diff options
| author | José Fonseca <[email protected]> | 2010-01-16 23:21:06 +0000 |
|---|---|---|
| committer | José Fonseca <[email protected]> | 2010-01-16 23:21:06 +0000 |
| commit | ba5d600c90f7075cf2c33a0c5c679ef822e8746e (patch) | |
| tree | ec6794fba9f2fe563a8bbc03ca06e288b7f51c28 /src/gallium/auxiliary | |
| parent | fdfe06ad804ea13e6e436d66c1bcafe0bde2f545 (diff) | |
| parent | 164fd16cfbc09970676c2e6866e062a5c9b410db (diff) | |
Merge remote branch 'origin/master' into lp-binning
Conflicts:
src/gallium/drivers/llvmpipe/lp_quad.h
src/gallium/drivers/llvmpipe/lp_setup.c
Diffstat (limited to 'src/gallium/auxiliary')
42 files changed, 969 insertions, 363 deletions
diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile index e3af41c6e04..8f937e3b4e9 100644 --- a/src/gallium/auxiliary/Makefile +++ b/src/gallium/auxiliary/Makefile @@ -111,6 +111,7 @@ C_SOURCES = \ util/u_math.c \ util/u_mm.c \ util/u_rect.c \ + util/u_ringbuffer.c \ util/u_simple_shaders.c \ util/u_snprintf.c \ util/u_stream_stdc.c \ diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript index 782eb533863..f957090b5fb 100644 --- a/src/gallium/auxiliary/SConscript +++ b/src/gallium/auxiliary/SConscript @@ -147,6 +147,7 @@ source = [ 'util/u_math.c', 'util/u_mm.c', 'util/u_rect.c', + 'util/u_ringbuffer.c', 'util/u_simple_shaders.c', 'util/u_snprintf.c', 'util/u_stream_stdc.c', diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 667aa46b208..e90dfc5aec4 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -95,6 +95,7 @@ void draw_destroy( struct draw_context *draw ) draw_pipeline_destroy( draw ); draw_pt_destroy( draw ); draw_vs_destroy( draw ); + draw_gs_destroy( draw ); FREE( draw ); } diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index b716209df29..8a64c06efcd 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -164,6 +164,14 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, void draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count); +void +draw_arrays_instanced(struct draw_context *draw, + unsigned mode, + unsigned start, + unsigned count, + unsigned startInstance, + unsigned instanceCount); + void draw_flush(struct draw_context *draw); diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c index 5db2e755423..daf8d071f12 100644 --- a/src/gallium/auxiliary/draw/draw_gs.c +++ b/src/gallium/auxiliary/draw/draw_gs.c @@ -59,6 +59,15 @@ draw_gs_init( struct draw_context *draw ) return TRUE; } +void draw_gs_destroy( struct draw_context *draw ) +{ + if (!draw->gs.machine) + return; + + align_free(draw->gs.machine->Primitives); + + tgsi_exec_machine_destroy(draw->gs.machine); +} void draw_gs_set_constants( struct draw_context *draw, const float (*constants)[4], diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index 205cda5eabe..51a6115ebf5 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -55,7 +55,7 @@ -struct clipper { +struct clip_stage { struct draw_stage stage; /**< base class */ /* Basically duplicate some of the flatshading logic here: @@ -70,9 +70,9 @@ struct clipper { /* This is a bit confusing: */ -static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) +static INLINE struct clip_stage *clip_stage( struct draw_stage *stage ) { - return (struct clipper *)stage; + return (struct clip_stage *)stage; } @@ -92,11 +92,12 @@ static void interp_attr( float *fdst, fdst[3] = LINTERP( t, fout[3], fin[3] ); } + static void copy_colors( struct draw_stage *stage, struct vertex_header *dst, const struct vertex_header *src ) { - const struct clipper *clipper = clipper_stage(stage); + const struct clip_stage *clipper = clip_stage(stage); uint i; for (i = 0; i < clipper->num_color_attribs; i++) { const uint attr = clipper->color_attribs[i]; @@ -108,7 +109,7 @@ static void copy_colors( struct draw_stage *stage, /* Interpolate between two vertices to produce a third. */ -static void interp( const struct clipper *clip, +static void interp( const struct clip_stage *clip, struct vertex_header *dst, float t, const struct vertex_header *out, @@ -179,7 +180,7 @@ static void emit_poly( struct draw_stage *stage, header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ if (i == n-1) - header.flags |= edge_last; + header.flags |= edge_last; if (0) { const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader; @@ -200,13 +201,14 @@ static void emit_poly( struct draw_stage *stage, } } + static INLINE float dot4(const float *a, const float *b) { - return (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); + return (a[0] * b[0] + + a[1] * b[1] + + a[2] * b[2] + + a[3] * b[3]); } @@ -217,7 +219,7 @@ do_clip_tri( struct draw_stage *stage, struct prim_header *header, unsigned clipmask ) { - struct clipper *clipper = clipper_stage( stage ); + struct clip_stage *clipper = clip_stage( stage ); struct vertex_header *a[MAX_CLIPPED_VERTICES]; struct vertex_header *b[MAX_CLIPPED_VERTICES]; struct vertex_header **inlist = a; @@ -280,6 +282,7 @@ do_clip_tri( struct draw_stage *stage, dp_prev = dp; } + /* swap in/out lists */ { struct vertex_header **tmp = inlist; inlist = outlist; @@ -291,15 +294,11 @@ do_clip_tri( struct draw_stage *stage, /* If flat-shading, copy color to new provoking vertex. */ if (clipper->flat && inlist[0] != header->v[2]) { - if (1) { - inlist[0] = dup_vert(stage, inlist[0], tmpnr++); - } + inlist[0] = dup_vert(stage, inlist[0], tmpnr++); copy_colors(stage, inlist[0], header->v[2]); } - - /* Emit the polygon as triangles to the setup stage: */ if (n >= 3) @@ -314,7 +313,7 @@ do_clip_line( struct draw_stage *stage, struct prim_header *header, unsigned clipmask ) { - const struct clipper *clipper = clipper_stage( stage ); + const struct clip_stage *clipper = clip_stage( stage ); struct vertex_header *v0 = header->v[0]; struct vertex_header *v1 = header->v[1]; const float *pos0 = v0->clip; @@ -416,13 +415,14 @@ clip_tri( struct draw_stage *stage, } } + /* Update state. Could further delay this until we hit the first * primitive that really requires clipping. */ static void clip_init_state( struct draw_stage *stage ) { - struct clipper *clipper = clipper_stage( stage ); + struct clip_stage *clipper = clip_stage( stage ); clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; @@ -488,7 +488,7 @@ static void clip_destroy( struct draw_stage *stage ) */ struct draw_stage *draw_clip_stage( struct draw_context *draw ) { - struct clipper *clipper = CALLOC_STRUCT(clipper); + struct clip_stage *clipper = CALLOC_STRUCT(clip_stage); if (clipper == NULL) goto fail; diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 1a5269c0de9..d40c0352401 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -138,7 +138,7 @@ emit_vertex( struct vbuf_stage *vbuf, /* Note: we really do want data[0] here, not data[pos]: */ vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); - vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); + vbuf->translate->run(vbuf->translate, 0, 1, 0, vbuf->vertex_ptr); if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); @@ -271,10 +271,12 @@ vbuf_start_prim( struct vbuf_stage *vbuf, uint prim ) emit_sz = 0; break; } - + + hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL; hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; hw_key.element[i].input_buffer = src_buffer; hw_key.element[i].input_offset = src_offset; + hw_key.element[i].instance_divisor = 0; hw_key.element[i].output_format = output_format; hw_key.element[i].output_offset = dst_offset; diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index e49041556bd..ef49e575366 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -172,6 +172,8 @@ struct draw_context boolean force_passthrough; /**< never clip or shade */ + boolean dump_vs; + double mrd; /**< minimum resolvable depth value, for polygon offset */ /* pipe state that we need: */ @@ -239,6 +241,8 @@ struct draw_context unsigned reduced_prim; + unsigned instance_id; + void *driver_private; }; @@ -265,6 +269,7 @@ boolean draw_gs_init( struct draw_context *draw ); void draw_gs_set_constants( struct draw_context *, const float (*constants)[4], unsigned size ); +void draw_gs_destroy( struct draw_context *draw ); /******************************************************************************* * Common shading code: diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 2801dbafe47..a5ddec52863 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -280,20 +280,33 @@ void draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { - unsigned reduced_prim = u_reduced_prim(prim); + draw_arrays_instanced(draw, prim, start, count, 0, 1); +} + +void +draw_arrays_instanced(struct draw_context *draw, + unsigned mode, + unsigned start, + unsigned count, + unsigned startInstance, + unsigned instanceCount) +{ + unsigned reduced_prim = u_reduced_prim(mode); + unsigned instance; + if (reduced_prim != draw->reduced_prim) { - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE); draw->reduced_prim = reduced_prim; } if (0) - draw_print_arrays(draw, prim, start, MIN2(count, 20)); + draw_print_arrays(draw, mode, start, MIN2(count, 20)); #if 0 { int i; - debug_printf("draw_arrays(prim=%u start=%u count=%u):\n", - prim, start, count); + debug_printf("draw_arrays(mode=%u start=%u count=%u):\n", + mode, start, count); tgsi_dump(draw->vs.vertex_shader->state.tokens, 0); debug_printf("Elements:\n"); for (i = 0; i < draw->pt.nr_vertex_elements; i++) { @@ -311,6 +324,8 @@ draw_arrays(struct draw_context *draw, unsigned prim, } #endif - /* drawing done here: */ - draw_pt_arrays(draw, prim, start, count); + for (instance = 0; instance < instanceCount; instance++) { + draw->instance_id = instance + startInstance; + draw_pt_arrays(draw, mode, start, count); + } } diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 20edf7a227e..d5e0d92a605 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -183,7 +183,8 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); struct pt_fetch; void draw_pt_fetch_prepare( struct pt_fetch *fetch, unsigned vertex_input_count, - unsigned vertex_size ); + unsigned vertex_size, + unsigned instance_id_index ); void draw_pt_fetch_run( struct pt_fetch *fetch, const unsigned *elts, diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 064e16c295c..4fb53276bbe 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -121,10 +121,12 @@ void draw_pt_emit_prepare( struct pt_emit *emit, emit_sz = 