diff options
Diffstat (limited to 'mesalib/src')
21 files changed, 1178 insertions, 291 deletions
diff --git a/mesalib/src/gallium/auxiliary/util/u_debug.c b/mesalib/src/gallium/auxiliary/util/u_debug.c index 004df439f..2d6193039 100644 --- a/mesalib/src/gallium/auxiliary/util/u_debug.c +++ b/mesalib/src/gallium/auxiliary/util/u_debug.c @@ -730,7 +730,7 @@ debug_dump_float_rgba_bmp(const char *filename, pixel.rgbRed = float_to_ubyte(ptr[x*4 + 0]); pixel.rgbGreen = float_to_ubyte(ptr[x*4 + 1]); pixel.rgbBlue = float_to_ubyte(ptr[x*4 + 2]); - pixel.rgbAlpha = 255; + pixel.rgbAlpha = float_to_ubyte(ptr[x*4 + 3]); os_stream_write(stream, &pixel, 4); } } diff --git a/mesalib/src/gallium/auxiliary/util/u_format.csv b/mesalib/src/gallium/auxiliary/util/u_format.csv index 347e2beb8..a3d2aae62 100644 --- a/mesalib/src/gallium/auxiliary/util/u_format.csv +++ b/mesalib/src/gallium/auxiliary/util/u_format.csv @@ -260,10 +260,10 @@ PIPE_FORMAT_R10G10B10X2_USCALED , plain, 1, 1, u10 , u10 , u10 , x2 , xyz1, r # A.k.a. D3DDECLTYPE_DEC3N PIPE_FORMAT_R10G10B10X2_SNORM , plain, 1, 1, sn10, sn10, sn10 , x2 , xyz1, rgb -PIPE_FORMAT_YV12 , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv -PIPE_FORMAT_YV16 , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv -PIPE_FORMAT_IYUV , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv -PIPE_FORMAT_NV12 , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv -PIPE_FORMAT_NV21 , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv -PIPE_FORMAT_IA44 , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv -PIPE_FORMAT_AI44 , subsampled, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_YV12 , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_YV16 , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_IYUV , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_NV12 , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_NV21 , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_IA44 , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv +PIPE_FORMAT_AI44 , other, 1, 1, x8 , x8 , x8 , x8 , xyzw, yuv diff --git a/mesalib/src/glsl/Makefile b/mesalib/src/glsl/Makefile index d1422c2a4..005b51d72 100644 --- a/mesalib/src/glsl/Makefile +++ b/mesalib/src/glsl/Makefile @@ -89,18 +89,32 @@ CXX_SOURCES = \ LIBS = \ $(TOP)/src/glsl/libglsl.a -APPS = glsl_compiler glcpp/glcpp +APPS = glsl_compiler glsl_test glcpp/glcpp GLSL2_C_SOURCES = \ ../mesa/program/hash_table.c \ ../mesa/program/symbol_table.c GLSL2_CXX_SOURCES = \ - main.cpp + main.cpp \ + standalone_scaffolding.cpp GLSL2_OBJECTS = \ $(GLSL2_C_SOURCES:.c=.o) \ $(GLSL2_CXX_SOURCES:.cpp=.o) +TEST_C_SOURCES = \ + ../mesa/program/hash_table.c \ + ../mesa/program/symbol_table.c + +TEST_CXX_SOURCES = \ + standalone_scaffolding.cpp \ + test.cpp \ + test_optpass.cpp + +TEST_OBJECTS = \ + $(TEST_C_SOURCES:.c=.o) \ + $(TEST_CXX_SOURCES:.cpp=.o) + ### Basic defines ### DEFINES += \ @@ -129,7 +143,9 @@ ALL_SOURCES = \ $(C_SOURCES) \ $(CXX_SOURCES) \ $(GLSL2_CXX_SOURCES) \ - $(GLSL2_C_SOURCES) + $(GLSL2_C_SOURCES) \ + $(TEST_CXX_SOURCES) \ + $(TEST_C_SOURCES) ##### TARGETS ##### @@ -151,7 +167,7 @@ depend: $(ALL_SOURCES) Makefile # Remove .o and backup files clean: clean-dricore - rm -f $(GLCPP_OBJECTS) $(GLSL2_OBJECTS) $(OBJECTS) lib$(LIBNAME).a depend depend.bak builtin_function.cpp builtin_function.o builtin_stubs.o builtin_compiler + rm -f $(GLCPP_OBJECTS) $(GLSL2_OBJECTS) $(TEST_OBJECTS) $(OBJECTS) lib$(LIBNAME).a depend depend.bak builtin_function.cpp builtin_function.o builtin_stubs.o builtin_compiler -rm -f $(APPS) clean-dricore: @@ -174,6 +190,9 @@ install-dricore: default glsl_compiler: $(GLSL2_OBJECTS) libglsl.a builtin_stubs.o $(APP_CXX) $(INCLUDES) $(CFLAGS) $(LDFLAGS) $(GLSL2_OBJECTS) builtin_stubs.o $(LIBS) -o $@ +glsl_test: $(TEST_OBJECTS) libglsl.a builtin_stubs.o + $(APP_CXX) $(INCLUDES) $(CFLAGS) $(LDFLAGS) $(TEST_OBJECTS) builtin_stubs.o $(LIBS) -o $@ + glcpp: glcpp/glcpp glcpp/glcpp: $(GLCPP_OBJECTS) $(APP_CC) $(INCLUDES) $(CFLAGS) $(LDFLAGS) $(GLCPP_OBJECTS) -o $@ diff --git a/mesalib/src/glsl/SConscript b/mesalib/src/glsl/SConscript index ea104abb8..1da58a91f 100644 --- a/mesalib/src/glsl/SConscript +++ b/mesalib/src/glsl/SConscript @@ -96,6 +96,7 @@ glsl_sources = [ 'opt_tree_grafting.cpp', 'ralloc.c', 's_expression.cpp', + 'standalone_scaffolding.cpp', 'strtod.c', ] diff --git a/mesalib/src/glsl/ir_optimization.h b/mesalib/src/glsl/ir_optimization.h index 59a040751..f7808bdda 100644 --- a/mesalib/src/glsl/ir_optimization.h +++ b/mesalib/src/glsl/ir_optimization.h @@ -69,3 +69,7 @@ bool lower_variable_index_to_cond_assign(exec_list *instructions, bool lower_input, bool lower_output, bool lower_temp, bool lower_uniform); bool lower_quadop_vector(exec_list *instructions, bool dont_lower_swz); bool optimize_redundant_jumps(exec_list *instructions); + +ir_rvalue * +compare_index_block(exec_list *instructions, ir_variable *index, + unsigned base, unsigned components, void *mem_ctx); diff --git a/mesalib/src/glsl/lower_variable_index_to_cond_assign.cpp b/mesalib/src/glsl/lower_variable_index_to_cond_assign.cpp index 8eb1612f0..f8e4a1de4 100644 --- a/mesalib/src/glsl/lower_variable_index_to_cond_assign.cpp +++ b/mesalib/src/glsl/lower_variable_index_to_cond_assign.cpp @@ -29,6 +29,21 @@ * * Pre-DX10 GPUs often don't have a native way to do this operation, * and this works around that. + * + * The lowering process proceeds as follows. Each non-constant index + * found in an r-value is converted to a canonical form \c array[i]. Each + * element of the array is conditionally assigned to a temporary by comparing + * \c i to a constant index. This is done by cloning the canonical form and + * replacing all occurances of \c i with a constant. Each remaining occurance + * of the canonical form in the IR is replaced with a dereference of the + * temporary variable. + * + * L-values with non-constant indices are handled similarly. In this case, + * the RHS of the assignment is assigned to a temporary. The non-constant + * index is replace with the canonical form (just like for r-values). The + * temporary is conditionally assigned to each element of the canonical form + * by comparing \c i with each index. The same clone-and-replace scheme is + * used. */ #include "ir.h" @@ -37,10 +52,140 @@ #include "glsl_types.h" #include "main/macros.h" +/** + * Generate a comparison value for a block of indices + * + * Lowering passes for non-constant indexing of arrays, matrices, or vectors + * can use this to generate blocks of index comparison values. + * + * \param instructions List where new instructions will be appended + * \param index \c ir_variable containing the desired index + * \param base Base value for this block of comparisons + * \param components Number of unique index values to compare. This must + * be on the range [1, 4]. + * \param mem_ctx ralloc memory context to be used for all allocations. + * + * \returns + * An \c ir_rvalue that \b must be cloned for each use in conditional + * assignments, etc. + */ +ir_rvalue * +compare_index_block(exec_list *instructions, ir_variable *index, + unsigned base, unsigned components, void *mem_ctx) +{ + ir_rvalue *broadcast_index = new(mem_ctx) ir_dereference_variable(index); + + assert(index->type->is_scalar()); + assert(index->type->base_type == GLSL_TYPE_INT); + assert(components >= 1 && components <= 4); + + if (components > 1) { + const ir_swizzle_mask