X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fstate_tracker%2Fst_program.c;h=0822509c4e582585e5306400361bb41d6389a368;hb=283d4d18e598793bbff7d9ba5a601bced9b36542;hp=80dcfd82743cd9e6108664d28bfde3516ed17e45;hpb=2c3f95d6aaab38cd66dd3dee1b089d5c91928eea;p=mesa.git diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 80dcfd82743..0822509c4e5 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -38,6 +38,8 @@ #include "program/prog_print.h" #include "program/programopt.h" +#include "compiler/nir/nir.h" + #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" @@ -51,8 +53,11 @@ #include "st_cb_bitmap.h" #include "st_cb_drawpixels.h" #include "st_context.h" +#include "st_tgsi_lower_yuv.h" #include "st_program.h" #include "st_mesa_to_tgsi.h" +#include "st_atifs_to_tgsi.h" +#include "st_nir.h" #include "cso_cache/cso_context.h" @@ -69,10 +74,10 @@ delete_vp_variant(struct st_context *st, struct st_vp_variant *vpv) if (vpv->draw_shader) draw_delete_vertex_shader( st->draw, vpv->draw_shader ); - - if (vpv->tgsi.tokens) + + if (((vpv->tgsi.type == PIPE_SHADER_IR_TGSI)) && vpv->tgsi.tokens) ureg_free_tokens(vpv->tgsi.tokens); - + free( vpv ); } @@ -95,7 +100,7 @@ st_release_vp_variants( struct st_context *st, stvp->variants = NULL; - if (stvp->tgsi.tokens) { + if ((stvp->tgsi.type == PIPE_SHADER_IR_TGSI) && stvp->tgsi.tokens) { tgsi_free_tokens(stvp->tgsi.tokens); stvp->tgsi.tokens = NULL; } @@ -132,7 +137,7 @@ st_release_fp_variants(struct st_context *st, struct st_fragment_program *stfp) stfp->variants = NULL; - if (stfp->tgsi.tokens) { + if ((stfp->tgsi.type == PIPE_SHADER_IR_TGSI) && stfp->tgsi.tokens) { ureg_free_tokens(stfp->tgsi.tokens); stfp->tgsi.tokens = NULL; } @@ -206,7 +211,7 @@ st_release_cp_variants(struct st_context *st, struct st_compute_program *stcp) for (v = *variants; v; ) { struct st_basic_variant *next = v->next; - delete_basic_variant(st, v, stcp->Base.Base.Target); + delete_basic_variant(st, v, stcp->Base.Target); v = next; } @@ -245,11 +250,12 @@ st_translate_vertex_program(struct st_context *st, * and TGSI generic input indexes, plus input attrib semantic info. */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { - if ((stvp->Base.Base.InputsRead & BITFIELD64_BIT(attr)) != 0) { + if ((stvp->Base.info.inputs_read & BITFIELD64_BIT(attr)) != 0) { input_to_index[attr] = stvp->num_inputs; stvp->index_to_input[stvp->num_inputs] = attr; stvp->num_inputs++; - if ((stvp->Base.Base.DoubleInputsRead & BITFIELD64_BIT(attr)) != 0) { + if ((stvp->Base.info.double_inputs_read & + BITFIELD64_BIT(attr)) != 0) { /* add placeholder for second part of a double attribute */ stvp->index_to_input[stvp->num_inputs] = ST_DOUBLE_ATTRIB_PLACEHOLDER; stvp->num_inputs++; @@ -263,7 +269,7 @@ st_translate_vertex_program(struct st_context *st, /* Compute mapping of vertex program outputs to slots. */ for (attr = 0; attr < VARYING_SLOT_MAX; attr++) { - if ((stvp->Base.Base.OutputsWritten & BITFIELD64_BIT(attr)) == 0) { + if ((stvp->Base.info.outputs_written & BITFIELD64_BIT(attr)) == 0) { stvp->result_to_output[attr] = ~0; } else { @@ -309,6 +315,11 @@ st_translate_vertex_program(struct st_context *st, output_semantic_name[slot] = TGSI_SEMANTIC_CLIPDIST; output_semantic_index[slot] = 1; break; + case VARYING_SLOT_CULL_DIST0: + case VARYING_SLOT_CULL_DIST1: + /* these should have been lowered by GLSL */ + assert(0); + break; case VARYING_SLOT_EDGE: assert(0); break; @@ -355,29 +366,59 @@ st_translate_vertex_program(struct st_context *st, output_semantic_name[num_outputs] = TGSI_SEMANTIC_EDGEFLAG; output_semantic_index[num_outputs] = 0; - if (!stvp->glsl_to_tgsi) - _mesa_remove_output_reads(&stvp->Base.Base, PROGRAM_OUTPUT); + /* ARB_vp: */ + if (!stvp->glsl_to_tgsi && !stvp->shader_program) { + _mesa_remove_output_reads(&stvp->Base, PROGRAM_OUTPUT); - ureg = ureg_create_with_screen(TGSI_PROCESSOR_VERTEX, st->pipe->screen); + /* This determines which states will be updated when the assembly + * shader is bound. + */ + stvp->affected_states = ST_NEW_VS_STATE | + ST_NEW_RASTERIZER | + ST_NEW_VERTEX_ARRAYS; + + if (stvp->Base.Parameters->NumParameters) + stvp->affected_states |= ST_NEW_VS_CONSTANTS; + + /* No samplers are allowed in ARB_vp. */ + } + + if (stvp->shader_program) { + nir_shader *nir = st_glsl_to_nir(st, &stvp->Base, stvp->shader_program, + MESA_SHADER_VERTEX); + + stvp->tgsi.type = PIPE_SHADER_IR_NIR; + stvp->tgsi.ir.nir = nir; + + st_translate_stream_output_info2(&stvp->shader_program->LinkedTransformFeedback, + stvp->result_to_output, + &stvp->tgsi.stream_output); + return true; + } + + ureg = ureg_create_with_screen(PIPE_SHADER_VERTEX, st->pipe->screen); if (ureg == NULL) return false; - if (stvp->Base.Base.ClipDistanceArraySize) + if (stvp->Base.ClipDistanceArraySize) ureg_property(ureg, TGSI_PROPERTY_NUM_CLIPDIST_ENABLED, - stvp->Base.Base.ClipDistanceArraySize); + stvp->Base.ClipDistanceArraySize); + if (stvp->Base.CullDistanceArraySize) + ureg_property(ureg, TGSI_PROPERTY_NUM_CULLDIST_ENABLED, + stvp->Base.CullDistanceArraySize); if (ST_DEBUG & DEBUG_MESA) { - _mesa_print_program(&stvp->Base.Base); - _mesa_print_program_parameters(st->ctx, &stvp->Base.Base); + _mesa_print_program(&stvp->Base); + _mesa_print_program_parameters(st->ctx, &stvp->Base); debug_printf("\n"); } if (stvp->glsl_to_tgsi) { error = st_translate_program(st->ctx, - TGSI_PROCESSOR_VERTEX, + PIPE_SHADER_VERTEX, ureg, stvp->glsl_to_tgsi, - &stvp->Base.Base, + &stvp->Base, /* inputs */ stvp->num_inputs, input_to_index, @@ -385,7 +426,6 @@ st_translate_vertex_program(struct st_context *st, NULL, /* input semantic name */ NULL, /* input semantic index */ NULL, /* interp mode */ - NULL, /* interp location */ /* outputs */ num_outputs, stvp->result_to_output, @@ -401,9 +441,9 @@ st_translate_vertex_program(struct st_context *st, stvp->glsl_to_tgsi = NULL; } else error = st_translate_mesa_program(st->ctx, - TGSI_PROCESSOR_VERTEX, + PIPE_SHADER_VERTEX, ureg, - &stvp->Base.Base, + &stvp->Base, /* inputs */ stvp->num_inputs, input_to_index, @@ -418,7 +458,7 @@ st_translate_vertex_program(struct st_context *st, if (error) { debug_printf("%s: failed to translate Mesa program:\n", __func__); - _mesa_print_program(&stvp->Base.Base); + _mesa_print_program(&stvp->Base); debug_assert(0); return false; } @@ -437,10 +477,27 @@ st_create_vp_variant(struct st_context *st, struct pipe_context *pipe = st->pipe; vpv->key = *key; - vpv->tgsi.tokens = tgsi_dup_tokens(stvp->tgsi.tokens); vpv->tgsi.stream_output = stvp->tgsi.stream_output; vpv->num_inputs = stvp->num_inputs; + if (stvp->tgsi.type == PIPE_SHADER_IR_NIR) { + vpv->tgsi.type = PIPE_SHADER_IR_NIR; + vpv->tgsi.ir.nir = nir_shader_clone(NULL, stvp->tgsi.ir.nir); + if (key->clamp_color) + NIR_PASS_V(vpv->tgsi.ir.nir, nir_lower_clamp_color_outputs); + if (key->passthrough_edgeflags) + NIR_PASS_V(vpv->tgsi.ir.nir, nir_lower_passthrough_edgeflags); + + st_finalize_nir(st, &stvp->Base, vpv->tgsi.ir.nir); + + vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->tgsi); + /* driver takes ownership of IR: */ + vpv->tgsi.ir.nir = NULL; + return vpv; + } + + vpv->tgsi.tokens = tgsi_dup_tokens(stvp->tgsi.tokens); + /* Emulate features. */ if (key->clamp_color || key->passthrough_edgeflags) { const struct tgsi_token *tokens; @@ -501,27 +558,6 @@ st_get_vp_variant(struct st_context *st, } -static unsigned -st_translate_interp(enum glsl_interp_qualifier glsl_qual, bool is_color) -{ - switch (glsl_qual) { - case INTERP_QUALIFIER_NONE: - if (is_color) - return TGSI_INTERPOLATE_COLOR; - return TGSI_INTERPOLATE_PERSPECTIVE; - case INTERP_QUALIFIER_SMOOTH: - return TGSI_INTERPOLATE_PERSPECTIVE; - case INTERP_QUALIFIER_FLAT: - return