0; break; } - + + hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL; hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; hw_key.element[i].input_buffer = src_buffer; hw_key.element[i].input_offset = src_offset; + hw_key.element[i].instance_divisor = 0; hw_key.element[i].output_format = output_format; hw_key.element[i].output_offset = dst_offset; @@ -204,6 +206,7 @@ void draw_pt_emit( struct pt_emit *emit, translate->run( translate, 0, vertex_count, + draw->instance_id, hw_verts ); render->unmap_vertices( render, @@ -263,6 +266,7 @@ void draw_pt_emit_linear(struct pt_emit *emit, translate->run(translate, 0, count, + draw->instance_id, hw_verts); if (0) { diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 305bfef4352..55e7a7b81ad 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -58,12 +58,14 @@ struct pt_fetch { */ void draw_pt_fetch_prepare( struct pt_fetch *fetch, unsigned vs_input_count, - unsigned vertex_size ) + unsigned vertex_size, + unsigned instance_id_index ) { struct draw_context *draw = fetch->draw; unsigned nr_inputs; - unsigned i, nr = 0; + unsigned i, nr = 0, ei = 0; unsigned dst_offset = 0; + unsigned num_extra_inputs = 0; struct translate_key key; fetch->vertex_size = vertex_size; @@ -78,9 +80,11 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, { /* Need to set header->vertex_id = 0xffff somehow. */ + key.element[nr].type = TRANSLATE_ELEMENT_NORMAL; key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT; key.element[nr].input_buffer = draw->pt.nr_vertex_buffers; key.element[nr].input_offset = 0; + key.element[nr].instance_divisor = 0; key.element[nr].output_format = PIPE_FORMAT_R32_FLOAT; key.element[nr].output_offset = dst_offset; dst_offset += 1 * sizeof(float); @@ -91,19 +95,36 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, */ dst_offset += 4 * sizeof(float); } - - assert( draw->pt.nr_vertex_elements >= vs_input_count ); - nr_inputs = MIN2( vs_input_count, draw->pt.nr_vertex_elements ); + if (instance_id_index != ~0) { + num_extra_inputs++; + } + + assert(draw->pt.nr_vertex_elements + num_extra_inputs >= vs_input_count); + + nr_inputs = MIN2(vs_input_count, draw->pt.nr_vertex_elements + num_extra_inputs); for (i = 0; i < nr_inputs; i++) { - key.element[nr].input_format = draw->pt.vertex_element[i].src_format; - key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index; - key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset; - key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - key.element[nr].output_offset = dst_offset; + if (i == instance_id_index) { + key.element[nr].type = TRANSLATE_ELEMENT_INSTANCE_ID; + key.element[nr].input_format = PIPE_FORMAT_R32_USCALED; + key.element[nr].output_format = PIPE_FORMAT_R32_USCALED; + key.element[nr].output_offset = dst_offset; + + dst_offset += sizeof(uint); + } else { + key.element[nr].type = TRANSLATE_ELEMENT_NORMAL; + key.element[nr].input_format = draw->pt.vertex_element[ei].src_format; + key.element[nr].input_buffer = draw->pt.vertex_element[ei].vertex_buffer_index; + key.element[nr].input_offset = draw->pt.vertex_element[ei].src_offset; + key.element[nr].instance_divisor = draw->pt.vertex_element[ei].instance_divisor; + key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + key.element[nr].output_offset = dst_offset; + + ei++; + dst_offset += 4 * sizeof(float); + } - dst_offset += 4 * sizeof(float); nr++; } @@ -158,6 +179,7 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, translate->run_elts( translate, elts, count, + draw->instance_id, verts ); } @@ -183,6 +205,7 @@ void draw_pt_fetch_run_linear( struct pt_fetch *fetch, translate->run( translate, start, count, + draw->instance_id, verts ); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index e7fe6b3b768..2a604470e9a 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -166,9 +166,11 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, continue; } + key.element[i].type = TRANSLATE_ELEMENT_NORMAL; key.element[i].input_format = input_format; key.element[i].input_buffer = input_buffer; key.element[i].input_offset = input_offset; + key.element[i].instance_divisor = src->instance_divisor; key.element[i].output_format = output_format; key.element[i].output_offset = dst_offset; @@ -256,6 +258,7 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, feme->translate->run_elts( feme->translate, fetch_elts, fetch_count, + draw->instance_id, hw_verts ); if (0) { @@ -314,6 +317,7 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, feme->translate->run( feme->translate, start, count, + draw->instance_id, hw_verts ); if (0) { @@ -374,6 +378,7 @@ static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, feme->translate->run( feme->translate, start, count, + draw->instance_id, hw_verts ); draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) ); 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 1a9df4cac5d..23da556f797 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -59,6 +59,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; struct draw_vertex_shader *vs = draw->vs.vertex_shader; + unsigned i; + unsigned instance_id_index = ~0; /* Add one to num_outputs because the pipeline occasionally tags on * an additional texcoord, eg for AA lines. @@ -66,6 +68,15 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned nr = MAX2( vs->info.num_inputs, vs->info.num_outputs + 1 ); + /* Scan for instanceID system value. + */ + for (i = 0; i < vs->info.num_inputs; i++) { + if (vs->info.input_semantic_name[i] == TGSI_SEMANTIC_INSTANCEID) { + instance_id_index = i; + break; + } + } + fpme->prim = prim; fpme->opt = opt; @@ -79,7 +90,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, draw_pt_fetch_prepare( fpme->fetch, vs->info.num_inputs, - fpme->vertex_size ); + fpme->vertex_size, + instance_id_index ); /* XXX: it's not really gl rasterization rules we care about here, * but gl vs dx9 clip spaces. */ diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 35536895326..e03ac8c2291 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -43,11 +43,11 @@ #include "translate/translate.h" #include "translate/translate_cache.h" +#include "tgsi/tgsi_dump.h" #include "tgsi/tgsi_exec.h" - void draw_vs_set_constants( struct draw_context *draw, const float (*constants)[4], unsigned size ) @@ -83,6 +83,10 @@ draw_create_vertex_shader(struct draw_context *draw, { struct draw_vertex_shader *vs; + if (draw->dump_vs) { + tgsi_dump(shader->tokens, 0); + } + vs = draw_create_vs_llvm( draw, shader ); if (!vs) { vs = draw_create_vs_sse( draw, shader ); @@ -152,6 +156,8 @@ draw_delete_vertex_shader(struct draw_context *draw, boolean draw_vs_init( struct draw_context *draw ) { + draw->dump_vs = debug_get_bool_option("GALLIUM_DUMP_VS", FALSE); + draw->vs.machine = tgsi_exec_machine_create(); if (!draw->vs.machine) return FALSE; diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index e3b807ebd0e..00036cfe68b 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -43,6 +43,7 @@ struct draw_varient_input enum pipe_format format; unsigned buffer; unsigned offset; + unsigned instance_divisor; }; struct draw_varient_output diff --git a/src/gallium/auxiliary/draw/draw_vs_ppc.c b/src/gallium/auxiliary/draw/draw_vs_ppc.c index ad184bd696d..da9f3e3d35c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_ppc.c +++ b/src/gallium/auxiliary/draw/draw_vs_ppc.c @@ -98,9 +98,9 @@ vs_ppc_run_linear( struct draw_vertex_shader *base, /* loop over verts */ for (i = 0; i < count; i += MAX_VERTICES) { const uint max_vertices = MIN2(MAX_VERTICES, count - i); - float inputs_soa[PIPE_MAX_SHADER_INPUTS][4][4] ALIGN16_ATTRIB; - float outputs_soa[PIPE_MAX_SHADER_OUTPUTS][4][4] ALIGN16_ATTRIB; - float temps_soa[TGSI_EXEC_NUM_TEMPS][4][4] ALIGN16_ATTRIB; + PIPE_ALIGN_VAR(16) float inputs_soa[PIPE_MAX_SHADER_INPUTS][4][4]; + PIPE_ALIGN_VAR(16) float outputs_soa[PIPE_MAX_SHADER_OUTPUTS][4][4]; + PIPE_ALIGN_VAR(16) float temps_soa[TGSI_EXEC_NUM_TEMPS][4][4]; uint attr; /* convert (up to) four input verts to SoA format */ diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index d16692584e5..9f40030f39f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -142,6 +142,7 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, vsvg->fetch->run_elts( vsvg->fetch, elts, count, + vsvg->draw->instance_id, temp_buffer ); vsvg->base.vs->run_linear( vsvg->base.vs, @@ -181,6 +182,7 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, vsvg->emit->run( vsvg->emit, 0, count, + vsvg->draw->instance_id, output_buffer ); FREE(temp_buffer); @@ -203,6 +205,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, vsvg->fetch->run( vsvg->fetch, start, count, + vsvg->draw->instance_id, temp_buffer ); vsvg->base.vs->run_linear( vsvg->base.vs, @@ -239,6 +242,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, vsvg->emit->run( vsvg->emit, 0, count, + vsvg->draw->instance_id, output_buffer ); FREE(temp_buffer); @@ -281,9 +285,11 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, fetch.nr_elements = key->nr_inputs; fetch.output_stride = vsvg->temp_vertex_stride; for (i = 0; i < key->nr_inputs; i++) { + fetch.element[i].type = TRANSLATE_ELEMENT_NORMAL; fetch.element[i].input_format = key->element[i].in.format; fetch.element[i].input_buffer = key->element[i].in.buffer; fetch.element[i].input_offset = key->element[i].in.offset; + fetch.element[i].instance_divisor = 0; fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; fetch.element[i].output_offset = i * 4 * sizeof(float); assert(fetch.element[i].output_offset < fetch.output_stride); @@ -295,17 +301,21 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, for (i = 0; i < key->nr_outputs; i++) { if (key->element[i].out.format != EMIT_1F_PSIZE) { + emit.element[i].type = TRANSLATE_ELEMENT_NORMAL; emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; emit.element[i].input_buffer = 0; emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float); + emit.element[i].instance_divisor = 0; emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format); emit.element[i].output_offset = key->element[i].out.offset; assert(emit.element[i].input_offset <= fetch.output_stride); } else { + emit.element[i].type = TRANSLATE_ELEMENT_NORMAL; emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT; emit.element[i].input_buffer = 1; emit.element[i].input_offset = 0; + emit.element[i].instance_divisor = 0; emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT; emit.element[i].output_offset = key->element[i].out.offset; } diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index