m = { 0, 0, 0, 0, components, false }; + broadcast_index = new(mem_ctx) ir_swizzle(broadcast_index, m); + } + + /* Compare the desired index value with the next block of four indices. + */ + ir_constant_data test_indices_data; + memset(&test_indices_data, 0, sizeof(test_indices_data)); + test_indices_data.i[0] = base; + test_indices_data.i[1] = base + 1; + test_indices_data.i[2] = base + 2; + test_indices_data.i[3] = base + 3; + + ir_constant *const test_indices = + new(mem_ctx) ir_constant(broadcast_index->type, + &test_indices_data); + + ir_rvalue *const condition_val = + new(mem_ctx) ir_expression(ir_binop_equal, + &glsl_type::bool_type[components - 1], + broadcast_index, + test_indices); + + ir_variable *const condition = + new(mem_ctx) ir_variable(condition_val->type, + "dereference_condition", + ir_var_temporary); + instructions->push_tail(condition); + + ir_rvalue *const cond_deref = + new(mem_ctx) ir_dereference_variable(condition); + instructions->push_tail(new(mem_ctx) ir_assignment(cond_deref, condition_val, 0)); + + return cond_deref; +} + +static inline bool +is_array_or_matrix(const ir_instruction *ir) +{ + return (ir->type->is_array() || ir->type->is_matrix()); +} + +/** + * Replace a dereference of a variable with a specified r-value + * + * Each time a dereference of the specified value is replaced, the r-value + * tree is cloned. + */ +class deref_replacer : public ir_rvalue_visitor { +public: + deref_replacer(const ir_variable *variable_to_replace, ir_rvalue *value) + : variable_to_replace(variable_to_replace), value(value), + progress(false) + { + assert(this->variable_to_replace != NULL); + assert(this->value != NULL); + } + + virtual void handle_rvalue(ir_rvalue **rvalue) + { + ir_dereference_variable *const dv = (*rvalue)->as_dereference_variable(); + + if ((dv != NULL) && (dv->var == this->variable_to_replace)) { + this->progress = true; + *rvalue = this->value->clone(ralloc_parent(*rvalue), NULL); + } + } + + const ir_variable *variable_to_replace; + ir_rvalue *value; + bool progress; +}; + +/** + * Find a variable index dereference of an array in an rvalue tree + */ +class find_variable_index : public ir_hierarchical_visitor { +public: + find_variable_index() + : deref(NULL) + { + /* empty */ + } + + virtual ir_visitor_status visit_enter(ir_dereference_array *ir) + { + if (is_array_or_matrix(ir->array) + && (ir->array_index->as_constant() == NULL)) { + this->deref = ir; + return visit_stop; + } + + return visit_continue; + } + + /** + * First array dereference found in the tree that has a non-constant index. + */ + ir_dereference_array *deref; +}; + struct assignment_generator { ir_instruction* base_ir; - ir_rvalue* array; + ir_dereference *rvalue; + ir_variable *old_index; bool is_write; unsigned int write_mask; ir_variable* var; @@ -55,18 +200,23 @@ struct assignment_generator * underlying variable. */ void *mem_ctx = ralloc_parent(base_ir); - ir_dereference *element = - new(mem_ctx) ir_dereference_array(this->array->clone(mem_ctx, NULL), - new(mem_ctx) ir_constant(i)); - ir_rvalue *variable = new(mem_ctx) ir_dereference_variable(this->var); - ir_assignment *assignment; - if (is_write) { - assignment = new(mem_ctx) ir_assignment(element, variable, condition, - write_mask); - } else { - assignment = new(mem_ctx) ir_assignment(variable, element, condition); - } + /* Clone the old r-value in its entirety. Then replace any occurances of + * the old variable index with the new constant index. + */ + ir_dereference *element = this->rvalue->clone(mem_ctx, NULL); + ir_constant *const index = new(mem_ctx) ir_constant(i); + deref_replacer r(this->old_index, index); + element->accept(&r); + assert(r.progress); + + /* Generate a conditional assignment to (or from) the constant indexed + * array dereference. + */ + ir_rvalue *variable = new(mem_ctx) ir_dereference_variable(this->var); + ir_assignment *const assignment = (is_write) + ? new(mem_ctx) ir_assignment(element, variable, condition, write_mask) + : new(mem_ctx) ir_assignment(variable, element, condition); list->push_tail(assignment); } @@ -118,54 +268,17 @@ struct switch_generator for (unsigned i = first; i < end; i += 4) { const unsigned comps = MIN2(condition_components, end - i); - ir_rvalue *broadcast_index = - new(this->mem_ctx) ir_dereference_variable(index); - - if (comps) { - const ir_swizzle_mask m = { 0, 0, 0, 0, comps, false }; - broadcast_index = new(this->mem_ctx) ir_swizzle(broadcast_index, m); - } - - /* Compare the desired index value with the next block of four indices. - */ - ir_constant_data test_indices_data; - memset(&test_indices_data, 0, sizeof(test_indices_data)); - test_indices_data.i[0] = i; - test_indices_data.i[1] = i + 1; - test_indices_data.i[2] = i + 2; - test_indices_data.i[3] = i + 3; - ir_constant *const test_indices = - new(this->mem_ctx) ir_constant(broadcast_index->type, - &test_indices_data); - - ir_rvalue *const condition_val = - new(this->mem_ctx) ir_expression(ir_binop_equal, - &glsl_type::bool_type[comps - 1], - broadcast_index, - test_indices); - - ir_variable *const condition = - new(this->mem_ctx) ir_variable(condition_val->type, - "dereference_array_condition", - ir_var_temporary); - list->push_tail(condition); - ir_rvalue *const cond_deref = - new(this->mem_ctx) ir_dereference_variable(condition); - list->push_tail(new(this->mem_ctx) ir_assignment(cond_deref, - condition_val, 0)); + compare_index_block(list, index, i, comps, this->mem_ctx); if (comps == 1) { - ir_rvalue *const cond_deref = - new(this->mem_ctx) ir_dereference_variable(condition); - - this->generator.generate(i, cond_deref, list); + this->generator.generate(i, cond_deref->clone(this->mem_ctx, NULL), + list); } else { for (unsigned j = 0; j < comps; j++) { - ir_rvalue *const cond_deref = - new(this->mem_ctx) ir_dereference_variable(condition); ir_rvalue *const cond_swiz = - new(this->mem_ctx) ir_swizzle(cond_deref, j, 0, 0, 0, 1); + new(this->mem_ctx) ir_swizzle(cond_deref->clone(this->mem_ctx, NULL), + j, 0, 0, 0, 1); this->generator.generate(i + j, cond_swiz, list); } @@ -233,21 +346,18 @@ public: bool lower_temps; bool lower_uniforms; - bool is_array_or_matrix(const ir_instruction *ir) const - { - return (ir->type->is_array() || ir->type->is_matrix()); - } - - bool needs_lowering(ir_dereference_array *deref) const + bool storage_type_needs_lowering(ir_dereference_array *deref) const { - if (deref == NULL || deref->array_index->as_constant() - || !is_array_or_matrix(deref->array)) - return false; - - if (deref->array->ir_type == ir_type_constant) + /* If a variable isn't eventually the target of this dereference, then + * it must be a constant or some sort of anonymous temporary storage. + * + * FINISHME: Is this correct? Most drivers treat arrays of constants as + * FINISHME: uniforms. It seems like this should do the same. + */ + const ir_variable *const var = deref->array->variable_referenced(); + if (var == NULL) return this->lower_temps; - const ir_variable *const var = deref->array->variable_referenced(); switch (var->mode) { case ir_var_auto: case ir_var_temporary: @@ -267,8 +377,18 @@ public: return false; } + bool needs_lowering(ir_dereference_array *deref) const + { + if (deref == NULL || deref->array_index->as_constant() + || !is_array_or_matrix(deref->array)) + return false; + + return this->storage_type_needs_lowering(deref); + } + ir_variable *convert_dereference_array(ir_dereference_array *orig_deref, - ir_assignment* orig_assign) + ir_assignment* orig_assign, + ir_dereference *orig_base) { assert(is_array_or_matrix(orig_deref->array)); @@ -314,9 +434,12 @@ public: new(mem_ctx) ir_assignment(lhs, orig_deref->array_index, NULL); base_ir->insert_before(assign); + orig_deref->array_index = lhs->clone(mem_ctx, NULL); + assignment_generator ag; - ag.array = orig_deref->array; + ag.rvalue = orig_base; ag.base_ir = base_ir; + ag.old_index = index; ag.var = var; if (orig_assign) { ag.is_write = true; @@ -327,21 +450,40 @@ public: switch_generator sg(ag, index, 4, 4); - exec_list list; - sg.generate(0, length, &list); - base_ir->insert_before(&list); + /* If the original assignment has a condition, respect that original + * condition! This is acomplished by wrapping the new conditional + * assignments in an if-statement that uses the original condition. + */ + if ((orig_assign != NULL) && (orig_assign->condition != NULL)) { + /* No need to clone the condition because the IR that it hangs on is + * going to be removed from the instruction sequence. + */ + ir_if *if_stmt = new(mem_ctx) ir_if(orig_assign->condition); + + sg.generate(0, length, &if_stmt->then_instructions); + base_ir->insert_before(if_stmt); + } else { + exec_list list; + + sg.generate(0, length, &list); + base_ir->insert_before(&list); + } return var; } virtual void handle_rvalue(ir_rvalue **pir) { + if (this->in_assignee) + return; + if (!*pir) return; ir_dereference_array* orig_deref = (*pir)->as_dereference_array(); if (needs_lowering(orig_deref)) { - ir_variable* var = convert_dereference_array(orig_deref, 0); + ir_variable *var = + convert_dereference_array(orig_deref, NULL, orig_deref); assert(var); *pir = new(ralloc_parent(base_ir)) ir_dereference_variable(var); this->progress = true; @@ -353,10 +495,11 @@ public: { ir_rvalue_visitor::visit_leave(ir); - ir_dereference_array *orig_deref = ir->lhs->as_dereference_array(); + find_variable_index f; + ir->lhs->accept(&f); - if (needs_lowering(orig_deref)) { - convert_dereference_array(orig_deref, ir); + if ((f.deref != NULL) && storage_type_needs_lowering(f.deref)) { + convert_dereference_array(f.deref, ir, ir->lhs); ir->remove(); this->progress = true; } @@ -377,7 +520,17 @@ lower_variable_index_to_cond_assign(exec_list *instructions, lower_temp, lower_uniform); - visit_list_elements(&v, instructions); - - return v.progress; + /* Continue lowering until no progress is made. If there are multiple + * levels of indirection (e.g., non-constant indexing of array elements and + * matrix columns of an array of matrix), each pass will only lower one + * level of indirection. + */ + bool progress_ever = false; + do { + v.progress = false; + visit_list_elements(&v, instructions); + progress_ever = v.progress || progress_ever; + } while (v.progress); + + return progress_ever; } diff --git a/mesalib/src/glsl/lower_vec_index_to_cond_assign.cpp b/mesalib/src/glsl/lower_vec_index_to_cond_assign.cpp index 3c4d93201..fce9c3424 100644 --- a/mesalib/src/glsl/lower_vec_index_to_cond_assign.cpp +++ b/mesalib/src/glsl/lower_vec_index_to_cond_assign.cpp @@ -71,8 +71,6 @@ ir_vec_index_to_cond_assign_visitor::convert_vec_index_to_cond_assign(ir_rvalue ir_assignment *assign; ir_variable *index, *var; ir_dereference *deref; - ir_expression *condition; - ir_swizzle *swizzle; int i; if (!orig_deref) @@ -86,39 +84,52 @@ ir_vec_index_to_cond_assign_visitor::convert_vec_index_to_cond_assign(ir_rvalue assert(orig_deref->array_index->type->base_type == GLSL_TYPE_INT); + exec_list list; + /* Store the index to a temporary to avoid reusing its tree. */ index = new(base_ir) ir_variable(glsl_type::int_type, "vec_index_tmp_i", ir_var_temporary); - base_ir->insert_before(index); + list.push_tail(index); deref = new(base_ir) ir_dereference_variable(index); assign = new(base_ir) ir_assignment(deref, orig_deref->array_index, NULL); - base_ir->insert_before(assign); + list.push_tail(assign); /* Temporary where we store whichever value we swizzle out. */ var = new(base_ir) ir_variable(ir->type, "vec_index_tmp_v", ir_var_temporary); - base_ir->insert_before(var); + list.push_tail(var); + + /* Generate a single comparison condition "mask" for all of the components + * in the vector. + */ + ir_rvalue *const cond_deref = + compare_index_block(&list, index, 0, + orig_deref->array->type->vector_elements, + mem_ctx); /* Generate a conditional move of each vector element to the temp. */ for (i = 0; i < orig_deref->array->type->vector_elements; i++) { - deref = new(base_ir) ir_dereference_variable(index); - condition = new(base_ir) ir_expression(ir_binop_equal, - glsl_type::bool_type, - deref, - new(base_ir) ir_constant(i)); + ir_rvalue *condition_swizzle = + new(base_ir) ir_swizzle(cond_deref->clone(ir, NULL), i, 0, 0, 0, 1); /* Just clone the rest of the deref chain when trying to get at the * underlying variable. */ - swizzle = new(base_ir) ir_swizzle(orig_deref->array->clone(mem_ctx, NULL), - i, 0, 0, 0, 1); + ir_rvalue *swizzle = + new(base_ir) ir_swizzle(orig_deref->array->clone(mem_ctx, NULL), + i, 0, 0, 0, 1); deref = new(base_ir) ir_dereference_variable(var); - assign = new(base_ir) ir_assignment(deref, swizzle, condition); - base_ir->insert_before(assign); + assign = new(base_ir) ir_assignment(deref, swizzle, condition_swizzle); + list.push_tail(assign); } + /* Put all of the new instructions in the IR stream before the old + * instruction. + */ + base_ir->insert_before(&list); + this->progress = true; return new(base_ir) ir_dereference_variable(var); } @@ -171,42 +182,66 @@ ir_vec_index_to_cond_assign_visitor::visit_leave(ir_assignment *ir) assert(orig_deref->array_index->type->base_type == GLSL_TYPE_INT); + exec_list list; + /* Store the index to a temporary to avoid reusing its tree. */ index = new(ir) ir_variable(glsl_type::int_type, "vec_index_tmp_i", ir_var_temporary); - ir->insert_before(index); + list.push_tail(index); deref = new(ir) ir_dereference_variable(index); assign = new(ir) ir_assignment(deref, orig_deref->array_index, NULL); - ir->insert_before(assign); + list.push_tail(assign); /* Store the RHS to a temporary to avoid reusing its tree. */ var = new(ir) ir_variable(ir->rhs->type, "vec_index_tmp_v", ir_var_temporary); - ir->insert_before(var); + list.push_tail(var); deref = new(ir) ir_dereference_variable(var); assign = new(ir) ir_assignment(deref, ir->rhs, NULL); - ir->insert_before(assign); + list.push_tail(assign); + + /* Generate a single comparison condition "mask" for all of the components + * in the vector. + */ + ir_rvalue *const cond_deref = + compare_index_block(&list, index, 0, + orig_deref->array->type->vector_elements, + mem_ctx); /* Generate a conditional move of each vector element to the temp. */ for (i = 0; i < orig_deref->array->type->vector_elements; i++) { - ir_rvalue *condition, *swizzle; + ir_rvalue *condition_swizzle = + new(ir) ir_swizzle(cond_deref->clone(ir, NULL), i, 0, 0, 0, 1); - deref = new(ir) ir_dereference_variable(index); - condition = new(ir) ir_expression(ir_binop_equal, - glsl_type::bool_type, - deref, - new(ir) ir_constant(i)); /* Just clone the rest of the deref chain when trying to get at the * underlying variable. */ - swizzle = new(ir) ir_swizzle(orig_deref->array->clone(mem_ctx, NULL), - i, 0, 0, 0, 1); + ir_rvalue *swizzle = + new(ir) ir_swizzle(orig_deref->array->clone(mem_ctx, NULL), + i, 0, 0, 0, 1); deref = new(ir) ir_dereference_variable(var); - assign = new(ir) ir_assignment(swizzle, deref, condition); - ir->insert_before(assign); + assign = new(ir) ir_assignment(swizzle, deref, condition_swizzle); + list.push_tail(assign); } + + /* If the original assignment has a condition, respect that original + * condition! This is acomplished by wrapping the new conditional + * assignments in an if-statement that uses the original condition. + */ + if (ir->condition != NULL) { + /* No need to clone the condition because the IR that it hangs on is + * going to be removed from the instruction sequence. + */ + ir_if *if_stmt = new(mem_ctx) ir_if(ir->condition); + + list.move_nodes_to(&if_stmt->then_instructions); + ir->insert_before(if_stmt); + } else { + ir->insert_before(&list); + } + ir->remove(); this->progress = true; diff --git a/mesalib/src/glsl/main.cpp b/mesalib/src/glsl/main.cpp index 7952bb1a3..9f85096e1 100644 --- a/mesalib/src/glsl/main.cpp +++ b/mesalib/src/glsl/main.cpp @@ -29,80 +29,26 @@ #include "ir_print_visitor.h" #include "program.h" #include "loop_analysis.h" - -extern "C" struct gl_shader * -_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type); - -extern "C" void -_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr, - struct gl_shader *sh); - -/* Copied from shader_api.c for the stand-alone compiler. - */ -void -_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr, - struct gl_shader *sh) -{ - *ptr = sh; -} - -struct gl_shader * -_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type) -{ - struct gl_shader *shader; - - (void) ctx; - - assert(type == GL_FRAGMENT_SHADER || type == GL_VERTEX_SHADER); - shader = rzalloc(NULL, struct gl_shader); - if (shader) { - shader->Type = type; - shader->Name = name; - shader->RefCount = 1; - } - return shader; -} +#include "standalone_scaffolding.h" static void initialize_context(struct gl_context *ctx, gl_api api) { - memset(ctx, 0, sizeof(*ctx)); - - ctx->API = api; - - ctx->Extensions.ARB_ES2_compatibility = GL_TRUE; - ctx->Extensions.ARB_draw_buffers = GL_TRUE; - ctx->Extensions.ARB_draw_instanced = GL_TRUE; - ctx->Extensions.ARB_fragment_coord_conventions = GL_TRUE; - ctx->Extensions.EXT_texture_array = GL_TRUE; - ctx->Extensions.NV_texture_rectangle = GL_TRUE; - ctx->Extensions.EXT_texture3D = GL_TRUE; + initialize_context_to_defaults(ctx, api); /* GLSL 1.30 isn't fully supported, but we need to advertise 1.30 so that * the built-in functions for 1.30 can be built. */ ctx->Const.GLSLVersion = 130; - /* 1.10 minimums. */ - ctx->Const.MaxLights = 8; ctx->Const.MaxClipPlanes = 8; - ctx->Const.MaxTextureUnits = 2; + ctx->Const.MaxDrawBuffers = 2; /* More than the 1.10 minimum to appease parser tests taken from * apps that (hopefully) already checked the number of coords. */ ctx->Const.MaxTextureCoordUnits = 4; - ctx->Const.VertexProgram.MaxAttribs = 16; - ctx->Const.VertexProgram.MaxUniformComponents = 512; - ctx->Const.MaxVarying = 8; - ctx->Const.MaxVertexTextureImageUnits = 0; - ctx->Const.MaxCombinedTextureImageUnits = 2; - ctx->Const.MaxTextureImageUnits = 2; - ctx->Const.FragmentProgram.MaxUniformComponents = 64; - - ctx->Const.MaxDrawBuffers = 2; - ctx->Driver.NewShader = _mesa_new_shader; } diff --git a/mesalib/src/glsl/standalone_scaffolding.cpp b/mesalib/src/glsl/standalone_scaffolding.cpp new file mode 100644 index 000000000..696ea757e --- /dev/null +++ b/mesalib/src/glsl/standalone_scaffolding.cpp @@ -0,0 +1,91 @@ +/* + * Copyright © 2011 Intel Corporation + * + * 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, sublicense, + * 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 NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + */ + +/* This file declares stripped-down versions of functions that + * normally exist outside of the glsl folder, so that they can be used + * when running the GLSL compiler standalone (for unit testing or + * compiling builtins). + */ + +#include "standalone_scaffolding.h" + +#include <assert.h> +#include <string.h> +#include "ralloc.h" + +void +_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr, + struct gl_shader *sh) +{ + *ptr = sh; +} + +struct gl_shader * +_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type) +{ + struct gl_shader *shader; + + (void) ctx; + + assert(type == GL_FRAGMENT_SHADER || type == GL_VERTEX_SHADER); + shader = rzalloc(NULL, struct gl_shader); + if (shader) { + shader->Type = type; + shader->Name = name; + shader->RefCount = 1; + } + return shader; +} + +void initialize_context_to_defaults(struct gl_context *ctx, gl_api api) +{ + memset(ctx, 0, sizeof(*ctx)); + + ctx->API = api; + + ctx->Extensions.ARB_ES2_compatibility = true; + ctx->Extensions.ARB_draw_buffers = true; + ctx->Extensions.ARB_draw_instanced = true; + ctx->Extensions.ARB_fragment_coord_conventions = true; + ctx->Extensions.EXT_texture_array = true; + ctx->Extensions.NV_texture_rectangle = true; + ctx->Extensions.EXT_texture3D = true; + + ctx->Const.GLSLVersion = 120; + + /* 1.20 minimums. */ + ctx->Const.MaxLights = 8; + ctx->Const.MaxClipPlanes = 6; + ctx->Const.MaxTextureUnits = 2; + ctx->Const.MaxTextureCoordUnits = 2; + ctx->Const.VertexProgram.MaxAttribs = 16; + + ctx->Const.VertexProgram.MaxUniformComponents = 512; + ctx->Const.MaxVarying = 8; /* == gl_MaxVaryingFloats / 4 */ + ctx->Const.MaxVertexTextureImageUnits = 0; + ctx->Const.MaxCombinedTextureImageUnits = 2; + ctx->Const.MaxTextureImageUnits = 2; + ctx->Const.FragmentProgram.MaxUniformComponents = 64; + + ctx->Const.MaxDrawBuffers = 1; +} diff --git a/mesalib/src/glsl/standalone_scaffolding.h b/mesalib/src/glsl/standalone_scaffolding.h new file mode 100644 index 000000000..877332006 --- /dev/null +++ b/mesalib/src/glsl/standalone_scaffolding.h @@ -0,0 +1,54 @@ +/* + * Copyright © 2011 Intel Corporation + * + * 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, sublicense, + * 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 NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + */ + +/* This file declares stripped-down versions of functions that + * normally exist outside of the glsl folder, so that they can be used + * when running the GLSL compiler standalone (for unit testing or + * compiling builtins). + */ + +#pragma once +#ifndef STANDALONE_SCAFFOLDING_H +#define STANDALONE_SCAFFOLDING_H + +#include "main/mtypes.h" + +extern "C" void +_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr, + struct gl_shader *sh); + +extern "C" struct gl_shader * +_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type); + +/** + * Initialize the given gl_context structure to a reasonable set of + * defaults representing the minimum capabilities required by the + * OpenGL spec. + * + * This is used when compiling builtin functions and in testing, when + * we don't have a connection to an actual driver. + */ +void initialize_context_to_defaults(struct gl_context *ctx, gl_api api); + + +#endif /* STANDALONE_SCAFFOLDING_H */ diff --git a/mesalib/src/glsl/test.cpp b/mesalib/src/glsl/test.cpp new file mode 100644 index 000000000..b1ff92ed1 --- /dev/null +++ b/mesalib/src/glsl/test.cpp @@ -0,0 +1,78 @@ +/* + * Copyright © 2011 Intel Corporation + * + * 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, sublicense, + * 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 NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + */ + +/** + * \file test.cpp + * + * Standalone tests for the GLSL