diff options
Diffstat (limited to 'src/gallium/drivers')
48 files changed, 2490 insertions, 2140 deletions
diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 7f656a9744..8df41c1d4c 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -104,7 +104,9 @@ struct cell_context uint num_textures; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + uint num_vertex_buffers; struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + uint num_vertex_elements; ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; ubyte *zsbuf_map; diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index b896252f81..6e08cf6fe8 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -101,17 +101,6 @@ cell_draw_elements(struct pipe_context *pipe, struct draw_context *draw = sp->draw; unsigned i; - /* first, check that the primitive is not malformed. It is the - * state tracker's responsibility to do send only correctly formed - * primitives down. It currently isn't doing that though... - */ -#if 1 - count = draw_trim_prim( mode, count ); -#else - if (!draw_validate_prim( mode, count )) - assert(0); -#endif - if (sp->dirty) cell_update_derived( sp ); @@ -123,14 +112,12 @@ cell_draw_elements(struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - void *buf = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size); - draw_set_mapped_vertex_buffer(draw, i, buf); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + void *buf = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size); + draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { @@ -151,11 +138,9 @@ cell_draw_elements(struct pipe_context *pipe, /* * unmap vertex/index buffers - will cause draw module to flush */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - draw_set_mapped_vertex_buffer(draw, i, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } if (indexBuffer) { draw_set_mapped_element_buffer(draw, 0, NULL); diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index fb2e940348..c3a3fbd066 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -30,33 +30,50 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "draw/draw_context.h" -#if 0 -#include "pipe/p_shader_tokens.h" -#include "gallivm/gallivm.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/exec/tgsi_sse2.h" -#endif +#include "tgsi/util/tgsi_parse.h" #include "cell_context.h" #include "cell_state.h" + +/** cast wrapper */ +static INLINE struct cell_fragment_shader_state * +cell_fragment_shader_state(void *shader) +{ + return (struct pipe_shader_state *) shader; +} + + +/** cast wrapper */ +static INLINE struct cell_vertex_shader_state * +cell_vertex_shader_state(void *shader) +{ + return (struct pipe_shader_state *) shader; +} + + + static void * cell_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { /*struct cell_context *cell = cell_context(pipe);*/ - struct cell_fragment_shader_state *state; + struct cell_fragment_shader_state *cfs; - state = CALLOC_STRUCT(cell_fragment_shader_state); - if (!state) + cfs = CALLOC_STRUCT(cell_fragment_shader_state); + if (!cfs) return NULL; - state->shader = *templ; + cfs->shader.tokens = tgsi_dup_tokens(templ->tokens); + if (!cfs->shader.tokens) { + FREE(cfs); + return NULL; + } - tgsi_scan_shader(templ->tokens, &state->info); + tgsi_scan_shader(templ->tokens, &cfs->info); - return state; + return cfs; } @@ -65,7 +82,7 @@ cell_bind_fs_state(struct pipe_context *pipe, void *fs) { struct cell_context *cell = cell_context(pipe); - cell->fs = (struct cell_fragment_shader_state *) fs; + cell->fs = cell_fragment_shader_state(fs); cell->dirty |= CELL_NEW_FS; } @@ -74,10 +91,10 @@ cell_bind_fs_state(struct pipe_context *pipe, void *fs) static void cell_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct cell_fragment_shader_state *state = - (struct cell_fragment_shader_state *) fs; + struct cell_fragment_shader_state *cfs = cell_fragment_shader_state(fs); - FREE( state ); + FREE((void *) cfs->shader.tokens); + FREE(cfs); } @@ -86,22 +103,28 @@ cell_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { struct cell_context *cell = cell_context(pipe); - struct cell_vertex_shader_state *state; + struct cell_vertex_shader_state *cvs; - state = CALLOC_STRUCT(cell_vertex_shader_state); - if (!state) + cvs = CALLOC_STRUCT(cell_vertex_shader_state); + if (!cvs) return NULL; - state->shader = *templ; - tgsi_scan_shader(templ->tokens, &state->info); + cvs->shader.tokens = tgsi_dup_tokens(templ->tokens); + if (!cvs->shader.tokens) { + FREE(cvs); + return NULL; + } - state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); - if (state->draw_data == NULL) { - FREE( state ); + tgsi_scan_shader(templ->tokens, &cvs->info); + + cvs->draw_data = draw_create_vertex_shader(cell->draw, &cvs->shader); + if (cvs->draw_data == NULL) { + FREE( (void *) cvs->shader.tokens ); + FREE( cvs ); return NULL; } - return state; + return cvs; } @@ -110,7 +133,7 @@ cell_bind_vs_state(struct pipe_context *pipe, void *vs) { struct cell_context *cell = cell_context(pipe); - cell->vs = (const struct cell_vertex_shader_state *) vs; + cell->vs = cell_vertex_shader_state(vs); draw_bind_vertex_shader(cell->draw, (cell->vs ? cell->vs->draw_data : NULL)); @@ -123,12 +146,11 @@ static void cell_delete_vs_state(struct pipe_context *pipe, void *vs) { struct cell_context *cell = cell_context(pipe); + struct cell_vertex_shader_state *cvs = cell_vertex_shader_state(vs); - struct cell_vertex_shader_state *state = - (struct cell_vertex_shader_state *) vs; - - draw_delete_vertex_shader(cell->draw, state->draw_data); - FREE( state ); + draw_delete_vertex_shader(cell->draw, cvs->draw_data); + FREE( (void *) cvs->shader.tokens ); + FREE( cvs ); } diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c index 6c83b8dc72..114684c2a3 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c +++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c @@ -45,6 +45,7 @@ cell_set_vertex_elements(struct pipe_context *pipe, assert(count <= PIPE_MAX_ATTRIBS); memcpy(cell->vertex_element, elements, count * sizeof(elements[0])); + cell->num_vertex_elements = count; cell->dirty |= CELL_NEW_VERTEX; @@ -62,6 +63,7 @@ cell_set_vertex_buffers(struct pipe_context *pipe, assert(count <= PIPE_MAX_ATTRIBS); memcpy(cell->vertex_buffer, buffers, count * sizeof(buffers[0])); + cell->num_vertex_buffers = count; cell->dirty |= CELL_NEW_VERTEX; diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index fee33d82de..4bef21619c 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -65,14 +65,12 @@ i915_draw_elements( struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (i915->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - i915->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } + for (i = 0; i < i915->num_vertex_buffers; i++) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + i915->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { @@ -96,11 +94,9 @@ i915_draw_elements( struct pipe_context *pipe, /* * unmap vertex/index buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (i915->vertex_buffer[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); - draw_set_mapped_vertex_buffer(draw, i, NULL); - } + for (i = 0; i < i915->num_vertex_buffers; i++) { + pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); } if (indexBuffer) { pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); @@ -146,7 +142,7 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, */ i915->draw = draw_create(); assert(i915->draw); - if (GETENV("I915_VBUF")) { + if (!GETENV("I915_NO_VBUF")) { draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); } else { diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 8e707ea574..38e6a34884 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -238,6 +238,8 @@ struct i915_context unsigned num_samplers;
unsigned num_textures;
+ unsigned num_vertex_elements;
+ unsigned num_vertex_buffers;
unsigned *batch_start;
diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c index 37a3508fe1..c024a051a5 100644 --- a/src/gallium/drivers/i915simple/i915_debug_fp.c +++ b/src/gallium/drivers/i915simple/i915_debug_fp.c @@ -333,12 +333,11 @@ void i915_disassemble_program(struct debug_stream *stream, const unsigned * program, unsigned sz) { - unsigned size = program[0] & 0x1ff; unsigned i; PRINTF(stream, "\t\tBEGIN\n"); - assert(size + 2 == sz); + assert((program[0] & 0x1ff) + 2 == sz); program++; for (i = 1; i < sz; i += 3, program += 3) { diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 7b4fca5db1..3ccf74c72c 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -33,6 +33,7 @@ #include "i915_fpc.h" #include "pipe/p_shader_tokens.h" +#include "util/u_string.h" #include "tgsi/util/tgsi_parse.h" #include "tgsi/util/tgsi_dump.h" @@ -122,7 +123,7 @@ i915_program_error(struct i915_fp_compile *p, const char *msg, ...) debug_printf("i915_program_error: "); va_start( args, msg ); - vsprintf( buffer, msg, args ); + util_vsnprintf( buffer, sizeof(buffer), msg, args ); va_end( args ); debug_printf(buffer); debug_printf("\n"); diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index b6fb0a6d88..9ffa460138 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "draw/draw_private.h" +#include "draw/draw_pipe.h" #include "pipe/p_util.h" #include "i915_context.h" @@ -78,9 +78,6 @@ emit_hw_vertex( struct i915_context *i915, const uint j = vinfo->src_index[i]; const float *attrib = vertex->data[j]; switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; case EMIT_1F: OUT_BATCH( fui(attrib[0]) ); count++; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 839b98c0ce..646cfd921d 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -28,6 +28,7 @@ #include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_string.h" #include "i915_reg.h" #include "i915_context.h" @@ -78,7 +79,7 @@ i915_get_name( struct pipe_screen *pscreen ) break; } - sprintf(buffer, "i915 (chipset: %s)", chipset); + util_snprintf(buffer, sizeof(buffer), "i915 (chipset: %s)", chipset); return buffer; } @@ -153,10 +154,10 @@ i915_is_format_supported( struct pipe_screen *screen, PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_R5G6B5_UNORM, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_A8_L8, + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_A8L8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV, PIPE_FORMAT_S8Z24_UNORM, @@ -181,6 +182,7 @@ i915_is_format_supported( struct pipe_screen *screen, break; default: assert(0); + return FALSE; } for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 4404bc4590..4adeb37e86 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -33,6 +33,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "tgsi/util/tgsi_parse.h" #include "i915_context.h" #include "i915_reg.h" @@ -436,7 +437,7 @@ i915_create_fs_state(struct pipe_context *pipe, if (!ifs) return NULL; - ifs->state = *templ; + ifs->state.tokens = tgsi_dup_tokens(templ->tokens); tgsi_scan_shader(templ->tokens, &ifs->info); @@ -465,6 +466,8 @@ void i915_delete_fs_state(struct pipe_context *pipe, void *shader) FREE(ifs->program); ifs->program_len = 0; + FREE(ifs->state.tokens); + FREE(ifs); } @@ -549,6 +552,9 @@ static void i915_set_sampler_textures(struct pipe_context *pipe, !memcmp(i915->texture, texture, num * sizeof(struct pipe_texture *))) return; + /* Fixes wrong texture in texobj with VBUF */ + draw_flush(i915->draw); + for (i = 0; i < num; i++) pipe_texture_reference((struct pipe_texture **) &i915->texture[i], texture[i]); @@ -694,6 +700,8 @@ static void i915_set_vertex_buffers(struct pipe_context *pipe, struct i915_context *i915 = i915_context(pipe); memcpy(i915->vertex_buffer, buffers, count * sizeof(buffers[0])); + i915->num_vertex_buffers = count; + /* pass-through to draw module */ draw_set_vertex_buffers(i915->draw, count, buffers); } @@ -703,6 +711,7 @@ static void i915_set_vertex_elements(struct pipe_context *pipe, const struct pipe_vertex_element *elements) { struct i915_context *i915 = i915_context(pipe); + i915->num_vertex_elements = count; /* pass-through to draw module */ draw_set_vertex_elements(i915->draw, count, elements); } diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 84f6529a3a..982eec4a1b 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -131,13 +131,13 @@ static uint translate_texture_format(enum pipe_format pipeFormat) { switch (pipeFormat) { - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: return MAPSURF_8BIT | MT_8BIT_L8; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: return MAPSURF_8BIT | MT_8BIT_I8; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: return MAPSURF_8BIT | MT_8BIT_A8; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: return MAPSURF_16BIT | MT_16BIT_AY88; case PIPE_FORMAT_R5G6B5_UNORM: return MAPSURF_16BIT | MT_16BIT_RGB565; diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index 5be369fe52..b700f7e4f5 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -28,6 +28,7 @@ #include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_string.h" #include "brw_context.h" #include "brw_screen.h" @@ -66,7 +67,7 @@ brw_get_name( struct pipe_screen *screen ) break; } - sprintf(buffer, "i965 (chipset: %s)", chipset); + util_snprintf(buffer, sizeof(buffer), "i965 (chipset: %s)", chipset); return buffer; } @@ -140,13 +141,13 @@ brw_is_format_supported( struct pipe_screen *screen, #if 0 /* XXX: This is broken -- rewrite if still needed. */ static const unsigned tex_supported[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_L8A8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV, PIPE_FORMAT_S8_Z24, @@ -156,16 +157,16 @@ brw_is_format_supported( struct pipe_screen *screen, /* Actually a lot more than this - add later: */ static const unsigned render_supported[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, }; /* */ static const unsigned z_stencil_supported[] = { - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_S8_Z24, + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_S8Z24_UNORM, }; switch (type) { diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 376f1487b2..ac243b7e4f 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -35,6 +35,7 @@ #include "pipe/p_inlines.h"
#include "pipe/p_shader_tokens.h"
#include "tgsi/util/tgsi_dump.h"
+#include "tgsi/util/tgsi_parse.h"
#include "brw_context.h"
#include "brw_defines.h"
@@ -182,9 +183,7 @@ static void * brw_create_fs_state(struct pipe_context *pipe, {
struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program);
- /* XXX: Do I have to duplicate the tokens as well??
