 6) { + tmp[j+0] = INDEX(i+0); + tmp[j+1] = INDEX(i+1); + tmp[j+2] = INDEX(i+3); + + tmp[j+3] = INDEX(i+2); + tmp[j+4] = INDEX(i+0); + tmp[j+5] = INDEX(i+3); + } + + if (j) + render->draw( render, tmp, j ); + break; + + case PIPE_PRIM_QUADS: + tmp = MALLOC( sizeof(int) * (length / 4 * 6) ); + + for (j = i = 0; i + 3 < length; i += 4, j += 6) { + tmp[j+0] = INDEX(i+0); + tmp[j+1] = INDEX(i+1); + tmp[j+2] = INDEX(i+3); + + tmp[j+3] = INDEX(i+1); + tmp[j+4] = INDEX(i+2); + tmp[j+5] = INDEX(i+3); + } + + if (j) + render->draw( render, tmp, j ); + break; + + default: + render->draw_arrays( render, + start, + length ); + break; + } + + if (tmp) + FREE(tmp); +} + + + +static boolean do_draw( struct fetch_shade_emit *fse, + unsigned start, unsigned count ) +{ + struct draw_context *draw = fse->draw; + + char *hw_verts = + draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)count ); + + if (!hw_verts) + return FALSE; + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping and viewport transformation are done on hardware. + */ + fse->active->run_linear( fse, + start, count, + hw_verts ); + + /* Draw arrays path to avoid re-emitting index list again and + * again. + */ + fse_render_linear( draw->render, + fse->prim, + 0, + count ); + + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + count ); + + return TRUE; +} + + +static void +fse_run(struct draw_pt_front_end *fe, + pt_elt_func elt_func, + const void *elt_ptr, + unsigned count) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)fe; + unsigned i = 0; + unsigned first, incr; + unsigned start = elt_func(elt_ptr, 0); + + //debug_printf("%s prim %d start %d count %d\n", __FUNCTION__, prim, start, count); + + draw_pt_split_prim(fse->prim, &first, &incr); + + count -= (count - first) % incr; + + while (i + first <= count) { + int nr = MIN2( count - i, 1024 ); + + /* snap to prim boundary + */ + nr -= (nr - first) % incr; + + if (!do_draw( fse, start + i, nr )) { + assert(0); + return ; + } + + /* increment allowing for repeated vertices + */ + i += nr - (first - incr); + } + + //return TRUE; +} + + +static void fse_finish( struct draw_pt_front_end *frontend ) +{ +} + + +static void +fse_destroy( struct draw_pt_front_end *frontend ) +{ + FREE(frontend); +} + +struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ) +{ + struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); + if (!fse) + return NULL; + + fse->base.prepare = fse_prepare; + fse->base.run = fse_run; + fse->base.finish = fse_finish; + fse->base.destroy = fse_destroy; + fse->draw = draw; + + fse->shader[0].run_linear = fetch_xyz_rgb_st; + fse->shader[0].key.nr_elements = 3; + fse->shader[0].key.output_stride = 12 * sizeof(float); + + fse->shader[0].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[0].input_buffer = 0; + fse->shader[0].key.element[0].input_offset = 0; + fse->shader[0].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[0].output_offset = 0; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 16; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 32; + + fse->shader[1].run_linear = fetch_xyz_rgb; + fse->shader[1].key.nr_elements = 2; + fse->shader[1].key.output_stride = 8 * sizeof(float); + + fse->shader[1].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[0].input_buffer = 0; + fse->shader[1].key.element[0].input_offset = 0; + fse->shader[1].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[0].output_offset = 0; + + fse->shader[1].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[1].input_buffer = 0; + fse->shader[1].key.element[1].input_offset = 0; + fse->shader[1].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[1].output_offset = 16; + + fse->shader[2].run_linear = fetch_xyz_rgb_psiz; + fse->shader[2].key.nr_elements = 3; + fse->shader[2].key.output_stride = 9 * sizeof(float); + + fse->shader[2].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[0].input_buffer = 0; + fse->shader[2].key.element[0].input_offset = 0; + fse->shader[2].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[0].output_offset = 0; + + fse->shader[2].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[1].input_buffer = 0; + fse->shader[2].key.element[1].input_offset = 0; + fse->shader[2].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[1].output_offset = 16; + + /* psize is special + * -- effectively add it here as another input!?! + * -- who knows how to add it as a buffer? + */ + fse->shader[2].key.element[2].input_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].input_buffer = 0; + fse->shader[2].key.element[2].input_offset = 0; + fse->shader[2].key.element[2].output_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].output_offset = 32; + + fse->nr_shaders = 3; + + return &fse->base; +} 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 4ec20493c4..dad54690a5 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -162,7 +162,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, fpme->vertex_size, draw_elts, draw_count ); - } + } else { draw_pt_emit( fpme->emit, (const float (*)[4])pipeline_verts->data, @@ -177,6 +177,79 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, } +static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count) +{ + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + struct draw_vertex_shader *shader = draw->vertex_shader; + unsigned opt = fpme->opt; + unsigned alloc_count = align_int( count, 4 ); + + struct vertex_header *pipeline_verts = + (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); + + if (!pipeline_verts) { + /* Not much we can do here - just skip the rendering. + */ + assert(0); + return; + } + + /* Fetch into our vertex buffer + */ + draw_pt_fetch_run_linear( fpme->fetch, + start, + count, + (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. + */ + if (opt & PT_SHADE) + { + shader->run_linear(shader, + (const float (*)[4])pipeline_verts->data, + ( float (*)[4])pipeline_verts->data, + (const float (*)[4])draw->pt.user.constants, + count, + fpme->vertex_size, + fpme->vertex_size); + } + + if (draw_pt_post_vs_run( fpme->post_vs, + pipeline_verts, + count, + fpme->vertex_size )) + { + opt |= PT_PIPELINE; + } + + /* Do we need to run the pipeline? + */ + if (opt & PT_PIPELINE) { + draw_pipeline_run_linear( fpme->draw, + fpme->prim, + pipeline_verts, + count, + fpme->vertex_size); + } + else { + draw_pt_emit_linear( fpme->emit, + (const float (*)[4])pipeline_verts->data, + count, + fpme->vertex_size, + 0, /*start*/ + count ); + } + + FREE(pipeline_verts); +} + + static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) { @@ -206,10 +279,11 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context * if (!fpme) goto fail; - fpme->base.prepare = fetch_pipeline_prepare; - fpme->base.run = fetch_pipeline_run; - fpme->base.finish = fetch_pipeline_finish; - fpme->base.destroy = fetch_pipeline_destroy; + fpme->base.prepare = fetch_pipeline_prepare; + fpme->base.run = fetch_pipeline_run; + fpme->base.run_linear = fetch_pipeline_linear_run; + fpme->base.finish = fetch_pipeline_finish; + fpme->base.destroy = fetch_pipeline_destroy; fpme->draw = draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_middle_fse.c b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c new file mode 100644 index 0000000000..cdb7d260da --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c @@ -0,0 +1,705 @@ +/************************************************************************** + * + * 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> + */ + + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "draw/draw_vs.h" + +#include "translate/translate.h" + +struct fetch_shade_emit; + +struct fse_shader { + struct translate_key key; + + void (*run_linear)( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ); + + void (*run_elts)( const struct fetch_shade_emit *fse, + const unsigned *fetch_elts, + unsigned fetch_count, + char *buffer ); + +}; + +/* Prototype fetch, shade, emit-hw-verts all in one go. + */ +struct fetch_shade_emit { + struct draw_pt_middle_end base; + struct draw_context *draw; + + struct translate_key key; + + /* Temporaries: + */ + const float *constants; + unsigned pitch[PIPE_MAX_ATTRIBS]; + const ubyte *src[PIPE_MAX_ATTRIBS]; + unsigned prim; + + /* Points to one of the three hardwired example shaders, below: + */ + struct fse_shader *active; + + /* Temporary: A list of hard-wired shaders. Of course the plan + * would be to generate these for a given (vertex-shader, + * translate-key) pair... + */ + struct fse_shader shader[10]; + int nr_shaders; +}; + + + +/* Not quite passthrough yet -- we're still running the 'shader' here, + * inlined into the vertex fetch function. + */ +static void shader0_run_linear( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = fse->constants; + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + const ubyte *st = fse->src[2] + start * fse->pitch[2]; + + float *out = (float *)buffer; + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + { + const float *in = (const float *)st; + out[8] = in[0]; + out[9] = in[1]; + out[10] = 0.0f; + out[11] = 1.0f; + st += fse->pitch[2]; + } + + out += 12; + } +} + + + +static void shader1_run_linear( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + float *out = (float *)buffer; + +// debug_printf("rgb %f %f %f\n", rgb[0], rgb[1], rgb[2]); + + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + out += 8; + } +} + + + + +static void shader2_run_linear( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + const float psiz = 1.0; + float *out = (float *)buffer; + + + assert(fse->pitch[1] == 0); + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + { + out[8] = psiz; + } + + out += 9; + } +} + + + + +static void shader0_run_elts( const struct fetch_shade_emit *fse, + const unsigned *elts, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = fse->constants; + float *out = (float *)buffer; + + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + unsigned elt = elts[i]; + { + const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + } + + { + const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + } + + { + const ubyte *st = fse->src[2] + elt * fse->pitch[2]; + const float *in = (const float *)st; + out[8] = in[0]; + out[9] = in[1]; + out[10] = 0.0f; + out[11] = 1.0f; + } + + out += 12; + } +} + + + +static void shader1_run_elts( const struct fetch_shade_emit *fse, + const unsigned *elts, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + float *out = (float *)buffer; + + for (i = 0; i < count; i++) { + unsigned elt = elts[i]; + + { + const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + out += 8; + } +} + + + + +static void shader2_run_elts( const struct fetch_shade_emit *fse, + const unsigned *elts, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + const float psiz = 1.0; + float *out = (float *)buffer; + + for (i = 0; i < count; i++) { + unsigned elt = elts[i]; + { + const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + } + + { + const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; + out[4] = rgb[0]; + out[5] = rgb[1]; + out[6] = rgb[2]; + out[7] = 1.0f; + } + + { + out[8] = psiz; + } + + out += 9; + } +} + + + +static void fse_prepare( struct