TGSI_INTERPOLATE_CONSTANT; - case INTERP_QUALIFIER_NOPERSPECTIVE: - return TGSI_INTERPOLATE_LINEAR; - default: - assert(0 && "unexpected interp mode in st_translate_interp()"); - return TGSI_INTERPOLATE_PERSPECTIVE; - } -} - - /** * Translate a Mesa fragment shader into a TGSI shader. */ @@ -529,11 +565,10 @@ bool st_translate_fragment_program(struct st_context *st, struct st_fragment_program *stfp) { - GLuint outputMapping[FRAG_RESULT_MAX]; + GLuint outputMapping[2 * FRAG_RESULT_MAX]; GLuint inputMapping[VARYING_SLOT_MAX]; GLuint inputSlotToAttr[VARYING_SLOT_MAX]; GLuint interpMode[PIPE_MAX_SHADER_INPUTS]; /* XXX size? */ - GLuint interpLocation[PIPE_MAX_SHADER_INPUTS]; GLuint attr; GLbitfield64 inputsRead; struct ureg_program *ureg; @@ -550,28 +585,43 @@ st_translate_fragment_program(struct st_context *st, memset(inputSlotToAttr, ~0, sizeof(inputSlotToAttr)); - if (!stfp->glsl_to_tgsi) { - _mesa_remove_output_reads(&stfp->Base.Base, PROGRAM_OUTPUT); + /* Non-GLSL programs: */ + if (!stfp->glsl_to_tgsi && !stfp->shader_program) { + _mesa_remove_output_reads(&stfp->Base, PROGRAM_OUTPUT); if (st->ctx->Const.GLSLFragCoordIsSysVal) - _mesa_program_fragment_position_to_sysval(&stfp->Base.Base); + _mesa_program_fragment_position_to_sysval(&stfp->Base); + + /* This determines which states will be updated when the assembly + * shader is bound. + * + * fragment.position and glDrawPixels always use constants. + */ + stfp->affected_states = ST_NEW_FS_STATE | + ST_NEW_SAMPLE_SHADING | + ST_NEW_FS_CONSTANTS; + + if (stfp->ati_fs) { + /* Just set them for ATI_fs unconditionally. */ + stfp->affected_states |= ST_NEW_FS_SAMPLER_VIEWS | + ST_NEW_RENDER_SAMPLERS; + } else { + /* ARB_fp */ + if (stfp->Base.SamplersUsed) + stfp->affected_states |= ST_NEW_FS_SAMPLER_VIEWS | + ST_NEW_RENDER_SAMPLERS; + } } /* * Convert Mesa program inputs to TGSI input register semantics. */ - inputsRead = stfp->Base.Base.InputsRead; + inputsRead = stfp->Base.info.inputs_read; for (attr = 0; attr < VARYING_SLOT_MAX; attr++) { if ((inputsRead & BITFIELD64_BIT(attr)) != 0) { const GLuint slot = fs_num_inputs++; inputMapping[attr] = slot; inputSlotToAttr[slot] = attr; - if (stfp->Base.IsCentroid & BITFIELD64_BIT(attr)) - interpLocation[slot] = TGSI_INTERPOLATE_LOC_CENTROID; - else if (stfp->Base.IsSample & BITFIELD64_BIT(attr)) - interpLocation[slot] = TGSI_INTERPOLATE_LOC_SAMPLE; - else - interpLocation[slot] = TGSI_INTERPOLATE_LOC_CENTER; switch (attr) { case VARYING_SLOT_POS: @@ -582,14 +632,14 @@ st_translate_fragment_program(struct st_context *st, case VARYING_SLOT_COL0: input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; input_semantic_index[slot] = 0; - interpMode[slot] = st_translate_interp(stfp->Base.InterpQualifier[attr], - TRUE); + interpMode[slot] = stfp->glsl_to_tgsi ? + TGSI_INTERPOLATE_COUNT : TGSI_INTERPOLATE_COLOR; break; case VARYING_SLOT_COL1: input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; input_semantic_index[slot] = 1; - interpMode[slot] = st_translate_interp(stfp->Base.InterpQualifier[attr], - TRUE); + interpMode[slot] = stfp->glsl_to_tgsi ? + TGSI_INTERPOLATE_COUNT : TGSI_INTERPOLATE_COLOR; break; case VARYING_SLOT_FOGC: input_semantic_name[slot] = TGSI_SEMANTIC_FOG; @@ -626,6 +676,11 @@ st_translate_fragment_program(struct st_context *st, input_semantic_index[slot] = 1; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; + case VARYING_SLOT_CULL_DIST0: + case VARYING_SLOT_CULL_DIST1: + /* these should have been lowered by GLSL */ + assert(0); + break; /* In most cases, there is nothing special about these * inputs, so adopt a convention to use the generic * semantic name and the mesa VARYING_SLOT_ number as the @@ -660,8 +715,8 @@ st_translate_fragment_program(struct st_context *st, if (st->needs_texcoord_semantic) { input_semantic_name[slot] = TGSI_SEMANTIC_TEXCOORD; input_semantic_index[slot] = attr - VARYING_SLOT_TEX0; - interpMode[slot] = - st_translate_interp(stfp->Base.InterpQualifier[attr], FALSE); + interpMode[slot] = stfp->glsl_to_tgsi ? + TGSI_INTERPOLATE_COUNT : TGSI_INTERPOLATE_PERSPECTIVE; break; } /* fall through */ @@ -683,9 +738,10 @@ st_translate_fragment_program(struct st_context *st, input_semantic_index[slot] = st_get_generic_varying_index(st, attr); if (attr == VARYING_SLOT_PNTC) interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - else - interpMode[slot] = st_translate_interp(stfp->Base.InterpQualifier[attr], - FALSE); + else { + interpMode[slot] = stfp->glsl_to_tgsi ? + TGSI_INTERPOLATE_COUNT : TGSI_INTERPOLATE_PERSPECTIVE; + } break; } } @@ -698,8 +754,7 @@ st_translate_fragment_program(struct st_context *st, * Semantics and mapping for outputs */ { - uint numColors = 0; - GLbitfield64 outputsWritten = stfp->Base.Base.OutputsWritten; + GLbitfield64 outputsWritten = stfp->Base.info.outputs_written; /* if z is written, emit that first */ if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) { @@ -727,9 +782,13 @@ st_translate_fragment_program(struct st_context *st, } /* handle remaining outputs (color) */ - for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { - if (outputsWritten & BITFIELD64_BIT(attr)) { - switch (attr) { + for (attr = 0; attr < ARRAY_SIZE(outputMapping); attr++) { + const GLbitfield64 written = attr < FRAG_RESULT_MAX ? outputsWritten : + stfp->Base.SecondaryOutputsWritten; + const unsigned loc = attr % FRAG_RESULT_MAX; + + if (written & BITFIELD64_BIT(loc)) { + switch (loc) { case FRAG_RESULT_DEPTH: case FRAG_RESULT_STENCIL: case FRAG_RESULT_SAMPLE_MASK: @@ -738,35 +797,55 @@ st_translate_fragment_program(struct st_context *st, break; case FRAG_RESULT_COLOR: write_all = GL_TRUE; /* fallthrough */ - default: - assert(attr == FRAG_RESULT_COLOR || - (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX)); + default: { + int index; + assert(loc == FRAG_RESULT_COLOR || + (FRAG_RESULT_DATA0 <= loc && loc < FRAG_RESULT_MAX)); + + index = (loc == FRAG_RESULT_COLOR) ? 0 : (loc - FRAG_RESULT_DATA0); + + if (attr >= FRAG_RESULT_MAX) { + /* Secondary color for dual source blending. */ + assert(index == 0); + index++; + } + fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR; - fs_output_semantic_index[fs_num_outputs] = numColors; + fs_output_semantic_index[fs_num_outputs] = index; outputMapping[attr] = fs_num_outputs; - numColors++; break; } + } fs_num_outputs++; } } } - ureg = ureg_create_with_screen(TGSI_PROCESSOR_FRAGMENT, st->pipe->screen); + if (stfp->shader_program) { + nir_shader *nir = st_glsl_to_nir(st, &stfp->Base, stfp->shader_program, + MESA_SHADER_FRAGMENT); + + stfp->tgsi.type = PIPE_SHADER_IR_NIR; + stfp->tgsi.ir.nir = nir; + + return true; + } + + ureg = ureg_create_with_screen(PIPE_SHADER_FRAGMENT, st->pipe->screen); if (ureg == NULL) return false; if (ST_DEBUG & DEBUG_MESA) { - _mesa_print_program(&stfp->Base.Base); - _mesa_print_program_parameters(st->ctx, &stfp->Base.Base); + _mesa_print_program(&stfp->Base); + _mesa_print_program_parameters(st->ctx, &stfp->Base); debug_printf("\n"); } if (write_all == GL_TRUE) ureg_property(ureg, TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS, 1); - if (stfp->Base.FragDepthLayout != FRAG_DEPTH_LAYOUT_NONE) { - switch (stfp->Base.FragDepthLayout) { + if (stfp->Base.info.fs.depth_layout != FRAG_DEPTH_LAYOUT_NONE) { + switch (stfp->Base.info.fs.depth_layout) { case FRAG_DEPTH_LAYOUT_ANY: ureg_property(ureg, TGSI_PROPERTY_FS_DEPTH_LAYOUT, TGSI_FS_DEPTH_LAYOUT_ANY); @@ -790,10 +869,10 @@ st_translate_fragment_program(struct st_context *st, if (stfp->glsl_to_tgsi) { st_translate_program(st->ctx, - TGSI_PROCESSOR_FRAGMENT, + PIPE_SHADER_FRAGMENT, ureg, stfp->glsl_to_tgsi, - &stfp->Base.Base, + &stfp->Base, /* inputs */ fs_num_inputs, inputMapping, @@ -801,7 +880,6 @@ st_translate_fragment_program(struct st_context *st, input_semantic_name, input_semantic_index, interpMode, - interpLocation, /* outputs */ fs_num_outputs, outputMapping, @@ -811,11 +889,26 @@ st_translate_fragment_program(struct st_context *st, free_glsl_to_tgsi_visitor(stfp->glsl_to_tgsi); stfp->glsl_to_tgsi = NULL; - } else + } else if (stfp->ati_fs) + st_translate_atifs_program(ureg, + stfp->ati_fs, + &stfp->Base, + /* inputs */ + fs_num_inputs, + inputMapping, + input_semantic_name, + input_semantic_index, + interpMode, + /* outputs */ + fs_num_outputs, + outputMapping, + fs_output_semantic_name, + fs_output_semantic_index); + else st_translate_mesa_program(st->ctx, - TGSI_PROCESSOR_FRAGMENT, + PIPE_SHADER_FRAGMENT, ureg, - &stfp->Base.Base, + &stfp->Base, /* inputs */ fs_num_inputs, inputMapping, @@ -841,14 +934,113 @@ st_create_fp_variant(struct st_context *st, struct pipe_context *pipe = st->pipe; struct st_fp_variant *variant = CALLOC_STRUCT(st_fp_variant); struct pipe_shader_state tgsi = {0}; + struct gl_program_parameter_list *params = stfp->Base.Parameters; + static const gl_state_index texcoord_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_CURRENT_ATTRIB, VERT_ATTRIB_TEX0 }; + static const gl_state_index scale_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_PT_SCALE }; + static const gl_state_index bias_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_PT_BIAS }; if (!variant) return NULL; + if (stfp->tgsi.type == PIPE_SHADER_IR_NIR) { + tgsi.type = PIPE_SHADER_IR_NIR; + tgsi.ir.nir = nir_shader_clone(NULL, stfp->tgsi.ir.nir); + + if (key->clamp_color) + NIR_PASS_V(tgsi.ir.nir, nir_lower_clamp_color_outputs); + + if (key->persample_shading) { + nir_shader *shader = tgsi.ir.nir; + nir_foreach_variable(var, &shader->inputs) + var->data.sample = true; + } + + assert(!(key->bitmap && key->drawpixels)); + + /* glBitmap */ + if (key->bitmap) { + nir_lower_bitmap_options options = {0}; + + variant->bitmap_sampler = ffs(~stfp->Base.SamplersUsed) - 1; + options.sampler = variant->bitmap_sampler; + options.swizzle_xxxx = (st->bitmap.tex_format == PIPE_FORMAT_L8_UNORM); + + NIR_PASS_V(tgsi.ir.nir, nir_lower_bitmap, &options); + } + + /* glDrawPixels (color only) */ + if (key->drawpixels) { + nir_lower_drawpixels_options options = {{0}}; + unsigned samplers_used = stfp->Base.SamplersUsed; + + /* Find the first unused slot. */ + variant->drawpix_sampler = ffs(~samplers_used) - 1; + options.drawpix_sampler = variant->drawpix_sampler; + samplers_used |= (1 << variant->drawpix_sampler); + + options.pixel_maps = key->pixelMaps; + if (key->pixelMaps) { + variant->pixelmap_sampler = ffs(~samplers_used) - 1; + options.pixelmap_sampler = variant->pixelmap_sampler; + } + + options.scale_and_bias = key->scaleAndBias; + if (key->scaleAndBias) { + _mesa_add_state_reference(params, scale_state); + memcpy(options.scale_state_tokens, scale_state, + sizeof(options.scale_state_tokens)); + _mesa_add_state_reference(params, bias_state); + memcpy(options.bias_state_tokens, bias_state, + sizeof(options.bias_state_tokens)); + } + + _mesa_add_state_reference(params, texcoord_state); + memcpy(options.texcoord_state_tokens, texcoord_state, + sizeof(options.texcoord_state_tokens)); + + NIR_PASS_V(tgsi.ir.nir, nir_lower_drawpixels, &options); + } + + if (unlikely(key->external.lower_nv12 || key->external.lower_iyuv)) { + nir_lower_tex_options options = {0}; + options.lower_y_uv_external = key->external.lower_nv12; + options.lower_y_u_v_external = key->external.lower_iyuv; + NIR_PASS_V(tgsi.ir.nir, nir_lower_tex, &options); + } + + st_finalize_nir(st, &stfp->Base, tgsi.ir.nir); + + if (unlikely(key->external.lower_nv12 || key->external.lower_iyuv)) { + /* This pass needs to happen *after* nir_lower_sampler */ + NIR_PASS_V(tgsi.ir.nir, st_nir_lower_tex_src_plane, + ~stfp->Base.SamplersUsed, + key->external.lower_nv12, + key->external.lower_iyuv); + } + + variant->driver_shader = pipe->create_fs_state(pipe, &tgsi); + variant->key = *key; + + return variant; + } + tgsi.tokens = stfp->tgsi.tokens; assert(!