ba6f7b15f9e..a4b78f14943 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -80,27 +80,11 @@ struct fenced_buffer_list */ struct fenced_buffer { - /* - * Immutable members. - */ - struct pb_buffer base; + struct pb_buffer *buffer; - struct fenced_buffer_list *list; - - /** - * Protected by fenced_buffer_list::mutex - */ - struct list_head head; - /** - * Following members are mutable and protected by this mutex. - * - * You may lock this mutex alone, or lock it with fenced_buffer_list::mutex - * held, but in order to prevent deadlocks you must never lock - * fenced_buffer_list::mutex with this mutex held. - */ - pipe_mutex mutex; + /* FIXME: protect access with mutex */ /** * A bitmask of PIPE_BUFFER_USAGE_CPU/GPU_READ/WRITE describing the current @@ -112,6 +96,9 @@ struct fenced_buffer struct pb_validate *vl; unsigned validation_flags; struct pipe_fence_handle *fence; + + struct list_head head; + struct fenced_buffer_list *list; }; @@ -123,24 +110,15 @@ fenced_buffer(struct pb_buffer *buf) } -/** - * Add the buffer to the fenced list. - * - * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this - * order, before calling this function. - * - * Reference count should be incremented before calling this function. - */ static INLINE void -fenced_buffer_add_locked(struct fenced_buffer_list *fenced_list, - struct fenced_buffer *fenced_buf) +_fenced_buffer_add(struct fenced_buffer *fenced_buf) { + struct fenced_buffer_list *fenced_list = fenced_buf->list; + assert(pipe_is_referenced(&fenced_buf->base.base.reference)); assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); assert(fenced_buf->fence); - /* TODO: Move the reference count increment here */ - #ifdef DEBUG LIST_DEL(&fenced_buf->head); assert(fenced_list->numUnfenced); @@ -152,16 +130,32 @@ fenced_buffer_add_locked(struct fenced_buffer_list *fenced_list, /** - * Remove the buffer from the fenced list. - * - * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this - * order before calling this function. - * - * Reference count should be decremented after calling this function. + * Actually destroy the buffer. */ static INLINE void -fenced_buffer_remove_locked(struct fenced_buffer_list *fenced_list, - struct fenced_buffer *fenced_buf) +_fenced_buffer_destroy(struct fenced_buffer *fenced_buf) +{ + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + assert(!pipe_is_referenced(&fenced_buf->base.base.reference)); + assert(!fenced_buf->fence); +#ifdef DEBUG + assert(fenced_buf->head.prev); + assert(fenced_buf->head.next); + LIST_DEL(&fenced_buf->head); + assert(fenced_list->numUnfenced); + --fenced_list->numUnfenced; +#else + (void)fenced_list; +#endif + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); +} + + +static INLINE void +_fenced_buffer_remove(struct fenced_buffer_list *fenced_list, + struct fenced_buffer *fenced_buf) { struct pb_fence_ops *ops = fenced_list->ops; @@ -183,56 +177,37 @@ fenced_buffer_remove_locked(struct fenced_buffer_list *fenced_list, ++fenced_list->numUnfenced; #endif - /* TODO: Move the reference count decrement and destruction here */ + /** + * FIXME!!! + */ + + if(!pipe_is_referenced(&fenced_buf->base.base.reference)) + _fenced_buffer_destroy(fenced_buf); } -/** - * Wait for the fence to expire, and remove it from the fenced list. - * - * fenced_buffer::mutex must be held. fenced_buffer_list::mutex must not be - * held -- it will be acquired internally. - */ static INLINE enum pipe_error -fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, - struct fenced_buffer *fenced_buf) +_fenced_buffer_finish(struct fenced_buffer *fenced_buf) { + struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pb_fence_ops *ops = fenced_list->ops; - enum pipe_error ret = PIPE_ERROR; #if 0 debug_warning("waiting for GPU"); #endif - assert(pipe_is_referenced(&fenced_buf->base.base.reference)); assert(fenced_buf->fence); - - /* - * Acquire the global lock. Must release buffer mutex first to preserve - * lock order. - */ - pipe_mutex_unlock(fenced_buf->mutex); - pipe_mutex_lock(fenced_list->mutex); - pipe_mutex_lock(fenced_buf->mutex); - if(fenced_buf->fence) { - if(ops->fence_finish(ops, fenced_buf->fence, 0) == 0) { - /* Remove from the fenced list */ - /* TODO: remove consequents */ - fenced_buffer_remove_locked(fenced_list, fenced_buf); - - p_atomic_dec(&fenced_buf->base.base.reference.count); - assert(pipe_is_referenced(&fenced_buf->base.base.reference)); - - fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; - - ret = PIPE_OK; + if(ops->fence_finish(ops, fenced_buf->fence, 0) != 0) { + return PIPE_ERROR; } + /* Remove from the fenced list */ + /* TODO: remove consequents */ + _fenced_buffer_remove(fenced_list, fenced_buf); } - pipe_mutex_unlock(fenced_list->mutex); - - return ret; + fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; + return PIPE_OK; } @@ -240,8 +215,8 @@ fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, * Free as many fenced buffers from the list head as possible. */ static void -fenced_buffer_list_check_free_locked(struct fenced_buffer_list *fenced_list, - int wait) +_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) { struct pb_fence_ops *ops = fenced_list->ops; struct list_head *curr, *next; @@ -254,29 +229,21 @@ fenced_buffer_list_check_free_locked(struct fenced_buffer_list *fenced_list, while(curr != &fenced_list->delayed) { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); - pipe_mutex_lock(fenced_buf->mutex); - if(fenced_buf->fence != prev_fence) { int signaled; if (wait) signaled = ops->fence_finish(ops, fenced_buf->fence, 0); else signaled = ops->fence_signalled(ops, fenced_buf->fence, 0); - if (signaled != 0) { - pipe_mutex_unlock(fenced_buf->mutex); + if (signaled != 0) break; - } prev_fence = fenced_buf->fence; } else { assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0); } - fenced_buffer_remove_locked(fenced_list, fenced_buf); - pipe_mutex_unlock(fenced_buf->mutex); - - pb_buf = &fenced_buf->base; - pb_reference(&pb_buf, NULL); + _fenced_buffer_remove(fenced_list, fenced_buf); curr = next; next = curr->next; @@ -290,25 +257,30 @@ fenced_buffer_destroy(struct pb_buffer *buf) struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; - assert(!pipe_is_referenced(&fenced_buf->base.base.reference)); - assert(!fenced_buf->fence); - -#ifdef DEBUG pipe_mutex_lock(fenced_list->mutex); - assert(fenced_buf->head.prev); - assert(fenced_buf->head.next); - LIST_DEL(&fenced_buf->head); - assert(fenced_list->numUnfenced); - --fenced_list->numUnfenced; + assert(!pipe_is_referenced(&fenced_buf->base.base.reference)); + if (fenced_buf->fence) { + struct pb_fence_ops *ops = fenced_list->ops; + if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) { + struct list_head *curr, *prev; + curr = &fenced_buf->head; + prev = curr->prev; + do { + fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0); + _fenced_buffer_remove(fenced_list, fenced_buf); + curr = prev; + prev = curr->prev; + } while (curr != &fenced_list->delayed); + } + else { + /* delay destruction */ + } + } + else { + _fenced_buffer_destroy(fenced_buf); + } pipe_mutex_unlock(fenced_list->mutex); -#else - (void)fenced_list; -#endif - - pb_reference(&fenced_buf->buffer, NULL); - - pipe_mutex_destroy(fenced_buf->mutex); - FREE(fenced_buf); } @@ -319,23 +291,24 @@ fenced_buffer_map(struct pb_buffer *buf, struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pb_fence_ops *ops = fenced_list->ops; - void *map = NULL; - - pipe_mutex_lock(fenced_buf->mutex); + void *map; assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE)); /* Serialize writes */ if((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) || ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) && (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) { - if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) && - ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) { + if(flags & PIPE_BUFFER_USAGE_DONTBLOCK) { /* Don't wait for the GPU to finish writing */ - goto done; + if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) + _fenced_buffer_remove(fenced_list, fenced_buf); + else + return NULL; + } + else { + /* Wait for the GPU to finish writing */ + _fenced_buffer_finish(fenced_buf); } - - /* Wait for the GPU to finish writing */ - fenced_buffer_finish_locked(fenced_list, fenced_buf); } #if 0 @@ -352,9 +325,6 @@ fenced_buffer_map(struct pb_buffer *buf, fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE; } -done: - pipe_mutex_unlock(fenced_buf->mutex); - return map; } @@ -363,9 +333,6 @@ static void fenced_buffer_unmap(struct pb_buffer *buf) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); - - pipe_mutex_lock(fenced_buf->mutex); - assert(fenced_buf->mapcount); if(fenced_buf->mapcount) { pb_unmap(fenced_buf->buffer); @@ -373,8 +340,6 @@ fenced_buffer_unmap(struct pb_buffer *buf) if(!fenced_buf->mapcount) fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE; } - - pipe_mutex_unlock(fenced_buf->mutex); } @@ -386,14 +351,11 @@ fenced_buffer_validate(struct pb_buffer *buf, struct fenced_buffer *fenced_buf = fenced_buffer(buf); enum pipe_error ret; - pipe_mutex_lock(fenced_buf->mutex); - if(!vl) { /* invalidate */ fenced_buf->vl = NULL; fenced_buf->validation_flags = 0; - ret = PIPE_OK; - goto done; + return PIPE_OK; } assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); @@ -401,17 +363,14 @@ fenced_buffer_validate(struct pb_buffer *buf, flags &= PIPE_BUFFER_USAGE_GPU_READ_WRITE; /* Buffer cannot be validated in two different lists */ - if(fenced_buf->vl && fenced_buf->vl != vl) { - ret = PIPE_ERROR_RETRY; - goto done; - } + if(fenced_buf->vl && fenced_buf->vl != vl) + return PIPE_ERROR_RETRY; #if 0 /* Do not validate if buffer is still mapped */ if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { /* TODO: wait for the thread that mapped the buffer to unmap it */ - ret = PIPE_ERROR_RETRY; - goto done; + return PIPE_ERROR_RETRY; } /* Final sanity checking */ assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE)); @@ -421,21 +380,17 @@ fenced_buffer_validate(struct pb_buffer *buf, if(fenced_buf->vl == vl && (fenced_buf->validation_flags & flags) == flags) { /* Nothing to do -- buffer already validated */ - ret = PIPE_OK; - goto done; + return PIPE_OK; } ret = pb_validate(fenced_buf->buffer, vl, flags); if (ret != PIPE_OK) - goto done; + return ret; fenced_buf->vl = vl; fenced_buf->validation_flags |= flags; -done: - pipe_mutex_unlock(fenced_buf->mutex); - - return ret; + return PIPE_OK; } @@ -450,36 +405,29 @@ fenced_buffer_fence(struct pb_buffer *buf, fenced_buf = fenced_buffer(buf); fenced_list = fenced_buf->list; ops = fenced_list->ops; - - pipe_mutex_lock(fenced_list->mutex); - pipe_mutex_lock(fenced_buf->mutex); - - assert(pipe_is_referenced(&fenced_buf->base.base.reference)); - - if(fence != fenced_buf->fence) { - assert(fenced_buf->vl); - assert(fenced_buf->validation_flags); - - if (fenced_buf->fence) { - fenced_buffer_remove_locked(fenced_list, fenced_buf); - p_atomic_dec(&fenced_buf->base.base.reference.count); - assert(pipe_is_referenced(&fenced_buf->base.base.reference)); - } - if (fence) { - ops->fence_reference(ops, &fenced_buf->fence, fence); - fenced_buf->flags |= fenced_buf->validation_flags; - p_atomic_inc(&fenced_buf->base.base.reference.count); - fenced_buffer_add_locked(fenced_list, fenced_buf); - } - - pb_fence(fenced_buf->buffer, fence); - fenced_buf->vl = NULL; - fenced_buf->validation_flags = 0; + if(fence == fenced_buf->fence) { + /* Nothing to do */ + return; } - pipe_mutex_unlock(fenced_buf->mutex); + assert(fenced_buf->vl); + assert(fenced_buf->validation_flags); + + pipe_mutex_lock(fenced_list->mutex); + if (fenced_buf->fence) + _fenced_buffer_remove(fenced_list, fenced_buf); + if (fence) { + ops->fence_reference(ops, &fenced_buf->fence, fence); + fenced_buf->flags |= fenced_buf->validation_flags; + _fenced_buffer_add(fenced_buf); + } pipe_mutex_unlock(fenced_list->mutex); + + pb_fence(fenced_buf->buffer, fence); + + fenced_buf->vl = NULL; + fenced_buf->validation_flags = 0; } @@ -489,7 +437,6 @@ fenced_buffer_get_base_buffer(struct pb_buffer *buf, pb_size *offset) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); - /* NOTE: accesses immutable members only -- mutex not necessary */ pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); } @@ -529,8 +476,6 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, buf->buffer = buffer; buf->list = fenced_list; - pipe_mutex_init(buf->mutex); - #ifdef DEBUG pipe_mutex_lock(fenced_list->mutex); LIST_ADDTAIL(&buf->head, &fenced_list->unfenced); @@ -572,7 +517,7 @@ fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, int wait) { pipe_mutex_lock(fenced_list->mutex); - fenced_buffer_list_check_free_locked(fenced_list, wait); + _fenced_buffer_list_check_free(fenced_list, wait); pipe_mutex_unlock(fenced_list->mutex); } @@ -594,13 +539,11 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) next = curr->next; while(curr != &fenced_list->unfenced) { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); - pipe_mutex_lock(fenced_buf->mutex); assert(!fenced_buf->fence); debug_printf("%10p %7u %7u\n", (void *) fenced_buf, fenced_buf->base.base.size, p_atomic_read(&fenced_buf->base.base.reference.count)); - pipe_mutex_unlock(fenced_buf->mutex); curr = next; next = curr->next; } @@ -610,7 +553,6 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) while(curr != &fenced_list->delayed) { int signaled; fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); - pipe_mutex_lock(fenced_buf->mutex); signaled = ops->fence_signalled(ops, fenced_buf->fence, 0); debug_printf("%10p %7u %7u %10p %s\n", (void *) fenced_buf, @@ -618,7 +560,6 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) p_atomic_read(&fenced_buf->base.base.reference.count), (void *) fenced_buf->fence, signaled == 0 ? "y" : "n"); - pipe_mutex_unlock(fenced_buf->mutex); curr = next; next = curr->next; } @@ -639,8 +580,8 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) #if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) sched_yield(); #endif + _fenced_buffer_list_check_free(fenced_list, 1); pipe_mutex_lock(fenced_list->mutex); - fenced_buffer_list_check_free_locked(fenced_list, 1); } #ifdef DEBUG @@ -648,7 +589,6 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) #endif pipe_mutex_unlock(fenced_list->mutex); - pipe_mutex_destroy(fenced_list->mutex); fenced_list->ops->destroy(fenced_list->ops); diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 1acf3c373eb..f675427d987 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -673,6 +673,13 @@ void x86_and( struct x86_function *p, emit_op_modrm( p, 0x23, 0x21, dst, src ); } +void x86_div( struct x86_function *p, + struct x86_reg src ) +{ + assert(src.file == file_REG32 && src.mod == mod_REG); + emit_op_modrm(p, 0xf7, 0, x86_make_reg(file_REG32, 6), src); +} + /*********************************************************************** diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 731a6517968..f7612d416a0 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -244,6 +244,7 @@ void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_sahf( struct x86_function *p ); +void x86_div( struct x86_function *p, struct x86_reg src ); void x86_cdecl_caller_push_regs( struct x86_function *p ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index e2e5394f86f..d7ff262f30a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -123,7 +123,8 @@ static const char *semantic_names[] = "NORMAL", "FACE", "EDGEFLAG", - "PRIM_ID" + "PRIM_ID", + "INSTANCEID" }; static const char *immediate_type_names[] = diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 2bcb33392a8..118a638ab48 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1041,11 +1041,19 @@ fetch_src_file_channel( default: assert( 0 ); + chan->u[0] = 0; + chan->u[1] = 0; + chan->u[2] = 0; + chan->u[3] = 0; } break; default: assert( 0 ); + chan->u[0] = 0; + chan->u[1] = 0; + chan->u[2] = 0; + chan->u[3] = 0; } } diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c index 138d2d095bb..ad553c71a57 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c @@ -51,7 +51,8 @@ * Since it's pretty much impossible to form PPC vector immediates, load * them from memory here: */ -const float ppc_builtin_constants[] ALIGN16_ATTRIB = { +PIPE_ALIGN_VAR(16) const float +ppc_builtin_constants[] = { 1.0f, -128.0f, 128.0, 0.0 }; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 7f1c8e5dd68..e1e4f97967d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -335,13 +335,13 @@ iter_instruction( fill_scan_register1d(ind_reg, inst->Src[i].Indirect.File, inst->Src[i].Indirect.Index); - if (!(reg->file == TGSI_FILE_ADDRESS || reg->file == TGSI_FILE_LOOP) || - reg->indices[0] != 0) { + if (!(ind_reg->file == TGSI_FILE_ADDRESS || ind_reg->file == TGSI_FILE_LOOP) || + ind_reg->indices[0] != 0) { report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]"); } check_register_usage( ctx, - reg, + ind_reg, "indirect", FALSE ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 2e13a7aaf9d..a85cc4659e0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -2146,40 +2146,50 @@ emit_instruction( break; case TGSI_OPCODE_XPD: + /* Note: we do all stores after all operands have been fetched + * to avoid src/dst register aliasing issues for an instruction + * such as: XPD TEMP[2].xyz, TEMP[0], TEMP[2]; + */ if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 1, 1, CHAN_Z ); - FETCH( func, *inst, 3, 0, CHAN_Z ); + FETCH( func, *inst, 1, 1, CHAN_Z ); /* xmm[1] = src[1].z */ + FETCH( func, *inst, 3, 0, CHAN_Z ); /* xmm[3] = src[0].z */ } if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 4, 1, CHAN_Y ); + FETCH( func, *inst, 0, 0, CHAN_Y ); /* xmm[0] = src[0].y */ + FETCH( func, *inst, 4, 1, CHAN_Y ); /* xmm[4] = src[1].y */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_MOV( func, 2, 0 ); - emit_mul( func, 2, 1 ); - emit_MOV( func, 5, 3 ); - emit_mul( func, 5, 4 ); - emit_sub( func, 2, 5 ); - STORE( func, *inst, 2, 0, CHAN_X ); + emit_MOV( func, 7, 0 ); /* xmm[7] = xmm[0] */ + emit_mul( func, 7, 1 ); /* xmm[7] = xmm[2] * xmm[1] */ + emit_MOV( func, 5, 3 ); /* xmm[5] = xmm[3] */ + emit_mul( func, 5, 4 ); /* xmm[5] = xmm[5] * xmm[4] */ + emit_sub( func, 7, 5 ); /* xmm[7] = xmm[2] - xmm[5] */ + /* store xmm[7] in dst.x below */ } if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 2, 1, CHAN_X ); - FETCH( func, *inst, 5, 0, CHAN_X ); + FETCH( func, *inst, 2, 1, CHAN_X ); /* xmm[2] = src[1].x */ + FETCH( func, *inst, 5, 0, CHAN_X ); /* xmm[5] = src[0].x */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - emit_mul( func, 3, 2 ); - emit_mul( func, 1, 5 ); - emit_sub( func, 3, 1 ); - STORE( func, *inst, 3, 0, CHAN_Y ); + emit_mul( func, 3, 2 ); /* xmm[3] = xmm[3] * xmm[2] */ + emit_mul( func, 1, 5 ); /* xmm[1] = xmm[1] * xmm[5] */ + emit_sub( func, 3, 1 ); /* xmm[3] = xmm[3] - xmm[1] */ + /* store xmm[3] in dst.y below */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - emit_mul( func, 5, 4 ); - emit_mul( func, 0, 2 ); - emit_sub( func, 5, 0 ); - STORE( func, *inst, 5, 0, CHAN_Z ); + emit_mul( func, 5, 4 ); /* xmm[5] = xmm[5] * xmm[4] */ + emit_mul( func, 0, 2 ); /* xmm[0] = xmm[0] * xmm[2] */ + emit_sub( func, 5, 0 ); /* xmm[5] = xmm[5] - xmm[0] */ + STORE( func, *inst, 5, 0, CHAN_Z ); /* dst.z = xmm[5] */ + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + STORE( func, *inst, 7, 0, CHAN_X ); /* dst.x = xmm[7] */ + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + STORE( func, *inst, 3, 0, CHAN_Y ); /* dst.y = xmm[3] */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { emit_tempf( diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c index e64e2b731df..8bd6f68dcc0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -40,6 +40,8 @@ union tgsi_any_token { struct tgsi_header header; struct tgsi_processor processor; struct tgsi_token token; + struct tgsi_property prop; + struct tgsi_property_data prop_data; struct tgsi_declaration decl; struct tgsi_declaration_range decl_range; struct tgsi_declaration_semantic decl_semantic; @@ -64,6 +66,7 @@ struct ureg_tokens { }; #define UREG_MAX_INPUT PIPE_MAX_ATTRIBS +#define UREG_MAX_SYSTEM_VALUE PIPE_MAX_ATTRIBS #define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS #define UREG_MAX_CONSTANT_RANGE 32 #define UREG_MAX_IMMEDIATE 32 @@ -95,6 +98,13 @@ struct ureg_program unsigned nr_gs_inputs; struct { + unsigned index; + unsigned semantic_name; + unsigned semantic_index; + } system_value[UREG_MAX_SYSTEM_VALUE]; + unsigned nr_system_values; + + struct { unsigned semantic_name; unsigned semantic_index; } output[UREG_MAX_OUTPUT]; @@ -123,6 +133,8 @@ struct ureg_program } constant_range[UREG_MAX_CONSTANT_RANGE]; unsigned nr_constant_ranges; + unsigned property_gs_input_prim; + unsigned nr_addrs; unsigned nr_preds; unsigned nr_loops; @@ -234,19 +246,28 @@ ureg_src_register( unsigned file, src.SwizzleY = TGSI_SWIZZLE_Y; src.SwizzleZ = TGSI_SWIZZLE_Z; src.SwizzleW = TGSI_SWIZZLE_W; - src.Pad = 0; src.Indirect = 0; src.IndirectIndex = 0; src.IndirectSwizzle = 0; src.Absolute = 0; src.Index = index; src.Negate = 0; + src.Dimension = 0; + src.DimensionIndex = 0; return src; } +void +ureg_property_gs_input_prim(struct ureg_program *ureg, + unsigned gs_input_prim) +{ + ureg->property_gs_input_prim = gs_input_prim; +} + + struct ureg_src ureg_DECL_fs_input( struct ureg_program *ureg, @@ -304,6 +325,25 @@ ureg_DECL_gs_input(struct ureg_program *ureg, } +struct ureg_src +ureg_DECL_system_value(struct ureg_program *ureg, + unsigned index, + unsigned semantic_name, + unsigned semantic_index) +{ + if (ureg->nr_system_values < UREG_MAX_SYSTEM_VALUE) { + ureg->system_value[ureg->nr_system_values].index = index; + ureg->system_value[ureg->nr_system_values].semantic_name = semantic_name; + ureg->system_value[ureg->nr_system_values].semantic_index = semantic_index; + ureg->nr_system_values++; + } else { + set_bad(ureg); + } + + return ureg_src_register(TGSI_FILE_SYSTEM_VALUE, index); +} + + struct ureg_dst ureg_DECL_output( struct ureg_program *ureg, unsigned name, @@ -628,7 +668,7 @@ void ureg_emit_src( struct ureg_program *ureg, struct ureg_src src ) { - unsigned size = 1 + (src.Indirect ? 