draw_pt_middle_end *middle, + unsigned prim, + unsigned opt ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + unsigned num_vs_inputs = draw->vertex_shader->info.num_inputs; + unsigned num_vs_outputs = draw->vertex_shader->info.num_outputs; + const struct vertex_info *vinfo; + unsigned i; + boolean need_psize = 0; + + + if (draw->pt.user.elts) { + assert(0); + return ; + } + + if (!draw->render->set_primitive( draw->render, + prim )) { + assert(0); + return; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + + + fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ + num_vs_inputs); /* inputs - fetch from api format */ + + fse->key.output_stride = vinfo->size * 4; + memset(fse->key.element, 0, + fse->key.nr_elements * sizeof(fse->key.element[0])); + + for (i = 0; i < num_vs_inputs; i++) { + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = src->src_format; + + /* Consider ignoring these at this point, ie make generated + * programs independent of this state: + */ + fse->key.element[i].input_buffer = 0; //src->vertex_buffer_index; + fse->key.element[i].input_offset = 0; //src->src_offset; + } + + + { + unsigned dst_offset = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned output_format = PIPE_FORMAT_NONE; + unsigned vs_output = vinfo->src_index[i]; + + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + need_psize = 1; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + vs_output = num_vs_outputs + 1; + + break; + default: + assert(0); + break; + } + + /* The elements in the key correspond to vertex shader output + * numbers, not to positions in the hw vertex description -- + * that's handled by the output_offset field. + */ + fse->key.element[vs_output].output_format = output_format; + fse->key.element[vs_output].output_offset = dst_offset; + + dst_offset += emit_sz; + assert(fse->key.output_stride >= dst_offset); + } + } + + /* To make psize work, really need to tell the vertex shader to + * copy that value from input->output. For 'translate' this was + * implicit for all elements. + */ +#if 0 + if (need_psize) { + unsigned input = num_vs_inputs + 1; + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; + fse->key.element[i].input_buffer = 0; //nr_buffers + 1; + fse->key.element[i].input_offset = 0; + + fse->key.nr_elements += 1; + + } +#endif + + fse->constants = draw->pt.user.constants; + + /* Would normally look up a vertex shader and peruse its list of + * varients somehow. We omitted that step and put all the + * hardcoded "shaders" into an array. We're just making the + * assumption that this happens to be a matching shader... ie + * you're running isosurf, aren't you? + */ + fse->active = NULL; + for (i = 0; i < fse->nr_shaders; i++) { + if (translate_key_compare( &fse->key, &fse->shader[i].key) == 0) + fse->active = &fse->shader[i]; + } + + if (!fse->active) { + assert(0); + return ; + } + + /* Now set buffer pointers: + */ + for (i = 0; i < num_vs_inputs; i++) { + unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; + + fse->src[i] = ((const ubyte *) draw->pt.user.vbuffer[buf] + + draw->pt.vertex_buffer[buf].buffer_offset + + draw->pt.vertex_element[i].src_offset); + + fse->pitch[i] = draw->pt.vertex_buffer[buf].pitch; + + } + + + //return TRUE; +} + + + + + + + +static void fse_run_linear( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + + char *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)count ); + + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping and viewport transformation are done elsewhere -- + * either by the API or on hardware, or for some other reason not + * required... + */ + fse->active->run_linear( fse, + start, count, + hw_verts ); + + /* 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, + hw_verts, + fse->key.output_stride, + count ); +} + + +static void +fse_run(struct draw_pt_middle_end *middle, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)fetch_count ); + if (!hw_verts) { + assert(0); + return; + } + + + /* Single routine to fetch vertices, run shader and emit HW verts. + */ + fse->active->run_elts( fse, + fetch_elts, + fetch_count, + hw_verts ); + + draw->render->draw( draw->render, + draw_elts, + draw_count ); + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + fetch_count ); + +} + + +static void fse_finish( struct draw_pt_middle_end *middle ) +{ +} + + +static void +fse_destroy( struct draw_pt_middle_end *middle ) +{ + FREE(middle); +} + +struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ) +{ + struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); + if (!fse) + return NULL; + + fse->base.prepare = fse_prepare; + fse->base.run = fse_run; + fse->base.run_linear = fse_run_linear; + fse->base.finish = fse_finish; + fse->base.destroy = fse_destroy; + fse->draw = draw; + + fse->shader[0].run_linear = shader0_run_linear; + fse->shader[0].run_elts = shader0_run_elts; + fse->shader[0].key.nr_elements = 3; + fse->shader[0].key.output_stride = 12 * sizeof(float); + + fse->shader[0].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[0].input_buffer = 0; + fse->shader[0].key.element[0].input_offset = 0; + fse->shader[0].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[0].output_offset = 0; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 16; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 32; + + fse->shader[1].run_linear = shader1_run_linear; + fse->shader[1].run_elts = shader1_run_elts; + fse->shader[1].key.nr_elements = 2; + fse->shader[1].key.output_stride = 8 * sizeof(float); + + fse->shader[1].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[0].input_buffer = 0; + fse->shader[1].key.element[0].input_offset = 0; + fse->shader[1].