diff options
Diffstat (limited to 'mesalib/src/glsl')
25 files changed, 1094 insertions, 733 deletions
diff --git a/mesalib/src/glsl/Makefile.sources b/mesalib/src/glsl/Makefile.sources index 2e81dedda..e69c1ac61 100644 --- a/mesalib/src/glsl/Makefile.sources +++ b/mesalib/src/glsl/Makefile.sources @@ -99,6 +99,7 @@ LIBGLSL_FILES = \ $(GLSL_SRCDIR)/opt_structure_splitting.cpp \ $(GLSL_SRCDIR)/opt_swizzle_swizzle.cpp \ $(GLSL_SRCDIR)/opt_tree_grafting.cpp \ + $(GLSL_SRCDIR)/opt_vectorize.cpp \ $(GLSL_SRCDIR)/s_expression.cpp \ $(GLSL_SRCDIR)/strtod.c diff --git a/mesalib/src/glsl/ast.h b/mesalib/src/glsl/ast.h index 344277803..502490200 100644 --- a/mesalib/src/glsl/ast.h +++ b/mesalib/src/glsl/ast.h @@ -280,6 +280,43 @@ private: bool cons; }; +class ast_array_specifier : public ast_node { +public: + /** Unsized array specifier ([]) */ + explicit ast_array_specifier(const struct YYLTYPE &locp) + : dimension_count(1), is_unsized_array(true) + { + set_location(locp); + } + + /** Sized array specifier ([dim]) */ + ast_array_specifier(const struct YYLTYPE &locp, ast_expression *dim) + : dimension_count(1), is_unsized_array(false) + { + set_location(locp); + array_dimensions.push_tail(&dim->link); + } + + void add_dimension(ast_expression *dim) + { + array_dimensions.push_tail(&dim->link); + dimension_count++; + } + + virtual void print(void) const; + + /* Count including sized and unsized dimensions */ + unsigned dimension_count; + + /* If true, this means that the array has an unsized outermost dimension. */ + bool is_unsized_array; + + /* This list contains objects of type ast_node containing the + * sized dimensions only, in outermost-to-innermost order. + */ + exec_list array_dimensions; +}; + /** * C-style aggregate initialization class * @@ -300,7 +337,16 @@ public: /* empty */ } - ast_type_specifier *constructor_type; + /** + * glsl_type of the aggregate, which is inferred from the LHS of whatever + * the aggregate is being used to initialize. This can't be inferred at + * parse time (since the parser deals with ast_type_specifiers, not + * glsl_types), so the parser leaves it NULL. However, the ast-to-hir + * conversion code makes sure to fill it in with the appropriate type + * before hir() is called. + */ + const glsl_type *constructor_type; + virtual ir_rvalue *hir(exec_list *instructions, struct _mesa_glsl_parse_state *state); }; @@ -329,14 +375,14 @@ public: class ast_declaration : public ast_node { public: - ast_declaration(const char *identifier, bool is_array, ast_expression *array_size, - ast_expression *initializer); + ast_declaration(const char *identifier, + ast_array_specifier *array_specifier, + ast_expression *initializer); virtual void print(void) const; const char *identifier; - - bool is_array; - ast_expression *array_size; + + ast_array_specifier *array_specifier; ast_expression *initializer; }; @@ -545,10 +591,10 @@ public: * Use only if the objects are allocated from the same context and will not * be modified. Zeros the inherited ast_node's fields. */ - ast_type_specifier(const ast_type_specifier *that, bool is_array, - ast_expression *array_size) + ast_type_specifier(const ast_type_specifier *that, + ast_array_specifier *array_specifier) : ast_node(), type_name(that->type_name), structure(that->structure), - is_array(is_array), array_size(array_size), + array_specifier(array_specifier), default_precision(that->default_precision) { /* empty */ @@ -556,8 +602,7 @@ public: /** Construct a type specifier from a type name */ ast_type_specifier(const char *name) - : type_name(name), structure(NULL), - is_array(false), array_size(NULL), + : type_name(name), structure(NULL), array_specifier(NULL), default_precision(ast_precision_none) { /* empty */ @@ -565,8 +610,7 @@ public: /** Construct a type specifier from a structure definition */ ast_type_specifier(ast_struct_specifier *s) - : type_name(s->name), structure(s), - is_array(false), array_size(NULL), + : type_name(s->name), structure(s), array_specifier(NULL), default_precision(ast_precision_none) { /* empty */ @@ -583,8 +627,7 @@ public: const char *type_name; ast_struct_specifier *structure; - bool is_array; - ast_expression *array_size; + ast_array_specifier *array_specifier; /** For precision statements, this is the given precision; otherwise none. */ unsigned default_precision:2; @@ -637,8 +680,7 @@ public: ast_parameter_declarator() : type(NULL), identifier(NULL), - is_array(false), - array_size(NULL), + array_specifier(NULL), formal_parameter(false), is_void(false) { @@ -652,8 +694,7 @@ public: ast_fully_specified_type *type; const char *identifier; - bool is_array; - ast_expression *array_size; + ast_array_specifier *array_specifier; static void parameters_to_hir(exec_list *ast_parameters, bool formal, exec_list *ir_parameters, @@ -900,13 +941,10 @@ class ast_interface_block : public ast_node { public: ast_interface_block(ast_type_qualifier layout, const char *instance_name, - bool is_array, - ast_expression *array_size) + ast_array_specifier *array_specifier) : layout(layout), block_name(NULL), instance_name(instance_name), - is_array(is_array), array_size(array_size) + array_specifier(array_specifier) { - if (!is_array) - assert(array_size == NULL); } virtual ir_rvalue *hir(exec_list *instructions, @@ -927,21 +965,12 @@ public: exec_list declarations; /** - * True if the block is declared as an array - * - * \note - * A block can only be an array if it also has an instance name. If this - * field is true, ::instance_name must also not be \c NULL. - */ - bool is_array; - - /** * Declared array size of the block instance * * If the block is not declared as an array or if the block instance array * is unsized, this field will be \c NULL. */ - ast_expression *array_size; + ast_array_specifier *array_specifier; }; @@ -982,9 +1011,8 @@ _mesa_ast_array_index_to_hir(void *mem_ctx, YYLTYPE &loc, YYLTYPE &idx_loc); extern void -_mesa_ast_set_aggregate_type(const ast_type_specifier *type, - ast_expression *expr, - _mesa_glsl_parse_state *state); +_mesa_ast_set_aggregate_type(const glsl_type *type, + ast_expression *expr); void emit_function(_mesa_glsl_parse_state *state, ir_function *f); diff --git a/mesalib/src/glsl/ast_array_index.cpp b/mesalib/src/glsl/ast_array_index.cpp index a5f23206a..f3b060ea6 100644 --- a/mesalib/src/glsl/ast_array_index.cpp +++ b/mesalib/src/glsl/ast_array_index.cpp @@ -25,6 +25,19 @@ #include "glsl_types.h" #include "ir.h" +void +ast_array_specifier::print(void) const +{ + if (this->is_unsized_array) { + printf("[ ] "); + } + + foreach_list_typed (ast_node, array_dimension, link, &this->array_dimensions) { + printf("[ "); + array_dimension->print(); + printf("] "); + } +} /** * If \c ir is a reference to an array for which we are tracking the max array diff --git a/mesalib/src/glsl/ast_function.cpp b/mesalib/src/glsl/ast_function.cpp index 2d05d0723..4c5b0e4aa 100644 --- a/mesalib/src/glsl/ast_function.cpp +++ b/mesalib/src/glsl/ast_function.cpp @@ -1687,14 +1687,12 @@ ast_aggregate_initializer::hir(exec_list *instructions, { void *ctx = state; YYLTYPE loc = this->get_location(); - const char *name; if (!this->constructor_type) { _mesa_glsl_error(&loc, state, "type of C-style initializer unknown"); return ir_rvalue::error_value(ctx); } - const glsl_type *const constructor_type = - this->constructor_type->glsl_type(&name, state); + const glsl_type *const constructor_type = this->constructor_type; if (!state->ARB_shading_language_420pack_enable) { _mesa_glsl_error(&loc, state, "C-style initialization requires the " @@ -1702,12 +1700,12 @@ ast_aggregate_initializer::hir(exec_list *instructions, return ir_rvalue::error_value(ctx); } - if (this->constructor_type->is_array) { + if (constructor_type->is_array()) { return process_array_constructor(instructions, constructor_type, &loc, &this->expressions, state); } - if (this->constructor_type->structure) { + if (constructor_type->is_record()) { return process_record_constructor(instructions, constructor_type, &loc, &this->expressions, state); } diff --git a/mesalib/src/glsl/ast_to_hir.cpp b/mesalib/src/glsl/ast_to_hir.cpp index 4cc8eb18d..1bfb4e531 100644 --- a/mesalib/src/glsl/ast_to_hir.cpp +++ b/mesalib/src/glsl/ast_to_hir.cpp @@ -740,14 +740,15 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state, { void *ctx = state; bool error_emitted = (lhs->type->is_error() || rhs->type->is_error()); + ir_rvalue *extract_channel = NULL; /* If the assignment LHS comes back as an ir_binop_vector_extract * expression, move it to the RHS as an ir_triop_vector_insert. */ if (lhs->ir_type == ir_type_expression) { - ir_expression *const expr = lhs->as_expression(); + ir_expression *const lhs_expr = lhs->as_expression(); - if (unlikely(expr->operation == ir_binop_vector_extract)) { + if (unlikely(lhs_expr->operation == ir_binop_vector_extract)) { ir_rvalue *new_rhs = validate_assignment(state, lhs_loc, lhs->type, rhs, is_initializer); @@ -755,12 +756,24 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state, if (new_rhs == NULL) { return