(key->bitmap && key->drawpixels)); + /* Fix texture targets and add fog for ATI_fs */ + if (stfp->ati_fs) { + const struct tgsi_token *tokens = st_fixup_atifs(tgsi.tokens, key); + + if (tokens) + tgsi.tokens = tokens; + else + fprintf(stderr, "mesa: cannot post-process ATI_fs\n"); + } + /* Emulate features. */ if (key->clamp_color || key->persample_shading) { const struct tgsi_token *tokens; @@ -858,9 +1050,11 @@ st_create_fp_variant(struct st_context *st, tokens = tgsi_emulate(tgsi.tokens, flags); - if (tokens) + if (tokens) { + if (tgsi.tokens != stfp->tgsi.tokens) + tgsi_free_tokens(tgsi.tokens); tgsi.tokens = tokens; - else + } else fprintf(stderr, "mesa: cannot emulate deprecated features\n"); } @@ -868,7 +1062,7 @@ st_create_fp_variant(struct st_context *st, if (key->bitmap) { const struct tgsi_token *tokens; - variant->bitmap_sampler = ffs(~stfp->Base.Base.SamplersUsed) - 1; + variant->bitmap_sampler = ffs(~stfp->Base.SamplersUsed) - 1; tokens = st_get_bitmap_shader(tgsi.tokens, st->internal_target, @@ -889,34 +1083,23 @@ st_create_fp_variant(struct st_context *st, if (key->drawpixels) { const struct tgsi_token *tokens; unsigned scale_const = 0, bias_const = 0, texcoord_const = 0; - struct gl_program_parameter_list *params = stfp->Base.Base.Parameters; /* Find the first unused slot. */ - variant->drawpix_sampler = ffs(~stfp->Base.Base.SamplersUsed) - 1; + variant->drawpix_sampler = ffs(~stfp->Base.SamplersUsed) - 1; if (key->pixelMaps) { - unsigned samplers_used = stfp->Base.Base.SamplersUsed | + unsigned samplers_used = stfp->Base.SamplersUsed | (1 << variant->drawpix_sampler); variant->pixelmap_sampler = ffs(~samplers_used) - 1; } if (key->scaleAndBias) { - static const gl_state_index scale_state[STATE_LENGTH] = - { STATE_INTERNAL, STATE_PT_SCALE }; - static const gl_state_index bias_state[STATE_LENGTH] = - { STATE_INTERNAL, STATE_PT_BIAS }; - scale_const = _mesa_add_state_reference(params, scale_state); bias_const = _mesa_add_state_reference(params, bias_state); } - { - static const gl_state_index state[STATE_LENGTH] = - { STATE_INTERNAL, STATE_CURRENT_ATTRIB, VERT_ATTRIB_TEX0 }; - - texcoord_const = _mesa_add_state_reference(params, state); - } + texcoord_const = _mesa_add_state_reference(params, texcoord_state); tokens = st_get_drawpix_shader(tgsi.tokens, st->needs_texcoord_semantic, @@ -934,6 +1117,25 @@ st_create_fp_variant(struct st_context *st, fprintf(stderr, "mesa: cannot create a shader for glDrawPixels\n"); } + if (unlikely(key->external.lower_nv12 || key->external.lower_iyuv)) { + const struct tgsi_token *tokens; + + /* samplers inserted would conflict, but this should be unpossible: */ + assert(!(key->bitmap || key->drawpixels)); + + tokens = st_tgsi_lower_yuv(tgsi.tokens, + ~stfp->Base.SamplersUsed, + key->external.lower_nv12, + key->external.lower_iyuv); + if (tokens) { + if (tgsi.tokens != stfp->tgsi.tokens) + tgsi_free_tokens(tgsi.tokens); + tgsi.tokens = tokens; + } else { + fprintf(stderr, "mesa: cannot create a shader for samplerExternalOES\n"); + } + } + if (ST_DEBUG & DEBUG_TGSI) { tgsi_dump(tgsi.tokens, 0); debug_printf("\n"); @@ -1016,12 +1218,15 @@ st_translate_program_common(struct st_context *st, if (prog->ClipDistanceArraySize) ureg_property(ureg, TGSI_PROPERTY_NUM_CLIPDIST_ENABLED, prog->ClipDistanceArraySize); + if (prog->CullDistanceArraySize) + ureg_property(ureg, TGSI_PROPERTY_NUM_CULLDIST_ENABLED, + prog->CullDistanceArraySize); /* * Convert Mesa program inputs to TGSI input register semantics. */ for (attr = 0; attr < VARYING_SLOT_MAX; attr++) { - if ((prog->InputsRead & BITFIELD64_BIT(attr)) != 