diff options
Diffstat (limited to 'src/gallium/drivers/svga')
41 files changed, 473 insertions, 354 deletions
diff --git a/src/gallium/drivers/svga/svga_context.c b/src/gallium/drivers/svga/svga_context.c index af99c9de37..d499ae6acc 100644 --- a/src/gallium/drivers/svga/svga_context.c +++ b/src/gallium/drivers/svga/svga_context.c @@ -26,7 +26,7 @@ #include "svga_cmd.h" #include "pipe/p_defines.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_screen.h" #include "util/u_memory.h" #include "util/u_bitmask.h" @@ -126,7 +126,8 @@ svga_is_buffer_referenced( struct pipe_context *pipe, } -struct pipe_context *svga_context_create( struct pipe_screen *screen ) +struct pipe_context *svga_context_create( struct pipe_screen *screen, + void *priv ) { struct svga_screen *svgascreen = svga_screen(screen); struct svga_context *svga = NULL; @@ -138,6 +139,7 @@ struct pipe_context *svga_context_create( struct pipe_screen *screen ) svga->pipe.winsys = screen->winsys; svga->pipe.screen = screen; + svga->pipe.priv = priv; svga->pipe.destroy = svga_destroy; svga->pipe.clear = svga_clear; @@ -215,7 +217,6 @@ struct pipe_context *svga_context_create( struct pipe_screen *screen ) svga->state.hw_draw.num_views = 0; svga->dirty = ~0; - svga->state.white_fs_id = SVGA3D_INVALID_ID; LIST_INITHEAD(&svga->dirty_buffers); diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h index e2a96034d1..f9a641c6df 100644 --- a/src/gallium/drivers/svga/svga_context.h +++ b/src/gallium/drivers/svga/svga_context.h @@ -326,10 +326,6 @@ struct svga_context unsigned texture_timestamp; - /* Internally generated shaders: - */ - unsigned white_fs_id; - /* */ struct svga_sw_state sw; @@ -429,6 +425,10 @@ void svga_context_flush( struct svga_context *svga, void svga_hwtnl_flush_retry( struct svga_context *svga ); +struct pipe_context * +svga_context_create(struct pipe_screen *screen, + void *priv); + /*********************************************************************** * Inline conversion functions. These are better-typed than the diff --git a/src/gallium/drivers/svga/svga_draw.c b/src/gallium/drivers/svga/svga_draw.c index ca73cf9d5a..f4d2d8992c 100644 --- a/src/gallium/drivers/svga/svga_draw.c +++ b/src/gallium/drivers/svga/svga_draw.c @@ -24,7 +24,7 @@ **********************************************************/ #include "pipe/p_compiler.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_memory.h" #include "util/u_math.h" diff --git a/src/gallium/drivers/svga/svga_draw_arrays.c b/src/gallium/drivers/svga/svga_draw_arrays.c index 75492dffca..6192aa96b1 100644 --- a/src/gallium/drivers/svga/svga_draw_arrays.c +++ b/src/gallium/drivers/svga/svga_draw_arrays.c @@ -25,8 +25,7 @@ #include "svga_cmd.h" -#include "pipe/p_inlines.h" -#include "util/u_prim.h" +#include "util/u_inlines.h" #include "indices/u_indices.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_draw_elements.c b/src/gallium/drivers/svga/svga_draw_elements.c index 167d817831..e8097d82f1 100644 --- a/src/gallium/drivers/svga/svga_draw_elements.c +++ b/src/gallium/drivers/svga/svga_draw_elements.c @@ -23,8 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" -#include "util/u_prim.h" +#include "util/u_inlines.h" #include "util/u_upload_mgr.h" #include "indices/u_indices.h" diff --git a/src/gallium/drivers/svga/svga_pipe_blend.c b/src/gallium/drivers/svga/svga_pipe_blend.c index 855d228755..9dd6fb068c 100644 --- a/src/gallium/drivers/svga/svga_pipe_blend.c +++ b/src/gallium/drivers/svga/svga_pipe_blend.c @@ -23,13 +23,12 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_hw_reg.h" @@ -182,15 +181,15 @@ svga_create_blend_state(struct pipe_context *pipe, } } else { - blend->rt[i].blend_enable = templ->blend_enable; + blend->rt[i].blend_enable = templ->rt[0].blend_enable; - if (templ->blend_enable) { - blend->rt[i].srcblend = svga_translate_blend_factor(templ->rgb_src_factor); - blend->rt[i].dstblend = svga_translate_blend_factor(templ->rgb_dst_factor); - blend->rt[i].blendeq = svga_translate_blend_func(templ->rgb_func); - blend->rt[i].srcblend_alpha = svga_translate_blend_factor(templ->alpha_src_factor); - blend->rt[i].dstblend_alpha = svga_translate_blend_factor(templ->alpha_dst_factor); - blend->rt[i].blendeq_alpha = svga_translate_blend_func(templ->alpha_func); + if (templ->rt[0].blend_enable) { + blend->rt[i].srcblend = svga_translate_blend_factor(templ->rt[0].rgb_src_factor); + blend->rt[i].dstblend = svga_translate_blend_factor(templ->rt[0].rgb_dst_factor); + blend->rt[i].blendeq = svga_translate_blend_func(templ->rt[0].rgb_func); + blend->rt[i].srcblend_alpha = svga_translate_blend_factor(templ->rt[0].alpha_src_factor); + blend->rt[i].dstblend_alpha = svga_translate_blend_factor(templ->rt[0].alpha_dst_factor); + blend->rt[i].blendeq_alpha = svga_translate_blend_func(templ->rt[0].alpha_func); if (blend->rt[i].srcblend_alpha != blend->rt[i].srcblend || blend->rt[i].dstblend_alpha != blend->rt[i].dstblend || @@ -201,7 +200,7 @@ svga_create_blend_state(struct pipe_context *pipe, } } - blend->rt[i].writemask = templ->colormask; + blend->rt[i].writemask = templ->rt[0].colormask; } return blend; diff --git a/src/gallium/drivers/svga/svga_pipe_constants.c b/src/gallium/drivers/svga/svga_pipe_constants.c index 10e7a12189..73a0cd6b3a 100644 --- a/src/gallium/drivers/svga/svga_pipe_constants.c +++ b/src/gallium/drivers/svga/svga_pipe_constants.c @@ -23,16 +23,12 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" -#include "util/u_memory.h" #include "tgsi/tgsi_parse.h" #include "svga_context.h" -#include "svga_state.h" -#include "svga_hw_reg.h" -#include "svga_cmd.h" /*********************************************************************** * Constant buffers @@ -49,7 +45,7 @@ struct svga_constbuf static void svga_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, - const struct pipe_constant_buffer *buf) + struct pipe_buffer *buf) { struct svga_context *svga = svga_context(pipe); @@ -57,7 +53,7 @@ static void svga_set_constant_buffer(struct pipe_context *pipe, assert(index == 0); pipe_buffer_reference( &svga->curr.cb[shader], - buf->buffer ); + buf ); if (shader == PIPE_SHADER_FRAGMENT) svga->dirty |= SVGA_NEW_FS_CONST_BUFFER; diff --git a/src/gallium/drivers/svga/svga_pipe_depthstencil.c b/src/gallium/drivers/svga/svga_pipe_depthstencil.c index df636c08a0..12bbd233a5 100644 --- a/src/gallium/drivers/svga/svga_pipe_depthstencil.c +++ b/src/gallium/drivers/svga/svga_pipe_depthstencil.c @@ -23,13 +23,12 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_pipe_draw.c b/src/gallium/drivers/svga/svga_pipe_draw.c index 0f24ef4ee8..f00cf23935 100644 --- a/src/gallium/drivers/svga/svga_pipe_draw.c +++ b/src/gallium/drivers/svga/svga_pipe_draw.c @@ -25,7 +25,7 @@ #include "svga_cmd.