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[0].output_offset = 0; + + fse->shader[1].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[1].input_buffer = 0; + fse->shader[1].key.element[1].input_offset = 0; + fse->shader[1].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[1].output_offset = 16; + + fse->shader[2].run_linear = shader2_run_linear; + fse->shader[2].run_elts = shader2_run_elts; + fse->shader[2].key.nr_elements = 3; + fse->shader[2].key.output_stride = 9 * sizeof(float); + + fse->shader[2].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[0].input_buffer = 0; + fse->shader[2].key.element[0].input_offset = 0; + fse->shader[2].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[0].output_offset = 0; + + fse->shader[2].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[1].input_buffer = 0; + fse->shader[2].key.element[1].input_offset = 0; + fse->shader[2].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[1].output_offset = 16; + + /* psize is special + * -- effectively add it here as another input!?! + * -- who knows how to add it as a buffer? + */ + fse->shader[2].key.element[2].input_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].input_buffer = 0; + fse->shader[2].key.element[2].input_offset = 0; + fse->shader[2].key.element[2].output_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].output_offset = 32; + + fse->nr_shaders = 3; + + return &fse->base; +} diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c new file mode 100644 index 0000000000..32c8a9632c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_util.c @@ -0,0 +1,103 @@ +/************************************************************************** + * + * 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> + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_pt.h" + +void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + break; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + break; + case PIPE_PRIM_LINE_STRIP: + case PIPE_PRIM_LINE_LOOP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + break; + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + break; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + break; + default: + assert(0); + *first = 0; + *incr = 1; /* set to one so that count % incr works */ + break; + } +} + + +unsigned draw_pt_reduced_prim(unsigned prim) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + return PIPE_PRIM_POINTS; + case PIPE_PRIM_LINES: + case PIPE_PRIM_LINE_STRIP: + case PIPE_PRIM_LINE_LOOP: + return PIPE_PRIM_LINES; + case PIPE_PRIM_TRIANGLES: + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + return PIPE_PRIM_TRIANGLES; + default: + assert(0); + return PIPE_PRIM_POINTS; + } +} + + diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 355093f945..06fd866ccd 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -43,6 +43,8 @@ struct varray_frontend { unsigned draw_count; unsigned fetch_count; + unsigned fetch_start; + struct draw_pt_middle_end *middle; unsigned input_prim; @@ -56,6 +58,11 @@ static void varray_flush(struct varray_frontend *varray) debug_printf("FLUSH fc = %d, dc = %d\n", varray->fetch_count, varray->draw_count); + debug_printf("\telt0 = %d, eltx = %d, draw0 = %d, drawx = %d\n", + varray->fetch_elts[0], + varray->fetch_elts[varray->fetch_count-1], + varray->draw_elts[0], + varray->draw_elts[varray->draw_count-1]); #endif varray->middle->run(varray->middle, varray->fetch_elts, @@ -68,20 +75,43 @@ static void varray_flush(struct varray_frontend *varray) varray->draw_count = 0; } -#if 0 -static void varray_check_flush(struct varray_frontend *varray) +static void varray_flush_linear(struct varray_frontend *varray, + unsigned start, unsigned count) { - if (varray->draw_count + 6 >= DRAW_MAX/* || - varray->fetch_count + 4 >= FETCH_MAX*/) { - varray_flush(varray); + if (count) { +#if 0 + debug_printf("FLUSH LINEAR start = %d, count = %d\n", + start, + count); +#endif + assert(varray->middle->run_linear); + varray->middle->run_linear(varray->middle, start, count); } } + +static INLINE void fetch_init(struct varray_frontend *varray, + unsigned count) +{ + unsigned idx; +#if 0 + debug_printf("FETCH INIT c = %d, fs = %d\n", + count, + varray->fetch_start); #endif + for (idx = 0; idx < count; ++idx) { + varray->fetch_elts[idx] = varray->fetch_start + idx; + } + varray->fetch_start += idx; + varray->fetch_count = idx; +} + + + static INLINE void add_draw_el(struct varray_frontend *varray, - int idx, ushort flags) + int idx) { - varray->draw_elts[varray->draw_count++] = idx | flags; + varray->draw_elts[varray->draw_count++] = idx; } @@ -90,106 +120,52 @@ static INLINE void varray_triangle( struct varray_frontend *varray, unsigned i1, unsigned i2 ) { - add_draw_el(varray, i0, 0); - add_draw_el(varray, i1, 0); - add_draw_el(varray, i2, 0); -} - -static INLINE void varray_triangle_flags( struct varray_frontend *varray, - ushort flags, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - add_draw_el(varray, i0, flags); - add_draw_el(varray, i1, 0); - add_draw_el(varray, i2, 0); + add_draw_el(varray, i0); + add_draw_el(varray, i1); + add_draw_el(varray, i2); } static INLINE void varray_line( struct varray_frontend *varray, unsigned i0, unsigned i1 ) { - add_draw_el(varray, i0, 0); - add_draw_el(varray, i1, 0); -} - - -static INLINE void varray_line_flags( struct varray_frontend *varray, - ushort flags, - unsigned i0, - unsigned i1 ) -{ - add_draw_el(varray, i0, flags); - add_draw_el(varray, i1, 0); + add_draw_el(varray, i0); + add_draw_el(varray, i1); } static