- */
- brw_fp->program = *shader;
+ brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens);
brw_fp->id = brw_context(pipe)->program_id++;
tgsi_scan_shader(shader->tokens, &brw_fp->info);
@@ -210,7 +209,10 @@ static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) static void brw_delete_fs_state(struct pipe_context *pipe, void *shader)
{
- FREE(shader);
+ struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader;
+
+ FREE((void *) brw_fp->program.tokens);
+ FREE(brw_fp);
}
@@ -223,9 +225,7 @@ static void *brw_create_vs_state(struct pipe_context *pipe, {
struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program);
- /* XXX: Do I have to duplicate the tokens as well??
- */
- brw_vp->program = *shader;
+ brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens);
brw_vp->id = brw_context(pipe)->program_id++;
tgsi_scan_shader(shader->tokens, &brw_vp->info);
@@ -251,7 +251,10 @@ static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) static void brw_delete_vs_state(struct pipe_context *pipe, void *shader)
{
- FREE(shader);
+ struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader;
+
+ FREE((void *) brw_vp->program.tokens);
+ FREE(brw_vp);
}
diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c index 853c743ccf..69e56dc8bd 100644 --- a/src/gallium/drivers/i965simple/brw_wm_surface_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c @@ -57,16 +57,16 @@ static unsigned translate_tex_target( enum pipe_texture_target target ) static unsigned translate_tex_format( enum pipe_format pipe_format ) { switch( pipe_format ) { - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: return BRW_SURFACEFORMAT_L8_UNORM; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: return BRW_SURFACEFORMAT_I8_UNORM; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: return BRW_SURFACEFORMAT_A8_UNORM; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: return BRW_SURFACEFORMAT_L8A8_UNORM; case PIPE_FORMAT_R8G8B8_UNORM: diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index f32db35d58..120bdfd9dd 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -17,7 +17,6 @@ C_SOURCES = \ sp_quad.c \ sp_quad_alpha_test.c \ sp_quad_blend.c \ - sp_quad_bufloop.c \ sp_quad_colormask.c \ sp_quad_coverage.c \ sp_quad_depth_test.c \ @@ -28,6 +27,7 @@ C_SOURCES = \ sp_quad_stencil.c \ sp_quad_stipple.c \ sp_screen.c \ + sp_setup.c \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index 88c21ee3b0..c1f7daa8ab 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -14,9 +14,9 @@ softpipe = env.ConvenienceLibrary( 'sp_flush.c', 'sp_prim_setup.c', 'sp_prim_vbuf.c', + 'sp_setup.c', 'sp_quad_alpha_test.c', 'sp_quad_blend.c', - 'sp_quad_bufloop.c', 'sp_quad.c', 'sp_quad_colormask.c', 'sp_quad_coverage.c', diff --git a/src/gallium/drivers/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c index 39aed151c7..1236706891 100644 --- a/src/gallium/drivers/softpipe/sp_clear.c +++ b/src/gallium/drivers/softpipe/sp_clear.c @@ -49,6 +49,9 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, struct softpipe_context *softpipe = softpipe_context(pipe); uint i; + if (softpipe->no_rast) + return; + #if 0 softpipe_update_derived(softpipe); /* not needed?? */ #endif diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index e298ed37c3..fe9cd8375e 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -98,7 +98,6 @@ static void softpipe_destroy( struct pipe_context *pipe ) softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); softpipe->quad.coverage->destroy( softpipe->quad.coverage ); - softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); softpipe->quad.blend->destroy( softpipe->quad.blend ); softpipe->quad.colormask->destroy( softpipe->quad.colormask ); softpipe->quad.output->destroy( softpipe->quad.output ); @@ -123,12 +122,12 @@ static void softpipe_destroy( struct pipe_context *pipe ) struct pipe_context * softpipe_create( struct pipe_screen *screen, struct pipe_winsys *pipe_winsys, - struct softpipe_winsys *softpipe_winsys ) + void *unused ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); uint i; -#if defined(__i386__) || defined(__386__) +#ifdef PIPE_ARCH_X86 softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; #else softpipe->use_sse = FALSE; @@ -179,6 +178,8 @@ softpipe_create( struct pipe_screen *screen, softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; + softpipe->pipe.set_edgeflags = softpipe_set_edgeflags; + softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; @@ -207,13 +208,10 @@ softpipe_create( struct pipe_screen *screen, softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); - softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); - softpipe->winsys = softpipe_winsys; - /* * Create drawing context and plug our rendering stage into it. */ @@ -221,6 +219,9 @@ softpipe_create( struct pipe_screen *screen, assert(softpipe->draw); softpipe->setup = sp_draw_render_stage(softpipe); + if (GETENV( "SP_NO_RAST" ) != NULL) + softpipe->no_rast = TRUE; + if (GETENV( "SP_VBUF" ) != NULL) { sp_init_vbuf(softpipe); } diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index dc9d0e6d5d..62eabfb30e 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -57,8 +57,6 @@ struct sp_vertex_shader; struct softpipe_context { struct pipe_context pipe; /**< base class */ - struct softpipe_winsys *winsys; /**< window system interface */ - /* The most recent drawing state as set by the driver: */ @@ -83,6 +81,10 @@ struct softpipe_context { unsigned num_samplers; unsigned num_textures; + unsigned num_vertex_elements; + unsigned num_vertex_buffers; + + boolean no_rast; /* Counter for occlusion queries. Note this supports overlapping * queries. @@ -103,6 +105,8 @@ struct softpipe_context { int psize_slot; + unsigned reduced_api_prim; /**< PIPE_PRIM_POINTS, _LINES or _TRIANGLES */ + #if 0 /* Stipple derived state: */ @@ -124,7 +128,6 @@ struct softpipe_context { struct quad_stage *depth_test; struct quad_stage *occlusion; struct quad_stage *coverage; - struct quad_stage *bufloop; struct quad_stage *blend; struct quad_stage *colormask; struct quad_stage *output; @@ -138,8 +141,6 @@ struct softpipe_context { struct draw_stage *vbuf; struct softpipe_vbuf_render *vbuf_render; - uint current_cbuf; /**< current color buffer being written to */ - struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; struct softpipe_tile_cache *zsbuf_cache; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index ab54050d3f..6c58f9909d 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -78,6 +78,20 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) } +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) @@ -104,17 +118,7 @@ softpipe_draw_elements(struct pipe_context *pipe, struct draw_context *draw = sp->draw; unsigned i; - /* first, check that the primitive is not malformed. It is the - * state tracker's responsibility to do send only correctly formed - * primitives down. It currently isn't doing that though... - */ -#if 1 - count = draw_trim_prim( mode, count ); -#else - if (!draw_validate_prim( mode, count )) - assert(0); -#endif - + sp->reduced_api_prim = reduced_prim[mode]; if (sp->dirty) softpipe_update_derived( sp ); @@ -125,14 +129,12 @@ softpipe_draw_elements(struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { @@ -153,11 +155,9 @@ softpipe_draw_elements(struct pipe_context *pipe, /* * unmap vertex/index buffers - will cause draw module to flush */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - draw_set_mapped_vertex_buffer(draw, i, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } if (indexBuffer) { draw_set_mapped_element_buffer(draw, 0, NULL); @@ -170,3 +170,12 @@ softpipe_draw_elements(struct pipe_context *pipe, return TRUE; } + + +void +softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags) +{ + struct softpipe_context *sp = softpipe_context(pipe); + draw_set_edgeflags(sp->draw, edgeflags); +} + diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index d5bd7a702f..0b199a2193 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -37,6 +37,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_parse.h" struct sp_exec_fragment_shader { struct sp_fragment_shader base; @@ -116,6 +117,7 @@ exec_run( const struct sp_fragment_shader *base, static void exec_delete( struct sp_fragment_shader *base ) { + FREE((void *) base->shader.tokens); FREE(base); } @@ -137,7 +139,8 @@ softpipe_create_fs_exec(struct softpipe_context *softpipe, if (!shader) return NULL; - shader->base.shader = *templ; + /* we need to keep a local copy of the tokens */ + shader->base.shader.tokens = tgsi_dup_tokens(templ->tokens); shader->base.prepare = exec_prepare; shader->base.run = exec_run; shader->base.delete = exec_delete; diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 07d058155f..6e1d9280bb 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -99,19 +99,19 @@ shade_quad_llvm(struct quad_stage *qs, allvmrt(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); for (i = 0; i < QUAD_SIZE; ++i) { - quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; - quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; - quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; - quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; + quad->outputs.color[0][0][i] = dests[i][qss->colorOutSlot][0]; + quad->outputs.color[0][1][i] = dests[i][qss->colorOutSlot][1]; + quad->outputs.color[0][2][i] = dests[i][qss->colorOutSlot][2]; + quad->outputs.color[0][3][i] = dests[i][qss->colorOutSlot][3]; } } #if DLLVM for (int i = 0; i < QUAD_SIZE; ++i) { debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, - quad->outputs.color[0][i], - quad->outputs.color[1][i], - quad->outputs.color[2][i], - quad->outputs.color[3][i]); + quad->outputs.color[0][0][i], + quad->outputs.color[0][1][i], + quad->outputs.color[0][2][i], + quad->outputs.color[0][3][i]); } #endif diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 8095d662ee..55741cc1df 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -40,7 +40,7 @@ #include "tgsi/exec/tgsi_sse2.h" -#if defined(__i386__) || defined(__386__) +#ifdef PIPE_ARCH_X86 #include "rtasm/rtasm_x86sse.h" @@ -49,9 +49,10 @@ typedef void (XSTDCALL *codegen_function)( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, - float (*constant)[4], + const float (*constant)[4], struct tgsi_exec_vector *temporary, - const struct tgsi_interp_coef *coef + const struct tgsi_interp_coef *coef, + float (*immediates)[4] //, const struct tgsi_exec_vector *quadPos ); @@ -60,14 +61,15 @@ struct sp_sse_fragment_shader { struct sp_fragment_shader base; struct x86_function sse2_program; codegen_function func; + float immediates[TGSI_EXEC_NUM_IMMEDIATES][4]; }; static void -fs_sse_prepare( struct sp_fragment_shader *base, - struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers ) +fs_sse_prepare( const struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers ) { } @@ -78,9 +80,9 @@ fs_sse_prepare( struct sp_fragment_shader *base, * TODO: process >1 quad at a time */ static unsigned -fs_sse_run( struct sp_fragment_shader *base, - struct tgsi_exec_machine *machine, - struct quad_header *quad ) +fs_sse_run( const struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct quad_header *quad ) { struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base; @@ -96,7 +98,8 @@ fs_sse_run( struct sp_fragment_shader *base, machine->Outputs, machine->Consts, machine->Temps, - machine->InterpCoefs + machine->InterpCoefs, + shader->immediates // , &machine->QuadPos ); @@ -129,13 +132,18 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, x86_init_func( &shader->sse2_program ); - if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program )) { + if (!tgsi_emit_sse2( templ->tokens, &shader->sse2_program, + shader->immediates, FALSE )) { FREE(shader); return NULL; } shader->func = (codegen_function) x86_get_func( &shader->sse2_program ); - assert(shader->func); + if (!shader->func) { + x86_release_func( &shader->sse2_program ); + FREE(shader); + return NULL; + } shader->base.shader = *templ; shader->base.prepare = fs_sse_prepare; diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h index 9cf8222133..3d9ede69bb 100644 --- a/src/gallium/drivers/softpipe/sp_headers.h +++ b/src/gallium/drivers/softpipe/sp_headers.h @@ -31,6 +31,7 @@ #ifndef SP_HEADERS_H #define SP_HEADERS_H +#include "pipe/p_state.h" #include "tgsi/exec/tgsi_exec.h" #define PRIM_POINT 1 @@ -66,7 +67,8 @@ struct quad_header { unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ struct { - float color[NUM_CHANNELS][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */ + float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE]; float depth[QUAD_SIZE]; } outputs; diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index c7eb12b3bb..1cf9ffa632 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -26,7 +26,9 @@ **************************************************************************/ /** - * \brief Primitive rasterization/rendering (points, lines, triangles) + * \brief A draw stage that drives our triangle setup routines from + * within the draw pipeline. One of two ways to drive setup, the + * other being in sp_prim_vbuf.c. * * \author Keith Whitwell <keith@tungstengraphics.com> * \author Brian Paul @@ -34,29 +36,12 @@ #include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" +#include "sp_setup.h" #include "sp_state.h" #include "sp_prim_setup.h" -#include "draw/draw_private.h" +#include "draw/draw_pipe.h" #include "draw/draw_vertex.h" #include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" - -#define DEBUG_VERTS 0 -#define DEBUG_FRAGS 0 - -/** - * Triangle edge info - */ -struct edge { - float dx; /**< X(v1) - X(v0), used only during setup */ - float dy; /**< Y(v1) - Y(v0), used only during setup */ - float dxdy; /**< dx/dy */ - float sx, sy; /**< first sample point coord */ - int lines; /**< number of lines on this edge */ -}; - /** * Triangle setup info (derived from draw_stage). @@ -65,39 +50,7 @@ struct edge { struct setup_stage { struct draw_stage stage; /**< This must be first (base class) */ - struct softpipe_context *softpipe; - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - float oneoverarea; - - struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; - struct tgsi_interp_coef posCoef; /* For Z, W */ - struct quad_header quad; - - struct { - int left[2]; /**< [0] = row0, [1] = row1 */ - int right[2]; - int y; - unsigned y_flags; - unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; - -#if DEBUG_FRAGS - uint numFragsEmitted; /**< per primitive */ - uint numFragsWritten; /**< per primitive */ -#endif + struct setup_context *setup; }; @@ -111,1112 +64,49 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) } -/** - * Clip setup->quad against the scissor/surface bounds. - */ -static INLINE void -quad_clip(struct setup_stage *setup) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - - if (setup->quad.x0 >= maxx || - setup->quad.y0 >= maxy || - setup->quad.x0 + 1 < minx || - setup->quad.y0 + 1 < miny) { - /* totally clipped */ - setup->quad.mask = 0x0; - return; - } - if (setup->quad.x0 < minx) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < miny) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup->quad.x0 == maxx - 1) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == maxy - 1) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); -} - - -/** - * Emit a quad (pass to next stage) with clipping. - */ -static INLINE void -clip_emit_quad(struct setup_stage *setup) -{ - quad_clip(setup); - if (setup->quad.mask) { - struct softpipe_context *sp = setup->softpipe; - sp->quad.first->run(sp->quad.first, &setup->quad); - } -} - - -/** - * Emit a quad (pass to next stage). No clipping is done. - */ -static INLINE void -emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) -{ - struct softpipe_context *sp = setup->softpipe; - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; -#if DEBUG_FRAGS - if (mask & 1) setup->numFragsEmitted++; - if (mask & 2) setup->numFragsEmitted++; - if (mask & 4) setup->numFragsEmitted++; - if (mask & 8) setup->numFragsEmitted++; -#endif - sp->quad.first->run(sp->quad.first, &setup->quad); -#if DEBUG_FRAGS - mask = setup->quad.mask; - if (mask & 1) setup->numFragsWritten++; - if (mask & 2) setup->numFragsWritten++; - if (mask & 4) setup->numFragsWritten++; - if (mask & 8) setup->numFragsWritten++; -#endif -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static INLINE int block( int x ) -{ - return x & ~1; -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static unsigned calculate_mask( struct setup_stage *setup, int x ) -{ - unsigned mask = 0x0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_TOP_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_BOTTOM_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_TOP_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_BOTTOM_RIGHT; - - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( struct setup_stage *setup ) -{ - int minleft, maxright; - int x; - - switch (setup->span.y_flags) { - case 0x3: - /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); - break; - - case 0x1: - /* only even line written (quad top row) */ - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; - break; - - case 0x2: - /* only odd line written (quad bottom row) */ - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; - break; - - default: - return; - } - - /* XXX this loop could be moved into the above switch cases and - * calculate_mask() could be simplified a bit... - */ - for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( setup, x, setup->span.y, - calculate_mask( setup, x ) ); - } - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -} - -#if DEBUG_VERTS -static void print_vertex(const struct setup_stage *setup, - const struct vertex_header *v) -{ - int i; - debug_printf("Vertex: (%p)\n", v); - for (i = 0; i < setup->quad.nr_attrs; i++) { - debug_printf(" %d: %f %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); - } -} -#endif - -static boolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; - -#if DEBUG_VERTS - debug_printf("Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); -#endif - - setup->vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - float y0 = v0->data[0][1]; - float y1 = v1->data[0][1]; - float y2 = v2->data[0][1]; - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; - } - } - } - - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - - /* - * Compute triangle's area. Use 1/area to compute partial - * derivatives of attributes later. - * - * The area will be the same as prim->det, but the sign may be - * different depending on how the vertices get sorted above. - * - * To determine whether the primitive is front or back facing we - * use the prim->det value because its sign is correct. - */ - { - const float area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); - - setup->oneoverarea = 1.0f / area; - /* - debug_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup->oneoverarea, area, prim->det ); - */ - } - - /* We need to know if this is a front or back-facing triangle for: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); - - return TRUE; -} - - -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - * The value value comes from vertex->data[slot][i]. - * The result will be put into setup->coef[slot].a0[i]. - * \param slot which attribute slot - * \param i which component of the slot (0..3) - */ -static void const_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - assert(i <= 3); - - coef->dadx[i] = 0; - coef->dady[i] = 0; - /* need provoking vertex info! - */ - coef->a0[i] = setup->vprovoke->data[vertSlot][i]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static void tri_linear_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - float botda = setup->vmid->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - float majda = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - float dadx = a * setup->oneoverarea; - float dady = b * setup->oneoverarea; - - assert(i <= 3); - - coef->dadx[i] = dadx; - coef->dady[i] = dady; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); - - /* - debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); - */ -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - * We basically multiply the vertex value by 1/w before computing - * the plane coefficients (a0, dadx, dady). - * Later, when we compute the value at a particular fragment position we'll - * divide the interpolated value by the interpolated W at that fragment. - */ -static void tri_persp_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - /* premultiply by 1/w (v->data[0][3] is always W): - */ - float mina = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - float mida = setup->vmid->data[vertSlot][i] * setup->vmid->data[0][3]; - float maxa = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; - float botda = mida - mina; - float majda = maxa - mina; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - float dadx = a * setup->oneoverarea; - float dady = b * setup->oneoverarea; - - /* - debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, - setup->vmin->data[vertSlot][i], - setup->vmid->data[vertSlot][i], - setup->vmax->data[vertSlot][i] - ); - */ - assert(i <= 3); - - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (mina - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Special coefficient setup for gl_FragCoord. - * X and Y are trivial, though Y has to be inverted for OpenGL. - * Z and W are copied from posCoef which should have already been computed. - * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. - */ static void -setup_fragcoord_coeff(struct setup_stage *setup, uint slot) -{ - /*X*/ - setup->coef[slot].a0[0] = 0; - setup->coef[slot].dadx[0] = 1.0; - setup->coef[slot].dady[0] = 0.0; - /*Y*/ - if (setup->softpipe->rasterizer->origin_lower_left) { - /* y=0=bottom */ - const int winHeight = setup->softpipe->framebuffer.height; - setup->coef[slot].a0[1] = (float) (winHeight - 1); - setup->coef[slot].dady[1] = -1.0; - } - else { - /* y=0=top */ - setup->coef[slot].a0[1] = 0.0; - setup->coef[slot].dady[1] = 1.0; - } - setup->coef[slot].dadx[1] = 0.0; - /*Z*/ - setup->coef[slot].a0[2] = setup->posCoef.a0[2]; - setup->coef[slot].dadx[2] = setup->posCoef.dadx[2]; - setup->coef[slot].dady[2] = setup->posCoef.dady[2]; - /*W*/ - setup->coef[slot].a0[3] = setup->posCoef.a0[3]; - setup->coef[slot].dadx[3] = setup->posCoef.dadx[3]; - setup->coef[slot].dady[3] = setup->posCoef.dady[3]; -} - - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients( struct setup_stage *setup ) -{ - struct softpipe_context *softpipe = setup->softpipe; - const struct sp_fragment_shader *spfs = softpipe->fs; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, &setup->posCoef, 0, 2); - tri_linear_coeff(setup, &setup->posCoef, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - setup_fragcoord_coeff(setup, fragSlot); - break; - default: - assert(0); - } - - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } -} - - - -static void setup_tri_edges( struct setup_stage *setup ) -{ - float vmin_x = setup->vmin->data[0][0] + 0.5f; - float vmid_x = setup->vmid->data[0][0] + 0.5f; - - float vmin_y = setup->vmin->data[0][1] - 0.5f; - float vmid_y = setup->vmid->data[0][1] - 0.5f; - float vmax_y = setup->vmax->data[0][1] - 0.5f; - - setup->emaj.sy = CEILF(vmin_y); - setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = CEILF(vmid_y); - setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = CEILF(vmin_y); - setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring/cliprect is applied here too. - */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, - struct edge *eright, - unsigned lines ) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - int y, start_y, finish_y; - int sy = (int)eleft->sy; - - assert((int)eleft->sy == (int) eright->sy); - - /* clip top/bottom */ - start_y = sy; - finish_y = sy + lines; - - if (start_y < miny) - start_y = miny; - - if (finish_y > maxy) - finish_y = maxy; - - start_y -= sy; - finish_y -= sy; - - /* - debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - */ - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - int left = (int)(eleft->sx + y * eleft->dxdy); - int right = (int)(eright->sx + y * eright->dxdy); - - /* clip left/right */ - if (left < minx) - left = minx; - if (right > maxx) - right = maxx; - - if (left < right) { - int _y = sy + y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Do setup for triangle rasterization, then render the triangle. - */ -static void setup_tri( struct draw_stage *stage, - struct prim_header *prim ) +do_tri(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - /* - debug_printf("%s\n", __FUNCTION__ ); - */ - -#if DEBUG_FRAGS - setup->numFragsEmitted = 0; - setup->numFragsWritten = 0; -#endif - - setup_sort_vertices( setup, prim ); - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); - - setup->quad.prim = PRIM_TRI; - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; - /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ - - /* init_constant_attribs( setup ); */ - - if (setup->oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); - } - - flush_spans( setup ); - -#if DEBUG_FRAGS - printf("Tri: %u frags emitted, %u written\n", - setup->numFragsEmitted, - setup->numFragsWritten); -#endif -} - - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a line. - */ -static void -line_linear_coeff(struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - const float da = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - const float dadx = da * setup->emaj.dx * setup->oneoverarea; - const float dady = da * setup->emaj.dy * setup->oneoverarea; - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a line. - */ -static void -line_persp_coeff(struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - /* XXX double-check/verify this arithmetic */ - const float a0 = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - const float a1 = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; - const float da = a1 - a0; - const float dadx = da * setup->emaj.dx * setup->oneoverarea; - const float dady = da * setup->emaj.dy * setup->oneoverarea; - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmax are initialized. - */ -static INLINE void -setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) -{ - struct softpipe_context *softpipe = setup->softpipe; - const struct sp_fragment_shader *spfs = softpipe->fs; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* use setup->vmin, vmax to point to vertices */ - setup->vprovoke = prim->v[1]; - setup->vmin = prim->v[0]; - setup->vmax = prim->v[1]; - - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); - - /* z and w are done by linear interpolation: - */ - line_linear_coeff(setup, &setup->posCoef, 0, 2); - line_linear_coeff(setup, &setup->posCoef, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - setup_fragcoord_coeff(setup, fragSlot); - break; - default: - assert(0); - } - - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } + setup_tri( setup->setup, + prim->v[0]->data, + prim->v[1]->data, + prim->v[2]->data ); } - -/** - * Plot a pixel in a line segment. - */ -static INLINE void -plot(struct setup_stage *setup, int x, int y) -{ - const int iy = y & 1; - const int ix = x & 1; - const int quadX = x - ix; - const int quadY = y - iy; - const int mask = (1 << ix) << (2 * iy); - - if (quadX != setup->quad.x0 || - quadY != setup->quad.y0) - { - /* flush prev quad, start new quad */ - - if (setup->quad.x0 != -1) - clip_emit_quad(setup); - - setup->quad.x0 = quadX; - setup->quad.y0 = quadY; - setup->quad.mask = 0x0; - } - - setup->quad.mask |= mask; -} - - -/** - * Do setup for line rasterization, then render the line. - * Single-pixel width, no stipple, etc. We rely on the 'draw' module - * to handle stippling and wide lines. - */ static void -setup_line(struct draw_stage *stage, struct prim_header *prim) +do_line(struct draw_stage *stage, struct prim_header *prim) { - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; struct setup_stage *setup = setup_stage( stage ); - int x0 = (int) v0->data[0][0]; - int x1 = (int) v1->data[0][0]; - int y0 = (int) v0->data[0][1]; - int y1 = (int) v1->data[0][1]; - int