1 : 0); + unsigned size = 1 + (src.Indirect ? 1 : 0) + (src.Dimension ? 1 : 0); union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); unsigned n = 0; @@ -660,6 +700,15 @@ ureg_emit_src( struct ureg_program *ureg, n++; } + if (src.Dimension) { + out[0].src.Dimension = 1; + out[n].dim.Indirect = 0; + out[n].dim.Dimension = 0; + out[n].dim.Padding = 0; + out[n].dim.Index = src.DimensionIndex; + n++; + } + assert(n == size); } @@ -1027,13 +1076,34 @@ emit_immediate( struct ureg_program *ureg, out[4].imm_data.Uint = v[3]; } +static void +emit_property(struct ureg_program *ureg, + unsigned name, + unsigned data) +{ + union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, 2); + out[0].value = 0; + out[0].prop.Type = TGSI_TOKEN_TYPE_PROPERTY; + out[0].prop.NrTokens = 2; + out[0].prop.PropertyName = name; + + out[1].prop_data.Data = data; +} static void emit_decls( struct ureg_program *ureg ) { unsigned i; + if (ureg->property_gs_input_prim != ~0) { + assert(ureg->processor == TGSI_PROCESSOR_GEOMETRY); + + emit_property(ureg, + TGSI_PROPERTY_GS_INPUT_PRIM, + ureg->property_gs_input_prim); + } + if (ureg->processor == TGSI_PROCESSOR_VERTEX) { for (i = 0; i < UREG_MAX_INPUT; i++) { if (ureg->vs_inputs[i/32] & (1 << (i%32))) { @@ -1058,6 +1128,15 @@ static void emit_decls( struct ureg_program *ureg ) } } + for (i = 0; i < ureg->nr_system_values; i++) { + emit_decl(ureg, + TGSI_FILE_SYSTEM_VALUE, + ureg->system_value[i].index, + ureg->system_value[i].semantic_name, + ureg->system_value[i].semantic_index, + TGSI_INTERPOLATE_CONSTANT); + } + for (i = 0; i < ureg->nr_outputs; i++) { emit_decl( ureg, TGSI_FILE_OUTPUT, @@ -1234,6 +1313,7 @@ struct ureg_program *ureg_create( unsigned processor ) return NULL; ureg->processor = processor; + ureg->property_gs_input_prim = ~0; return ureg; } diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h index 6f11273320a..03eaf24854c 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -47,13 +47,14 @@ struct ureg_src unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned Pad : 1; /* BOOL */ unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ unsigned Absolute : 1; /* BOOL */ - int Index : 16; /* SINT */ unsigned Negate : 1; /* BOOL */ + int Index : 16; /* SINT */ int IndirectIndex : 16; /* SINT */ int IndirectSwizzle : 2; /* TGSI_SWIZZLE_ */ + int DimensionIndex : 16; /* SINT */ }; /* Very similar to a tgsi_dst_register, removing unsupported fields @@ -118,6 +119,14 @@ ureg_create_shader_and_destroy( struct ureg_program *p, } +/*********************************************************************** + * Build shader properties: + */ + +void +ureg_property_gs_input_prim(struct ureg_program *ureg, + unsigned gs_input_prim); + /*********************************************************************** * Build shader declarations: @@ -137,6 +146,12 @@ struct ureg_src ureg_DECL_gs_input(struct ureg_program *, unsigned index); +struct ureg_src +ureg_DECL_system_value(struct ureg_program *, + unsigned index, + unsigned semantic_name, + unsigned semantic_index); + struct ureg_dst ureg_DECL_output( struct ureg_program *, unsigned semantic_name, @@ -760,6 +775,15 @@ ureg_src_indirect( struct ureg_src reg, struct ureg_src addr ) return reg; } +static INLINE struct ureg_src +ureg_src_dimension( struct ureg_src reg, int index ) +{ + assert(reg.File != TGSI_FILE_NULL); + reg.Dimension = 1; + reg.DimensionIndex = index; + return reg; +} + static INLINE struct ureg_dst ureg_dst( struct ureg_src src ) { @@ -792,13 +816,14 @@ ureg_src( struct ureg_dst dst ) src.SwizzleY = TGSI_SWIZZLE_Y; src.SwizzleZ = TGSI_SWIZZLE_Z; src.SwizzleW = TGSI_SWIZZLE_W; - src.Pad = 0; src.Indirect = dst.Indirect; src.IndirectIndex = dst.IndirectIndex; src.IndirectSwizzle = dst.IndirectSwizzle; src.Absolute = 0; src.Index = dst.Index; src.Negate = 0; + src.Dimension = 0; + src.DimensionIndex = 0; return src; } @@ -837,13 +862,14 @@ ureg_src_undef( void ) src.SwizzleY = 0; src.SwizzleZ = 0; src.SwizzleW = 0; - src.Pad = 0; src.Indirect = 0; src.IndirectIndex = 0; src.IndirectSwizzle = 0; src.Absolute = 0; src.Index = 0; src.Negate = 0; + src.Dimension = 0; + src.DimensionIndex = 0; return src; } diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index 34526eb0617..54ed2c1a4be 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -44,12 +44,19 @@ #include "pipe/p_format.h" #include "pipe/p_state.h" +enum translate_element_type { + TRANSLATE_ELEMENT_NORMAL, + TRANSLATE_ELEMENT_INSTANCE_ID +}; + struct translate_element { + enum translate_element_type type; enum pipe_format input_format; enum pipe_format output_format; unsigned input_buffer:8; unsigned input_offset:24; + unsigned instance_divisor; unsigned output_offset; }; @@ -74,11 +81,13 @@ struct translate { void (PIPE_CDECL *run_elts)( struct translate *, const unsigned *elts, unsigned count, + unsigned instance_id, void *output_buffer); void (PIPE_CDECL *run)( struct translate *, unsigned start, unsigned count, + unsigned instance_id, void *output_buffer); }; @@ -103,8 +112,13 @@ static INLINE int translate_keysize( const struct translate_key *key ) static INLINE int translate_key_compare( const struct translate_key *a, const struct translate_key *b ) { - int keysize = translate_keysize(a); - return memcmp(a, b, keysize); + int keysize_a = translate_keysize(a); + int keysize_b = translate_keysize(b); + + if (keysize_a != keysize_b) { + return keysize_a - keysize_b; + } + return memcmp(a, b, keysize_a); } diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 266e7ee81e6..24727d49888 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -46,9 +46,12 @@ struct translate_generic { struct translate translate; struct { + enum translate_element_type type; + fetch_func fetch; unsigned buffer; unsigned input_offset; + unsigned instance_divisor; emit_func emit; unsigned output_offset; @@ -568,6 +571,7 @@ static emit_func get_emit_func( enum pipe_format format ) static void PIPE_CDECL generic_run_elts( struct translate *translate, const unsigned *elts, unsigned count, + unsigned instance_id, void *output_buffer ) { struct translate_generic *tg = translate_generic(translate); @@ -583,13 +587,20 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate, for (attr = 0; attr < nr_attrs; attr++) { float data[4]; - - const char *src = (tg->attrib[attr].input_ptr + - tg->attrib[attr].input_stride * elt); + const char *src; char *dst = (vert + tg->attrib[attr].output_offset); + if (tg->attrib[attr].instance_divisor) { + src = tg->attrib[attr].input_ptr + + tg->attrib[attr].input_stride * + (instance_id / tg->attrib[attr].instance_divisor); + } else { + src = tg->attrib[attr].input_ptr + + tg->attrib[attr].input_stride * elt; + } + tg->attrib[attr].fetch( src, data ); if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n", @@ -607,6 +618,7 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate, static void PIPE_CDECL generic_run( struct translate *translate, unsigned start, unsigned count, + unsigned instance_id, void *output_buffer ) { struct translate_generic *tg = translate_generic(translate); @@ -623,13 +635,25 @@ static void PIPE_CDECL generic_run( struct translate *translate, for (attr = 0; attr < nr_attrs; attr++) { float data[4]; - const char *src = (tg->attrib[attr].input_ptr + - tg->attrib[attr].input_stride * elt); - char *dst = (vert + tg->attrib[attr].output_offset); - tg->attrib[attr].fetch( src, data ); + if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) { + const char *src; + + if (tg->attrib[attr].instance_divisor) { + src = tg->attrib[attr].input_ptr + + tg->attrib[attr].input_stride * + (instance_id / tg->attrib[attr].instance_divisor); + } else { + src = tg->attrib[attr].input_ptr + + tg->attrib[attr].input_stride * elt; + } + + tg->attrib[attr].fetch( src, data ); + } else { + data[0] = (float)instance_id; + } if (0) debug_printf("vert %d attr %d: %f %f %f %f\n", i, attr, data[0], data[1], data[2], data[3]); @@ -683,10 +707,12 @@ struct translate *translate_generic_create( const struct translate_key *key ) tg->translate.run = generic_run; for (i = 0; i < key->nr_elements; i++) { + tg->attrib[i].type = key->element[i].type; tg->attrib[i].fetch = get_fetch_func(key->element[i].input_format); tg->attrib[i].buffer = key->element[i].input_buffer; tg->attrib[i].input_offset = key->element[i].input_offset; + tg->attrib[i].instance_divisor = key->element[i].instance_divisor; tg->attrib[i].emit = get_emit_func(key->element[i].output_format); tg->attrib[i].output_offset = key->element[i].output_offset; diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index b62db8d8f33..c13e7427387 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -49,19 +49,29 @@ typedef void (PIPE_CDECL *run_func)( struct translate *translate, unsigned start, unsigned count, - void *output_buffer ); + unsigned instance_id, + void *output_buffer); typedef void (PIPE_CDECL *run_elts_func)( struct translate *translate, const unsigned *elts, unsigned count, - void *output_buffer ); + unsigned instance_id, + void *output_buffer); struct translate_buffer { const void *base_ptr; unsigned stride; - void *ptr; /* updated per vertex */ }; +struct translate_buffer_varient { + unsigned buffer_index; + unsigned instance_divisor; + void *ptr; /* updated either per vertex or per instance */ +}; + + +#define ELEMENT_BUFFER_INSTANCE_ID 1001 + struct translate_sse { struct translate translate; @@ -81,6 +91,16 @@ struct translate_sse { struct translate_buffer buffer[PIPE_MAX_ATTRIBS]; unsigned nr_buffers; + /* Multiple buffer varients can map to a single buffer. */ + struct translate_buffer_varient buffer_varient[PIPE_MAX_ATTRIBS]; + unsigned nr_buffer_varients; + + /* Multiple elements can map to a single buffer varient. */ + unsigned element_to_buffer_varient[PIPE_MAX_ATTRIBS]; + + boolean use_instancing; + unsigned instance_id; + run_func gen_run; run_elts_func gen_run_elts; @@ -359,32 +379,61 @@ static boolean init_inputs( struct translate_sse *p, boolean linear ) { unsigned i; - if (linear) { - for (i = 0; i < p->nr_buffers; i++) { + struct x86_reg instance_id = x86_make_disp(p->machine_EDX, + get_offset(p, &p->instance_id)); + + for (i = 0; i < p->nr_buffer_varients; i++) { + struct translate_buffer_varient *varient = &p->buffer_varient[i]; + struct translate_buffer *buffer = &p->buffer[varient->buffer_index]; + + if (linear || varient->instance_divisor) { struct x86_reg buf_stride = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[i].stride)); + get_offset(p, &buffer->stride)); struct x86_reg buf_ptr = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[i].ptr)); + get_offset(p, &varient->ptr)); struct x86_reg buf_base_ptr = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[i].base_ptr)); + get_offset(p, &buffer->base_ptr)); struct x86_reg elt = p->idx_EBX; - struct x86_reg tmp = p->tmp_EAX; - + struct x86_reg tmp_EAX = p->tmp_EAX; /* Calculate pointer to first attrib: + * base_ptr + stride * index, where index depends on instance divisor */ - x86_mov(p->func, tmp, buf_stride); - x86_imul(p->func, tmp, elt); - x86_add(p->func, tmp, buf_base_ptr); + if (varient->instance_divisor) { + /* Our index is instance ID divided by instance divisor. + */ + x86_mov(p->func, tmp_EAX, instance_id); + + if (varient->instance_divisor != 1) { + struct x86_reg tmp_EDX = p->machine_EDX; + struct x86_reg tmp_ECX = p->outbuf_ECX; + + /* TODO: Add x86_shr() to rtasm and use it whenever + * instance divisor is power of two. + */ + + x86_push(p->func, tmp_EDX); + x86_push(p->func, tmp_ECX); + x86_xor(p->func, tmp_EDX, tmp_EDX); + x86_mov_reg_imm(p->func, tmp_ECX, varient->instance_divisor); + x86_div(p->func, tmp_ECX); /* EAX = EDX:EAX / ECX */ + x86_pop(p->func, tmp_ECX); + x86_pop(p->func, tmp_EDX); + } + } else { + x86_mov(p->func, tmp_EAX, elt); + } + x86_imul(p->func, tmp_EAX, buf_stride); + x86_add(p->func, tmp_EAX, buf_base_ptr); /* In the linear case, keep the buffer pointer instead of the * index number. */ - if (p->nr_buffers == 1) - x86_mov( p->func, elt, tmp ); + if (linear && p->nr_buffer_varients == 1) + x86_mov(p->func, elt, tmp_EAX); else - x86_mov( p->func, buf_ptr, tmp ); + x86_mov(p->func, buf_ptr, tmp_EAX); } } @@ -394,31 +443,36 @@ static boolean init_inputs( struct translate_sse *p, static struct x86_reg get_buffer_ptr( struct translate_sse *p, boolean linear, - unsigned buf_idx, + unsigned var_idx, struct x86_reg elt ) { - if (linear && p->nr_buffers == 1) { + if (var_idx == ELEMENT_BUFFER_INSTANCE_ID) { + return x86_make_disp(p->machine_EDX, + get_offset(p, &p->instance_id)); + } + if (linear && p->nr_buffer_varients == 1) { return p->idx_EBX; } - else if (linear) { + else if (linear || p->buffer_varient[var_idx].instance_divisor) { struct x86_reg ptr = p->tmp_EAX; struct x86_reg buf_ptr = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[buf_idx].ptr)); + get_offset(p, &p->buffer_varient[var_idx].ptr)); x86_mov(p->func, ptr, buf_ptr); return ptr; } else { struct x86_reg ptr = p->tmp_EAX; + const struct translate_buffer_varient *varient = &p->buffer_varient[var_idx]; struct x86_reg buf_stride = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[buf_idx].stride)); + get_offset(p, &p->buffer[varient->buffer_index].stride)); struct x86_reg buf_base_ptr = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[buf_idx].base_ptr)); + get_offset(p, &p->buffer[varient->buffer_index].base_ptr)); @@ -436,28 +490,33 @@ static struct x86_reg get_buffer_ptr( struct translate_sse *p, static boolean incr_inputs( struct translate_sse *p, boolean linear ) { - if (linear && p->nr_buffers == 1) { + if (linear && p->nr_buffer_varients == 1) { struct x86_reg stride = x86_make_disp(p->machine_EDX, get_offset(p, &p->buffer[0].stride)); - x86_add(p->func, p->idx_EBX, stride); - sse_prefetchnta(p->func, x86_make_disp(p->idx_EBX, 192)); + if (p->buffer_varient[0].instance_divisor == 0) { + x86_add(p->func, p->idx_EBX, stride); + sse_prefetchnta(p->func, x86_make_disp(p->idx_EBX, 192)); + } } else if (linear) { unsigned i; /* Is this worthwhile?? */ - for (i = 0; i < p->nr_buffers; i++) { + for (i = 0; i < p->nr_buffer_varients; i++) { + struct translate_buffer_varient *varient = &p->buffer_varient[i]; struct x86_reg buf_ptr = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[i].ptr)); + get_offset(p, &varient->ptr)); struct x86_reg buf_stride = x86_make_disp(p->machine_EDX, - get_offset(p, &p->buffer[i].stride)); + get_offset(p, &p->buffer[varient->buffer_index].stride)); - x86_mov(p->func, p->tmp_EAX, buf_ptr); - x86_add(p->func, p->tmp_EAX, buf_stride); - if (i == 0) sse_prefetchnta(p->func, x86_make_disp(p->tmp_EAX, 192)); - x86_mov(p->func, buf_ptr, p->tmp_EAX); + if (varient->instance_divisor == 0) { + x86_mov(p->func, p->tmp_EAX, buf_ptr); + x86_add(p->func, p->tmp_EAX, buf_stride); + if (i == 0) sse_prefetchnta(p->func, x86_make_disp(p->tmp_EAX, 192)); + x86_mov(p->func, buf_ptr, p->tmp_EAX); + } } } else { @@ -514,7 +573,18 @@ static boolean build_vertex_emit( struct translate_sse *p, x86_mov(p->func, p->machine_EDX, x86_fn_arg(p->func, 1)); x86_mov(p->func, p->idx_EBX, x86_fn_arg(p->func, 2)); x86_mov(p->func, p->count_ESI, x86_fn_arg(p->func, 3)); - x86_mov(p->func, p->outbuf_ECX, x86_fn_arg(p->func, 4)); + x86_mov(p->func, p->outbuf_ECX, x86_fn_arg(p->func, 5)); + + /* Load instance ID. + */ + if (p->use_instancing) { + x86_mov(p->func, + p->tmp_EAX, + x86_fn_arg(p->func, 4)); + x86_mov(p->func, + x86_make_disp(p->machine_EDX, get_offset(p, &p->instance_id)), + p->tmp_EAX); + } /* Get vertex count, compare to zero */ @@ -531,17 +601,18 @@ static boolean build_vertex_emit( struct translate_sse *p, label = x86_get_label(p->func); { struct x86_reg elt = linear ? p->idx_EBX : x86_deref(p->idx_EBX); - int last_vb = -1; + int last_varient = -1; struct x86_reg vb; for (j = 0; j < p->translate.key.nr_elements; j++) { const struct translate_element *a = &p->translate.key.element[j]; + unsigned varient = p->element_to_buffer_varient[j]; /* Figure out source pointer address: */ - if (a->input_buffer != last_vb) { - last_vb = a->input_buffer; - vb = get_buffer_ptr(p, linear, a->input_buffer, elt); + if (varient != last_varient) { + last_varient = varient; + vb = get_buffer_ptr(p, linear, varient, elt); } if (!translate_attr( p, a, @@ -624,6 +695,7 @@ static void translate_sse_release( struct translate *translate ) static void PIPE_CDECL translate_sse_run_elts( struct translate *translate, const unsigned *elts, unsigned count, + unsigned instance_id, void *output_buffer ) { struct translate_sse *p = (struct translate_sse *)translate; @@ -631,12 +703,14 @@ static void PIPE_CDECL translate_sse_run_elts( struct translate *translate, p->gen_run_elts( translate, elts, count, - output_buffer ); + instance_id, + output_buffer); } static void PIPE_CDECL translate_sse_run( struct translate *translate, unsigned start, unsigned count, + unsigned instance_id, void *output_buffer ) { struct translate_sse *p = (struct translate_sse *)translate; @@ -644,7 +718,8 @@ static void PIPE_CDECL translate_sse_run( struct translate *translate, p->gen_run( translate, start, count, - output_buffer ); + instance_id, + output_buffer); } @@ -666,8 +741,37 @@ struct translate *translate_sse2_create( const struct translate_key *key ) p->translate.run_elts = translate_sse_run_elts; p->translate.run = translate_sse_run; - for (i = 0; i < key->nr_elements; i++) - p->nr_buffers = MAX2( p->nr_buffers, key->element[i].input_buffer + 1 ); + for (i = 0; i < key->nr_elements; i++) { + if (key->element[i].type == TRANSLATE_ELEMENT_NORMAL) { + unsigned j; + + p->nr_buffers = MAX2(p->nr_buffers, key->element[i].input_buffer + 1); + + if (key->element[i].instance_divisor) { + p->use_instancing = TRUE; + } + + /* + * Map vertex element to vertex buffer varient. + */ + for (j = 0; j < p->nr_buffer_varients; j++) { + if (p->buffer_varient[j].buffer_index == key->element[i].input_buffer && + p->buffer_varient[j].instance_divisor == key->element[i].instance_divisor) { + break; + } + } + if (j == p->nr_buffer_varients) { + p->buffer_varient[j].buffer_index = key->element[i].input_buffer; + p->buffer_varient[j].instance_divisor = key->element[i].instance_divisor; + p->nr_buffer_varients++; + } + p->element_to_buffer_varient[i] = j; + } else { + assert(key->element[i].type == TRANSLATE_ELEMENT_INSTANCE_ID); + + p->element_to_buffer_varient[i] = ELEMENT_BUFFER_INSTANCE_ID; + } + } if (0) debug_printf("nr_buffers: %d\n", p->nr_buffers); diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c index cef3b69e46d..249a0375fc5 100644 --- a/src/gallium/auxiliary/util/u_blitter.c +++ b/src/gallium/auxiliary/util/u_blitter.c @@ -163,6 +163,7 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe) rs_state.cull_mode = PIPE_WINDING_NONE; rs_state.bypass_vs_clip_and_viewport = 1; rs_state.gl_rasterization_rules = 1; + rs_state.flatshade = 1; ctx->rs_state = pipe->create_rasterizer_state(pipe, &rs_state); /* fragment shaders are created on-demand */ @@ -378,9 +379,16 @@ static void blitter_set_texcoords_cube(struct blitter_context_priv *ctx, float t1 = y1 / (float)surf->height; float s2 = x2 / (float)surf->width; float t2 = y2 / (float)surf->height; - const float st[4][2] = { - {s1, t1}, {s2, t1}, {s2, t2}, {s1, t2} - }; + float st[4][2]; + + st[0][0] = s1; + st[0][1] = t1; + st[1][0] = s2; + st[1][1] = t1; + st[2][0] = s2; + st[2][1] = t2; + st[3][0] = s1; + st[3][1] = t2; util_map_texcoords2d_onto_cubemap(surf->face, /* pointer, stride