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "util/u_prim.h" #include "util/u_time.h" #include "indices/u_indices.h" @@ -33,7 +33,6 @@ #include "svga_hw_reg.h" #include "svga_context.h" #include "svga_screen.h" -#include "svga_winsys.h" #include "svga_draw.h" #include "svga_state.h" #include "svga_swtnl.h" @@ -217,11 +216,6 @@ svga_draw_range_elements( struct pipe_context *pipe, } if (SVGA_DEBUG & DEBUG_FLUSH) { - static unsigned id; - debug_printf("%s %d\n", __FUNCTION__, id++); - if (id > 1300) - util_time_sleep( 2000 ); - svga_hwtnl_flush_retry( svga ); svga_context_flush(svga, NULL); } diff --git a/src/gallium/drivers/svga/svga_pipe_flush.c b/src/gallium/drivers/svga/svga_pipe_flush.c index 0becb0765a..7fa2205ae5 100644 --- a/src/gallium/drivers/svga/svga_pipe_flush.c +++ b/src/gallium/drivers/svga/svga_pipe_flush.c @@ -27,14 +27,8 @@ #include "svga_screen.h" #include "svga_screen_texture.h" #include "svga_context.h" -#include "svga_winsys.h" -#include "svga_draw.h" #include "svga_debug.h" -#include "svga_hw_reg.h" - - - static void svga_flush( struct pipe_context *pipe, unsigned flags, diff --git a/src/gallium/drivers/svga/svga_pipe_fs.c b/src/gallium/drivers/svga/svga_pipe_fs.c index 5f1213e46a..b71bc66552 100644 --- a/src/gallium/drivers/svga/svga_pipe_fs.c +++ b/src/gallium/drivers/svga/svga_pipe_fs.c @@ -23,20 +23,17 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_bitmask.h" #include "tgsi/tgsi_parse.h" -#include "tgsi/tgsi_text.h" #include "svga_screen.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_tgsi.h" #include "svga_hw_reg.h" #include "svga_cmd.h" -#include "svga_draw.h" #include "svga_debug.h" diff --git a/src/gallium/drivers/svga/svga_pipe_misc.c b/src/gallium/drivers/svga/svga_pipe_misc.c index 58cb1e6e23..49b43bebc2 100644 --- a/src/gallium/drivers/svga/svga_pipe_misc.c +++ b/src/gallium/drivers/svga/svga_pipe_misc.c @@ -25,14 +25,10 @@ #include "svga_cmd.h" +#include "util/u_inlines.h" + #include "svga_context.h" #include "svga_screen_texture.h" -#include "svga_state.h" -#include "svga_winsys.h" - -#include "svga_hw_reg.h" - - static void svga_set_scissor_state( struct pipe_context *pipe, diff --git a/src/gallium/drivers/svga/svga_pipe_query.c b/src/gallium/drivers/svga/svga_pipe_query.c index 01336b0a2c..08283e3731 100644 --- a/src/gallium/drivers/svga/svga_pipe_query.c +++ b/src/gallium/drivers/svga/svga_pipe_query.c @@ -32,7 +32,6 @@ #include "svga_screen.h" #include "svga_screen_buffer.h" #include "svga_winsys.h" -#include "svga_draw.h" #include "svga_debug.h" diff --git a/src/gallium/drivers/svga/svga_pipe_rasterizer.c b/src/gallium/drivers/svga/svga_pipe_rasterizer.c index 09ccb71884..3571778867 100644 --- a/src/gallium/drivers/svga/svga_pipe_rasterizer.c +++ b/src/gallium/drivers/svga/svga_pipe_rasterizer.c @@ -24,13 +24,12 @@ **********************************************************/ #include "draw/draw_context.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_pipe_sampler.c b/src/gallium/drivers/svga/svga_pipe_sampler.c index 460a101f8c..b70081343d 100644 --- a/src/gallium/drivers/svga/svga_pipe_sampler.c +++ b/src/gallium/drivers/svga/svga_pipe_sampler.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -32,9 +32,6 @@ #include "svga_context.h" #include "svga_screen_texture.h" -#include "svga_state.h" - -#include "svga_hw_reg.h" #include "svga_debug.h" diff --git a/src/gallium/drivers/svga/svga_pipe_vertex.c b/src/gallium/drivers/svga/svga_pipe_vertex.c index 42f290d162..ffc0f99565 100644 --- a/src/gallium/drivers/svga/svga_pipe_vertex.c +++ b/src/gallium/drivers/svga/svga_pipe_vertex.c @@ -23,19 +23,14 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" -#include "util/u_memory.h" #include "tgsi/tgsi_parse.h" #include "svga_screen.h" #include "svga_screen_buffer.h" #include "svga_context.h" -#include "svga_state.h" -#include "svga_winsys.h" - -#include "svga_hw_reg.h" static void svga_set_vertex_buffers(struct pipe_context *pipe, diff --git a/src/gallium/drivers/svga/svga_pipe_vs.c b/src/gallium/drivers/svga/svga_pipe_vs.c index 7e6ab576ad..de8c919e12 100644 --- a/src/gallium/drivers/svga/svga_pipe_vs.c +++ b/src/gallium/drivers/svga/svga_pipe_vs.c @@ -24,7 +24,7 @@ **********************************************************/ #include "draw/draw_context.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_bitmask.h" @@ -33,7 +33,6 @@ #include "svga_screen.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_tgsi.h" #include "svga_hw_reg.h" #include "svga_cmd.h" diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c index 45b5d85ae2..8143be5024 100644 --- a/src/gallium/drivers/svga/svga_screen.c +++ b/src/gallium/drivers/svga/svga_screen.c @@ -24,7 +24,7 @@ **********************************************************/ #include "util/u_memory.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "util/u_string.h" #include "util/u_math.h" @@ -33,10 +33,8 @@ #include "svga_screen.h" #include "svga_screen_texture.h" #include "svga_screen_buffer.h" -#include "svga_cmd.h" #include "svga_debug.h" -#include "svga_hw_reg.h" #include "svga3d_shaderdefs.h" @@ -146,6 +144,13 @@ svga_get_paramf(struct pipe_screen *screen, int param) case PIPE_CAP_BLEND_EQUATION_SEPARATE: /* req. for GL 1.5 */ return 1; + case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT: + case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER: + return 1; + case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT: + case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER: + return 0; + default: return 0; } @@ -361,6 +366,7 @@ svga_screen_create(struct svga_winsys_screen *sws) screen->get_param = svga_get_param; screen->get_paramf = svga_get_paramf; screen->is_format_supported = svga_is_format_supported; + screen->context_create = svga_context_create; screen->fence_reference = svga_fence_reference; screen->fence_signalled = svga_fence_signalled; screen->fence_finish = svga_fence_finish; @@ -393,8 +399,6 @@ svga_screen_create(struct svga_winsys_screen *sws) pipe_mutex_init(svgascreen->tex_mutex); pipe_mutex_init(svgascreen->swc_mutex); - LIST_INITHEAD(&svgascreen->cached_buffers); - svga_screen_cache_init(svgascreen); return