dx = x1 - x0; - int dy = y1 - y0; - int xstep, ystep; - if (dx == 0 && dy == 0) - return; - - setup_line_coefficients(setup, prim); - - if (dx < 0) { - dx = -dx; /* make positive */ - xstep = -1; - } - else { - xstep = 1; - } - - if (dy < 0) { - dy = -dy; /* make positive */ - ystep = -1; - } - else { - ystep = 1; - } - - assert(dx >= 0); - assert(dy >= 0); - - setup->quad.x0 = setup->quad.y0 = -1; - setup->quad.mask = 0x0; - setup->quad.prim = PRIM_LINE; - /* XXX temporary: set coverage to 1.0 so the line appears - * if AA mode happens to be enabled. - */ - setup->quad.coverage[0] = - setup->quad.coverage[1] = - setup->quad.coverage[2] = - setup->quad.coverage[3] = 1.0; - - if (dx > dy) { - /*** X-major line ***/ - int i; - const int errorInc = dy + dy; - int error = errorInc - dx; - const int errorDec = error - dx; - - for (i = 0; i < dx; i++) { - plot(setup, x0, y0); - - x0 += xstep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - y0 += ystep; - } - } - } - else { - /*** Y-major line ***/ - int i; - const int errorInc = dx + dx; - int error = errorInc - dy; - const int errorDec = error - dy; - - for (i = 0; i < dy; i++) { - plot(setup, x0, y0); - - y0 += ystep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - x0 += xstep; - } - } - } - - /* draw final quad */ - if (setup->quad.mask) { - clip_emit_quad(setup); - } + setup_line( setup->setup, + prim->v[0]->data, + prim->v[1]->data ); } - static void -point_persp_coeff(struct setup_stage *setup, - const struct vertex_header *vert, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - assert(i <= 3); - coef->dadx[i] = 0.0F; - coef->dady[i] = 0.0F; - coef->a0[i] = vert->data[vertSlot][i] * vert->data[0][3]; -} - - -/** - * Do setup for point rasterization, then render the point. - * Round or square points... - * XXX could optimize a lot for 1-pixel points. - */ -static void -setup_point(struct draw_stage *stage, struct prim_header *prim) +do_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - struct softpipe_context *softpipe = setup->softpipe; - const struct sp_fragment_shader *spfs = softpipe->fs; - const struct vertex_header *v0 = prim->v[0]; - const int sizeAttr = setup->softpipe->psize_slot; - const float size - = sizeAttr > 0 ? v0->data[sizeAttr][0] - : setup->softpipe->rasterizer->point_size; - const float halfSize = 0.5F * size; - const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; - const float x = v0->data[0][0]; /* Note: data[0] is always position */ - const float y = v0->data[0][1]; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* For points, all interpolants are constant-valued. - * However, for point sprites, we'll need to setup texcoords appropriately. - * XXX: which coefficients are the texcoords??? - * We may do point sprites as textured quads... - * - * KW: We don't know which coefficients are texcoords - ultimately - * the choice of what interpolation mode to use for each attribute - * should be determined by the fragment program, using - * per-attribute declaration statements that include interpolation - * mode as a parameter. So either the fragment program will have - * to be adjusted for pointsprite vs normal point behaviour, or - * otherwise a special interpolation mode will have to be defined - * which matches the required behaviour for point sprites. But - - * the latter is not a feature of normal hardware, and as such - * probably should be ruled out on that basis. - */ - setup->vprovoke = prim->v[0]; - - /* setup Z, W */ - const_coeff(setup, &setup->posCoef, 0, 2); - const_coeff(setup, &setup->posCoef, 0, 3); - - for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - /* fall-through */ - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - point_persp_coeff(setup, setup->vprovoke, - &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - setup_fragcoord_coeff(setup, fragSlot); - break; - default: - assert(0); - } - - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } - - setup->quad.prim = PRIM_POINT; - - if (halfSize <= 0.5 && !round) { - /* special case for 1-pixel points */ - const int ix = ((int) x) & 1; - const int iy = ((int) y) & 1; - setup->quad.x0 = (int) x - ix; - setup->quad.y0 = (int) y - iy; - setup->quad.mask = (1 << ix) << (2 * iy); - clip_emit_quad(setup); - } - else { - if (round) { - /* rounded points */ - const int ixmin = block((int) (x - halfSize)); - const int ixmax = block((int) (x + halfSize)); - const int iymin = block((int) (y - halfSize)); - const int iymax = block((int) (y + halfSize)); - const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ - const float rmax = halfSize + 0.7071F; - const float rmin2 = MAX2(0.0F, rmin * rmin); - const float rmax2 = rmax * rmax; - const float cscale = 1.0F / (rmax2 - rmin2); - int ix, iy; - - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { - float dx, dy, dist2, cover; - - setup->quad.mask = 0x0; - - dx = (ix + 0.5f) - x; - dy = (iy + 0.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_LEFT; - } - - dx = (ix + 1.5f) - x; - dy = (iy + 0.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_RIGHT; - } - dx = (ix + 0.5f) - x; - dy = (iy + 1.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_LEFT; - } - - dx = (ix + 1.5f) - x; - dy = (iy + 1.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_RIGHT; - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); - } - } - } - } - else { - /* square points */ - const int xmin = (int) (x + 0.75 - halfSize); - const int ymin = (int) (y + 0.25 - halfSize); - const int xmax = xmin + (int) size; - const int ymax = ymin + (int) size; - /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ - const int ixmin = block(xmin); - const int ixmax = block(xmax - 1); - const int iymin = block(ymin); - const int iymax = block(ymax - 1); - int ix, iy; - - /* - debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); - */ - for (iy = iymin; iy <= iymax; iy += 2) { - uint rowMask = 0xf; - if (iy < ymin) { - /* above the top edge */ - rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - if (iy + 1 >= ymax) { - /* below the bottom edge */ - rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - } - - for (ix = ixmin; ix <= ixmax; ix += 2) { - uint mask = rowMask; - - if (ix < xmin) { - /* fragment is past left edge of point, turn off left bits */ - mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - } - if (ix + 1 >= xmax) { - /* past the right edge */ - mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - } - - setup->quad.mask = mask; - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); - } - } - } - } + setup_point( setup->setup, + prim->v[0]->data ); } + static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); - struct softpipe_context *sp = setup->softpipe; - const struct sp_fragment_shader *fs = setup->softpipe->fs; - uint i; - - if (sp->dirty) { - softpipe_update_derived(sp); - } - /* Mark surfaces as defined now */ - for (i = 0; i < sp->framebuffer.num_cbufs; i++){ - if (sp->framebuffer.cbufs[i]) { - sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; - } - } - if (sp->framebuffer.zsbuf) { - sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; - } + setup_prepare( setup->setup ); - setup->quad.nr_attrs = fs->info.num_inputs; - - sp->quad.first->begin(sp->quad.first); - - stage->point = setup_point; - stage->line = setup_line; - stage->tri = setup_tri; + stage->point = do_point; + stage->line = do_line; + stage->tri = do_tri; } @@ -1260,6 +150,8 @@ static void reset_stipple_counter( struct draw_stage *stage ) static void render_destroy( struct draw_stage *stage ) { + struct setup_stage *ssetup = setup_stage(stage); + setup_destroy_context(ssetup->setup); FREE( stage ); } @@ -1269,19 +161,29 @@ static void render_destroy( struct draw_stage *stage ) */ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) { - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + struct setup_stage *sstage = CALLOC_STRUCT(setup_stage); - setup->softpipe = softpipe; - setup->stage.draw = softpipe->draw; - setup->stage.point = setup_first_point; - setup->stage.line = setup_first_line; - setup->stage.tri = setup_first_tri; - setup->stage.flush = setup_flush; - setup->stage.reset_stipple_counter = reset_stipple_counter; - setup->stage.destroy = render_destroy; + sstage->setup = setup_create_context(softpipe); + sstage->stage.draw = softpipe->draw; + sstage->stage.point = setup_first_point; + sstage->stage.line = setup_first_line; + sstage->stage.tri = setup_first_tri; + sstage->stage.flush = setup_flush; + sstage->stage.reset_stipple_counter = reset_stipple_counter; + sstage->stage.destroy = render_destroy; - setup->quad.coef = setup->coef; - setup->quad.posCoef = &setup->posCoef; + return (struct draw_stage *)sstage; +} - return &setup->stage; +struct setup_context * +sp_draw_setup_context( struct draw_stage *stage ) +{ + struct setup_stage *ssetup = setup_stage(stage); + return ssetup->setup; +} + +void +sp_draw_flush( struct draw_stage *stage ) +{ + stage->flush( stage, 0 ); } diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.h b/src/gallium/drivers/softpipe/sp_prim_setup.h index f3e8a79dd9..49bdd98ed8 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.h +++ b/src/gallium/drivers/softpipe/sp_prim_setup.h @@ -69,6 +69,12 @@ typedef void (*vbuf_draw_func)( struct pipe_context *pipe, extern struct draw_stage * sp_draw_render_stage( struct softpipe_context *softpipe ); +extern struct setup_context * +sp_draw_setup_context( struct draw_stage * ); + +extern void +sp_draw_flush( struct draw_stage * ); + extern struct draw_stage * sp_draw_vbuf_stage( struct draw_context *draw_context, diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index d940718ed2..e063fe82ef 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ /** @@ -30,7 +30,7 @@ * softpipe rendering pipeline. * Probably not desired in general, but useful for testing/debuggin. * Enabled/Disabled with SP_VBUF env var. - * + * * Authors * Brian Paul */ @@ -39,14 +39,16 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_prim_vbuf.h" +#include "sp_prim_setup.h" +#include "sp_setup.h" #include "draw/draw_context.h" -#include "draw/draw_private.h" #include "draw/draw_vbuf.h" #define SP_MAX_VBUF_INDEXES 1024 #define SP_MAX_VBUF_SIZE 4096 +typedef const float (*cptrf4)[4]; /** * Subclass of vbuf_render. @@ -69,7 +71,6 @@ softpipe_vbuf_render(struct vbuf_render *vbr) } - static const struct vertex_info * sp_vbuf_get_vertex_info(struct vbuf_render *vbr) { @@ -91,7 +92,7 @@ sp_vbuf_allocate_vertices(struct vbuf_render *vbr, static void -sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, +sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, unsigned vertex_size, unsigned vertices_used) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); @@ -105,6 +106,16 @@ static boolean sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + + /* XXX: break this dependency - make setup_context live under + * softpipe, rename the old "setup" draw stage to something else. + */ + struct setup_context *setup_ctx = sp_draw_setup_context(cvbr->softpipe->setup); + + setup_prepare( setup_ctx ); + + + if (prim == PIPE_PRIM_TRIANGLES || prim == PIPE_PRIM_LINES || prim == PIPE_PRIM_POINTS) { @@ -114,31 +125,10 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) else { return FALSE; } - + } -/** - * Recalculate prim's determinant. - * XXX is this needed? - */ -static void -calc_det(struct prim_header *header) -{ - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - - /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; -} static void @@ -146,49 +136,59 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; - struct draw_stage *setup = softpipe->setup; - struct prim_header prim; unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i, j; void *vertex_buffer = cvbr->vertex_buffer; + cptrf4 v[3]; + + /* XXX: break this dependency - make setup_context live under + * softpipe, rename the old "setup" draw stage to something else. + */ + struct draw_stage *setup = softpipe->setup; + struct setup_context *setup_ctx = sp_draw_setup_context(softpipe->setup); - prim.det = 0; - prim.reset_line_stipple = 0; - prim.edgeflags = 0; - prim.pad = 0; switch (cvbr->prim) { case PIPE_PRIM_TRIANGLES: for (i = 0; i < nr_indices; i += 3) { - for (j = 0; j < 3; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - calc_det(&prim); - setup->tri( setup, &prim ); + for (j = 0; j < 3; j++) + v[j] = (cptrf4)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + setup_tri( setup_ctx, + v[0], + v[1], + v[2]); } break; case PIPE_PRIM_LINES: for (i = 0; i < nr_indices; i += 2) { - for (j = 0; j < 2; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - setup->line( setup, &prim ); + for (j = 0; j < 2; j++) + v[j] = (cptrf4)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + setup_line( setup_ctx, + v[0], + v[1] ); } break; case PIPE_PRIM_POINTS: for (i = 0; i < nr_indices; i++) { - prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + - indices[i] * vertex_size); - setup->point( setup, &prim ); + v[0] = (cptrf4)((char *)vertex_buffer + + indices[i] * vertex_size); + + setup_point( setup_ctx, + v[0] ); } break; } - setup->flush( setup, 0 ); + /* XXX: why are we calling this??? If we had to call something, it + * would be a function in sp_setup.c: + */ + sp_draw_flush( setup ); } @@ -202,111 +202,123 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; - struct prim_header prim; const void *vertex_buffer = cvbr->vertex_buffer; const unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i; - - prim.det = 0; - prim.reset_line_stipple = 0; - prim.edgeflags = 0; - prim.pad = 0; + struct setup_context *setup_ctx = sp_draw_setup_context(setup); + cptrf4 v[3]; #define VERTEX(I) \ - (struct vertex_header *) ((char *) vertex_buffer + (I) * vertex_size) + (cptrf4) ((char *) vertex_buffer + (I) * vertex_size) switch (cvbr->prim) { case PIPE_PRIM_POINTS: for (i = 0; i < nr; i++) { - prim.v[0] = VERTEX(i); - setup->point( setup, &prim ); + v[0] = VERTEX(i); + setup_point( setup_ctx, v[0] ); } break; case PIPE_PRIM_LINES: assert(nr % 2 == 0); for (i = 0; i < nr; i += 2) { - prim.v[0] = VERTEX(i); - prim.v[1] = VERTEX(i + 