INLINE void varray_point( struct varray_frontend *varray, unsigned i0 ) { - add_draw_el(varray, i0, 0); -} - -static INLINE void varray_quad( struct varray_frontend *varray, - unsigned i0, - unsigned i1, - unsigned i2, - unsigned i3 ) -{ - varray_triangle( varray, i0, i1, i3 ); - varray_triangle( varray, i1, i2, i3 ); -} - -static INLINE void varray_ef_quad( struct varray_frontend *varray, - unsigned i0, - unsigned i1, - unsigned i2, - unsigned i3 ) -{ - const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; - const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; - - varray_triangle_flags( varray, - DRAW_PIPE_RESET_STIPPLE | omitEdge1, - i0, i1, i3 ); - - varray_triangle_flags( varray, - omitEdge2, - i1, i2, i3 ); + add_draw_el(varray, i0); } -/* At least for now, we're back to using a template include file for - * this. The two paths aren't too different though - it may be - * possible to reunify them. - */ -#define TRIANGLE(vc,flags,i0,i1,i2) varray_triangle_flags(vc,flags,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) varray_ef_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) varray_line_flags(vc,flags,i0,i1) -#define POINT(vc,i0) varray_point(vc,i0) -#define FUNC varray_run_extras -#include "draw_pt_varray_tmp.h" -#define TRIANGLE(vc,flags,i0,i1,i2) varray_triangle(vc,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) varray_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) varray_line(vc,i0,i1) +#if 0 +#define TRIANGLE(flags,i0,i1,i2) varray_triangle(varray,i0,i1,i2) +#define LINE(flags,i0,i1) varray_line(varray,i0,i1) +#define POINT(i0) varray_point(varray,i0) +#define FUNC varray_decompose +#include "draw_pt_decompose.h" +#else +#define TRIANGLE(vc,i0,i1,i2) varray_triangle(vc,i0,i1,i2) +#define LINE(vc,i0,i1) varray_line(vc,i0,i1) #define POINT(vc,i0) varray_point(vc,i0) #define FUNC varray_run -#include "draw_pt_varray_tmp.h" - - +#include "draw_pt_varray_tmp_linear.h" +#endif -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { +static unsigned decompose_prim[PIPE_PRIM_POLYGON + 1] = { PIPE_PRIM_POINTS, PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, + PIPE_PRIM_LINE_STRIP, + PIPE_PRIM_LINES, /* decomposed */ PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES + PIPE_PRIM_TRIANGLE_STRIP, + PIPE_PRIM_TRIANGLES, /* decomposed */ + PIPE_PRIM_QUADS, + PIPE_PRIM_QUAD_STRIP, + PIPE_PRIM_TRIANGLES /* decomposed */ }; @@ -201,17 +177,10 @@ static void varray_prepare(struct draw_pt_front_end *frontend, { struct varray_frontend *varray = (struct varray_frontend *)frontend; - if (opt & PT_PIPELINE) - { - varray->base.run = varray_run_extras; - } - else - { - varray->base.run = varray_run; - } + varray->base.run = varray_run; varray->input_prim = prim; - varray->output_prim = reduced_prim[prim]; + varray->output_prim = decompose_prim[prim]; varray->middle = middle; middle->prepare(middle, varray->output_prim, opt); diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index b9a319b253..6979f6b544 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -10,32 +10,44 @@ static void FUNC(struct draw_pt_front_end *frontend, boolean flatfirst = (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first); - unsigned i, flags; + unsigned i, j, flags; + unsigned first, incr; + + varray->fetch_start = start; + + draw_pt_split_prim(varray->input_prim, &first, &incr); #if 0 - debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count); -#endif -#if 0 - debug_printf("INPUT PRIM = %d (start = %d, count = %d)\n", varray->input_prim, + debug_printf("%s (%d) %d/%d\n", __FUNCTION__, + varray->input_prim, start, count); #endif - for (i = 0; i < count; ++i) { - varray->fetch_elts[i] = start + i; - } - varray->fetch_count = count; - switch (varray->input_prim) { case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - POINT(varray, i + 0); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i < end; i++) { + POINT(varray, i + 0); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - LINE(varray, DRAW_PIPE_RESET_STIPPLE, - i + 0, i + 1); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+1 < end; i += 2) { + LINE(varray, DRAW_PIPE_RESET_STIPPLE, + i + 0, i + 1); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } break; @@ -43,38 +55,81 @@ static void FUNC(struct draw_pt_front_end *frontend, if (count >= 2) { flags = DRAW_PIPE_RESET_STIPPLE; - for (i = 1; i < count; i++, flags = 0) { - LINE(varray, flags, i - 1, i); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + LINE(varray, flags, i - 1, 0); + i = end; + fetch_init(varray, end); + varray_flush(varray); } - LINE(varray, flags, i - 1, 0); } break; case PIPE_PRIM_LINE_STRIP: flags = DRAW_PIPE_RESET_STIPPLE; - for (i = 1; i < count; i++, flags = 0) { - LINE(varray, flags, i - 1, i); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0, i + 1, i + 2); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i += 3) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1, i + 2); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_TRIANGLE_STRIP: if (flatfirst) { - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0, i + 1 + (i&1), i + 2 - (i&1)); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1 + (i&1), i + 2 - (i&1)); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } } else { - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0 + (i&1), i + 1 - (i&1), i + 2); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i + 2 < end; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0 + (i&1), i + 1 - (i&1), i + 2); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } } break; @@ -83,51 +138,89 @@ static void FUNC(struct draw_pt_front_end *frontend, if (count >= 3) { if (flatfirst) { flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, flags, i + 1, i + 2, 0); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } } else { flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, flags, 0, i + 1, i + 2); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, flags, 0, i + 1, i + 2); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } } } break; case PIPE_PRIM_QUADS: - for (i = 0; i+3 < count; i += 4) { - QUAD(varray, i + 0, i + 1, i + 2, i + 3); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+3 < end; i += 4) { + QUAD(varray, i + 0, i + 1, i + 2, i + 3); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - QUAD(varray, i + 2, i + 0, i + 1, i + 3); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+3 < end; i += 2) { + QUAD(varray, i + 2, i + 0, i + 1, i + 3); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } break; case PIPE_PRIM_POLYGON: { - /* These bitflags look a little odd because we submit the - * vertices as (1,2,0) to satisfy flatshade requirements. - */ - const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; - const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; - - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; - - for (i = 0; i+2 < count; i++, flags = edge_middle) { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++, flags = edge_middle) { if (i + 3 == count) flags |= edge_last; - TRIANGLE(varray, flags, i + 1, i + 2, 0); - } + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); } - break; + } + break; default: assert(0); diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h new file mode 100644 index 0000000000..114ed371a0 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -0,0 +1,94 @@ +static unsigned trim( unsigned count, unsigned first, unsigned incr ) +{ + return count - (count - first) % incr; +} + +static void FUNC(struct draw_pt_front_end *frontend, + pt_elt_func get_elt, + const void *elts, + unsigned count) +{ + struct varray_frontend *varray = (struct varray_frontend *)frontend; + unsigned start = (unsigned)elts; + + unsigned i, j; + unsigned first, incr; + + varray->fetch_start = start; + + draw_pt_split_prim(varray->input_prim, &first, &incr); + + /* Sanitize primitive length: + */ + count = trim(count, first, incr); + if (count < first) + return; + +#if 0 + debug_printf("%s (%d) %d/%d\n", __FUNCTION__, + varray->input_prim, + start, count); +#endif + + switch (varray->input_prim) { + case PIPE_PRIM_POINTS: + case PIPE_PRIM_LINES: + case PIPE_PRIM_TRIANGLES: + case PIPE_PRIM_LINE_STRIP: + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + for (j = 0; j < count;) { + unsigned remaining = count - j; + unsigned nr = trim( MIN2(FETCH_MAX, remaining), first, incr ); + varray_flush_linear(varray, start + j, nr); + j += nr; + if (nr != remaining) + j -= (first - incr); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++) { + LINE(varray, i - 1, i); + } + LINE(varray, i - 1, 0); + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + } + break; + + + case PIPE_PRIM_POLYGON: + case PIPE_PRIM_TRIANGLE_FAN: + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 2; i < end; i++) { + TRIANGLE(varray, 0, i - 1, i); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + break; + + default: + assert(0); + break; + } + + varray_flush(varray); +} + +#undef TRIANGLE +#undef QUAD +#undef POINT +#undef LINE +#undef FUNC diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 6b3fb1406b..6c17edba34 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -104,23 +104,11 @@ static INLINE void vcache_elt( struct vcache_frontend *vcache, static void vcache_triangle( struct vcache_frontend *vcache, + ushort flags, unsigned i0, unsigned i1, unsigned i2 ) { - vcache_elt(vcache, i0, 0); - vcache_elt(vcache, i1, 0); - vcache_elt(vcache, i2, 0); - vcache_check_flush(vcache); -} - - -static void vcache_triangle_flags( struct vcache_frontend *vcache, - ushort flags, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ vcache_elt(vcache, i0, flags); vcache_elt(vcache, i1, 0); vcache_elt(vcache, i2, 0); @@ -128,20 +116,10 @@ static void vcache_triangle_flags( struct vcache_frontend *vcache, } static void vcache_line( struct vcache_frontend *vcache, + ushort flags, unsigned i0, unsigned i1 ) { - vcache_elt(vcache, i0, 0); - vcache_elt(vcache, i1, 0); - vcache_check_flush(vcache); -} - - -static void vcache_line_flags( struct vcache_frontend *vcache, - ushort flags, - unsigned i0, - unsigned i1 ) -{ vcache_elt(vcache, i0, flags); vcache_elt(vcache, i1, 0); vcache_check_flush(vcache); @@ -161,61 +139,65 @@ static void vcache_quad( struct vcache_frontend *vcache, unsigned i2, unsigned i3 ) { - vcache_triangle( vcache, i0, i1, i3 ); - vcache_triangle( vcache, i1, i2, i3 ); -} - -static void vcache_ef_quad( struct vcache_frontend *vcache, - unsigned i0, - unsigned i1, - unsigned i2, - unsigned i3 ) -{ - const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; - const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; - - vcache_triangle_flags( vcache, - DRAW_PIPE_RESET_STIPPLE | omitEdge1, - i0, i1, i3 ); - - vcache_triangle_flags( vcache, - omitEdge2, - i1, i2, i3 ); + vcache_triangle( vcache, + ( DRAW_PIPE_RESET_STIPPLE | + DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_2 ), + i0, i1, i3 ); + + vcache_triangle( vcache, + ( DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_1 ), + i1, i2, i3 ); } /* At least for now, we're back to using a template include file for * this. The two paths aren't too different though - it