screen; diff --git a/src/gallium/drivers/svga/svga_screen.h b/src/gallium/drivers/svga/svga_screen.h index b94ca7fc1c..9dc229b0a8 100644 --- a/src/gallium/drivers/svga/svga_screen.h +++ b/src/gallium/drivers/svga/svga_screen.h @@ -28,7 +28,7 @@ #include "pipe/p_screen.h" -#include "pipe/p_thread.h" +#include "os/os_thread.h" #include "util/u_double_list.h" @@ -68,12 +68,6 @@ struct svga_screen pipe_mutex tex_mutex; pipe_mutex swc_mutex; /* Protects the use of swc and dirty_buffers */ - /** - * List of buffers with cached GMR. Ordered from the most recently used to - * the least recently used - */ - struct list_head cached_buffers; - struct svga_host_surface_cache cache; }; diff --git a/src/gallium/drivers/svga/svga_screen_buffer.c b/src/gallium/drivers/svga/svga_screen_buffer.c index 58a1aba464..c9e9bef540 100644 --- a/src/gallium/drivers/svga/svga_screen_buffer.c +++ b/src/gallium/drivers/svga/svga_screen_buffer.c @@ -27,8 +27,8 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_thread.h" +#include "util/u_inlines.h" +#include "os/os_thread.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -113,68 +113,9 @@ svga_buffer_destroy_hw_storage(struct svga_screen *ss, struct svga_buffer *sbuf) if(sbuf->hw.buf) { sws->buffer_destroy(sws, sbuf->hw.buf); sbuf->hw.buf = NULL; - assert(sbuf->head.prev && sbuf->head.next); - LIST_DEL(&sbuf->head); -#ifdef DEBUG - sbuf->head.next = sbuf->head.prev = NULL; -#endif } } -static INLINE enum pipe_error -svga_buffer_backup(struct svga_screen *ss, struct svga_buffer *sbuf) -{ - if (sbuf->hw.buf && sbuf->hw.num_ranges) { - void *src; - - if (!sbuf->swbuf) - sbuf->swbuf = align_malloc(sbuf->base.size, sbuf->base.alignment); - if (!sbuf->swbuf) - return PIPE_ERROR_OUT_OF_MEMORY; - - src = ss->sws->buffer_map(ss->sws, sbuf->hw.buf, - PIPE_BUFFER_USAGE_CPU_READ); - if (!src) - return PIPE_ERROR; - - memcpy(sbuf->swbuf, src, sbuf->base.size); - ss->sws->buffer_unmap(ss->sws, sbuf->hw.buf); - } - - return PIPE_OK; -} - -/** - * Try to make GMR space available by freeing the hardware storage of - * unmapped - */ -boolean -svga_buffer_free_cached_hw_storage(struct svga_screen *ss) -{ - struct list_head *curr; - struct svga_buffer *sbuf; - enum pipe_error ret = PIPE_OK; - - curr = ss->cached_buffers.prev; - - /* free the least recently used buffer's hw storage which is not mapped */ - do { - if(curr == &ss->cached_buffers) - return FALSE; - - sbuf = LIST_ENTRY(struct svga_buffer, curr, head); - - curr = curr->prev; - if (sbuf->map.count == 0) - ret = svga_buffer_backup(ss, sbuf); - - } while(sbuf->map.count != 0 || ret != PIPE_OK); - - svga_buffer_destroy_hw_storage(ss, sbuf); - - return TRUE; -} - struct svga_winsys_buffer * svga_winsys_buffer_create( struct svga_screen *ss, unsigned alignment, @@ -195,12 +136,6 @@ svga_winsys_buffer_create( struct svga_screen *ss, svga_screen_flush(ss, NULL); buf = sws->buffer_create(sws, alignment, usage, size); - SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "evicting buffers to find %d bytes GMR\n", - size); - - /* Try evicing all buffer storage */ - while(!buf && svga_buffer_free_cached_hw_storage(ss)) - buf = sws->buffer_create(sws, alignment, usage, size); } return buf; @@ -226,8 +161,6 @@ svga_buffer_create_hw_storage(struct svga_screen *ss, return PIPE_ERROR_OUT_OF_MEMORY; assert(!sbuf->needs_flush); - assert(!sbuf->head.prev && !sbuf->head.next); - LIST_ADD(&sbuf->head, &ss->cached_buffers); } return PIPE_OK; @@ -311,7 +244,6 @@ static void svga_buffer_upload_flush(struct svga_context *svga, struct svga_buffer *sbuf) { - struct svga_screen *ss = svga_screen(svga->pipe.screen); SVGA3dCopyBox *boxes; unsigned i; @@ -348,13 +280,16 @@ svga_buffer_upload_flush(struct svga_context *svga, assert(sbuf->head.prev && sbuf->head.next); LIST_DEL(&sbuf->head); +#ifdef DEBUG + sbuf->head.next = sbuf->head.prev = NULL; +#endif sbuf->needs_flush = FALSE; - /* XXX: do we care about cached_buffers any more ?*/ - LIST_ADD(&sbuf->head, &ss->cached_buffers); sbuf->hw.svga = NULL; sbuf->hw.boxes = NULL; + sbuf->host_written = TRUE; + /* Decrement reference count */ pipe_reference(&(sbuf->base.reference), NULL); sbuf = NULL; @@ -437,17 +372,17 @@ svga_buffer_map_range( struct pipe_screen *screen, } else { if(!sbuf->hw.buf) { - struct svga_winsys_surface *handle = sbuf->handle; - if(svga_buffer_create_hw_storage(ss, sbuf) != PIPE_OK) return NULL; /* Populate the hardware storage if the host surface pre-existed */ - if((usage & PIPE_BUFFER_USAGE_CPU_READ) && handle) { + if(sbuf->host_written) { SVGA3dSurfaceDMAFlags flags; enum pipe_error ret; struct pipe_fence_handle *fence = NULL; + assert(sbuf->handle); + SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "dma from sid %p (buffer), bytes %u - %u\n", sbuf->handle, 0, sbuf->base.size); @@ -478,17 +413,6 @@ svga_buffer_map_range( struct pipe_screen *screen, sws->fence_reference(sws, &fence, NULL); } } - else { - if((usage & PIPE_BUFFER_USAGE_CPU_READ) && !sbuf->needs_flush) { - /* We already had the hardware storage but we would have to issue - * a download if we hadn't, so move the buffer to the begginning - * of the LRU list. - */ - assert(sbuf->head.prev && sbuf->head.next); - LIST_DEL(&sbuf->head); - LIST_ADD(&sbuf->head, &ss->cached_buffers); - } - } map = sws->buffer_map(sws, sbuf->hw.buf, usage); } @@ -572,10 +496,8 @@ svga_buffer_destroy( struct pipe_buffer *buf ) assert(!sbuf->needs_flush); - if(sbuf->handle) { - SVGA_DBG(DEBUG_DMA, "release sid %p sz %d\n", sbuf->handle, sbuf->base.size); - svga_screen_surface_destroy(ss, &sbuf->key, &sbuf->handle); - } + if(sbuf->handle) + svga_buffer_destroy_host_surface(ss, sbuf); if(sbuf->hw.buf) svga_buffer_destroy_hw_storage(ss, sbuf); @@ -595,6 +517,9 @@ svga_buffer_create(struct pipe_screen *screen, struct svga_screen *ss = svga_screen(screen); struct svga_buffer *sbuf; + assert(size); + assert(alignment); + sbuf = CALLOC_STRUCT(svga_buffer); if(!sbuf) goto error1; @@ -755,8 +680,7 @@ svga_buffer_handle(struct svga_context *svga, assert(sbuf->hw.svga == svga); sbuf->needs_flush = TRUE; - assert(sbuf->head.prev && sbuf->head.next); - LIST_DEL(&sbuf->head); + assert(!sbuf->head.prev && !sbuf->head.next); LIST_ADDTAIL(&sbuf->head, &svga->dirty_buffers); } diff --git a/src/gallium/drivers/svga/svga_screen_buffer.h b/src/gallium/drivers/svga/svga_screen_buffer.h index 5d7af5a7c5..448ac107c7 100644 --- a/src/gallium/drivers/svga/svga_screen_buffer.h +++ b/src/gallium/drivers/svga/svga_screen_buffer.h @@ -135,6 +135,11 @@ struct svga_buffer */ struct svga_winsys_surface *handle; + /** + * Whether the host has been ever written. + */ + boolean host_written; + struct { unsigned count; boolean writing; @@ -178,9 +183,6 @@ svga_buffer_handle(struct svga_context *svga, void svga_context_flush_buffers(struct svga_context *svga); -boolean -svga_buffer_free_cached_hw_storage(struct svga_screen *ss); - struct svga_winsys_buffer * svga_winsys_buffer_create(struct svga_screen *ss, unsigned alignment, diff --git a/src/gallium/drivers/svga/svga_screen_cache.h b/src/gallium/drivers/svga/svga_screen_cache.h index f5aa740d40..62156e3f52 100644 --- a/src/gallium/drivers/svga/svga_screen_cache.h +++ b/src/gallium/drivers/svga/svga_screen_cache.h @@ -31,7 +31,7 @@ #include "svga_reg.h" #include "svga3d_reg.h" -#include "pipe/p_thread.h" +#include "os/os_thread.h" #include "util/u_double_list.h" diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c index 2224c2d394..12f3531a1d 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.c +++ b/src/gallium/drivers/svga/svga_screen_texture.c @@ -27,8 +27,8 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_thread.h" +#include "util/u_inlines.h" +#include "os/os_thread.h" #include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -205,7 +205,7 @@ svga_transfer_dma(struct svga_transfer *st, if(transfer == SVGA3D_READ_HOST_VRAM) { svga_screen_flush(screen, &fence); sws->fence_finish(sws, fence, 0); - //sws->fence_reference(sws, &fence, NULL); + sws->fence_reference(sws, &fence, NULL); } } else { @@ -235,7 +235,7 @@ svga_transfer_dma(struct svga_transfer *st, if(y) { svga_screen_flush(screen, &fence); sws->fence_finish(sws, fence, 0); - //sws->fence_reference(sws, &fence, NULL); + sws->fence_reference(sws, &fence, NULL); } hw = sws->buffer_map(sws, st->hwbuf, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -306,11 +306,19 @@ svga_texture_create(struct pipe_screen *screen, tex->key.numFaces = 1; } + tex->key.cachable = 1; + if(templat->tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) tex->key.flags |= SVGA3D_SURFACE_HINT_TEXTURE; - if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) + if(templat->tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + tex->key.cachable = 0; + } + + if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) { tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT; + tex->key.cachable = 0; + } /* * XXX: Never pass the SVGA3D_SURFACE_HINT_RENDERTARGET hint. Mesa cannot @@ -333,8 +341,6 @@ svga_texture_create(struct pipe_screen *screen, if(tex->key.format == SVGA3D_FORMAT_INVALID) goto error2; - tex->key.cachable = 1; - SVGA_DBG(DEBUG_DMA, "surface_create for texture\n", tex->handle); tex->handle = svga_screen_surface_create(svgascreen, &tex->key); if (tex->handle) @@ -416,6 +422,62 @@ svga_texture_blanket(struct pipe_screen * screen, } +struct pipe_texture * +svga_screen_texture_wrap_surface(struct pipe_screen *screen, + struct pipe_texture *base, + enum SVGA3dSurfaceFormat format, + struct svga_winsys_surface *srf) +{ + struct svga_texture *tex; + assert(screen); + + /* Only supports one type */ + if (base->target != PIPE_TEXTURE_2D || + base->last_level != 0 || + base->depth0 != 1) { + return NULL; + } + + if (!srf) + return NULL; + + if (svga_translate_format(base->format) != format) { + unsigned f1 = svga_translate_format(base->format); + unsigned f2 = format; + + /* It's okay for XRGB and ARGB or depth with/out stencil to get mixed up */ + if ( !( (f1 == SVGA3D_X8R8G8B8 && f2 == SVGA3D_A8R8G8B8) || + (f1 == SVGA3D_A8R8G8B8 && f2 == SVGA3D_X8R8G8B8) || + (f1 == SVGA3D_Z_D24X8 && f2 == SVGA3D_Z_D24S8) ) ) { + debug_printf("%s wrong format %u != %u\n", __FUNCTION__, f1, f2); + return NULL; + } + } + + tex = CALLOC_STRUCT(svga_texture); + if (!tex) + return NULL; + + tex->base = *base; + + + if (format == 1) + tex->base.format = PIPE_FORMAT_X8R8G8B8_UNORM; + else if (format == 2) + tex->base.format = PIPE_FORMAT_A8R8G8B8_UNORM; + + pipe_reference_init(&tex->base.reference, 1); + tex->base.screen = screen; + + SVGA_DBG(DEBUG_DMA, "wrap surface sid %p\n", srf); + + tex->key.cachable = 0; + tex->handle = srf; + + return &tex->base; +} + + static void svga_texture_destroy(struct pipe_texture *pt) { diff --git a/src/gallium/drivers/svga/svga_screen_texture.h b/src/gallium/drivers/svga/svga_screen_texture.h index 89ae24219f..43853d48f8 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.h +++ b/src/gallium/drivers/svga/svga_screen_texture.h @@ -29,6 +29,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_state.h" +#include "util/u_inlines.h" #include "svga_screen_cache.h" struct pipe_context; diff --git a/src/gallium/drivers/svga/svga_state_constants.c b/src/gallium/drivers/svga/svga_state_constants.c index 6b0e511cec..bb92f818ea 100644 --- a/src/gallium/drivers/svga/svga_state_constants.c +++ b/src/gallium/drivers/svga/svga_state_constants.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "svga_context.h" diff --git a/src/gallium/drivers/svga/svga_state_framebuffer.c b/src/gallium/drivers/svga/svga_state_framebuffer.c index cfdcae4ee4..b4cafb8f21 100644 --- a/src/gallium/drivers/svga/svga_state_framebuffer.c +++ b/src/gallium/drivers/svga/svga_state_framebuffer.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" @@ -32,8 +32,6 @@ #include "svga_cmd.h" #include "svga_debug.h" -#include "svga_hw_reg.h" - /*********************************************************************** * Hardware state update diff --git a/src/gallium/drivers/svga/svga_state_fs.c b/src/gallium/drivers/svga/svga_state_fs.c index d29f3762d2..2973444d0a 100644 --- a/src/gallium/drivers/svga/svga_state_fs.c +++ b/src/gallium/drivers/svga/svga_state_fs.