in floats */ @@ -567,45 +575,29 @@ void util_blitter_clear(struct blitter_context *blitter, blitter_restore_CSOs(ctx); } -void util_blitter_copy(struct blitter_context *blitter, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height, - boolean ignore_stencil) +static boolean +is_overlap(unsigned sx1, unsigned sx2, unsigned sy1, unsigned sy2, + unsigned dx1, unsigned dx2, unsigned dy1, unsigned dy2) +{ + if (sx1 >= dx2 || sx2 <= dx1 || sy1 >= dy2 || sy2 <= dy1) { + return FALSE; + } else { + return TRUE; + } +} + +static void util_blitter_do_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height, + boolean is_depth) { struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; struct pipe_context *pipe = ctx->pipe; - struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb_state; - boolean is_stencil, is_depth; - unsigned dst_tex_usage; - - /* give up if textures are not set */ - assert(dst->texture && src->texture); - if (!dst->texture || !src->texture) - return; - - is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; - is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0; - dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL : - PIPE_TEXTURE_USAGE_RENDER_TARGET; - /* check if we can sample from and render to the surfaces */ - /* (assuming copying a stencil buffer is not possible) */ - if ((!ignore_stencil && is_stencil) || - !screen->is_format_supported(screen, dst->format, dst->texture->target, - dst_tex_usage, 0) || - !screen->is_format_supported(screen, src->format, src->texture->target, - PIPE_TEXTURE_USAGE_SAMPLER, 0)) { - util_surface_copy(pipe, FALSE, dst, dstx, dsty, src, srcx, srcy, - width, height); - return; - } - - /* check whether the states are properly saved */ - blitter_check_saved_CSOs(ctx); assert(blitter->saved_fb_state.nr_cbufs != ~0); assert(blitter->saved_num_textures != ~0); assert(blitter->saved_num_sampler_states != ~0); @@ -663,6 +655,108 @@ void util_blitter_copy(struct blitter_context *blitter, blitter_set_rectangle(ctx, dstx, dsty, dstx+width, dsty+height, 0); blitter_draw_quad(ctx); + +} + +static void util_blitter_overlap_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + + struct pipe_texture texTemp; + struct pipe_texture *texture; + struct pipe_surface *tex_surf; + + /* check whether the states are properly saved */ + blitter_check_saved_CSOs(ctx); + + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = dst->texture->format; /* XXX verify supported by driver! */ + texTemp.last_level = 0; + texTemp.width0 = width; + texTemp.height0 = height; + texTemp.depth0 = 1; + + texture = screen->texture_create(screen, &texTemp); + if (!texture) + return; + + tex_surf = screen->get_tex_surface(screen, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* blit from the src to the temp */ + util_blitter_do_copy(blitter, tex_surf, 0, 0, + src, srcx, srcy, + width, height, + FALSE); + util_blitter_do_copy(blitter, dst, dstx, dsty, + tex_surf, 0, 0, + width, height, + FALSE); + pipe_surface_reference(&tex_surf, NULL); + pipe_texture_reference(&texture, NULL); + blitter_restore_CSOs(ctx); +} + +void util_blitter_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height, + boolean ignore_stencil) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + boolean is_stencil, is_depth; + unsigned dst_tex_usage; + + /* give up if textures are not set */ + assert(dst->texture && src->texture); + if (!dst->texture || !src->texture) + return; + + if (dst->texture == src->texture) { + if (is_overlap(srcx, srcx + width, srcy, srcy + height, + dstx, dstx + width, dsty, dsty + height)) { + util_blitter_overlap_copy(blitter, dst, dstx, dsty, src, srcx, srcy, + width, height); + return; + } + } + + is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; + is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0; + dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL : + PIPE_TEXTURE_USAGE_RENDER_TARGET; + + /* check if we can sample from and render to the surfaces */ + /* (assuming copying a stencil buffer is not possible) */ + if ((!ignore_stencil && is_stencil) || + !screen->is_format_supported(screen, dst->format, dst->texture->target, + dst_tex_usage, 0) || + !screen->is_format_supported(screen, src->format, src->texture->target, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { + util_surface_copy(pipe, FALSE, dst, dstx, dsty, src, srcx, srcy, + width, height); + return; + } + + /* check whether the states are properly saved */ + blitter_check_saved_CSOs(ctx); + util_blitter_do_copy(blitter, + dst, dstx, dsty, + src, srcx, srcy, + width, height, is_depth); blitter_restore_CSOs(ctx); } diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c index 4110485fb19..e2e23c3cdd8 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.c +++ b/src/gallium/auxiliary/util/u_draw_quad.c @@ -61,6 +61,7 @@ util_draw_vertex_buffer(struct pipe_context *pipe, /* tell pipe about the vertex attributes */ for (i = 0; i < num_attribs; i++) { velements[i].src_offset = i * 4 * sizeof(float); + velements[i].instance_divisor = 0; velements[i].vertex_buffer_index = 0; velements[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; velements[i].nr_components = 4; diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv index 9f16b42944e..01f7931aed1 100644 --- a/src/gallium/auxiliary/util/u_format.csv +++ b/src/gallium/auxiliary/util/u_format.csv @@ -62,10 +62,10 @@ PIPE_FORMAT_R16G16_SSCALED , array , 1, 1, s16 , s16 , , , xy01, PIPE_FORMAT_R16G16B16_SSCALED , array , 1, 1, s16 , s16 , s16 , , xyz1, rgb PIPE_FORMAT_R16G16B16A16_SSCALED , array , 1, 1, s16 , s16 , s16 , s16 , xyzw, rgb PIPE_FORMAT_R8_UNORM , array , 1, 1, un8 , , , , x001, rgb -PIPE_FORMAT_R8G8_UNORM , array , 1, 1, un8 , un8 , , , xy01, rgb -PIPE_FORMAT_R8G8B8_UNORM , array , 1, 1, un8 , un8 , un8 , , xyz1, rgb -PIPE_FORMAT_R8G8B8A8_UNORM , array , 1, 1, un8 , un8 , un8 , un8 , xyzw, rgb -PIPE_FORMAT_R8G8B8X8_UNORM , array , 1, 1, un8 , un8 , un8 , un8 , xyz1, rgb +PIPE_FORMAT_R8G8_UNORM , array , 1, 1, un8 , un8 , , , yx01, rgb +PIPE_FORMAT_R8G8B8_UNORM , array , 1, 1, un8 , un8 , un8 , , zyx1, rgb +PIPE_FORMAT_R8G8B8A8_UNORM , array , 1, 1, un8 , un8 , un8 , un8 , wzyx, rgb +PIPE_FORMAT_R8G8B8X8_UNORM , array , 1, 1, un8 , un8 , un8 , un8 , wzy1, rgb PIPE_FORMAT_R8_USCALED , array , 1, 1, u8 , , , , x001, rgb PIPE_FORMAT_R8G8_USCALED , array , 1, 1, u8 , u8 , , , xy01, rgb PIPE_FORMAT_R8G8B8_USCALED , array , 1, 1, u8 , u8 , u8 , , xyz1, rgb diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 43eb0153ee7..0ab53c75dd6 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -425,6 +425,8 @@ util_pack_z(enum pipe_format format, double z) if (z == 1.0) return 0xffffffff; return (uint) (z * 0xffffffff); + case PIPE_FORMAT_Z32_FLOAT: + return (uint)z; case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: if (z == 1.0) diff --git a/src/gallium/auxiliary/util/u_ringbuffer.c b/src/gallium/auxiliary/util/u_ringbuffer.c new file mode 100644 index 00000000000..3f43a19e018 --- /dev/null +++ b/src/gallium/auxiliary/util/u_ringbuffer.c @@ -0,0 +1,145 @@ + +#include "pipe/p_thread.h" +#include "pipe/p_defines.h" +#include "util/u_ringbuffer.h" +#include "util/u_math.h" +#include "util/u_memory.h" + +/* Generic ringbuffer: + */ +struct util_ringbuffer +{ + struct util_packet *buf; + unsigned mask; + + /* Can this be done with atomic variables?? + */ + unsigned head; + unsigned tail; + pipe_condvar change; + pipe_mutex mutex; +}; + + +struct util_ringbuffer *util_ringbuffer_create( unsigned dwords ) +{ + struct util_ringbuffer *ring = CALLOC_STRUCT(util_ringbuffer); + if (ring == NULL) + return NULL; + + assert(util_is_power_of_two(dwords)); + + ring->buf = MALLOC( dwords * sizeof(unsigned) ); + if (ring->buf == NULL) + goto fail; + + ring->mask = dwords - 1; + + pipe_condvar_init(ring->change); + pipe_mutex_init(ring->mutex); + return ring; + +fail: + FREE(ring->buf); + FREE(ring); + return NULL; +} + +void util_ringbuffer_destroy( struct util_ringbuffer *ring ) +{ + pipe_condvar_destroy(ring->change); + pipe_mutex_destroy(ring->mutex); + FREE(ring->buf); + FREE(ring); +} + +static INLINE unsigned util_ringbuffer_space( const struct util_ringbuffer *ring ) +{ + return (ring->tail - (ring->head + 1)) & ring->mask; +} + +void util_ringbuffer_enqueue( struct util_ringbuffer *ring, + const struct util_packet *packet ) +{ + unsigned i; + + /* XXX: over-reliance on mutexes, etc: + */ + pipe_mutex_lock(ring->mutex); + + /* Wait for free space: + */ + while (util_ringbuffer_space(ring) < packet->dwords) + pipe_condvar_wait(ring->change, ring->mutex); + + /* Copy data to ring: + */ + for (i = 0; i < packet->dwords; i++) { + + /* Copy all dwords of the packet. Note we're abusing the + * typesystem a little - we're being passed a pointer to + * something, but probably not an array of packet structs: + */ + ring->buf[ring->head] = packet[i]; + ring->head++; + ring->head &= ring->mask; + } + + /* Signal change: + */ + pipe_condvar_signal(ring->change); + pipe_mutex_unlock(ring->mutex); +} + +enum pipe_error util_ringbuffer_dequeue( struct util_ringbuffer *ring, + struct util_packet *packet, + unsigned max_dwords, + boolean wait ) +{ + const struct util_packet *ring_packet; + unsigned i; + int ret = PIPE_OK; + + /* XXX: over-reliance on mutexes, etc: + */ + pipe_mutex_lock(ring->mutex); + + /* Wait for free space: + */ + if (wait) { + while (util_ringbuffer_space(ring) == 0) + pipe_condvar_wait(ring->change, ring->mutex); + } + else { + if (util_ringbuffer_space(ring) == 0) { + ret = PIPE_ERROR_OUT_OF_MEMORY; + goto out; + } + } + + ring_packet = &ring->buf[ring->tail]; + + /* Both of these are considered bugs. Raise an assert on debug builds. + */ + if (ring_packet->dwords > ring->mask + 1 - util_ringbuffer_space(ring) || + ring_packet->dwords > max_dwords) { + assert(0); + ret = PIPE_ERROR_BAD_INPUT; + goto out; + } + + /* Copy data from ring: + */ + for (i = 0; i < ring_packet->dwords; i++) { + packet[i] = ring->buf[ring->tail]; + ring->tail++; + ring->tail &= ring->mask; + } + +out: + /* Signal change: + */ + pipe_condvar_signal(ring->change); + pipe_mutex_unlock(ring->mutex); + return ret; +} diff --git a/src/gallium/auxiliary/util/u_ringbuffer.h b/src/gallium/auxiliary/util/u_ringbuffer.h new file mode 100644 index 00000000000..85f0ad6c1f6 --- /dev/null +++ b/src/gallium/auxiliary/util/u_ringbuffer.h @@ -0,0 +1,29 @@ + +#ifndef UTIL_RINGBUFFER_H +#define UTIL_RINGBUFFER_H + +#include "pipe/p_compiler.h" +#include "pipe/p_defines.h" /* only for pipe_error! */ + +/* Generic header + */ +struct util_packet { + unsigned dwords:8; + unsigned data24:24; +}; + +struct util_ringbuffer; + +struct util_ringbuffer *util_ringbuffer_create( unsigned dwords ); + +void util_ringbuffer_destroy( struct util_ringbuffer *ring ); + +void util_ringbuffer_enqueue( struct util_ringbuffer *ring, + const struct util_packet *packet ); + +enum pipe_error util_ringbuffer_dequeue( struct util_ringbuffer *ring, + struct util_packet *packet, + unsigned max_dwords, + boolean wait ); + +#endif diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index 1ba82bb21f0..f9936eb1cb2 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -1357,7 +1357,10 @@ pipe_put_tile_rgba(struct pipe_transfer *pt, /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; default: - debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(format)); + util_format_write_4f(format, + p, src_stride * sizeof(float), + packed, util_format_get_stride(format, w), + 0, 0, w, h); } pipe_put_tile_raw(pt, x, y, w, h, packed, 0); diff --git a/src/gallium/auxiliary/vl/vl_compositor.c b/src/gallium/auxiliary/vl/vl_compositor.c index fc2a1c59a6b..a524e2fdfb3 100644 --- a/src/gallium/auxiliary/vl/vl_compositor.c +++ b/src/gallium/auxiliary/vl/vl_compositor.c @@ -316,6 +316,7 @@ init_buffers(struct vl_compositor *c) pipe_buffer_unmap(c->pipe->screen, c->vertex_bufs[0].buffer); c->vertex_elems[0].src_offset = 0; + c->vertex_elems[0].instance_divisor = 0; c->vertex_elems[0].vertex_buffer_index = 0; c->vertex_elems[0].nr_components = 2; c->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT; @@ -345,6 +346,7 @@ init_buffers(struct vl_compositor *c) pipe_buffer_unmap(c->pipe->screen, c->vertex_bufs[1].buffer); c->vertex_elems[1].src_offset = 0; + c->vertex_elems[1].instance_divisor = 0; c->vertex_elems[1].vertex_buffer_index = 1; c->vertex_elems[1].nr_components = 2; c->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT; @@ -353,7 +355,7 @@ init_buffers(struct vl_compositor *c) * Create our vertex shader's constant buffer * Const buffer contains scaling and translation vectors */ - c->vs_const_buf.buffer = pipe_buffer_create + c->vs_const_buf = pipe_buffer_create ( c->pipe->screen, 1, @@ -365,7 +367,7 @@ init_buffers(struct vl_compositor *c) * Create our fragment shader's constant buffer * Const buffer contains the color conversion matrix and bias vectors */ - c->fs_const_buf.buffer = pipe_buffer_create + c->fs_const_buf = pipe_buffer_create ( c->pipe->screen, 1, @@ -390,8 +392,8 @@ cleanup_buffers(struct vl_compositor *c) for (i = 0; i < 2; ++i) pipe_buffer_reference(&c->vertex_bufs[i].buffer, NULL); - pipe_buffer_reference(&c->vs_const_buf.buffer, NULL); - pipe_buffer_reference(&c->fs_const_buf.buffer, NULL); + pipe_buffer_reference(&c->vs_const_buf, NULL); + pipe_buffer_reference(&c->fs_const_buf, NULL); } bool vl_compositor_init(struct vl_compositor *compositor, struct pipe_context *pipe) @@ -483,13 +485,13 @@ void vl_compositor_render(struct vl_compositor *compositor, compositor->pipe->bind_fs_state(compositor->pipe, compositor->fragment_shader); compositor->pipe->set_vertex_buffers(compositor->pipe, 2, compositor->vertex_bufs); compositor->pipe->set_vertex_elements(compositor->pipe, 2, compositor->vertex_elems); - compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_VERTEX, 0, &compositor->vs_const_buf); - compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_FRAGMENT, 0, &compositor->fs_const_buf); + compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_VERTEX, 0, compositor->vs_const_buf); + compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_FRAGMENT, 0, compositor->fs_const_buf); vs_consts = pipe_buffer_map ( compositor->pipe->screen, - compositor->vs_const_buf.buffer, + compositor->vs_const_buf, PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD ); @@ -511,7 +513,7 @@ void vl_compositor_render(struct vl_compositor *compositor, vs_consts->src_trans.z = 0; vs_consts->src_trans.w = 0; - pipe_buffer_unmap(compositor->pipe->screen, compositor->vs_const_buf.buffer); + pipe_buffer_unmap(compositor->pipe->screen, compositor->vs_const_buf); compositor->pipe->draw_arrays(compositor->pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4); compositor->pipe->flush(compositor->pipe, PIPE_FLUSH_RENDER_CACHE, fence); @@ -525,10 +527,10 @@ void vl_compositor_set_csc_matrix(struct vl_compositor *compositor, const float memcpy ( - pipe_buffer_map(compositor->pipe->screen, compositor->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + pipe_buffer_map(compositor->pipe->screen, compositor->fs_const_buf, PIPE_BUFFER_USAGE_CPU_WRITE), mat, sizeof(struct fragment_shader_consts) ); - pipe_buffer_unmap(compositor->pipe->screen, compositor->fs_const_buf.buffer); + pipe_buffer_unmap(compositor->pipe->screen, compositor->fs_const_buf); } diff --git a/src/gallium/auxiliary/vl/vl_compositor.h b/src/gallium/auxiliary/vl/vl_compositor.h index f441901a751..6a9a3fd7af1 100644 --- a/src/gallium/auxiliary/vl/vl_compositor.h +++ b/src/gallium/auxiliary/vl/vl_compositor.h @@ -47,7 +47,7 @@ struct vl_compositor struct pipe_scissor_state scissor; struct pipe_vertex_buffer vertex_bufs[2]; struct pipe_vertex_element vertex_elems[2]; - struct pipe_constant_buffer vs_const_buf, fs_const_buf; + struct pipe_buffer *vs_const_buf, *fs_const_buf; }; bool vl_compositor_init(struct vl_compositor *compositor, struct pipe_context *pipe); diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c index caf581aca60..e43187545c5 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c @@ -891,53 +891,61 @@ init_buffers(struct vl_mpeg12_mc_renderer *r) /* Position element */ r->vertex_elems[0].src_offset = 0; + r->vertex_elems[0].instance_divisor = 0; r->vertex_elems[0].vertex_buffer_index = 0; r->vertex_elems[0].nr_components = 2; r->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT; /* Luma, texcoord element */ r->vertex_elems[1].src_offset = sizeof(struct vertex2f); + r->vertex_elems[1].instance_divisor = 0; r->vertex_elems[1].vertex_buffer_index = 0; r->vertex_elems[1].nr_components = 2; r->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT; /* Chroma Cr texcoord element */ r->vertex_elems[2].src_offset = sizeof(struct vertex2f) * 2; + r->vertex_elems[2].instance_divisor = 0; r->vertex_elems[2].vertex_buffer_index = 0; r->vertex_elems[2].nr_components = 2; r->vertex_elems[2].src_format = PIPE_FORMAT_R32G32_FLOAT; /* Chroma Cb texcoord element */ r->vertex_elems[3].src_offset = sizeof(struct vertex2f) * 3; + r->vertex_elems[3].instance_divisor = 0; r->vertex_elems[3].vertex_buffer_index = 0; r->vertex_elems[3].nr_components = 2; r->vertex_elems[3].src_format = PIPE_FORMAT_R32G32_FLOAT; /* First ref surface top field texcoord element */ r->vertex_elems[4].src_offset = 0; + r->vertex_elems[4].instance_divisor = 0; r->vertex_elems[4].vertex_buffer_index = 1; r->vertex_elems[4].nr_components = 2; r->vertex_elems[4].src_format = PIPE_FORMAT_R32G32_FLOAT; /* First ref surface bottom field texcoord element */ r->vertex_elems[5].src_offset = sizeof(struct vertex2f); + r->vertex_elems[5].instance_divisor = 0; r->vertex_elems[5].vertex_buffer_index = 1; r->vertex_elems[5].nr_components = 2; r->vertex_elems[5].src_format = PIPE_FORMAT_R32G32_FLOAT; /* Second ref surface top field texcoord element */ r->vertex_elems[6].src_offset = 0; + r->vertex_elems[6].instance_divisor = 0; r->vertex_elems[6].vertex_buffer_index = 2; r->vertex_elems[6].nr_components = 2; r->vertex_elems[6].src_format = PIPE_FORMAT_R32G32_FLOAT; /* Second ref surface bottom field texcoord element */ r->vertex_elems[7].src_offset = sizeof(struct vertex2f); + r->vertex_elems[7].instance_divisor = 0; r->vertex_elems[7].vertex_buffer_index = 2; r->vertex_elems[7].nr_components = 2; r->vertex_elems[7].src_format = PIPE_FORMAT_R32G32_FLOAT; - r->vs_const_buf.buffer = pipe_buffer_create + r->vs_const_buf = pipe_buffer_create ( r->pipe->screen, DEFAULT_BUF_ALIGNMENT, @@ -945,7 +953,7 @@ init_buffers(struct vl_mpeg12_mc_renderer *r) sizeof(struct vertex_shader_consts) ); - r->fs_const_buf.buffer = pipe_buffer_create + r->fs_const_buf = pipe_buffer_create ( r->pipe->screen, DEFAULT_BUF_ALIGNMENT, @@ -954,11 +962,11 @@ init_buffers(struct vl_mpeg12_mc_renderer *r) memcpy ( - pipe_buffer_map(r->pipe->screen, r->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + pipe_buffer_map(r->pipe->screen, r->fs_const_buf, PIPE_BUFFER_USAGE_CPU_WRITE), &fs_consts, sizeof(struct fragment_shader_consts) ); - pipe_buffer_unmap(r->pipe->screen, r->fs_const_buf.buffer); + pipe_buffer_unmap(r->pipe->screen, r->fs_const_buf); return true; } @@ -970,8 +978,8 @@ cleanup_buffers(struct vl_mpeg12_mc_renderer *r) assert(r); - pipe_buffer_reference(&r->vs_const_buf.buffer, NULL); - pipe_buffer_reference(&r->fs_const_buf.buffer, NULL); + pipe_buffer_reference(&r->vs_const_buf, NULL); + pipe_buffer_reference(&r->fs_const_buf, NULL); for (i = 0; i < 3; ++i) pipe_buffer_reference(&r->vertex_bufs.all[i].buffer, NULL); @@ -1284,19 +1292,19 @@ flush(struct vl_mpeg12_mc_renderer *r) vs_consts = pipe_buffer_map ( - r->pipe->screen, r->vs_const_buf.buffer, + r->pipe->screen, r->vs_const_buf, PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD ); vs_consts->denorm.x = r->surface->width0; vs_consts->denorm.y = r->surface->height0; - pipe_buffer_unmap(r->pipe->screen, r->vs_const_buf.buffer); + pipe_buffer_unmap(r->pipe->screen, r->vs_const_buf); r->pipe->set_constant_buffer(r->pipe, PIPE_SHADER_VERTEX, 0, - &r->vs_const_buf); + r->vs_const_buf); r->pipe->set_constant_buffer(r->pipe, PIPE_SHADER_FRAGMENT, 0, - &r->fs_const_buf); + r->fs_const_buf); if (num_macroblocks[MACROBLOCK_TYPE_INTRA] > 0) { r->pipe->set_vertex_buffers(r->pipe, 1, r->vertex_bufs.all); diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h index 64184337a06..f00b8c7b8b1 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h @@ -63,8 +63,8 @@ struct vl_mpeg12_mc_renderer struct pipe_viewport_state viewport; struct pipe_scissor_state scissor; - struct pipe_constant_buffer vs_const_buf; - struct pipe_constant_buffer fs_const_buf; + struct pipe_buffer *vs_const_buf; + struct pipe_buffer *fs_const_buf; struct pipe_framebuffer_state fb_state; struct pipe_vertex_element vertex_elems[8]; |