may be * possible to reunify them. */ -#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) vcache_ef_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) vcache_line_flags(vc,flags,i0,i1) -#define POINT(vc,i0) vcache_point(vc,i0) -#define FUNC vcache_run_extras -#include "draw_pt_vcache_tmp.h" - -#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) vcache_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) vcache_line(vc,i0,i1) -#define POINT(vc,i0) vcache_point(vc,i0) +#define TRIANGLE(flags,i0,i1,i2) \ + vcache_triangle(vcache, \ + flags, \ + get_elt(elts,i0), \ + get_elt(elts,i1), \ + get_elt(elts,i2)) + +#define QUAD(i0,i1,i2,i3) \ + vcache_quad(vcache, \ + get_elt(elts,i0), \ + get_elt(elts,i1), \ + get_elt(elts,i2), \ + get_elt(elts,i3)) + +#define LINE(flags,i0,i1) \ + vcache_line(vcache, \ + flags, \ + get_elt(elts,i0), \ + get_elt(elts,i1)) + +#define POINT(i0) \ + vcache_point(vcache, \ + get_elt(elts,i0)) + #define FUNC vcache_run -#include "draw_pt_vcache_tmp.h" +#define ARGS \ + struct draw_pt_front_end *frontend, \ + pt_elt_func get_elt, \ + const void *elts +#define LOCAL_VARS \ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; \ + struct draw_context *draw = vcache->draw; \ + boolean flatfirst = (draw->rasterizer->flatshade && \ + draw->rasterizer->flatshade_first); \ + unsigned prim = vcache->input_prim; \ + unsigned i, flags; +#define FLUSH vcache_flush( vcache ) + +#include "draw_pt_decompose.h" -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; @@ -226,17 +208,9 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - if (opt & PT_PIPELINE) - { - vcache->base.run = vcache_run_extras; - } - else - { - vcache->base.run = vcache_run; - } - + vcache->base.run = vcache_run; vcache->input_prim = prim; - vcache->output_prim = reduced_prim[prim]; + vcache->output_prim = draw_pt_reduced_prim(prim); vcache->middle = middle; middle->prepare( middle, vcache->output_prim, opt ); diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index fe9cd8375e..a48e546139 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -88,7 +88,8 @@ static void softpipe_destroy( struct pipe_context *pipe ) struct pipe_winsys *ws = pipe->winsys; uint i; - draw_destroy( softpipe->draw ); + if (softpipe->draw) + draw_destroy( softpipe->draw ); softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); @@ -216,8 +217,12 @@ softpipe_create( struct pipe_screen *screen, * Create drawing context and plug our rendering stage into it. */ softpipe->draw = draw_create(); - assert(softpipe->draw); + if (!softpipe->draw) + goto fail; + softpipe->setup = sp_draw_render_stage(softpipe); + if (!softpipe->setup) + goto fail; if (GETENV( "SP_NO_RAST" ) != NULL) softpipe->no_rast = TRUE; @@ -241,4 +246,8 @@ softpipe_create( struct pipe_screen *screen, sp_init_surface_functions(softpipe); return &softpipe->pipe; + + fail: + softpipe_destroy(&softpipe->pipe); + return NULL; } diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 1cf9ffa632..941ab62e00 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -64,16 +64,17 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) } +typedef const float (*cptrf4)[4]; static void do_tri(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - + setup_tri( setup->setup, - prim->v[0]->data, - prim->v[1]->data, - prim->v[2]->data ); + (cptrf4)prim->v[0]->data, + (cptrf4)prim->v[1]->data, + (cptrf4)prim->v[2]->data ); } static void @@ -82,8 +83,8 @@ do_line(struct draw_stage *stage, struct prim_header *prim) struct setup_stage *setup = setup_stage( stage ); setup_line( setup->setup, - prim->v[0]->data, - prim->v[1]->data ); + (cptrf4)prim->v[0]->data, + (cptrf4)prim->v[1]->data ); } static void @@ -92,7 +93,7 @@ do_point(struct draw_stage *stage, struct prim_header *prim) struct setup_stage *setup = setup_stage( stage ); setup_point( setup->setup, - prim->v[0]->data ); + (cptrf4)prim->v[0]->data ); } diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index e063fe82ef..e9fae951e0 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -116,30 +116,28 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) - if (prim == PIPE_PRIM_TRIANGLES || - prim == PIPE_PRIM_LINES || - prim == PIPE_PRIM_POINTS) { - cvbr->prim = prim; - return TRUE; - } - else { - return FALSE; - } + cvbr->prim = prim; + return TRUE; } +static INLINE cptrf4 get_vert( const void *vertex_buffer, + int index, + int stride ) +{ + return (cptrf4)((char *)vertex_buffer + index * stride); +} static void -sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) +sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; - unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); - unsigned i, j; - void *vertex_buffer = cvbr->vertex_buffer; - cptrf4 v[3]; + unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); + unsigned i; + const void *vertex_buffer = cvbr->vertex_buffer; /* XXX: break this dependency - make setup_context live under * softpipe, rename the old "setup" draw stage to something else. @@ -149,40 +147,98 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) switch (cvbr->prim) { - case PIPE_PRIM_TRIANGLES: - for (i = 0; i < nr_indices; i += 3) { - for (j = 0; j < 3; j++) - v[j] = (cptrf4)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - setup_tri( setup_ctx, - v[0], - v[1], - v[2]); + case PIPE_PRIM_POINTS: + for (i = 0; i < nr; i++) { + setup_point( setup_ctx, + get_vert(vertex_buffer, indices[i-0], stride) ); } break; case PIPE_PRIM_LINES: - for (i = 0; i < nr_indices; i += 2) { - for (j = 0; j < 2; j++) - v[j] = (cptrf4)((char *)vertex_buffer + - indices[i+j] * vertex_size); + for (i = 