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" #include "util/u_bitmask.h" @@ -81,8 +81,10 @@ static enum pipe_error compile_fs( struct svga_context *svga, } result->id = util_bitmask_add(svga->fs_bm); - if(result->id == UTIL_BITMASK_INVALID_INDEX) + if(result->id == UTIL_BITMASK_INVALID_INDEX) { + ret = PIPE_ERROR_OUT_OF_MEMORY; goto fail; + } ret = SVGA3D_DefineShader(svga->swc, result->id, @@ -106,70 +108,6 @@ fail: return ret; } -/* The blend workaround for simulating logicop xor behaviour requires - * that the incoming fragment color be white. This change achieves - * that by hooking up a hard-wired fragment shader that just emits - * color 1,1,1,1 - * - * This is a slightly incomplete solution as it assumes that the - * actual bound shader has no other effects beyond generating a - * fragment color. In particular shaders containing TEXKIL and/or - * depth-write will not have the correct behaviour, nor will those - * expecting to use alphatest. - * - * These are avoidable issues, but they are not much worse than the - * unavoidable ones associated with this technique, so it's not clear - * how much effort should be expended trying to resolve them - the - * ultimate result will still not be correct in most cases. - * - * Shader below was generated with: - * SVGA_DEBUG=tgsi ./mesa/progs/fp/fp-tri white.txt - */ -static int emit_white_fs( struct svga_context *svga ) -{ - int ret = PIPE_ERROR; - - /* ps_3_0 - * def c0, 1.000000, 0.000000, 0.000000, 1.000000 - * mov oC0, c0.x - * end - */ - static const unsigned white_tokens[] = { - 0xffff0300, - 0x05000051, - 0xa00f0000, - 0x3f800000, - 0x00000000, - 0x00000000, - 0x3f800000, - 0x02000001, - 0x800f0800, - 0xa0000000, - 0x0000ffff, - }; - - assert(SVGA3D_INVALID_ID == UTIL_BITMASK_INVALID_INDEX); - svga->state.white_fs_id = util_bitmask_add(svga->fs_bm); - if(svga->state.white_fs_id == SVGA3D_INVALID_ID) - goto no_fs_id; - - ret = SVGA3D_DefineShader(svga->swc, - svga->state.white_fs_id, - SVGA3D_SHADERTYPE_PS, - white_tokens, - sizeof(white_tokens)); - if (ret) - goto no_definition; - - return 0; - -no_definition: - util_bitmask_clear(svga->fs_bm, svga->state.white_fs_id); - svga->state.white_fs_id = SVGA3D_INVALID_ID; -no_fs_id: - return ret; -} - /* SVGA_NEW_TEXTURE_BINDING * SVGA_NEW_RAST @@ -197,6 +135,23 @@ static int make_fs_key( const struct svga_context *svga, PIPE_WINDING_CW); } + /* The blend workaround for simulating logicop xor behaviour + * requires that the incoming fragment color be white. This change + * achieves that by creating a varient of the current fragment + * shader that overrides all output colors with 1,1,1,1 + * + * This will work for most shaders, including those containing + * TEXKIL and/or depth-write. However, it will break on the + * combination of xor-logicop plus alphatest. + * + * Ultimately, we could implement alphatest in the shader using + * texkil prior to overriding the outgoing fragment color. + * + * SVGA_NEW_BLEND + */ + if (svga->curr.blend->need_white_fragments) { + key->white_fragments = 1; + } /* XXX: want to limit this to the textures that the shader actually * refers to. @@ -236,40 +191,29 @@ static int emit_hw_fs( struct svga_context *svga, unsigned id = SVGA3D_INVALID_ID; int ret = 0; + struct svga_fragment_shader *fs = svga->curr.fs; + struct svga_fs_compile_key key; + /* SVGA_NEW_BLEND + * SVGA_NEW_TEXTURE_BINDING + * SVGA_NEW_RAST + * SVGA_NEW_NEED_SWTNL + * SVGA_NEW_SAMPLER */ - if (svga->curr.blend->need_white_fragments) { - if (svga->state.white_fs_id == SVGA3D_INVALID_ID) { - ret = emit_white_fs( svga ); - if (ret) - return ret; - } - id = svga->state.white_fs_id; - } - else { - struct svga_fragment_shader *fs = svga->curr.fs; - struct svga_fs_compile_key key; - - /* SVGA_NEW_TEXTURE_BINDING - * SVGA_NEW_RAST - * SVGA_NEW_NEED_SWTNL - * SVGA_NEW_SAMPLER - */ - ret = make_fs_key( svga, &key ); + ret = make_fs_key( svga, &key ); + if (ret) + return ret; + + result = search_fs_key( fs, &key ); + if (!result) { + ret = compile_fs( svga, fs, &key, &result ); if (ret) return ret; - - result = search_fs_key( fs, &key ); - if (!result) { - ret = compile_fs( svga, fs, &key, &result ); - if (ret) - return ret; - } - - assert (result); - id = result->id; } + assert (result); + id = result->id; + assert(id != SVGA3D_INVALID_ID); if (result != svga->state.hw_draw.fs) { diff --git a/src/gallium/drivers/svga/svga_state_need_swtnl.c b/src/gallium/drivers/svga/svga_state_need_swtnl.c index 3c35a8579f..dd13a89d24 100644 --- a/src/gallium/drivers/svga/svga_state_need_swtnl.c +++ b/src/gallium/drivers/svga/svga_state_need_swtnl.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_state.h" diff --git a/src/gallium/drivers/svga/svga_state_rss.c b/src/gallium/drivers/svga/svga_state_rss.c index c582e44524..5ce9b4ef4f 100644 --- a/src/gallium/drivers/svga/svga_state_rss.c +++ b/src/gallium/drivers/svga/svga_state_rss.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" @@ -31,9 +31,6 @@ #include "svga_state.h" #include "svga_cmd.h" -#include "svga_hw_reg.h" - - struct rs_queue { unsigned rs_count; diff --git a/src/gallium/drivers/svga/svga_state_tss.c b/src/gallium/drivers/svga/svga_state_tss.c index b313794520..17b4785978 100644 --- a/src/gallium/drivers/svga/svga_state_tss.c +++ b/src/gallium/drivers/svga/svga_state_tss.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" @@ -33,8 +33,6 @@ #include "svga_state.h" #include "svga_cmd.h" -#include "svga_hw_reg.h" - void svga_cleanup_tss_binding(struct svga_context *svga) { diff --git a/src/gallium/drivers/svga/svga_state_vdecl.c b/src/gallium/drivers/svga/svga_state_vdecl.c index c534308f50..d1066ce13b 100644 --- a/src/gallium/drivers/svga/svga_state_vdecl.c +++ b/src/gallium/drivers/svga/svga_state_vdecl.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_math.h" #include "util/u_upload_mgr.h" diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c index ae1e77e7d4..d7999fe53d 100644 --- a/src/gallium/drivers/svga/svga_state_vs.c +++ b/src/gallium/drivers/svga/svga_state_vs.c @@ -23,7 +23,7 @@ * **********************************************************/ -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_defines.h" #include "util/u_format.h" #include "util/u_math.h" @@ -71,7 +71,7 @@ static enum pipe_error compile_vs( struct svga_context *svga, struct svga_shader_result **out_result ) { struct svga_shader_result *result; - enum pipe_error ret = PIPE_OK; + enum pipe_error ret = PIPE_ERROR; result = svga_translate_vertex_program( vs, key ); if (result == NULL) { @@ -80,8 +80,10 @@ static enum pipe_error compile_vs( struct svga_context *svga, } result->id = util_bitmask_add(svga->vs_bm); - if(result->id == UTIL_BITMASK_INVALID_INDEX) + if(result->id == UTIL_BITMASK_INVALID_INDEX) { + ret = PIPE_ERROR_OUT_OF_MEMORY; goto fail; + } ret = SVGA3D_DefineShader(svga->swc, result->id, @@ -200,10 +202,12 @@ static int update_zero_stride( struct svga_context *svga, key.output_stride = 4 * sizeof(float); key.nr_elements = 1; + key.element[0].type = TRANSLATE_ELEMENT_NORMAL; key.element[0].input_format = vel->src_format; key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; key.element[0].input_buffer = vel->vertex_buffer_index; key.element[0].input_offset = vel->src_offset; + key.element[0].instance_divisor = vel->instance_divisor; key.element[0].output_offset = const_idx * 4 * sizeof(float); translate_key_sanitize(&key); @@ -222,7 +226,7 @@ static int update_zero_stride( struct svga_context *svga, translate->set_buffer(translate, vel->vertex_buffer_index, mapped_buffer, vbuffer->stride); - translate->run(translate, 0, 1, + translate->run(translate, 0, 1, 0, svga->curr.zero_stride_constants); pipe_buffer_unmap(svga->pipe.screen, diff --git a/src/gallium/drivers/svga/svga_swtnl_backend.c b/src/gallium/drivers/svga/svga_swtnl_backend.c index b4f757a47a..e9d7942fb5 100644 --- a/src/gallium/drivers/svga/svga_swtnl_backend.c +++ b/src/gallium/drivers/svga/svga_swtnl_backend.c @@ -28,10 +28,9 @@ #include "draw/draw_vertex.h" #include "util/u_debug.