compiler. + * + * This file provides a standalone executable which can be used to + * test components of the GLSL. + * + * Each test is a function with the same signature as main(). The + * main function interprets its first argument as the name of the test + * to run, strips out that argument, and then calls the test function. + */ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "test_optpass.h" + +/** + * Print proper usage and exit with failure. + */ +static void +usage_fail(const char *name) +{ + printf("*** usage: %s <command> <options>\n", name); + printf("\n"); + printf("Possible commands are:\n"); + printf(" optpass: test an optimization pass in isolation\n"); + exit(EXIT_FAILURE); +} + +static const char *extract_command_from_argv(int *argc, char **argv) +{ + if (*argc < 2) { + usage_fail(argv[0]); + } + const char *command = argv[1]; + --*argc; + memmove(&argv[1], &argv[2], (*argc) * sizeof(argv[1])); + return command; +} + +int main(int argc, char **argv) +{ + const char *command = extract_command_from_argv(&argc, argv); + if (strcmp(command, "optpass") == 0) { + return test_optpass(argc, argv); + } else { + usage_fail(argv[0]); + } + + /* Execution should never reach here. */ + return EXIT_FAILURE; +} diff --git a/mesalib/src/glsl/test_optpass.cpp b/mesalib/src/glsl/test_optpass.cpp new file mode 100644 index 000000000..89b7f8338 --- /dev/null +++ b/mesalib/src/glsl/test_optpass.cpp @@ -0,0 +1,273 @@ +/* + * Copyright © 2011 Intel Corporation + * + * 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, sublicense, + * 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 NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + */ + +/** + * \file test_optpass.cpp + * + * Standalone test for optimization passes. + * + * This file provides the "optpass" command for the standalone + * glsl_test app. It accepts either GLSL or high-level IR as input, + * and performs the optimiation passes specified on the command line. + * It outputs the IR, both before and after optimiations. + */ + +#include <string> +#include <iostream> +#include <sstream> +#include <getopt.h> + +#include "ast.h" +#include "ir_optimization.h" +#include "ir_print_visitor.h" +#include "program.h" +#include "ir_reader.h" +#include "standalone_scaffolding.h" + +using namespace std; + +static string read_stdin_to_eof() +{ + stringbuf sb; + cin.get(sb, '\0'); + return sb.str(); +} + +static GLboolean +do_optimization(struct exec_list *ir, const char *optimization) +{ + int int_0; + int int_1; + int int_2; + int int_3; + int int_4; + + if (sscanf(optimization, "do_common_optimization ( %d , %d ) ", + &int_0, &int_1) == 2) { + return do_common_optimization(ir, int_0 != 0, int_1); + } else if (strcmp(optimization, "do_algebraic") == 0) { + return do_algebraic(ir); + } else if (strcmp(optimization, "do_constant_folding") == 0) { + return do_constant_folding(ir); + } else if (strcmp(optimization, "do_constant_variable") == 0) { + return do_constant_variable(ir); + } else if (strcmp(optimization, "do_constant_variable_unlinked") == 0) { + return do_constant_variable_unlinked(ir); + } else if (strcmp(optimization, "do_copy_propagation") == 0) { + return do_copy_propagation(ir); + } else if (strcmp(optimization, "do_copy_propagation_elements") == 0) { + return do_copy_propagation_elements(ir); + } else if (strcmp(optimization, "do_constant_propagation") == 0) { + return do_constant_propagation(ir); + } else if (strcmp(optimization, "do_dead_code") == 0) { + return do_dead_code(ir); + } else if (strcmp(optimization, "do_dead_code_local") == 0) { + return do_dead_code_local(ir); + } else if (strcmp(optimization, "do_dead_code_unlinked") == 0) { + return do_dead_code_unlinked(ir); + } else if (strcmp(optimization, "do_dead_functions") == 0) { + return do_dead_functions(ir); + } else if (strcmp(optimization, "do_function_inlining") == 0) { + return do_function_inlining(ir); + } else if (sscanf(optimization, + "do_lower_jumps ( %d , %d , %d , %d , %d ) ", + &int_0, &int_1, &int_2, &int_3, &int_4) == 5) { + return do_lower_jumps(ir, int_0 != 0, int_1 != 0, int_2 != 0, + int_3 != 0, int_4 != 0); + } else if (strcmp(optimization, "do_lower_texture_projection") == 0) { + return do_lower_texture_projection(ir); + } else if (strcmp(optimization, "do_if_simplification") == 0) { + return do_if_simplification(ir); + } else if (strcmp(optimization, "do_discard_simplification") == 0) { + return do_discard_simplification(ir); + } else if (sscanf(optimization, "lower_if_to_cond_assign ( %d ) ", + &int_0) == 1) { + return lower_if_to_cond_assign(ir, int_0); + } else if (strcmp(optimization, "do_mat_op_to_vec") == 0) { + return do_mat_op_to_vec(ir); + } else if (strcmp(optimization, "do_noop_swizzle") == 0) { + return do_noop_swizzle(ir); + } else if (strcmp(optimization, "do_structure_splitting") == 0) { + return do_structure_splitting(ir); + } else if (strcmp(optimization, "do_swizzle_swizzle") == 0) { + return do_swizzle_swizzle(ir); + } else if (strcmp(optimization, "do_tree_grafting") == 0) { + return do_tree_grafting(ir); + } else if (strcmp(optimization, "do_vec_index_to_cond_assign") == 0) { + return do_vec_index_to_cond_assign(ir); + } else if (strcmp(optimization, "do_vec_index_to_swizzle") == 0) { + return do_vec_index_to_swizzle(ir); + } else if (strcmp(optimization, "lower_discard") == 0) { + return lower_discard(ir); + } else if (sscanf(optimization, "lower_instructions ( %d ) ", + &int_0) == 1) { + return lower_instructions(ir, int_0); + } else if (strcmp(optimization, "lower_noise") == 0) { + return lower_noise(ir); + } else if (sscanf(optimization, "lower_variable_index_to_cond_assign " + "( %d , %d , %d , %d ) ", &int_0, &int_1, &int_2, + &int_3) == 4) { + return lower_variable_index_to_cond_assign(ir, int_0 != 0, int_1 != 0, + int_2 != 0, int_3 != 0); + } else if (sscanf(optimization, "lower_quadop_vector ( %d ) ", + &int_0) == 1) { + return lower_quadop_vector(ir, int_0 != 0); + } else if (strcmp(optimization, "optimize_redundant_jumps") == 0) { + return optimize_redundant_jumps(ir); + } else { + printf("Unrecognized optimization %s\n", optimization); + exit(EXIT_FAILURE); + return false; + } +} + +static GLboolean +do_optimization_passes(struct exec_list *ir, char **optimizations, + int num_optimizations, bool quiet) +{ + GLboolean overall_progress = false; + + for (int i = 0; i < num_optimizations; ++i) { + const char *optimization = optimizations[i]; + if (!quiet) { + printf("*** Running optimization %s...", optimization); + } + GLboolean progress = do_optimization(ir, optimization); + if (!quiet) { + printf("%s\n", progress ? "progress" : "no progress"); + } + validate_ir_tree(ir); + + overall_progress = overall_progress || progress; + } + + return overall_progress; +} + +int test_optpass(int argc, char **argv) +{ + int input_format_ir = 0; /* 0=glsl, 1=ir */ + int loop = 0; + int shader_type = GL_VERTEX_SHADER; + int quiet = 0; + + const struct option optpass_opts[] = { + { "input-ir", no_argument, &input_format_ir, 1 }, + { "input-glsl", no_argument, &input_format_ir, 0 }, + { "loop", no_argument, &loop, 1 }, + { "vertex-shader", no_argument, &shader_type, GL_VERTEX_SHADER }, + { "fragment-shader", no_argument, &shader_type, GL_FRAGMENT_SHADER }, + { "quiet", no_argument, &quiet, 1 }, + { NULL, 0, NULL, 0 } + }; + + int idx = 0; + int c; + while ((c = getopt_long(argc, argv, "", optpass_opts, &idx)) != -1) { + if (c != 0) { + printf("*** usage: %s optpass <optimizations> <options>\n", argv[0]); + printf("\n"); + printf("Possible options are:\n"); + printf(" --input-ir: input format is IR\n"); + printf(" --input-glsl: input format is GLSL (the default)\n"); + printf(" --loop: run optimizations repeatedly until no progress\n"); + printf(" --vertex-shader: test with a vertex shader (the default)\n"); + printf(" --fragment-shader: test with a fragment shader\n"); + exit(EXIT_FAILURE); + } + } + + struct gl_context local_ctx; + struct gl_context *ctx = &local_ctx; + initialize_context_to_defaults(ctx, API_OPENGL); + + ctx->Driver.NewShader = _mesa_new_shader; + + struct gl_shader *shader = rzalloc(NULL, struct gl_shader); + shader->Type = shader_type; + + string input = read_stdin_to_eof(); + + struct _mesa_glsl_parse_state *state + = new(shader) _mesa_glsl_parse_state(ctx, shader->Type, shader); + + if (input_format_ir) { + shader->ir = new(shader) exec_list; + _mesa_glsl_initialize_types(state); + _mesa_glsl_read_ir(state, shader->ir, input.c_str(), true); + } else { + shader->Source = input.c_str(); + const char *source = shader->Source; + state->error = preprocess(state, &source, &state->info_log, + state->extensions, ctx->API) != 0; + + if (!state->error) { + _mesa_glsl_lexer_ctor(state, source); + _mesa_glsl_parse(state); + _mesa_glsl_lexer_dtor(state); + } + + shader->ir = new(shader) exec_list; + if (!state->error && !state->translation_unit.is_empty()) + _mesa_ast_to_hir(shader->ir, state); + } + + /* Print out the initial IR */ + if (!state->error && !quiet) { + printf("*** pre-optimization IR:\n"); + _mesa_print_ir(shader->ir, state); + printf("\n--\n"); + } + + /* Optimization passes */ + if (!state->error) { + GLboolean progress; + do { + progress = do_optimization_passes(shader->ir, &argv[optind], + argc - optind, quiet != 0); + } while (loop && progress); + } + + /* Print out the resulting IR */ + if (!state->error) { + if (!quiet) { + printf("*** resulting IR:\n"); + } + _mesa_print_ir(shader->ir, state); + if (!quiet) { + printf("\n--\n"); + } + } + + if (state->error) { + printf("*** error(s) occurred:\n"); + printf("%s\n", state->info_log); + printf("--\n"); + } + + ralloc_free(state); + ralloc_free(shader); + + return state->error; +} + diff --git a/mesalib/src/glsl/test_optpass.h b/mesalib/src/glsl/test_optpass.h new file mode 100644 index 000000000..923ccf3de --- /dev/null +++ b/mesalib/src/glsl/test_optpass.h @@ -0,0 +1,30 @@ +/* + * Copyright © 2011 Intel Corporation + * + * 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, sublicense, + * 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 NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + */ + +#pragma once +#ifndef TEST_OPTPASS_H +#define TEST_OPTPASS_H + +int test_optpass(int argc, char **argv); + +#endif /* TEST_OPTPASS_H */ diff --git a/mesalib/src/mesa/drivers/common/meta.c b/mesalib/src/mesa/drivers/common/meta.c index fa78674e4..26c895196 100644 --- a/mesalib/src/mesa/drivers/common/meta.c +++ b/mesalib/src/mesa/drivers/common/meta.c @@ -90,13 +90,14 @@ #define META_SCISSOR 0x100 #define META_SHADER 0x200 #define META_STENCIL_TEST 0x400 -#define META_TRANSFORM 0x800 /**< modelview, projection, clip planes */ +#define META_TRANSFORM 0x800 /**< modelview/projection matrix state */ #define META_TEXTURE 0x1000 #define META_VERTEX 0x2000 #define META_VIEWPORT 0x4000 #define META_CLAMP_FRAGMENT_COLOR 0x8000 #define META_CLAMP_VERTEX_COLOR 0x10000 #define META_CONDITIONAL_RENDER 0x20000 +#define META_CLIP 0x40000 /*@}*/ @@ -165,6 +166,8 @@ struct save_state GLfloat ModelviewMatrix[16]; GLfloat ProjectionMatrix[16]; GLfloat TextureMatrix[16]; + + /** META_CLIP */ GLbitfield ClipPlanesEnabled; /** META_TEXTURE */ @@ -547,6 +550,9 @@ _mesa_meta_begin(struct gl_context *ctx, GLbitfield state) _mesa_Ortho(0.0, ctx->DrawBuffer->Width, 0.0, ctx->DrawBuffer->Height, -1.0, 1.0); + } + + if (state & META_CLIP) { save->ClipPlanesEnabled = ctx->Transform.ClipPlanesEnabled; if (ctx->Transform.ClipPlanesEnabled) { GLuint i; @@ -846,7 +852,9 @@ _mesa_meta_end(struct gl_context *ctx) _mesa_LoadMatrixf(save->ProjectionMatrix); _mesa_MatrixMode(save->MatrixMode); + } + if (state & META_CLIP) { if (save->ClipPlanesEnabled) { GLuint i; for (i = 0; i < ctx->Const.MaxClipPlanes; i++) { @@ -1669,6 +1677,7 @@ _mesa_meta_glsl_Clear(struct gl_context *ctx, GLbitfield buffers) META_STENCIL_TEST | META_VERTEX | META_VIEWPORT | + META_CLIP | META_CLAMP_FRAGMENT_COLOR); if (!(buffers & BUFFER_BITS_COLOR)) { @@ -1783,6 +1792,7 @@ _mesa_meta_CopyPixels(struct gl_context *ctx, GLint srcX, GLint srcY, META_SHADER | META_TEXTURE | META_TRANSFORM | + META_CLIP | META_VERTEX | META_VIEWPORT)); @@ -2104,6 +2114,7 @@ _mesa_meta_DrawPixels(struct gl_context *ctx, META_SHADER | META_TEXTURE | META_TRANSFORM | + META_CLIP | META_VERTEX | META_VIEWPORT | META_CLAMP_FRAGMENT_COLOR | @@ -2313,6 +2324,7 @@ _mesa_meta_Bitmap(struct gl_context *ctx, META_SHADER | META_TEXTURE | META_TRANSFORM | + META_CLIP | META_VERTEX | META_VIEWPORT)); diff --git a/mesalib/src/mesa/main/compiler.h b/mesalib/src/mesa/main/compiler.h index 743841be4..d736fdfc5 100644 --- a/mesalib/src/mesa/main/compiler.h +++ b/mesalib/src/mesa/main/compiler.h @@ -139,26 +139,28 @@ extern "C" { /** * Function inlining */ -#if defined(__GNUC__) -# define INLINE __inline__ -#elif defined(__MSC__) -# define INLINE __inline -#elif defined(_MSC_VER) -# define INLINE __inline -#elif defined(__ICL) -# define INLINE __inline -#elif defined(__INTEL_COMPILER) -# define INLINE inline -#elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100) -# define INLINE __inline -#elif defined(__SUNPRO_C) && defined(__C99FEATURES__) -# define INLINE inline -# define __inline inline -# define __inline__ inline -#elif (__STDC_VERSION__ >= 199901L) /* C99 */ -# define INLINE inline -#else -# define INLINE +#ifndef INLINE +# if defined(__GNUC__) +# define INLINE __inline__ +# elif defined(__MSC__) +# define INLINE __inline +# elif defined(_MSC_VER) +# define INLINE __inline +# elif defined(__ICL) +# define INLINE __inline +# elif defined(__INTEL_COMPILER) +# define INLINE inline +# elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100) +# define INLINE __inline +# elif defined(__SUNPRO_C) && defined(__C99FEATURES__) +# define INLINE inline +# define __inline inline +# define __inline__ inline +# elif (__STDC_VERSION__ >= 199901L) /* C99 */ +# define INLINE inline +# else +# define INLINE +# endif #endif diff --git a/mesalib/src/mesa/main/get.c b/mesalib/src/mesa/main/get.c index 0492e1585..d32c68a53 100644 --- a/mesalib/src/mesa/main/get.c +++ b/mesalib/src/mesa/main/get.c @@ -1569,11 +1569,11 @@ find_custom_value(struct gl_context *ctx, const struct value_desc *d, union valu break; case GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB: - v->value_int = _mesa_get_compressed_formats(ctx, NULL, GL_FALSE); + v->value_int = _mesa_get_compressed_formats(ctx, NULL); break; case GL_COMPRESSED_TEXTURE_FORMATS_ARB: v->value_int_n.n = - _mesa_get_compressed_formats(ctx, v->value_int_n.ints, GL_FALSE); + _mesa_get_compressed_formats(ctx, v->value_int_n.ints); ASSERT(v->value_int_n.n <= 100); break; diff --git a/mesalib/src/mesa/main/texcompress.c b/mesalib/src/mesa/main/texcompress.c index d820ae927..42bd1eee5 100644 --- a/mesalib/src/mesa/main/texcompress.c +++ b/mesalib/src/mesa/main/texcompress.c @@ -40,19 +40,192 @@ /** + * Get the GL base format of a specified GL compressed texture format + * + * From page 232 of the OpenGL 3.3 (Compatiblity Profile) spec: + * + * "Compressed Internal Format Base Internal Format Type + * --------------------------- -------------------- --------- + * COMPRESSED_ALPHA ALPHA Generic + * COMPRESSED_LUMINANCE LUMINANCE Generic + * COMPRESSED_LUMINANCE_ALPHA LUMINANCE_ALPHA Generic + * COMPRESSED_INTENSITY INTENSITY Generic + * COMPRESSED_RED RED Generic + * COMPRESSED_RG RG Generic + * COMPRESSED_RGB RGB Generic + * COMPRESSED_RGBA RGBA Generic + * COMPRESSED_SRGB RGB Generic + * COMPRESSED_SRGB_ALPHA RGBA Generic + * COMPRESSED_SLUMINANCE LUMINANCE Generic + * COMPRESSED_SLUMINANCE_ALPHA LUMINANCE_ALPHA Generic + * COMPRESSED_RED_RGTC1 RED Specific + * COMPRESSED_SIGNED_RED_RGTC1 RED Specific + * COMPRESSED_RG_RGTC2 RG Specific + * COMPRESSED_SIGNED_RG_RGTC2 RG Specific" + * + * \return + * The base format of \c format if \c format is a compressed format (either + * generic or specific. Otherwise 0 is returned. + */ +GLenum +_mesa_gl_compressed_format_base_format(GLenum format) +{ + switch (format) { + case GL_COMPRESSED_RED: + case GL_COMPRESSED_RED_RGTC1: + case GL_COMPRESSED_SIGNED_RED_RGTC1: + return GL_RED; + + case GL_COMPRESSED_RG: + case GL_COMPRESSED_RG_RGTC2: + case GL_COMPRESSED_SIGNED_RG_RGTC2: + return GL_RG; + + case GL_COMPRESSED_RGB: + case GL_COMPRESSED_SRGB: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + case GL_COMPRESSED_RGB_FXT1_3DFX: + case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT: + return GL_RGB; + + case GL_COMPRESSED_RGBA: + case GL_COMPRESSED_SRGB_ALPHA: + case GL_COMPRESSED_RGBA_BPTC_UNORM_ARB: + case GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM_ARB: + case GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB: + case GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB: + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + case GL_COMPRESSED_RGBA_FXT1_3DFX: + case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT: + case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT: + case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT: + return GL_RGBA; + + case GL_COMPRESSED_ALPHA: + return GL_ALPHA; + + case GL_COMPRESSED_LUMINANCE: + case GL_COMPRESSED_SLUMINANCE: + case GL_COMPRESSED_LUMINANCE_LATC1_EXT: + case GL_COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT: + return GL_LUMINANCE; + + case GL_COMPRESSED_LUMINANCE_ALPHA: + case GL_COMPRESSED_SLUMINANCE_ALPHA: + case GL_COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT: + case GL_COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT: + case GL_COMPRESSED_LUMINANCE_ALPHA_3DC_ATI: + return GL_LUMINANCE_ALPHA; + + case GL_COMPRESSED_INTENSITY: + return GL_INTENSITY; + + default: + return 0; + } +} + +/** * Return list of (and count of) all specific texture compression * formats that are supported. * + * Some formats are \b not returned by this function. The + * \c GL_COMPRESSED_TEXTURE_FORMATS query only returns formats that are + * "suitable for general-purpose usage." All texture compression extensions + * have taken this to mean either linear RGB or linear RGBA. + * + * The GL_ARB_texture_compress_rgtc spec says: + * + * "19) Should the GL_NUM_COMPRESSED_TEXTURE_FORMATS and + * GL_COMPRESSED_TEXTURE_FORMATS queries return the RGTC formats? + * + * RESOLVED: No. + * + * The OpenGL 2.1 specification says "The only values returned + * by this query [GL_COMPRESSED_TEXTURE_FORMATS"] are those + * corresponding to formats suitable for general-purpose usage. + * The renderer will not enumerate formats with restrictions that + * need to be specifically understood prior to use." + * + * Compressed textures with just red or red-green components are + * not general-purpose so should not be returned by these queries + * because they have restrictions. + * + * Applications that seek to use the RGTC formats should do so + * by looking for this extension's name in the string returned by + * glGetString(GL_EXTENSIONS) rather than + * what GL_NUM_COMPRESSED_TEXTURE_FORMATS and + * GL_COMPRESSED_TEXTURE_FORMATS return." + * + * There is nearly identical wording in the GL_EXT_texture_compression_rgtc + * spec. + * + * The GL_EXT_texture_rRGB spec says: + * + * "22) Should the new COMPRESSED_SRGB_* formats be listed in an + * implementation's GL_COMPRESSED_TEXTURE_FORMATS list? + * + * RESOLVED: No. Section 3.8.1 says formats listed by + * GL_COMPRESSED_TEXTURE_FORMATS are "suitable for general-purpose + * usage." The non-linear distribution of red, green, and + * blue for these sRGB compressed formats makes them not really + * general-purpose." + * + * The GL_EXT_texture_compression_latc spec says: + * + * "16) Should the GL_NUM_COMPRESSED_TEXTURE_FORMATS and + * GL_COMPRESSED_TEXTURE_FORMATS queries return the LATC formats? + * + * RESOLVED: No. + * + * The OpenGL 2.1 specification says "The only values returned + * by this query [GL_COMPRESSED_TEXTURE_FORMATS"] are those + * corresponding to formats suitable for general-purpose usage. + * The renderer will not enumerate formats with restrictions that + * need to be specifically understood prior to use." + * + * Historically, OpenGL implementation have advertised the RGB and + * RGBA versions of the S3TC extensions compressed format tokens + * through this mechanism. + * + * The specification is not sufficiently clear about what "suitable + * for general-purpose usage" means. Historically that seems to mean + * unsigned RGB or unsigned RGBA. The DXT1 format supporting alpha + * (GL_COMPRESSED_RGBA_S3TC_DXT1_EXT) is not exposed in the list (at + * least for NVIDIA drivers) because the alpha is always 1.0 expect + * when it is 0.0 when RGB is required to be black. NVIDIA's even + * limits itself to true linear RGB or RGBA formats, specifically + * not including EXT_texture_sRGB's sRGB S3TC compressed formats. + * + * Adding luminance and luminance-alpha texture formats (and + * certainly signed versions of luminance and luminance-alpha + * formats!) invites potential comptaibility problems with old + * applications using this mechanism since old applications are + * unlikely to expect non-RGB or non-RGBA formats to be advertised + * through this mechanism. However no specific misinteractions + * with old applications is known. + * + * Applications that seek to use the LATC formats should do so + * by looking for this extension's name in the string returned by + * glGetString(GL_EXTENSIONS) rather than + * what GL_NUM_COMPRESSED_TEXTURE_FORMATS and + * GL_COMPRESSED_TEXTURE_FORMATS return." + * + * There is no formal spec for GL_ATI_texture_compression_3dc. Since the + * formats added by this extension are luminance-alpha formats, it is + * reasonable to expect them to follow the same rules as + * GL_EXT_texture_compression_latc. At the very least, Catalyst 11.6 does not + * expose the 3dc formats through this mechanism. + * * \param ctx the GL context * \param formats the resulting format list (may be NULL). - * \param all if true return all formats, even those with some kind - * of restrictions/limitations (See GL_ARB_texture_compression - * spec for more info). * * \return number of formats. */ GLuint -_mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats, GLboolean all) +_mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats) { GLuint n = 0; if (ctx->Extensions.TDFX_texture_compression_FXT1) { @@ -64,24 +237,15 @@ _mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats, GLboolean a n += 2; } } - /* don't return RGTC - ARB_texture_compression_rgtc query 19 */ + if (ctx->Extensions.EXT_texture_compression_s3tc) { if (formats) { formats[n++] = GL_COMPRESSED_RGB_S3TC_DXT1_EXT; - /* This format has some restrictions/limitations and so should - * not be returned via the GL_COMPRESSED_TEXTURE_FORMATS query. - * Specifically, all transparent pixels become black. NVIDIA - * omits this format too. - */ - if (all) - formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT1_EXT; formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT3_EXT; formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT5_EXT; } else { n += 3; - if (all) - n += 1; } } if (ctx->Extensions.S3_s3tc) { @@ -95,19 +259,6 @@ _mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats, GLboolean a n += 4; } } -#if FEATURE_EXT_texture_sRGB - if (ctx->Extensions.EXT_texture_sRGB) { - if (formats) { - formats[n++] = GL_COMPRESSED_SRGB_S3TC_DXT1_EXT; - formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT; - formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT; - formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT; - } - else { - n += 4; - } - } -#endif /* FEATURE_EXT_texture_sRGB */ return n; #if FEATURE_ES1 || FEATURE_ES2 diff --git a/mesalib/src/mesa/main/texcompress.h b/mesalib/src/mesa/main/texcompress.h index fd42f6402..375cf90c8 100644 --- a/mesalib/src/mesa/main/texcompress.h +++ b/mesalib/src/mesa/main/texcompress.h @@ -1,59 +1,62 @@ -/*
- * Mesa 3-D graphics library
- * Version: 6.5.1
- *
- * Copyright (C) 1999-2006 Brian Paul 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, sublicense,
- * 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 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 NONINFRINGEMENT. IN NO EVENT SHALL
- * BRIAN PAUL 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 TEXCOMPRESS_H
-#define TEXCOMPRESS_H
-
-#include "formats.h"
-#include "glheader.h"
-#include "mfeatures.h"
-
-struct gl_context;
-
-#if _HAVE_FULL_GL
-
-extern GLuint
-_mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats, GLboolean all);
-
-extern gl_format
-_mesa_glenum_to_compressed_format(GLenum format);
-
-extern GLenum
-_mesa_compressed_format_to_glenum(struct gl_context *ctx, GLuint mesaFormat);
-
-extern GLubyte *
-_mesa_compressed_image_address(GLint col, GLint row, GLint img,
- gl_format mesaFormat,
- GLsizei width, const GLubyte *image);
-
-#else /* _HAVE_FULL_GL */
-
-/* no-op macros */
-#define _mesa_get_compressed_formats( c, f ) 0
-#define _mesa_compressed_image_address(c, r, i, f, w, i2 ) 0
-#define _mesa_compress_teximage( c, w, h, sF, s, sRS, dF, d, drs ) ((void)0)
-
-#endif /* _HAVE_FULL_GL */
-
-#endif /* TEXCOMPRESS_H */
+/* + * Mesa 3-D graphics library + * Version: 6.5.1 + * + * Copyright (C) 1999-2006 Brian Paul 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, sublicense, + * 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 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 NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL 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 TEXCOMPRESS_H +#define TEXCOMPRESS_H + +#include "formats.h" +#include "glheader.h" +#include "mfeatures.h" + +struct gl_context; + +#if _HAVE_FULL_GL + +extern GLenum +_mesa_gl_compressed_format_base_format(GLenum format); + +extern GLuint +_mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats); + +extern gl_format +_mesa_glenum_to_compressed_format(GLenum format); + +extern GLenum +_mesa_compressed_format_to_glenum(struct gl_context *ctx, GLuint mesaFormat); + +extern GLubyte * +_mesa_compressed_image_address(GLint col, GLint row, GLint img, + gl_format mesaFormat, + GLsizei width, const GLubyte *image); + +#else /* _HAVE_FULL_GL */ + +/* no-op macros */ +#define _mesa_get_compressed_formats( c, f ) 0 +#define _mesa_compressed_image_address(c, r, i, f, w, i2 ) 0 +#define _mesa_compress_teximage( c, w, h, sF, s, sRS, dF, d, drs ) ((void)0) + +#endif /* _HAVE_FULL_GL */ + +#endif /* TEXCOMPRESS_H */ diff --git a/mesalib/src/mesa/main/texparam.c b/mesalib/src/mesa/main/texparam.c index 4b9dcb5d3..c4ec29533 100644 --- a/mesalib/src/mesa/main/texparam.c +++ b/mesalib/src/mesa/main/texparam.c @@ -915,9 +915,23 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level, *params = _mesa_compressed_format_to_glenum(ctx, texFormat); } else { - /* return the user's requested internal format */ - *params = img->InternalFormat; - } + /* If the true internal format is not compressed but the user + * requested a generic compressed format, we have to return the + * generic base format that matches. + * + * From page 119 (page 129 of the PDF) of the OpenGL 1.3 spec: + * + * "If no specific compressed format is available, + * internalformat is instead replaced by the corresponding base + * internal format." + * + * Otherwise just return the user's requested internal format + */ + const GLenum f = + _mesa_gl_compressed_format_base_format(img->InternalFormat); + + *params = (f != 0) ? f : img->InternalFormat; + } break; case GL_TEXTURE_BORDER: *params = img->Border; diff --git a/mesalib/src/mesa/program/ir_to_mesa.cpp b/mesalib/src/mesa/program/ir_to_mesa.cpp index 738e97ca5..8b4a535b7 100644 --- a/mesalib/src/mesa/program/ir_to_mesa.cpp +++ b/mesalib/src/mesa/program/ir_to_mesa.cpp @@ -134,7 +134,7 @@ src_reg::src_reg(dst_reg reg) this->index = reg.index; this->swizzle = SWIZZLE_XYZW; this->negate = 0; - this->reladdr = NULL; + this->reladdr = reg.reladdr; } dst_reg::dst_reg(src_reg reg) @@ -1415,9 +1415,9 @@ ir_to_mesa_visitor::visit(ir_dereference_variable *ir) case ir_var_in: case ir_var_inout: /* The linker assigns locations for varyings and attributes, - * including deprecated builtins (like gl_Color), user-assign - * generic attributes (glBindVertexLocation), and - * user-defined varyings. + * including deprecated builtins (like gl_Color), + * user-assigned generic attributes (glBindVertexLocation), + * and user-defined varyings. * * FINISHME: We would hit this path for function arguments. Fix! */ @@ -1496,6 +1496,18 @@ ir_to_mesa_visitor::visit(ir_dereference_array *ir) this->result, src_reg_for_float(element_size)); } + /* If there was already a relative address register involved, add the + * new and the old together to get the new offset. + */ + if (src.reladdr != NULL) { + src_reg accum_reg = get_temp(glsl_type::float_type); + + emit(ir, OPCODE_ADD, dst_reg(accum_reg), + index_reg, *src.reladdr); + + index_reg = accum_reg; + } + src.reladdr = ralloc(mem_ctx, src_reg); memcpy(src.reladdr, &index_reg, sizeof(index_reg)); } diff --git a/mesalib/src/mesa/program/prog_optimize.c b/mesalib/src/mesa/program/prog_optimize.c index 8a40fa69e..f4a7a638d 100644 --- a/mesalib/src/mesa/program/prog_optimize.c +++ b/mesalib/src/mesa/program/prog_optimize.c @@ -1319,6 +1319,15 @@ _mesa_simplify_cmp(struct gl_program * program) inst->Opcode = OPCODE_MOV; inst->SrcReg[0] = inst->SrcReg[1]; + + /* Unused operands are expected to have the file set to + * PROGRAM_UNDEFINED. This is how _mesa_init_instructions initializes + * all of the sources. + */ + inst->SrcReg[1].File = PROGRAM_UNDEFINED; + inst->SrcReg[1].Swizzle = SWIZZLE_NOOP; + inst->SrcReg[2].File = PROGRAM_UNDEFINED; + inst->SrcReg[2].Swizzle = SWIZZLE_NOOP; } } if (dbg) { |