1); - setup->line( setup, &prim ); + v[0] = VERTEX(i); + v[1] = VERTEX(i + 1); + setup_line( setup_ctx, v[0], v[1] ); } break; case PIPE_PRIM_LINE_STRIP: for (i = 1; i < nr; i++) { - prim.v[0] = VERTEX(i - 1); - prim.v[1] = VERTEX(i); - setup->line( setup, &prim ); + v[0] = VERTEX(i - 1); + v[1] = VERTEX(i); + setup_line( setup_ctx, v[0], v[1] ); } break; case PIPE_PRIM_TRIANGLES: assert(nr % 3 == 0); for (i = 0; i < nr; i += 3) { - prim.v[0] = VERTEX(i + 0); - prim.v[1] = VERTEX(i + 1); - prim.v[2] = VERTEX(i + 2); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i + 0); + v[1] = VERTEX(i + 1); + v[2] = VERTEX(i + 2); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_TRIANGLE_STRIP: assert(nr >= 3); for (i = 2; i < nr; i++) { - prim.v[0] = VERTEX(i - 2); - prim.v[1] = VERTEX(i - 1); - prim.v[2] = VERTEX(i); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i - 2); + v[1] = VERTEX(i - 1); + v[2] = VERTEX(i); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_TRIANGLE_FAN: assert(nr >= 3); for (i = 2; i < nr; i++) { - prim.v[0] = VERTEX(0); - prim.v[1] = VERTEX(i - 1); - prim.v[2] = VERTEX(i); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(0); + v[1] = VERTEX(i - 1); + v[2] = VERTEX(i); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_QUADS: assert(nr % 4 == 0); for (i = 0; i < nr; i += 4) { - prim.v[0] = VERTEX(i + 0); - prim.v[1] = VERTEX(i + 1); - prim.v[2] = VERTEX(i + 2); - calc_det(&prim); - setup->tri( setup, &prim ); - - prim.v[0] = VERTEX(i + 0); - prim.v[1] = VERTEX(i + 2); - prim.v[2] = VERTEX(i + 3); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i + 0); + v[1] = VERTEX(i + 1); + v[2] = VERTEX(i + 2); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); + + v[0] = VERTEX(i + 0); + v[1] = VERTEX(i + 2); + v[2] = VERTEX(i + 3); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_QUAD_STRIP: assert(nr >= 4); for (i = 2; i < nr; i += 2) { - prim.v[0] = VERTEX(i - 2); - prim.v[1] = VERTEX(i); - prim.v[2] = VERTEX(i + 1); - calc_det(&prim); - setup->tri( setup, &prim ); - - prim.v[0] = VERTEX(i - 2); - prim.v[1] = VERTEX(i + 1); - prim.v[2] = VERTEX(i - 1); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i - 2); + v[1] = VERTEX(i); + v[2] = VERTEX(i + 1); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); + + v[0] = VERTEX(i - 2); + v[1] = VERTEX(i + 1); + v[2] = VERTEX(i - 1); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_POLYGON: /* draw as tri fan */ for (i = 2; i < nr; i++) { - prim.v[0] = VERTEX(0); - prim.v[1] = VERTEX(i - 1); - prim.v[2] = VERTEX(i); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(0); + v[1] = VERTEX(i - 1); + v[2] = VERTEX(i); + setup_tri( setup_ctx, + v[0], + v[1], + v[2] ); } break; default: diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c index 8603c1a367..bc83d78ea1 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -76,15 +76,6 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp_push_quad_first( sp, sp->quad.blend ); } - if (sp->framebuffer.num_cbufs == 1) { - /* the usual case: write to exactly one colorbuf */ - sp->current_cbuf = 0; - } - else { - /* insert bufloop stage */ - sp_push_quad_first( sp, sp->quad.bufloop ); - } - if (sp->depth_stencil->depth.occlusion_count) { sp_push_quad_first( sp, sp->quad.occlusion ); } diff --git a/src/gallium/drivers/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h index f1e0281764..08513cb95f 100644 --- a/src/gallium/drivers/softpipe/sp_quad.h +++ b/src/gallium/drivers/softpipe/sp_quad.h @@ -58,7 +58,6 @@ struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c index 318916fe30..7a42b08ef5 100644 --- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -16,7 +16,8 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_context *softpipe = qs->softpipe; const float ref = softpipe->depth_stencil->alpha.ref; unsigned passMask = 0x0, j; - const float *aaaa = quad->outputs.color[3]; + const uint cbuf = 0; /* only output[0].alpha is tested */ + const float *aaaa = quad->outputs.color[cbuf][3]; switch (softpipe->depth_stencil->alpha.func) { case PIPE_FUNC_NEVER: @@ -25,7 +26,7 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_FUNC_LESS: /* * If mask were an array [4] we could do this SIMD-style: - * passMask = (quad->outputs.color[3] <= vec4(ref)); + * passMask = (quad->outputs.color[0][3] <= vec4(ref)); */ for (j = 0; j < QUAD_SIZE; j++) { if (aaaa[j] < ref) { diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 0b79cfa1ed..74c6bff84a 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -101,114 +101,119 @@ static void logicop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - float dest[4][QUAD_SIZE]; - ubyte src[4][4], dst[4][4], res[4][4]; - uint *src4 = (uint *) src; - uint *dst4 = (uint *) dst; - uint *res4 = (uint *) res; - struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float dest[4][QUAD_SIZE]; + ubyte src[4][4], dst[4][4], res[4][4]; + uint *src4 = (uint *) src; + uint *dst4 = (uint *) dst; + uint *res4 = (uint *) res; + struct softpipe_cached_tile * + tile = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } } - } - /* convert to ubyte */ - for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ - - UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ - } + /* convert to ubyte */ + for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ + + UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ + } - switch (softpipe->blend->logicop_func) { - case PIPE_LOGICOP_CLEAR: - for (j = 0; j < 4; j++) - res4[j] = 0; - break; - case PIPE_LOGICOP_NOR: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] | dst4[j]); - break; - case PIPE_LOGICOP_AND_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j] & dst4[j]; - break; - case PIPE_LOGICOP_COPY_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j]; - break; - case PIPE_LOGICOP_AND_REVERSE: - for (j = 0; j < 4; j++) - res4[j] = src4[j] & ~dst4[j]; - break; - case PIPE_LOGICOP_INVERT: - for (j = 0; j < 4; j++) - res4[j] = ~dst4[j]; - break; - case PIPE_LOGICOP_XOR: - for (j = 0; j < 4; j++) - res4[j] = dst4[j] ^ src4[j]; - break; - case PIPE_LOGICOP_NAND: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] & dst4[j]); - break; - case PIPE_LOGICOP_AND: - for (j = 0; j < 4; j++) - res4[j] = src4[j] & dst4[j]; - break; - case PIPE_LOGICOP_EQUIV: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] ^ dst4[j]); - break; - case PIPE_LOGICOP_NOOP: - for (j = 0; j < 4; j++) - res4[j] = dst4[j]; - break; - case PIPE_LOGICOP_OR_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j] | dst4[j]; - break; - case PIPE_LOGICOP_COPY: - for (j = 0; j < 4; j++) - res4[j] = src4[j]; - break; - case PIPE_LOGICOP_OR_REVERSE: - for (j = 0; j < 4; j++) - res4[j] = src4[j] | ~dst4[j]; - break; - case PIPE_LOGICOP_OR: - for (j = 0; j < 4; j++) - res4[j] = src4[j] | dst4[j]; - break; - case PIPE_LOGICOP_SET: - for (j = 0; j < 4; j++) - res4[j] = ~0; - break; - default: - assert(0); - } + switch (softpipe->blend->logicop_func) { + case PIPE_LOGICOP_CLEAR: + for (j = 0; j < 4; j++) + res4[j] = 0; + break; + case PIPE_LOGICOP_NOR: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] | dst4[j]); + break; + case PIPE_LOGICOP_AND_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_COPY_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j]; + break; + case PIPE_LOGICOP_AND_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & ~dst4[j]; + break; + case PIPE_LOGICOP_INVERT: + for (j = 0; j < 4; j++) + res4[j] = ~dst4[j]; + break; + case PIPE_LOGICOP_XOR: + for (j = 0; j < 4; j++) + res4[j] = dst4[j] ^ src4[j]; + break; + case PIPE_LOGICOP_NAND: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] & dst4[j]); + break; + case PIPE_LOGICOP_AND: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_EQUIV: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] ^ dst4[j]); + break; + case PIPE_LOGICOP_NOOP: + for (j = 0; j < 4; j++) + res4[j] = dst4[j]; + break; + case PIPE_LOGICOP_OR_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_COPY: + for (j = 0; j < 4; j++) + res4[j] = src4[j]; + break; + case PIPE_LOGICOP_OR_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | ~dst4[j]; + break; + case PIPE_LOGICOP_OR: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_SET: + for (j = 0; j < 4; j++) + res4[j] = ~0; + break; + default: + assert(0); + } - for (j = 0; j < 4; j++) { - quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); - quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); - quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); - quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); + for (j = 0; j < 4; j++) { + quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); + quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); + quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); + quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); + } } /* pass quad to next stage */ @@ -221,504 +226,508 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { - struct softpipe_context *softpipe = qs->softpipe; static const float zero[4] = { 0, 0, 0, 0 }; static const float one[4] = { 1, 1, 1, 1 }; - float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; + + struct softpipe_context *softpipe = qs->softpipe; + uint cbuf; if (softpipe->blend->logicop_enable) { logicop_quad(qs, quad); return; } - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } } - } - /* - * Compute src/first term RGB - */ - switch (softpipe->blend->rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[0], quadColor[0]); /* R */ - VEC4_COPY(source[1], quadColor[1]); /* G */ - VEC4_COPY(source[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ - VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ - VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const float *alpha = quadColor[3]; - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ - VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ - VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - { - const float *alpha = dest[3]; - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[0], alpha, diff); /* R */ - VEC4_MIN(source[1], alpha, diff); /* G */ - VEC4_MIN(source[2], alpha, diff); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float alpha[4]; - VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_SRC1_ALPHA: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[0], zero); /* R */ - VEC4_COPY(source[1], zero); /* G */ - VEC4_COPY(source[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quadColor[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - float inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(source[0], quadColor[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(source[1], quadColor[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(source[2], quadColor[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_alpha[4]; - VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - assert(0); /* to do */ - break; - default: - assert(0); - } - - /* - * Compute src/first term A - */ - switch (softpipe->blend->alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[3], quadColor[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const float *alpha = quadColor[3]; - VEC4_MUL(source[3], quadColor[3], alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[3], alpha, diff); /* A */ - } - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(source[3], quadColor[3], comp); /* A */ + /* + * Compute src/first term RGB + */ + switch (softpipe->blend->rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[0], quadColor[0]); /* R */ + VEC4_COPY(source[1], quadColor[1]); /* G */ + VEC4_COPY(source[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const float *alpha = quadColor[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + { + const float *alpha = dest[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + { + const float *alpha = quadColor[3]; + float diff[4], temp[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(temp, alpha, diff); + VEC4_MUL(source[0], quadColor[0], temp); /* R */ + VEC4_MUL(source[1], quadColor[1], temp); /* G */ + VEC4_MUL(source[2], quadColor[2], temp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float alpha[4]; + VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[0], zero); /* R */ + VEC4_COPY(source[1], zero); /* G */ + VEC4_COPY(source[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(source[0], quadColor[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(source[1], quadColor[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(source[2], quadColor[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_alpha[4]; + VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + assert(0); /* to do */ + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[3], zero); /* A */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quadColor[3]); - VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - /* A */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[3], quadColor[3], inv_comp); - } - break; - default: - assert(0); - } - - - /* - * Compute dest/second term RGB - */ - switch (softpipe->blend->rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ - VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ - VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ - VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ - VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* illegal */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(dest[0], dest[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(dest[1], dest[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(dest[2], dest[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(dest[0], dest[0], comp); /* R */ - VEC4_MUL(dest[1], dest[1], comp); /* G */ - VEC4_MUL(dest[2], dest[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[0], zero); /* R */ - VEC4_COPY(dest[1], zero); /* G */ - VEC4_COPY(dest[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - case PIPE_BLENDFACTOR_SRC1_ALPHA: - /* XXX what are these? */ - assert(0); - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ - VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ - VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ - VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quadColor[3]); - VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ - VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ - VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ - VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - float inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(dest[0], dest[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(dest[1], dest[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(dest[2], dest[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(dest[0], dest[0], inv_comp); - VEC4_MUL(dest[1], dest[1], inv_comp); - VEC4_MUL(dest[2], dest[2], inv_comp); + + /* + * Compute src/first term A + */ + switch (softpipe->blend->alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[3], quadColor[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const float *alpha = quadColor[3]; + VEC4_MUL(source[3], quadColor[3], alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + /* multiply alpha by 1.0 */ + VEC4_COPY(source[3], quadColor[3]); /* A */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(source[3], quadColor[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + /* A */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[3], quadColor[3], inv_comp); + } + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - /* XXX what are these? */ - assert(0); - break; - default: - assert(0); - } - - /* - * Compute dest/second term A - */ - switch (softpipe->blend->alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ - break; - case PIPE_BLENDFACTOR_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* illegal */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(dest[3], dest[3], comp); /* A */ + + + /* + * Compute dest/second term RGB + */ + switch (softpipe->blend->rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ + VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ + VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ + VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ + VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[0], zero); /* R */ + VEC4_COPY(dest[1], zero); /* G */ + VEC4_COPY(dest[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quadColor[3]); + VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ + VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ + VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[3]); /* A */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(dest[0], dest[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(dest[1], dest[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[0], dest[0], inv_comp); + VEC4_MUL(dest[1], dest[1], inv_comp); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[3], zero); /* A */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quadColor[3]); - VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ + + /* + * Compute dest/second term A + */ + switch (softpipe->blend->alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[3], dest[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quadColor[3]); + VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[3]); /* A */ + VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[3], dest[3], inv_comp); + } + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[3]); /* A */ - VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ + + /* + * Combine RGB terms + */ + switch (softpipe->blend->rgb_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ + VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ + VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ + VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ + VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ + VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ + VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(dest[3], dest[3], inv_comp); + + /* + * Combine A terms + */ + switch (softpipe->blend->alpha_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ + break; + default: + assert(0); } - break; - default: - assert(0); - } - - /* - * Combine RGB terms - */ - switch (softpipe->blend->rgb_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ - VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ - VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ - VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ - VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ - VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ - VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ - VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ - VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ - VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ - VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ - break; - default: - assert(0); - } - - /* - * Combine A terms - */ - switch (softpipe->blend->alpha_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ - break; - default: - assert(0); - } - + + } /* cbuf loop */ + /* pass blended quad to next stage */ qs->next->run(qs->next, quad); } diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c index 2ae4e22a7d..b3db428ef1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_bufloop.c +++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c @@ -13,7 +13,7 @@ static void cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - float tmp[4][QUAD_SIZE]; + float tmp[PIPE_MAX_COLOR_BUFS][4][QUAD_SIZE]; unsigned i; assert(sizeof(quad->outputs.color) == sizeof(tmp)); @@ -30,7 +30,9 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { /* set current cbuffer */ +#if 0 /* obsolete & going away */ softpipe->current_cbuf = i; +#endif /* pass blended quad to next stage */ qs->next->run(qs->next, quad); diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c index 1f09d900ca..7fe080990b 100644 --- a/src/gallium/drivers/softpipe/sp_quad_colormask.c +++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c @@ -47,38 +47,43 @@ static void colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - float dest[4][QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float dest[4][QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } } - } - /* R */ - if (!(softpipe->blend->colormask & PIPE_MASK_R)) - COPY_4V(quadColor[0], dest[0]); + /* R */ + if (!(softpipe->blend->colormask & PIPE_MASK_R)) + COPY_4V(quadColor[0], dest[0]); - /* G */ - if (!(softpipe->blend->colormask & PIPE_MASK_G)) - COPY_4V(quadColor[1], dest[1]); + /* G */ + if (!(softpipe->blend->colormask & PIPE_MASK_G)) + COPY_4V(quadColor[1], dest[1]); - /* B */ - if (!(softpipe->blend->colormask & PIPE_MASK_B)) - COPY_4V(quadColor[2], dest[2]); + /* B */ + if (!(softpipe->blend->colormask & PIPE_MASK_B)) + COPY_4V(quadColor[2], dest[2]); - /* A */ - if (!(softpipe->blend->colormask & PIPE_MASK_A)) - COPY_4V(quadColor[3], dest[3]); + /* A */ + if (!(softpipe->blend->colormask & PIPE_MASK_A)) + COPY_4V(quadColor[3], dest[3]); + } /* pass quad to next stage */ qs->next->run(qs->next, quad); diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c index b3d3fae22f..dd5ebb2296 100644 --- a/src/gallium/drivers/softpipe/sp_quad_coverage.c +++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c @@ -50,12 +50,17 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) || (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) || (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) { - float (*quadColor)[4] = quad->outputs.color; - unsigned j; - for (j = 0; j < QUAD_SIZE; j++) { - assert(quad->coverage[j] >= 0.0); - assert(quad->coverage[j] <= 1.0); + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float (*quadColor)[4] = quad->outputs.color[cbuf]; + unsigned j; + for (j = 0; j < QUAD_SIZE; j++) { + assert(quad->coverage[j] >= 0.0); + assert(quad->coverage[j] <= 1.0); quadColor[3][j] *= quad->coverage[j]; + } } } diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index a73df31383..8c88c192f8 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -53,7 +53,6 @@ struct quad_shade_stage struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; - int colorOutSlot, depthOutSlot; }; @@ -77,7 +76,8 @@ shade_quad( struct quad_shade_stage *qss = quad_shade_stage( qs ); struct softpipe_context *softpipe = qs->softpipe; struct tgsi_exec_machine *machine = &qss->machine; - + boolean z_written; + /* Consts do not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; @@ -88,31 +88,37 @@ shade_quad( &qss->machine, quad ); - /* store result color */ - if (qss->colorOutSlot >= 0) { - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color, - &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); - } - - /* - * XXX the following code for updating quad->outputs.depth - * isn't really needed if we did early z testing. - */ - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ + /* store outputs */ + z_written = FALSE; + { + const ubyte *sem_name = softpipe->fs->info.output_semantic_name; + const ubyte *sem_index = softpipe->fs->info.output_semantic_index; + const uint n = qss->stage.softpipe->fs->info.num_outputs; uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + for (i = 0; i < n; i++) { + switch (sem_name[i]) { + case TGSI_SEMANTIC_COLOR: + { + uint cbuf = sem_index[i]; + memcpy(quad->outputs.color[cbuf], + &machine->Outputs[i].xyzw[0].f[0], + sizeof(quad->outputs.color[0]) ); + } + break; + case TGSI_SEMANTIC_POSITION: + { + uint j; + for (j = 0; j < 4; j++) { + quad->outputs.depth[j] = machine->Outputs[0].xyzw[2].f[j]; + } + z_written = TRUE; + } + break; + } } } - else { + + if (!z_written) { /* compute Z values now, as in the quad earlyz stage */ /* XXX we should really only do this if the earlyz stage is not used */ const float fx = (float) quad->x0; @@ -149,20 +155,6 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].texture = softpipe->texture[i]; } - /* find output slots for depth, color */ - qss->colorOutSlot = -1; - qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { - switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - qss->depthOutSlot = i; - break; - case TGSI_SEMANTIC_COLOR: - qss->colorOutSlot = i; - break; - } - } - softpipe->fs->prepare( softpipe->fs, &qss->machine, qss->samplers ); diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c index cfe8f11808..40083138a4 100644 --- a/src/gallium/drivers/softpipe/sp_quad_output.c +++ b/src/gallium/drivers/softpipe/sp_quad_output.c @@ -34,31 +34,36 @@ /** - * Write quad to framebuffer, taking mask into account. - * - * Note that surfaces support only full quad reads and writes. + * Last step of quad processing: write quad colors to the framebuffer, + * taking mask into account. */ static void output_quad(struct quad_stage *qs, struct quad_header *quad) { - struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); /* in-tile pos: */ const int itx = quad->x0 % TILE_SIZE; const int ity = quad->y0 % TILE_SIZE; - float (*quadColor)[4] = quad->outputs.color; - int i, j; - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - int x = itx + (j & 1); - int y = ity + (j >> 1); - for (i = 0; i < 4; i++) { /* loop over color chans */ - tile->data.color[y][x][i] = quadColor[i][j]; + struct softpipe_context *softpipe = qs->softpipe; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + int i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + int x = itx + (j & 1); + int y = ity + (j >> 1); + for (i = 0; i < 4; i++) { /* loop over color chans */ + tile->data.color[y][x][i] = quadColor[i][j]; + } } } } diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c index 92a0da0083..b4c7e942fa 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stencil.c +++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c @@ -243,7 +243,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) stencilVals[j] = tile->data.depth32[y][x] & 0xff; } break; - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); @@ -311,7 +311,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = z24s8; } break; - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 1850a1ced3..7dacb1c461 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -68,7 +68,7 @@ softpipe_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_POINT_SPRITE: return 1; case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; + return PIPE_MAX_COLOR_BUFS; case PIPE_CAP_OCCLUSION_QUERY: return 1; case PIPE_CAP_TEXTURE_SHADOW_MAP: diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c new file mode 100644 index 0000000000..543d86a5cb --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -0,0 +1,1308 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Primitive rasterization/rendering (points, lines, triangles) + * + * \author Keith Whitwell <keith@tungstengraphics.com> + * \author Brian Paul + */ + +#include "sp_setup.h" + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_state.h" +#include "sp_prim_setup.h" +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#define DEBUG_VERTS 0 +#define DEBUG_FRAGS 0 + +/** + * Triangle edge info + */ +struct edge { + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ +}; + + +/** + * Triangle setup info (derived from draw_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_context { + struct softpipe_context *softpipe; + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const float (*vmax)[4]; + const float (*vmid)[4]; + const float (*vmin)[4]; + const float (*vprovoke)[4]; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + float oneoverarea; + + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; + struct tgsi_interp_coef posCoef; /* For Z, W */ + struct quad_header quad; + + struct { + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; + +#if DEBUG_FRAGS + uint numFragsEmitted; /**< per primitive */ + uint numFragsWritten; /**< per primitive */ +#endif + + unsigned winding; /* which winding to cull */ +}; + + + + + +static boolean cull_tri( struct setup_context *setup, + float det ) +{ + if (det != 0) + { + /* if (det < 0 then Z points toward camera and triangle is + * counter-clockwise winding. + */ + unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + + if ((winding & setup->winding) == 0) + return FALSE; + } + + /* Culled: + */ + return TRUE; +} + + + +/** + * Clip