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" -#include "util/u_simple_shaders.h" #include "svga_context.h" #include "svga_state.h" @@ -87,13 +86,13 @@ svga_vbuf_render_allocate_vertices( struct vbuf_render *render, if (!svga_render->vbuf) { svga_render->vbuf_size = MAX2(size, svga_render->vbuf_alloc_size); svga_render->vbuf = pipe_buffer_create(screen, - 0, + 16, PIPE_BUFFER_USAGE_VERTEX, svga_render->vbuf_size); if(!svga_render->vbuf) { svga_context_flush(svga, NULL); svga_render->vbuf = pipe_buffer_create(screen, - 0, + 16, PIPE_BUFFER_USAGE_VERTEX, svga_render->vbuf_size); assert(svga_render->vbuf); @@ -123,7 +122,9 @@ svga_vbuf_render_map_vertices( struct vbuf_render *render ) char *ptr = (char*)pipe_buffer_map(screen, svga_render->vbuf, PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_FLUSH_EXPLICIT); + PIPE_BUFFER_USAGE_FLUSH_EXPLICIT | + PIPE_BUFFER_USAGE_DISCARD | + PIPE_BUFFER_USAGE_UNSYNCHRONIZED); return ptr + svga_render->vbuf_offset; } @@ -259,14 +260,14 @@ svga_vbuf_render_draw( struct vbuf_render *render, if (!svga_render->ibuf) { svga_render->ibuf_size = MAX2(size, svga_render->ibuf_alloc_size); svga_render->ibuf = pipe_buffer_create(screen, - 0, + 2, PIPE_BUFFER_USAGE_VERTEX, svga_render->ibuf_size); svga_render->ibuf_offset = 0; } - pipe_buffer_write(screen, svga_render->ibuf, - svga_render->ibuf_offset, 2 * nr_indices, indices); + pipe_buffer_write_nooverlap(screen, svga_render->ibuf, + svga_render->ibuf_offset, 2 * nr_indices, indices); /* off to hardware */ diff --git a/src/gallium/drivers/svga/svga_swtnl_draw.c b/src/gallium/drivers/svga/svga_swtnl_draw.c index 7655121bec..da15be155c 100644 --- a/src/gallium/drivers/svga/svga_swtnl_draw.c +++ b/src/gallium/drivers/svga/svga_swtnl_draw.c @@ -25,9 +25,8 @@ #include "draw/draw_context.h" #include "draw/draw_vbuf.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_state.h" -#include "util/u_memory.h" #include "svga_context.h" #include "svga_swtnl.h" @@ -90,7 +89,7 @@ svga_swtnl_draw_range_elements(struct svga_context *svga, PIPE_BUFFER_USAGE_CPU_READ); assert(map); draw_set_mapped_constant_buffer( - draw, PIPE_SHADER_VERTEX, + draw, PIPE_SHADER_VERTEX, 0, map, svga->curr.cb[PIPE_SHADER_VERTEX]->size); } diff --git a/src/gallium/drivers/svga/svga_swtnl_state.c b/src/gallium/drivers/svga/svga_swtnl_state.c index 94b6ccc62d..35f36a828f 100644 --- a/src/gallium/drivers/svga/svga_swtnl_state.c +++ b/src/gallium/drivers/svga/svga_swtnl_state.c @@ -25,9 +25,8 @@ #include "draw/draw_context.h" #include "draw/draw_vbuf.h" -#include "pipe/p_inlines.h" +#include "util/u_inlines.h" #include "pipe/p_state.h" -#include "util/u_memory.h" #include "svga_context.h" #include "svga_swtnl.h" diff --git a/src/gallium/drivers/svga/svga_tgsi.h b/src/gallium/drivers/svga/svga_tgsi.h index 737a2213af..063c9cf422 100644 --- a/src/gallium/drivers/svga/svga_tgsi.h +++ b/src/gallium/drivers/svga/svga_tgsi.h @@ -49,6 +49,7 @@ struct svga_fs_compile_key { unsigned light_twoside:1; unsigned front_cw:1; + unsigned white_fragments:1; unsigned num_textures:8; unsigned num_unnormalized_coords:8; struct { diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c index 23b3ace7f3..1ae9906761 100644 --- a/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c +++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c @@ -29,9 +29,6 @@ #include "util/u_memory.h" #include "svga_tgsi_emit.h" -#include "svga_context.h" - - static boolean ps20_input( struct svga_shader_emitter *emit, diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c index d1c7336dec..73102a72a8 100644 --- a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c +++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c @@ -29,7 +29,6 @@ #include "util/u_memory.h" #include "svga_tgsi_emit.h" -#include "svga_context.h" static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semantic, unsigned *usage, @@ -195,8 +194,19 @@ static boolean ps30_output( struct svga_shader_emitter *emit, switch (semantic.Name) { case TGSI_SEMANTIC_COLOR: - emit->output_map[idx] = dst_register( SVGA3DREG_COLOROUT, - semantic.Index ); + if (emit->unit == PIPE_SHADER_FRAGMENT && + emit->key.fkey.white_fragments) { + + emit->output_map[idx] = dst_register( SVGA3DREG_TEMP, + emit->nr_hw_temp++ ); + emit->temp_col[idx] = emit->output_map[idx]; + emit->true_col[idx] = dst_register( SVGA3DREG_COLOROUT, + semantic.Index ); + } + else { + emit->output_map[idx] = dst_register( SVGA3DREG_COLOROUT, + semantic.Index ); + } break; case TGSI_SEMANTIC_POSITION: emit->output_map[idx] = dst_register( SVGA3DREG_TEMP, diff --git a/src/gallium/drivers/svga/svga_tgsi_emit.h b/src/gallium/drivers/svga/svga_tgsi_emit.h index 2557824293..e8f75485d5 100644 --- a/src/gallium/drivers/svga/svga_tgsi_emit.h +++ b/src/gallium/drivers/svga/svga_tgsi_emit.h @@ -79,6 +79,8 @@ struct svga_shader_emitter int ps30_input_count; + int dynamic_branching_level; + boolean in_main_func; boolean created_zero_immediate; @@ -199,6 +201,23 @@ static INLINE boolean emit_op3( struct svga_shader_emitter *emit, } +static INLINE boolean emit_op4( struct svga_shader_emitter *emit, + SVGA3dShaderInstToken inst, + SVGA3dShaderDestToken dest, + struct src_register src0, + struct src_register src1, + struct src_register src2, + struct src_register src3) +{ + return (emit_instruction( emit, inst ) && + emit_dst( emit, dest ) && + emit_src( emit, src0 ) && + emit_src( emit, src1 ) && + emit_src( emit, src2 ) && + emit_src( emit, src3 )); +} + + #define TRANSLATE_SWIZZLE(x,y,z,w) ((x) | ((y) << 2) | ((z) << 4) | ((w) << 6)) #define SWIZZLE_XYZW \ TRANSLATE_SWIZZLE(TGSI_SWIZZLE_X,TGSI_SWIZZLE_Y,TGSI_SWIZZLE_Z,TGSI_SWIZZLE_W) diff --git a/src/gallium/drivers/svga/svga_tgsi_insn.c b/src/gallium/drivers/svga/svga_tgsi_insn.c index dc5eb8fc60..be821e9821 100644 --- a/src/gallium/drivers/svga/svga_tgsi_insn.c +++ b/src/gallium/drivers/svga/svga_tgsi_insn.c @@ -46,8 +46,6 @@ translate_opcode( case TGSI_OPCODE_ABS: return SVGA3DOP_ABS; case TGSI_OPCODE_ADD: return SVGA3DOP_ADD; case TGSI_OPCODE_BREAKC: return SVGA3DOP_BREAKC; - case TGSI_OPCODE_DDX: return SVGA3DOP_DSX; - case TGSI_OPCODE_DDY: return SVGA3DOP_DSY; case TGSI_OPCODE_DP2A: return SVGA3DOP_DP2ADD; case TGSI_OPCODE_DP3: return SVGA3DOP_DP3; case TGSI_OPCODE_DP4: return SVGA3DOP_DP4; @@ -415,6 +413,88 @@ static boolean submit_op3( struct svga_shader_emitter *emit, } + + +/* SVGA shaders may not refer to >1 constant register in a single + * instruction. This function checks for that usage and inserts a + * move to temporary if detected. + */ +static boolean submit_op4( struct svga_shader_emitter *emit, + SVGA3dShaderInstToken inst, + SVGA3dShaderDestToken dest, + struct src_register src0, + struct src_register src1, + struct src_register src2, + struct src_register src3) +{ + SVGA3dShaderDestToken temp0; + SVGA3dShaderDestToken temp3; + boolean need_temp0 = FALSE; + boolean need_temp3 = FALSE; + SVGA3dShaderRegType type0, type1, type2, type3; + + temp0.value = 0; + temp3.value = 0; + type0 = SVGA3dShaderGetRegType( src0.base.value ); + type1 = SVGA3dShaderGetRegType( src1.base.value ); + type2 = SVGA3dShaderGetRegType( src2.base.value ); + type3 = SVGA3dShaderGetRegType( src2.base.value ); + + /* Make life a little easier - this is only used by the TXD + * instruction which is guaranteed not to have a constant/input reg + * in one slot at least: + */ + assert(type1 == SVGA3DREG_SAMPLER); + + if (type0 == SVGA3DREG_CONST && + ((type3 == SVGA3DREG_CONST && src0.base.num != src3.base.num) || + (type2 == SVGA3DREG_CONST && src0.base.num != src2.base.num))) + need_temp0 = TRUE; + + if (type3 == SVGA3DREG_CONST && + (type2 == SVGA3DREG_CONST && src3.base.num != src2.base.num)) + need_temp3 = TRUE; + + if (type0 == SVGA3DREG_INPUT && + ((type3 == SVGA3DREG_INPUT && src0.base.num != src3.base.num) || + (type2 == SVGA3DREG_INPUT && src0.base.num != src2.base.num))) + need_temp0 = TRUE; + + if (type3 == SVGA3DREG_INPUT && + (type2 == SVGA3DREG_INPUT && src3.base.num != src2.base.num)) + need_temp3 = TRUE; + + if (need_temp0) + { + temp0 = get_temp( emit ); + + if (!emit_op1( emit, inst_token( SVGA3DOP_MOV ), temp0, src0 )) + return FALSE; + + src0 = src( temp0 ); + } + + if (need_temp3) + { + temp3 = get_temp( emit ); + + if (!emit_op1( emit, inst_token( SVGA3DOP_MOV ), temp3, src3 )) + return FALSE; + + src3 = src( temp3 ); + } + + if (!emit_op4( emit, inst, dest, src0, src1, src2, src3 )) + return FALSE; + + if (need_temp3) + release_temp( emit, temp3 ); + if (need_temp0) + release_temp( emit, temp0 ); + return TRUE; +} + + static boolean emit_def_const( struct svga_shader_emitter *emit, SVGA3dShaderConstType type, unsigned idx, @@ -660,6 +740,8 @@ static boolean emit_if(struct svga_shader_emitter *emit, if_token.control = SVGA3DOPCOMPC_NE; zero = scalar(zero, TGSI_SWIZZLE_X); + emit->dynamic_branching_level++; + return (emit_instruction( emit, if_token ) && emit_src( emit, src ) && emit_src( emit, zero ) ); @@ -668,6 +750,8 @@ static boolean emit_if(struct svga_shader_emitter *emit, static boolean emit_endif(struct svga_shader_emitter *emit, const struct tgsi_full_instruction *insn) { + emit->dynamic_branching_level--; + return (emit_instruction( emit, inst_token( SVGA3DOP_ENDIF ))); } @@ -1011,10 +1095,10 @@ static boolean emit_kilp(struct svga_shader_emitter *emit, { SVGA3dShaderInstToken inst; SVGA3dShaderDestToken temp; - struct src_register one = get_zero_immediate( emit ); + struct src_register one = scalar( get_zero_immediate( emit ), + TGSI_SWIZZLE_W ); inst = inst_token( SVGA3DOP_TEXKILL ); - one = scalar( one, TGSI_SWIZZLE_W ); /* texkill doesn't allow negation on the operand so lets move * negation of {1} to a temp register */ @@ -1169,41 +1253,79 @@ static boolean emit_tex2(struct svga_shader_emitter *emit, SVGA3dShaderDestToken dst ) { SVGA3dShaderInstToken inst; - struct src_register src0; - struct src_register src1; - + struct src_register texcoord; + struct src_register sampler; + SVGA3dShaderDestToken tmp; + inst.value = 0; - inst.op = SVGA3DOP_TEX; switch (insn->Instruction.Opcode) { case TGSI_OPCODE_TEX: + inst.op = SVGA3DOP_TEX; break; case TGSI_OPCODE_TXP: + inst.op = SVGA3DOP_TEX; inst.control = SVGA3DOPCONT_PROJECT; break; case TGSI_OPCODE_TXB: + inst.op = SVGA3DOP_TEX; inst.control = SVGA3DOPCONT_BIAS; break; + case TGSI_OPCODE_TXL: + inst.op = SVGA3DOP_TEXLDL; + break; default: assert(0); return FALSE; } - src0 = translate_src_register( emit, &insn->Src[0] ); - src1 = translate_src_register( emit, &insn->Src[1] ); + texcoord = translate_src_register( emit, &insn->Src[0] ); + sampler = translate_src_register( emit, &insn->Src[1] ); - if (emit->key.fkey.tex[src1.base.num].unnormalized) { - struct src_register wh = get_tex_dimensions( emit, src1.base.num ); - SVGA3dShaderDestToken tmp = get_temp( emit ); + if (emit->key.fkey.tex[sampler.base.num].unnormalized || + emit->dynamic_branching_level > 0) + tmp = get_temp( emit ); + + /* Can't do mipmapping inside dynamic branch constructs. Force LOD + * zero in that case. + */ + if (emit->dynamic_branching_level > 0 && + inst.op == SVGA3DOP_TEX && + SVGA3dShaderGetRegType(texcoord.base.value) == SVGA3DREG_TEMP) { + struct src_register zero = get_zero_immediate( emit ); + + /* MOV tmp, texcoord */ + if (!submit_op1( emit, + inst_token( SVGA3DOP_MOV ), + tmp, + texcoord )) + return FALSE; + + /* MOV tmp.w, zero */ + if (!submit_op1( emit, + inst_token( SVGA3DOP_MOV ), + writemask( tmp, TGSI_WRITEMASK_W ), + scalar( zero, TGSI_SWIZZLE_X ))) + return FALSE; + + texcoord = src( tmp ); + inst.op = SVGA3DOP_TEXLDL; + } + + /* Explicit normalization of texcoords: + */ + if (emit->key.fkey.tex[sampler.base.num].unnormalized) { + struct src_register wh = get_tex_dimensions( emit, sampler.base.num ); /* MUL tmp, SRC0, WH */ if (!submit_op2( emit, inst_token( SVGA3DOP_MUL ), - tmp, src0, wh )) + tmp, texcoord, wh )) return FALSE; - src0 = src( tmp ); + + texcoord = src( tmp ); } - return submit_op2( emit, inst, dst, src0, src1 ); + return submit_op2( emit, inst, dst, texcoord, sampler ); } @@ -1211,31 +1333,33 @@ static boolean emit_tex2(struct svga_shader_emitter *emit, /* Translate texture instructions to SVGA3D representation. */ -static boolean emit_tex3(struct svga_shader_emitter *emit, +static boolean emit_tex4(struct svga_shader_emitter *emit, const struct tgsi_full_instruction *insn, SVGA3dShaderDestToken dst ) { SVGA3dShaderInstToken inst; - struct src_register src0; - struct src_register src1; - struct src_register src2; + struct src_register texcoord; + struct src_register ddx; + struct src_register ddy; + struct src_register sampler; + + texcoord = translate_src_register( emit, &insn->Src[0] ); + ddx = translate_src_register( emit, &insn->Src[1] ); + ddy = translate_src_register( emit, &insn->Src[2] ); + sampler = translate_src_register( emit, &insn->Src[3] ); inst.value = 0; switch (insn->Instruction.Opcode) { case TGSI_OPCODE_TXD: - inst.op = SVGA3DOP_TEXLDD; - break; - case TGSI_OPCODE_TXL: - inst.op = SVGA3DOP_TEXLDL; + inst.op = SVGA3DOP_TEXLDD; /* 4 args! */ break; + default: + assert(0); + return FALSE; } - src0 = translate_src_register( emit, &insn->Src[0] ); - src1 = translate_src_register( emit, &insn->Src[1] ); - src2 = translate_src_register( emit, &insn->Src[2] ); - - return submit_op3( emit, inst, dst, src0, src1, src2 ); + return submit_op4( emit, inst, dst, texcoord, sampler, ddx, ddy ); } @@ -1271,12 +1395,12 @@ static boolean emit_tex(struct svga_shader_emitter *emit, case TGSI_OPCODE_TEX: case TGSI_OPCODE_TXB: case TGSI_OPCODE_TXP: + case TGSI_OPCODE_TXL: if (!emit_tex2( emit, insn, tex_result )) return FALSE; break; - case TGSI_OPCODE_TXL: case TGSI_OPCODE_TXD: - if (!emit_tex3( emit, insn, tex_result )) + if (!emit_tex4( emit, insn, tex_result )) return FALSE; break; default: @@ -1330,6 +1454,8 @@ static boolean emit_bgnloop2( struct svga_shader_emitter *emit, struct src_register loop_reg = src_register( SVGA3DREG_LOOP, 0 ); struct src_register const_int = get_loop_const( emit ); + emit->dynamic_branching_level++; + return (emit_instruction( emit, inst ) && emit_src( emit, loop_reg ) && emit_src( emit, const_int ) ); @@ -1339,6 +1465,9 @@ static boolean emit_endloop2( struct svga_shader_emitter *emit, const struct tgsi_full_instruction *insn ) { SVGA3dShaderInstToken inst = inst_token( SVGA3DOP_ENDLOOP ); + + emit->dynamic_branching_level--; + return emit_instruction( emit, inst ); } @@ -1398,6 +1527,46 @@ static boolean emit_simple_instruction(struct svga_shader_emitter *emit, } } + +static boolean emit_deriv(struct svga_shader_emitter *emit, + const struct tgsi_full_instruction *insn ) +{ + if (emit->dynamic_branching_level > 0 && + insn->Src[0].Register.File == TGSI_FILE_TEMPORARY) + { + struct src_register zero = get_zero_immediate( emit ); + SVGA3dShaderDestToken dst = + translate_dst_register( emit, insn, 0 ); + + /* Deriv opcodes not valid inside dynamic branching, workaround + * by zeroing out the destination. + */ + if (!submit_op1(emit, + inst_token( SVGA3DOP_MOV ), + dst, + scalar(zero, TGSI_SWIZZLE_X))) + return FALSE; + + return TRUE; + } + else { + unsigned opcode; + + switch (insn->Instruction.Opcode) { + case TGSI_OPCODE_DDX: + opcode = SVGA3DOP_DSX; + break; + case TGSI_OPCODE_DDY: + opcode = SVGA3DOP_DSY; + break; + default: + return FALSE; + } + + return emit_simple_instruction( emit, opcode, insn ); + } +} + static boolean emit_arl(struct svga_shader_emitter *emit, const struct tgsi_full_instruction *insn) { @@ -2002,6 +2171,10 @@ static boolean svga_emit_instruction( struct svga_shader_emitter *emit, case TGSI_OPCODE_TXD: return emit_tex( emit, insn ); + case TGSI_OPCODE_DDX: + case TGSI_OPCODE_DDY: + return emit_deriv( emit, insn ); + case TGSI_OPCODE_BGNSUB: return emit_bgnsub( emit, position, insn ); @@ -2254,11 +2427,28 @@ static boolean emit_ps_postamble( struct svga_shader_emitter *emit ) for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { if (SVGA3dShaderGetRegType(emit->true_col[i].value) != 0) { - if (!submit_op1( emit, - inst_token(SVGA3DOP_MOV), - emit->true_col[i], - src(emit->temp_col[i]) )) - return FALSE; + /* Potentially override output colors with white for XOR + * logicop workaround. + */ + if (emit->unit == PIPE_SHADER_FRAGMENT && + emit->key.fkey.white_fragments) { + + struct src_register one = scalar( get_zero_immediate( emit ), + TGSI_SWIZZLE_W ); + + if (!submit_op1( emit, + inst_token(SVGA3DOP_MOV), + emit->true_col[i], + one )) + return FALSE; + } + else { + if (!submit_op1( emit, + inst_token(SVGA3DOP_MOV), + emit->true_col[i], + src(emit->temp_col[i]) )) + return FALSE; + } } } @@ -2467,6 +2657,9 @@ needs_to_create_zero( struct svga_shader_emitter *emit ) if (emit->key.fkey.light_twoside) return TRUE; + if (emit->key.fkey.white_fragments) + return TRUE; + if (emit->emit_frontface) return TRUE; @@ -2476,6 +2669,10 @@ needs_to_create_zero( struct svga_shader_emitter *emit ) } if (emit->info.opcode_count[TGSI_OPCODE_IF] >= 1 || + emit->info.opcode_count[TGSI_OPCODE_BGNLOOP] >= 1 || + emit->info.opcode_count[TGSI_OPCODE_BGNFOR] >= 1 || + emit->info.opcode_count[TGSI_OPCODE_DDX] >= 1 || + emit->info.opcode_count[TGSI_OPCODE_DDY] >= 1 || emit->info.opcode_count[TGSI_OPCODE_SGE] >= 1 || emit->info.opcode_count[TGSI_OPCODE_SGT] >= 1 || emit->info.opcode_count[TGSI_OPCODE_SLE] >= 1 || @@ -2702,6 +2899,8 @@ boolean svga_shader_emit_instructions( struct svga_shader_emitter *emit, goto done; } + assert(emit->dynamic_branching_level == 0); + /* Need to terminate the whole shader: */ ret = emit_instruction( emit, inst_token( SVGA3DOP_END ) ); diff --git a/src/gallium/drivers/svga/svga_winsys.h b/src/gallium/drivers/svga/svga_winsys.h index 59f299c185..b4e3af0eaf 100644 --- a/src/gallium/drivers/svga/svga_winsys.h +++ b/src/gallium/drivers/svga/svga_winsys.h @@ -272,9 +272,6 @@ struct svga_winsys_screen }; -struct pipe_context * -svga_context_create(struct pipe_screen *screen); - struct pipe_screen * svga_screen_create(struct svga_winsys_screen *sws); @@ -296,4 +293,10 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture, struct pipe_buffer **buffer, unsigned *stride); +struct pipe_texture * +svga_screen_texture_wrap_surface(struct pipe_screen *screen, + struct pipe_texture *base, + enum SVGA3dSurfaceFormat format, + struct svga_winsys_surface *srf); + #endif /* SVGA_WINSYS_H_ */ |