1; i < nr; i += 2) { + setup_line( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } + break; + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < nr; i ++) { setup_line( setup_ctx, - v[0], - v[1] ); + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); } break; - case PIPE_PRIM_POINTS: - for (i = 0; i < nr_indices; i++) { - v[0] = (cptrf4)((char *)vertex_buffer + - indices[i] * vertex_size); + case PIPE_PRIM_LINE_LOOP: + for (i = 1; i < nr; i ++) { + setup_line( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } + if (nr) { + setup_line( setup_ctx, + get_vert(vertex_buffer, indices[nr-1], stride), + get_vert(vertex_buffer, indices[0], stride) ); + } + break; - setup_point( setup_ctx, - v[0] ); + + case PIPE_PRIM_TRIANGLES: + for (i = 2; i < nr; i += 3) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 2; i < nr; i += 1) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i+(i&1)-2], stride), + get_vert(vertex_buffer, indices[i-(i&1)-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + for (i = 2; i < nr; i += 1) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[0], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); } break; + case PIPE_PRIM_QUADS: + for (i = 3; i < nr; i += 4) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + case PIPE_PRIM_QUAD_STRIP: + for (i = 3; i < nr; i += 2) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + default: + assert(0); } /* XXX: why are we calling this??? If we had to call something, it @@ -202,131 +258,107 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; - const void *vertex_buffer = cvbr->vertex_buffer; - const unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); + const void *vertex_buffer = NULL; + const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i; struct setup_context *setup_ctx = sp_draw_setup_context(setup); - cptrf4 v[3]; -#define VERTEX(I) \ - (cptrf4) ((char *) vertex_buffer + (I) * vertex_size) + vertex_buffer = (void *)get_vert(cvbr->vertex_buffer, start, stride); switch (cvbr->prim) { case PIPE_PRIM_POINTS: for (i = 0; i < nr; i++) { - v[0] = VERTEX(i); - setup_point( setup_ctx, v[0] ); + setup_point( setup_ctx, + get_vert(vertex_buffer, i-0, stride) ); } break; + case PIPE_PRIM_LINES: - assert(nr % 2 == 0); - for (i = 0; i < nr; i += 2) { - v[0] = VERTEX(i); - v[1] = VERTEX(i + 1); - setup_line( setup_ctx, v[0], v[1] ); + for (i = 1; i < nr; i += 2) { + setup_line( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); } break; + case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < nr; i++) { - v[0] = VERTEX(i - 1); - v[1] = VERTEX(i); - setup_line( setup_ctx, v[0], v[1] ); + for (i = 1; i < nr; i ++) { + setup_line( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); } break; + + case PIPE_PRIM_LINE_LOOP: + for (i = 1; i < nr; i ++) { + setup_line( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); + } + if (nr) { + setup_line( setup_ctx, + get_vert(vertex_buffer, nr-1, stride), + get_vert(vertex_buffer, 0, stride) ); + } + break; + + case PIPE_PRIM_TRIANGLES: - assert(nr % 3 == 0); - for (i = 0; i < nr; i += 3) { - v[0] = VERTEX(i + 0); - v[1] = VERTEX(i + 1); - v[2] = VERTEX(i + 2); + for (i = 2; i < nr; i += 3) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; + case PIPE_PRIM_TRIANGLE_STRIP: - assert(nr >= 3); - for (i = 2; i < nr; i++) { - v[0] = VERTEX(i - 2); - v[1] = VERTEX(i - 1); - v[2] = VERTEX(i); + for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i+(i&1)-2, stride), + get_vert(vertex_buffer, i-(i&1)-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; + case PIPE_PRIM_TRIANGLE_FAN: - assert(nr >= 3); - for (i = 2; i < nr; i++) { - v[0] = VERTEX(0); - v[1] = VERTEX(i - 1); - v[2] = VERTEX(i); + case PIPE_PRIM_POLYGON: + for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, 0, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; case PIPE_PRIM_QUADS: - assert(nr % 4 == 0); - for (i = 0; i < nr; i += 4) { - v[0] = VERTEX(i + 0); - v[1] = VERTEX(i + 1); - v[2] = VERTEX(i + 2); + for (i = 3; i < nr; i += 4) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-0, stride)); - v[0] = VERTEX(i + 0); - v[1] = VERTEX(i + 2); - v[2] = VERTEX(i + 3); setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; case PIPE_PRIM_QUAD_STRIP: - assert(nr >= 4); - for (i = 2; i < nr; i += 2) { - v[0] = VERTEX(i - 2); - v[1] = VERTEX(i); - v[2] = VERTEX(i + 1); + for (i = 3; i < nr; i += 2) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-0, stride)); - v[0] = VERTEX(i - 2); - v[1] = VERTEX(i + 1); - v[2] = VERTEX(i - 1); setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); - } - break; - case PIPE_PRIM_POLYGON: - /* draw as tri fan */ - for (i = 2; i < nr; i++) { - v[0] = VERTEX(0); - v[1] = VERTEX(i - 1); - v[2] = VERTEX(i); - setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-0, stride)); } break; default: - /* XXX finish remaining prim types */ assert(0); } - -#undef VERTEX } diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index 26b722f343..8a32c54349 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -797,8 +797,14 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) pipe = xmesa_create_i965simple(xmesa_get_pipe_winsys_aub(v)); } + if (pipe == NULL) + goto fail; + c->st = st_create_context(pipe, &v->mesa_visual, share_list ? share_list->st : NULL); + if (c->st == NULL) + goto fail; + mesaCtx = c->st->ctx; c->st->ctx->DriverCtx = c; @@ -818,6 +824,14 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) #endif return c; + + fail: + if (c->st) + st_destroy_context(c->st); + if (pipe) + pipe->destroy(pipe); + FREE(c); + return NULL; } |