0) { + if ((prog->info.inputs_read & BITFIELD64_BIT(attr)) != 0) { const GLuint slot = num_inputs++; inputMapping[attr] = slot; @@ -1029,7 +1234,7 @@ st_translate_program_common(struct st_context *st, switch (attr) { case VARYING_SLOT_PRIMITIVE_ID: - assert(tgsi_processor == TGSI_PROCESSOR_GEOMETRY); + assert(tgsi_processor == PIPE_SHADER_GEOMETRY); input_semantic_name[slot] = TGSI_SEMANTIC_PRIMID; input_semantic_index[slot] = 0; break; @@ -1061,6 +1266,11 @@ st_translate_program_common(struct st_context *st, input_semantic_name[slot] = TGSI_SEMANTIC_CLIPDIST; input_semantic_index[slot] = 1; break; + case VARYING_SLOT_CULL_DIST0: + case VARYING_SLOT_CULL_DIST1: + /* these should have been lowered by GLSL */ + assert(0); + break; case VARYING_SLOT_PSIZ: input_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; input_semantic_index[slot] = 0; @@ -1093,7 +1303,7 @@ st_translate_program_common(struct st_context *st, /* Also add patch inputs. */ for (attr = 0; attr < 32; attr++) { - if (prog->PatchInputsRead & (1 << attr)) { + if (prog->info.patch_inputs_read & (1u << attr)) { GLuint slot = num_inputs++; GLuint patch_attr = VARYING_SLOT_PATCH0 + attr; @@ -1115,7 +1325,7 @@ st_translate_program_common(struct st_context *st, * mapping and the semantic information for each output. */ for (attr = 0; attr < VARYING_SLOT_MAX; attr++) { - if (prog->OutputsWritten & BITFIELD64_BIT(attr)) { + if (prog->info.outputs_written & BITFIELD64_BIT(attr)) { GLuint slot = num_outputs++; outputMapping[attr] = slot; @@ -1163,6 +1373,11 @@ st_translate_program_common(struct st_context *st, output_semantic_name[slot] = TGSI_SEMANTIC_CLIPDIST; output_semantic_index[slot] = 1; break; + case VARYING_SLOT_CULL_DIST0: + case VARYING_SLOT_CULL_DIST1: + /* these should have been lowered by GLSL */ + assert(0); + break; case VARYING_SLOT_LAYER: output_semantic_name[slot] = TGSI_SEMANTIC_LAYER; output_semantic_index[slot] = 0; @@ -1212,7 +1427,7 @@ st_translate_program_common(struct st_context *st, /* Also add patch outputs. */ for (attr = 0; attr < 32; attr++) { - if (prog->PatchOutputsWritten & (1 << attr)) { + if (prog->info.patch_outputs_written & (1u << attr)) { GLuint slot = num_outputs++; GLuint patch_attr = VARYING_SLOT_PATCH0 + attr; @@ -1235,7 +1450,6 @@ st_translate_program_common(struct st_context *st, input_semantic_name, input_semantic_index, NULL, - NULL, /* outputs */ num_outputs, outputMapping, @@ -1271,18 +1485,21 @@ st_translate_geometry_program(struct st_context *st, { struct ureg_program *ureg; - ureg = ureg_create_with_screen(TGSI_PROCESSOR_GEOMETRY, st->pipe->screen); + ureg = ureg_create_with_screen(PIPE_SHADER_GEOMETRY, st->pipe->screen); if (ureg == NULL) return false; - ureg_property(ureg, TGSI_PROPERTY_GS_INPUT_PRIM, stgp->Base.InputType); - ureg_property(ureg, TGSI_PROPERTY_GS_OUTPUT_PRIM, stgp->Base.OutputType); + ureg_property(ureg, TGSI_PROPERTY_GS_INPUT_PRIM, + stgp->Base.info.gs.input_primitive); + ureg_property(ureg, TGSI_PROPERTY_GS_OUTPUT_PRIM, + stgp->Base.info.gs.output_primitive); ureg_property(ureg, TGSI_PROPERTY_GS_MAX_OUTPUT_VERTICES, - stgp->Base.VerticesOut); - ureg_property(ureg, TGSI_PROPERTY_GS_INVOCATIONS, stgp->Base.Invocations); + stgp->Base.info.gs.vertices_out); + ureg_property(ureg, TGSI_PROPERTY_GS_INVOCATIONS, + stgp->Base.info.gs.invocations); - st_translate_program_common(st, &stgp->Base.Base, stgp->glsl_to_tgsi, ureg, - TGSI_PROCESSOR_GEOMETRY, &stgp->tgsi); + st_translate_program_common(st, &stgp->Base, stgp->glsl_to_tgsi, ureg, + PIPE_SHADER_GEOMETRY, &stgp->tgsi); free_glsl_to_tgsi_visitor(stgp->glsl_to_tgsi); stgp->glsl_to_tgsi = NULL; @@ -1355,15 +1572,15 @@ st_translate_tessctrl_program(struct st_context *st, { struct ureg_program *ureg; - ureg = ureg_create_with_screen(TGSI_PROCESSOR_TESS_CTRL, st->pipe->screen); + ureg = ureg_create_with_screen(PIPE_SHADER_TESS_CTRL, st->pipe->screen); if (ureg == NULL) return false; ureg_property(ureg, TGSI_PROPERTY_TCS_VERTICES_OUT, - sttcp->Base.VerticesOut); + sttcp->Base.info.tcs.vertices_out); - st_translate_program_common(st, &sttcp->Base.Base, sttcp->glsl_to_tgsi, - ureg, TGSI_PROCESSOR_TESS_CTRL, &sttcp->tgsi); + st_translate_program_common(st, &sttcp->Base, sttcp->glsl_to_tgsi, ureg, + PIPE_SHADER_TESS_CTRL, &sttcp->tgsi); free_glsl_to_tgsi_visitor(sttcp->glsl_to_tgsi); sttcp->glsl_to_tgsi = NULL; @@ -1380,16 +1597,17 @@ st_translate_tesseval_program(struct st_context *st, { struct ureg_program *ureg; - ureg = ureg_create_with_screen(TGSI_PROCESSOR_TESS_EVAL, st->pipe->screen); + ureg = ureg_create_with_screen(PIPE_SHADER_TESS_EVAL, st->pipe->screen); if (ureg == NULL) return false; - if (sttep->Base.PrimitiveMode == GL_ISOLINES) + if (sttep->Base.info.tes.primitive_mode == GL_ISOLINES) ureg_property(ureg, TGSI_PROPERTY_TES_PRIM_MODE, GL_LINES); else - ureg_property(ureg, TGSI_PROPERTY_TES_PRIM_MODE, sttep->Base.PrimitiveMode); + ureg_property(ureg, TGSI_PROPERTY_TES_PRIM_MODE, + sttep->Base.info.tes.primitive_mode); - switch (sttep->Base.Spacing) { + switch (sttep->Base.info.tes.spacing) { case GL_EQUAL: ureg_property(ureg, TGSI_PROPERTY_TES_SPACING, PIPE_TESS_SPACING_EQUAL); break; @@ -1406,11 +1624,12 @@ st_translate_tesseval_program(struct st_context *st, } ureg_property(ureg, TGSI_PROPERTY_TES_VERTEX_ORDER_CW, - sttep->Base.VertexOrder == GL_CW); - ureg_property(ureg, TGSI_PROPERTY_TES_POINT_MODE, sttep->Base.PointMode); + sttep->Base.info.tes.vertex_order == GL_CW); + ureg_property(ureg, TGSI_PROPERTY_TES_POINT_MODE, + sttep->Base.info.tes.point_mode); - st_translate_program_common(st, &sttep->Base.Base, sttep->glsl_to_tgsi, - ureg, TGSI_PROCESSOR_TESS_EVAL, &sttep->tgsi); + st_translate_program_common(st, &sttep->Base, sttep->glsl_to_tgsi, + ureg, PIPE_SHADER_TESS_EVAL, &sttep->tgsi); free_glsl_to_tgsi_visitor(sttep->glsl_to_tgsi); sttep->glsl_to_tgsi = NULL; @@ -1428,15 +1647,16 @@ st_translate_compute_program(struct st_context *st, struct ureg_program *ureg; struct pipe_shader_state prog; - ureg = ureg_create_with_screen(TGSI_PROCESSOR_COMPUTE, st->pipe->screen); + ureg = ureg_create_with_screen(PIPE_SHADER_COMPUTE, st->pipe->screen); if (ureg == NULL) return false; - st_translate_program_common(st, &stcp->Base.Base, stcp->glsl_to_tgsi, ureg, - TGSI_PROCESSOR_COMPUTE, &prog); + st_translate_program_common(st, &stcp->Base, stcp->glsl_to_tgsi, ureg, + PIPE_SHADER_COMPUTE, &prog); + stcp->tgsi.ir_type = PIPE_SHADER_IR_TGSI; stcp->tgsi.prog = prog.tokens; - stcp->tgsi.req_local_mem = stcp->Base.SharedSize; + stcp->tgsi.req_local_mem = stcp->Base.info.cs.shared_size; stcp->tgsi.req_private_mem = 0; stcp->tgsi.req_input_mem = 0; @@ -1593,10 +1813,6 @@ destroy_shader_program_variants_cb(GLuint key, void *data, void *userData) struct gl_shader_program *shProg = (struct gl_shader_program *) data; GLuint i; - for (i = 0; i < shProg->NumShaders; i++) { - destroy_program_variants(st, shProg->Shaders[i]->Program); - } - for (i = 0; i < ARRAY_SIZE(shProg->_LinkedShaders); i++) { if (shProg->_LinkedShaders[i]) destroy_program_variants(st, shProg->_LinkedShaders[i]->Program); @@ -1609,9 +1825,6 @@ destroy_shader_program_variants_cb(GLuint key, void *data, void *userData) case GL_TESS_CONTROL_SHADER: case GL_TESS_EVALUATION_SHADER: case GL_COMPUTE_SHADER: - { - destroy_program_variants(st, shader->Program); - } break; default: assert(0); @@ -1669,7 +1882,7 @@ st_print_current_vertex_program(void) (struct st_vertex_program *) ctx->VertexProgram._Current; struct st_vp_variant *stv; - debug_printf("Vertex program %u\n", stvp->Base.Base.Id); + debug_printf("Vertex program %u\n", stvp->Base.Id); for (stv = stvp->variants; stv; stv = stv->next) { debug_printf("variant %p\n", stv);