setup->quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_context *setup) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup->quad.x0 >= maxx || + setup->quad.y0 >= maxy || + setup->quad.x0 + 1 < minx || + setup->quad.y0 + 1 < miny) { + /* totally clipped */ + setup->quad.mask = 0x0; + return; + } + if (setup->quad.x0 < minx) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup->quad.y0 < miny) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup->quad.x0 == maxx - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup->quad.y0 == maxy - 1) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); +} + + +/** + * Emit a quad (pass to next stage) with clipping. + */ +static INLINE void +clip_emit_quad(struct setup_context *setup) +{ + quad_clip(setup); + if (setup->quad.mask) { + struct softpipe_context *sp = setup->softpipe; + sp->quad.first->run(sp->quad.first, &setup->quad); + } +} + + +/** + * Emit a quad (pass to next stage). No clipping is done. + */ +static INLINE void +emit_quad( struct setup_context *setup, int x, int y, unsigned mask ) +{ + struct softpipe_context *sp = setup->softpipe; + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; +#if DEBUG_FRAGS + if (mask & 1) setup->numFragsEmitted++; + if (mask & 2) setup->numFragsEmitted++; + if (mask & 4) setup->numFragsEmitted++; + if (mask & 8) setup->numFragsEmitted++; +#endif + sp->quad.first->run(sp->quad.first, &setup->quad); +#if DEBUG_FRAGS + mask = setup->quad.mask; + if (mask & 1) setup->numFragsWritten++; + if (mask & 2) setup->numFragsWritten++; + if (mask & 4) setup->numFragsWritten++; + if (mask & 8) setup->numFragsWritten++; +#endif +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static INLINE int block( int x ) +{ + return x & ~1; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_context *setup ) +{ + const int xleft0 = setup->span.left[0]; + const int xleft1 = setup->span.left[1]; + const int xright0 = setup->span.right[0]; + const int xright1 = setup->span.right[1]; + int minleft, maxright; + int x; + + switch (setup->span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ + minleft = block(MIN2(xleft0, xleft1)); + maxright = block(MAX2(xright0, xright1)); + for (x = minleft; x <= maxright; x += 2) { + /* determine which of the four pixels is inside the span bounds */ + uint mask = 0x0; + if (x >= xleft0 && x < xright0) + mask |= MASK_TOP_LEFT; + if (x >= xleft1 && x < xright1) + mask |= MASK_BOTTOM_LEFT; + if (x+1 >= xleft0 && x+1 < xright0) + mask |= MASK_TOP_RIGHT; + if (x+1 >= xleft1 && x+1 < xright1) + mask |= MASK_BOTTOM_RIGHT; + emit_quad( setup, x, setup->span.y, mask ); + } + break; + + case 0x1: + /* only even line written (quad top row) */ + minleft = block(xleft0); + maxright = block(xright0); + for (x = minleft; x <= maxright; x += 2) { + uint mask = 0x0; + if (x >= xleft0 && x < xright0) + mask |= MASK_TOP_LEFT; + if (x+1 >= xleft0 && x+1 < xright0) + mask |= MASK_TOP_RIGHT; + emit_quad( setup, x, setup->span.y, mask ); + } + break; + + case 0x2: + /* only odd line written (quad bottom row) */ + minleft = block(xleft1); + maxright = block(xright1); + for (x = minleft; x <= maxright; x += 2) { + uint mask = 0x0; + if (x >= xleft1 && x < xright1) + mask |= MASK_BOTTOM_LEFT; + if (x+1 >= xleft1 && x+1 < xright1) + mask |= MASK_BOTTOM_RIGHT; + emit_quad( setup, x, setup->span.y, mask ); + } + break; + + default: + return; + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + + +#if DEBUG_VERTS +static void print_vertex(const struct setup_context *setup, + const float (*v)[4]) +{ + int i; + debug_printf("Vertex: (%p)\n", v); + for (i = 0; i < setup->quad.nr_attrs; i++) { + debug_printf(" %d: %f %f %f %f\n", i, + v[i][0], v[i][1], v[i][2], v[i][3]); + } +} +#endif + +static boolean setup_sort_vertices( struct setup_context *setup, + float det, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + float y0 = v0[0][1]; + float y1 = v1[0][1]; + float y2 = v2[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid[0][0] - setup->vmin[0][0]; + setup->ebot.dy = setup->vmid[0][1] - setup->vmin[0][1]; + setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0]; + setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1]; + setup->etop.dx = setup->vmax[0][0] - setup->vmid[0][0]; + setup->etop.dy = setup->vmax[0][1] - setup->vmid[0][1]; + + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. + * + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. + */ + { + const float area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0f / area; + /* + debug_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup->oneoverarea, area, det ); + */ + } + + /* We need to know if this is a front or back-facing triangle for: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); + + return TRUE; +} + + +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex[slot][i]. + * The result will be put into setup->coef[slot].a0[i]. + * \param slot which attribute slot + * \param i which component of the slot (0..3) + */ +static void const_coeff( struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + assert(i <= 3); + + coef->dadx[i] = 0; + coef->dady[i] = 0; + + /* need provoking vertex info! + */ + coef->a0[i] = setup->vprovoke[vertSlot][i]; +} + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + float botda = setup->vmid[vertSlot][i] - setup->vmin[vertSlot][i]; + float majda = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + + assert(i <= 3); + + coef->dadx[i] = dadx; + coef->dady[i] = dady; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + coef->a0[i] = (setup->vmin[vertSlot][i] - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); + + /* + debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); + */ +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. + */ +static void tri_persp_coeff( struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + /* premultiply by 1/w (v[0][3] is always W): + */ + float mina = setup->vmin[vertSlot][i] * setup->vmin[0][3]; + float mida = setup->vmid[vertSlot][i] * setup->vmid[0][3]; + float maxa = setup->vmax[vertSlot][i] * setup->vmax[0][3]; + float botda = mida - mina; + float majda = maxa - mina; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + + /* + debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, + setup->vmin[vertSlot][i], + setup->vmid[vertSlot][i], + setup->vmax[vertSlot][i] + ); + */ + assert(i <= 3); + + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (mina - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); +} + + +/** + * Special coefficient setup for gl_FragCoord. + * X and Y are trivial, though Y has to be inverted for OpenGL. + * Z and W are copied from posCoef which should have already been computed. + * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. + */ +static void +setup_fragcoord_coeff(struct setup_context *setup, uint slot) +{ + /*X*/ + setup->coef[slot].a0[0] = 0; + setup->coef[slot].dadx[0] = 1.0; + setup->coef[slot].dady[0] = 0.0; + /*Y*/ + if (setup->softpipe->rasterizer->origin_lower_left) { + /* y=0=bottom */ + const int winHeight = setup->softpipe->framebuffer.height; + setup->coef[slot].a0[1] = (float) (winHeight - 1); + setup->coef[slot].dady[1] = -1.0; + } + else { + /* y=0=top */ + setup->coef[slot].a0[1] = 0.0; + setup->coef[slot].dady[1] = 1.0; + } + setup->coef[slot].dadx[1] = 0.0; + /*Z*/ + setup->coef[slot].a0[2] = setup->posCoef.a0[2]; + setup->coef[slot].dadx[2] = setup->posCoef.dadx[2]; + setup->coef[slot].dady[2] = setup->posCoef.dady[2]; + /*W*/ + setup->coef[slot].a0[3] = setup->posCoef.a0[3]; + setup->coef[slot].dadx[3] = setup->posCoef.dadx[3]; + setup->coef[slot].dady[3] = setup->posCoef.dady[3]; +} + + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_context *setup ) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct sp_fragment_shader *spfs = softpipe->fs; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, &setup->posCoef, 0, 2); + tri_linear_coeff(setup, &setup->posCoef, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + setup_fragcoord_coeff(setup, fragSlot); + break; + default: + assert(0); + } + + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } +} + + + +static void setup_tri_edges( struct setup_context *setup ) +{ + float vmin_x = setup->vmin[0][0] + 0.5f; + float vmid_x = setup->vmid[0][0] + 0.5f; + + float vmin_y = setup->vmin[0][1] - 0.5f; + float vmid_y = setup->vmid[0][1] - 0.5f; + float vmax_y = setup->vmax[0][1] - 0.5f; + + setup->emaj.sy = CEILF(vmin_y); + setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = CEILF(vmid_y); + setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = CEILF(vmin_y); + setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring/cliprect is applied here too. + */ +static void subtriangle( struct setup_context *setup, + struct edge *eleft, + struct edge *eright, + unsigned lines ) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + int y, start_y, finish_y; + int sy = (int)eleft->sy; + + assert((int)eleft->sy == (int) eright->sy); + + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; + + if (start_y < miny) + start_y = miny; + + if (finish_y > maxy) + finish_y = maxy; + + start_y -= sy; + finish_y -= sy; + + /* + debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); + + /* clip left/right */ + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; + + if (left < right) { + int _y = sy + y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Recalculate prim's determinant. This is needed as we don't have + * get this information through the vbuf_render interface & we must + * calculate it here. + */ +static float +calc_det( const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0][0] - v2[0][0]; + const float ey = v0[0][1] - v2[0][1]; + const float fx = v1[0][0] - v2[0][0]; + const float fy = v1[0][1] - v2[0][1]; + + /* det = cross(e,f).z */ + return ex * fy - ey * fx; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +void setup_tri( struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ + float det; + +#if DEBUG_VERTS + debug_printf("Setup triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + + if (setup->softpipe->no_rast) + return; + + det = calc_det(v0, v1, v2); + /* + debug_printf("%s\n", __FUNCTION__ ); + */ + +#if DEBUG_FRAGS + setup->numFragsEmitted = 0; + setup->numFragsWritten = 0; +#endif + + if (cull_tri( setup, det )) + return; + + setup_sort_vertices( setup, det, v0, v1, v2 ); + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); + + setup->quad.prim = PRIM_TRI; + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; + /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ + + /* init_constant_attribs( setup ); */ + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); + +#if DEBUG_FRAGS + printf("Tri: %u frags emitted, %u written\n", + setup->numFragsEmitted, + setup->numFragsWritten); +#endif +} + + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a line. + */ +static void +line_linear_coeff(struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + const float da = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i]; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin[vertSlot][i] - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a line. + */ +static void +line_persp_coeff(struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + /* XXX double-check/verify this arithmetic */ + const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3]; + const float a1 = setup->vmax[vertSlot][i] * setup->vmax[0][3]; + const float da = a1 - a0; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin[vertSlot][i] - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); +} + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmax are initialized. + */ +static INLINE void +setup_line_coefficients(struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4]) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct sp_fragment_shader *spfs = softpipe->fs; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* use setup->vmin, vmax to point to vertices */ + setup->vprovoke = v1; + setup->vmin = v0; + setup->vmax = v1; + + setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0]; + setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1]; + /* NOTE: this is not really 1/area */ + setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); + + /* z and w are done by linear interpolation: + */ + line_linear_coeff(setup, &setup->posCoef, 0, 2); + line_linear_coeff(setup, &setup->posCoef, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + setup_fragcoord_coeff(setup, fragSlot); + break; + default: + assert(0); + } + + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } +} + + +/** + * Plot a pixel in a line segment. + */ +static INLINE void +plot(struct setup_context *setup, int x, int y) +{ + const int iy = y & 1; + const int ix = x & 1; + const int quadX = x - ix; + const int quadY = y - iy; + const int mask = (1 << ix) << (2 * iy); + + if (quadX != setup->quad.x0 || + quadY != setup->quad.y0) + { + /* flush prev quad, start new quad */ + + if (setup->quad.x0 != -1) + clip_emit_quad(setup); + + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; + setup->quad.mask = 0x0; + } + + setup->quad.mask |= mask; +} + + +/** + * Do setup for line rasterization, then render the line. + * Single-pixel width, no stipple, etc. We rely on the 'draw' module + * to handle stippling and wide lines. + */ +void +setup_line(struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4]) +{ + int x0 = (int) v0[0][0]; + int x1 = (int) v1[0][0]; + int y0 = (int) v0[0][1]; + int y1 = (int) v1[0][1]; + int dx = x1 - x0; + int dy = y1 - y0; + int xstep, ystep; + +#if DEBUG_VERTS + debug_printf("Setup line:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); +#endif + + if (setup->softpipe->no_rast) + return; + + if (dx == 0 && dy == 0) + return; + + setup_line_coefficients(setup, v0, v1); + + if (dx < 0) { + dx = -dx; /* make positive */ + xstep = -1; + } + else { + xstep = 1; + } + + if (dy < 0) { + dy = -dy; /* make positive */ + ystep = -1; + } + else { + ystep = 1; + } + + assert(dx >= 0); + assert(dy >= 0); + + setup->quad.x0 = setup->quad.y0 = -1; + setup->quad.mask = 0x0; + setup->quad.prim = PRIM_LINE; + /* XXX temporary: set coverage to 1.0 so the line appears + * if AA mode happens to be enabled. + */ + setup->quad.coverage[0] = + setup->quad.coverage[1] = + setup->quad.coverage[2] = + setup->quad.coverage[3] = 1.0; + + if (dx > dy) { + /*** X-major line ***/ + int i; + const int errorInc = dy + dy; + int error = errorInc - dx; + const int errorDec = error - dx; + + for (i = 0; i < dx; i++) { + plot(setup, x0, y0); + + x0 += xstep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + y0 += ystep; + } + } + } + else { + /*** Y-major line ***/ + int i; + const int errorInc = dx + dx; + int error = errorInc - dy; + const int errorDec = error - dy; + + for (i = 0; i < dy; i++) { + plot(setup, x0, y0); + + y0 += ystep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + x0 += xstep; + } + } + } + + /* draw final quad */ + if (setup->quad.mask) { + clip_emit_quad(setup); + } +} + + +static void +point_persp_coeff(struct setup_context *setup, + const float (*vert)[4], + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + assert(i <= 3); + coef->dadx[i] = 0.0F; + coef->dady[i] = 0.0F; + coef->a0[i] = vert[vertSlot][i] * vert[0][3]; +} + + +/** + * Do setup for point rasterization, then render the point. + * Round or square points... + * XXX could optimize a lot for 1-pixel points. + */ +void +setup_point( struct setup_context *setup, + const float (*v0)[4] ) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct sp_fragment_shader *spfs = softpipe->fs; + const int sizeAttr = setup->softpipe->psize_slot; + const float size + = sizeAttr > 0 ? v0[sizeAttr][0] + : setup->softpipe->rasterizer->point_size; + const float halfSize = 0.5F * size; + const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; + const float x = v0[0][0]; /* Note: data[0] is always position */ + const float y = v0[0][1]; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + +#if DEBUG_VERTS + debug_printf("Setup point:\n"); + print_vertex(setup, v0); +#endif + + if (softpipe->no_rast) + return; + + /* For points, all interpolants are constant-valued. + * However, for point sprites, we'll need to setup texcoords appropriately. + * XXX: which coefficients are the texcoords??? + * We may do point sprites as textured quads... + * + * KW: We don't know which coefficients are texcoords - ultimately + * the choice of what interpolation mode to use for each attribute + * should be determined by the fragment program, using + * per-attribute declaration statements that include interpolation + * mode as a parameter. So either the fragment program will have + * to be adjusted for pointsprite vs normal point behaviour, or + * otherwise a special interpolation mode will have to be defined + * which matches the required behaviour for point sprites. But - + * the latter is not a feature of normal hardware, and as such + * probably should be ruled out on that basis. + */ + setup->vprovoke = v0; + + /* setup Z, W */ + const_coeff(setup, &setup->posCoef, 0, 2); + const_coeff(setup, &setup->posCoef, 0, 3); + + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + /* fall-through */ + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + point_persp_coeff(setup, setup->vprovoke, + &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + setup_fragcoord_coeff(setup, fragSlot); + break; + default: + assert(0); + } + + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } + + setup->quad.prim = PRIM_POINT; + + if (halfSize <= 0.5 && !round) { + /* special case for 1-pixel points */ + const int ix = ((int) x) & 1; + const int iy = ((int) y) & 1; + setup->quad.x0 = (int) x - ix; + setup->quad.y0 = (int) y - iy; + setup->quad.mask = (1 << ix) << (2 * iy); + clip_emit_quad(setup); + } + else { + if (round) { + /* rounded points */ + const int ixmin = block((int) (x - halfSize)); + const int ixmax = block((int) (x + halfSize)); + const int iymin = block((int) (y - halfSize)); + const int iymax = block((int) (y + halfSize)); + const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ + const float rmax = halfSize + 0.7071F; + const float rmin2 = MAX2(0.0F, rmin * rmin); + const float rmax2 = rmax * rmax; + const float cscale = 1.0F / (rmax2 - rmin2); + int ix, iy; + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + float dx, dy, dist2, cover; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5f) - x; + dy = (iy + 0.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_LEFT; + } + + dx = (ix + 1.5f) - x; + dy = (iy + 0.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_RIGHT; + } + + dx = (ix + 0.5f) - x; + dy = (iy + 1.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_LEFT; + } + + dx = (ix + 1.5f) - x; + dy = (iy + 1.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_RIGHT; + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); + } + } + } + } + else { + /* square points */ + const int xmin = (int) (x + 0.75 - halfSize); + const int ymin = (int) (y + 0.25 - halfSize); + const int xmax = xmin + (int) size; + const int ymax = ymin + (int) size; + /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ + const int ixmin = block(xmin); + const int ixmax = block(xmax - 1); + const int iymin = block(ymin); + const int iymax = block(ymax - 1); + int ix, iy; + + /* + debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); + */ + for (iy = iymin; iy <= iymax; iy += 2) { + uint rowMask = 0xf; + if (iy < ymin) { + /* above the top edge */ + rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + if (iy + 1 >= ymax) { + /* below the bottom edge */ + rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + } + + for (ix = ixmin; ix <= ixmax; ix += 2) { + uint mask = rowMask; + + if (ix < xmin) { + /* fragment is past left edge of point, turn off left bits */ + mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + } + if (ix + 1 >= xmax) { + /* past the right edge */ + mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + } + + setup->quad.mask = mask; + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); + } + } + } + } +} + +void setup_prepare( struct setup_context *setup ) +{ + struct softpipe_context *sp = setup->softpipe; + unsigned i; + + if (sp->dirty) { + softpipe_update_derived(sp); + } + + /* Mark surfaces as defined now */ + for (i = 0; i < sp->framebuffer.num_cbufs; i++){ + if (sp->framebuffer.cbufs[i]) { + sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; + } + } + if (sp->framebuffer.zsbuf) { + sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; + } + + { + const struct sp_fragment_shader *fs = setup->softpipe->fs; + setup->quad.nr_attrs = fs->info.num_inputs; + sp->quad.first->begin(sp->quad.first); + } + + if (sp->reduced_api_prim == PIPE_PRIM_TRIANGLES && + sp->rasterizer->fill_cw == PIPE_POLYGON_MODE_FILL && + sp->rasterizer->fill_ccw == PIPE_POLYGON_MODE_FILL) { + /* we'll do culling */ + setup->winding = sp->rasterizer->cull_mode; + } + else { + /* 'draw' will do culling */ + setup->winding = PIPE_WINDING_NONE; + } +} + + + +void setup_destroy_context( struct setup_context *setup ) +{ + FREE( setup ); +} + + +/** + * Create a new primitive setup/render stage. + */ +struct setup_context *setup_create_context( struct softpipe_context *softpipe ) +{ + struct setup_context *setup = CALLOC_STRUCT(setup_context); + + setup->softpipe = softpipe; + + setup->quad.coef = setup->coef; + setup->quad.posCoef = &setup->posCoef; + + return setup; +} diff --git a/src/gallium/drivers/softpipe/sp_setup.h b/src/gallium/drivers/softpipe/sp_setup.h new file mode 100644 index 0000000000..d54f334428 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_setup.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +#ifndef SP_SETUP_H +#define SP_SETUP_H + +struct setup_context; +struct softpipe_context; + +void +setup_tri( struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ); + +void +setup_line(struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4]); + +void +setup_point( struct setup_context *setup, + const float (*v0)[4] ); + + +struct setup_context *setup_create_context( struct softpipe_context *softpipe ); +void setup_prepare( struct setup_context *setup ); +void setup_destroy_context( struct setup_context *setup ); + +#endif diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 6e6501f5bc..452e51fa79 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -54,9 +54,11 @@ struct tgsi_sampler; struct tgsi_exec_machine; +struct vertex_info; -/** Subclass of pipe_shader_state (though it doesn't really need to be). +/** + * Subclass of pipe_shader_state (though it doesn't really need to be). * * This is starting to look an awful lot like a quad pipeline stage... */ @@ -80,11 +82,10 @@ struct sp_fragment_shader { void (*delete)( struct sp_fragment_shader * ); }; -struct vertex_info; /** Subclass of pipe_shader_state */ struct sp_vertex_shader { - struct pipe_shader_state shader; + struct pipe_shader_state shader; /* Note: this field not actually used */ struct draw_vertex_shader *draw_data; }; @@ -171,6 +172,9 @@ boolean softpipe_draw_elements(struct pipe_context *pipe, unsigned indexSize, unsigned mode, unsigned start, unsigned count); +void +softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); + void softpipe_map_surfaces(struct softpipe_context *sp); diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 14abb20eeb..f10a1fa471 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -71,25 +71,16 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) * simply emit the whole post-xform vertex as-is: */ struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; -#if 0 - vinfo_vbuf->num_attribs = 0; - /* special-case to allow memcpy of whole vertex */ - draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); - /* size in dwords or floats */ - vinfo_vbuf->size = 4 * draw_num_vs_outputs(softpipe->draw) - + sizeof(struct vertex_header) / 4; -#else - /* for pass-through mode, we need a more explicit list of attribs */ const uint num = draw_num_vs_outputs(softpipe->draw); uint i; + /* No longer any need to try and emit draw vertex_header info. + */ vinfo_vbuf->num_attribs = 0; - draw_emit_vertex_attr(vinfo_vbuf, EMIT_HEADER, INTERP_NONE, 0); for (i = 0; i < num; i++) { draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i); } draw_compute_vertex_size(vinfo_vbuf); -#endif } /* diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 4eefd1d61f..24b91fbc79 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -83,6 +83,8 @@ void softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { struct sp_fragment_shader *state = fs; + + assert(fs != softpipe_context(pipe)->fs); state->delete( state ); } @@ -100,10 +102,7 @@ softpipe_create_vs_state(struct pipe_context *pipe, return NULL; } - state->shader = *templ; - - state->draw_data = draw_create_vertex_shader(softpipe->draw, - &state->shader); + state->draw_data = draw_create_vertex_shader(softpipe->draw, templ); if (state->draw_data == NULL) { FREE( state ); return NULL; diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c index e0230e16a4..46b6991195 100644 --- a/src/gallium/drivers/softpipe/sp_state_vertex.c +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -47,6 +47,7 @@ softpipe_set_vertex_elements(struct pipe_context *pipe, memcpy(softpipe->vertex_element, attribs, count * sizeof(struct pipe_vertex_element)); + softpipe->num_vertex_elements = count; softpipe->dirty |= SP_NEW_VERTEX; @@ -64,6 +65,7 @@ softpipe_set_vertex_buffers(struct pipe_context *pipe, assert(count <= PIPE_MAX_ATTRIBS); memcpy(softpipe->vertex_buffer, buffers, count * sizeof(buffers[0])); + softpipe->num_vertex_buffers = count; softpipe->dirty |= SP_NEW_VERTEX; diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 34da6356d7..be0b57d9fa 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -651,7 +651,9 @@ shadow_compare(uint compare_func, k = 0; break; default: + k = 0; assert(0); + break; } rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; @@ -1049,5 +1051,19 @@ sp_get_samples(struct tgsi_sampler *sampler, default: assert(0); } + +#if 0 /* DEBUG */ + { + int i; + printf("Sampled at %f, %f, %f:\n", s[0], t[0], p[0]); + for (i = 0; i < 4; i++) { + printf("Frag %d: %f %f %f %f\n", i, + rgba[0][i], + rgba[1][i], + rgba[2][i], + rgba[3][i]); + } + } +#endif } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 19f71887e7..1117c0ad4c 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -39,7 +39,7 @@ #include "sp_surface.h" #include "sp_tile_cache.h" -#define NUM_ENTRIES 30 +#define NUM_ENTRIES 32 /** XXX move these */ @@ -131,7 +131,7 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) uint pos; for (pos = 0; pos < NUM_ENTRIES; pos++) { - //assert(tc->entries[pos].x < 0); + /*assert(tc->entries[pos].x < 0);*/ } if (tc->surface) { pipe_surface_reference(&tc->surface, NULL); @@ -168,7 +168,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || ps->format == PIPE_FORMAT_Z32_UNORM || - ps->format == PIPE_FORMAT_U_S8); + ps->format == PIPE_FORMAT_S8_UNORM); } } @@ -332,8 +332,8 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { pipe_put_tile_raw(pipe, ps, - x, y, TILE_SIZE, TILE_SIZE, - tc->tile.data.color32, 0/*STRIDE*/); + x, y, TILE_SIZE, TILE_SIZE, + tc->tile.data.color32, 0/*STRIDE*/); /* do this? */ clear_clear_flag(tc->clear_flags, x, y); @@ -367,8 +367,8 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, if (tile->x >= 0) { if (tc->depth_stencil) { pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); } else { pipe_put_tile_rgba(pipe, ps, @@ -385,7 +385,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, #endif } else if (tc->texture) { - /* caching a texture, mark all entries as embpy */ + /* caching a texture, mark all entries as empty */ for (pos = 0; pos < NUM_ENTRIES; pos++) { tc->entries[pos].x = -1; } diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h index 291825dfe2..4ab666486c 100644 --- a/src/gallium/drivers/softpipe/sp_winsys.h +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -59,7 +59,7 @@ struct pipe_context; struct pipe_context *softpipe_create( struct pipe_screen *, struct pipe_winsys *, - struct softpipe_winsys * ); + void *unused ); struct pipe_screen * |