lhs; } else { + /* This converts: + * - LHS: (expression float vector_extract <vec> <channel>) + * - RHS: <scalar> + * into: + * - LHS: <vec> + * - RHS: (expression vec2 vector_insert <vec> <channel> <scalar>) + * + * The LHS type is now a vector instead of a scalar. Since GLSL + * allows assignments to be used as rvalues, we need to re-extract + * the channel from assignment_temp when returning the rvalue. + */ + extract_channel = lhs_expr->operands[1]; rhs = new(ctx) ir_expression(ir_triop_vector_insert, - expr->operands[0]->type, - expr->operands[0], + lhs_expr->operands[0]->type, + lhs_expr->operands[0], new_rhs, - expr->operands[1]); - lhs = expr->operands[0]->clone(ctx, NULL); + extract_channel); + lhs = lhs_expr->operands[0]->clone(ctx, NULL); } } } @@ -830,8 +843,10 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state, rhs->type->array_size()); d->type = var->type; } - mark_whole_array_access(rhs); - mark_whole_array_access(lhs); + if (lhs->type->is_array()) { + mark_whole_array_access(rhs); + mark_whole_array_access(lhs); + } } /* Most callers of do_assignment (assign, add_assign, pre_inc/dec, @@ -854,6 +869,11 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state, if (!error_emitted) instructions->push_tail(new(ctx) ir_assignment(lhs, deref_var)); + if (extract_channel) { + return new(ctx) ir_expression(ir_binop_vector_extract, + new(ctx) ir_dereference_variable(var), + extract_channel->clone(ctx, NULL)); + } return new(ctx) ir_dereference_variable(var); } @@ -1771,64 +1791,108 @@ ast_compound_statement::hir(exec_list *instructions, return NULL; } +/** + * Evaluate the given exec_node (which should be an ast_node representing + * a single array dimension) and return its integer value. + */ +static const unsigned +process_array_size(exec_node *node, + struct _mesa_glsl_parse_state *state) +{ + exec_list dummy_instructions; + + ast_node *array_size = exec_node_data(ast_node, node, link); + ir_rvalue *const ir = array_size->hir(& dummy_instructions, + state); + YYLTYPE loc = array_size->get_location(); + + if (ir == NULL) { + _mesa_glsl_error(& loc, state, + "array size could not be resolved"); + return 0; + } + + if (!ir->type->is_integer()) { + _mesa_glsl_error(& loc, state, + "array size must be integer type"); + return 0; + } + + if (!ir->type->is_scalar()) { + _mesa_glsl_error(& loc, state, + "array size must be scalar type"); + return 0; + } + + ir_constant *const size = ir->constant_expression_value(); + if (size == NULL) { + _mesa_glsl_error(& loc, state, "array size must be a " + "constant valued expression"); + return 0; + } + + if (size->value.i[0] <= 0) { + _mesa_glsl_error(& loc, state, "array size must be > 0"); + return 0; + } + + assert(size->type == ir->type); + + /* If the array size is const (and we've verified that + * it is) then no instructions should have been emitted + * when we converted it to HIR. If they were emitted, + * then either the array size isn't const after all, or + * we are emitting unnecessary instructions. + */ + assert(dummy_instructions.is_empty()); + + return size->value.u[0]; +} static const glsl_type * -process_array_type(YYLTYPE *loc, const glsl_type *base, ast_node *array_size, - struct _mesa_glsl_parse_state *state) +process_array_type(YYLTYPE *loc, const glsl_type *base, + ast_array_specifier *array_specifier, + struct _mesa_glsl_parse_state *state) { - unsigned length = 0; + const glsl_type *array_type = base; - if (base == NULL) - return glsl_type::error_type; + if (array_specifier != NULL) { + if (base->is_array()) { - /* From page 19 (page 25) of the GLSL 1.20 spec: - * - * "Only one-dimensional arrays may be declared." - */ - if (base->is_array()) { - _mesa_glsl_error(loc, state, - "invalid array of `%s' (only one-dimensional arrays " - "may be declared)", - base->name); - return glsl_type::error_type; - } + /* From page 19 (page 25) of the GLSL 1.20 spec: + * + * "Only one-dimensional arrays may be declared." + */ + if (!state->ARB_arrays_of_arrays_enable) { + _mesa_glsl_error(loc, state, + "invalid array of `%s'" + "GL_ARB_arrays_of_arrays " + "required for defining arrays of arrays", + base->name); + return glsl_type::error_type; + } - if (array_size != NULL) { - exec_list dummy_instructions; - ir_rvalue *const ir = array_size->hir(& dummy_instructions, state); - YYLTYPE loc = array_size->get_location(); + if (base->length == 0) { + _mesa_glsl_error(loc, state, + "only the outermost array dimension can " + "be unsized", + base->name); + return glsl_type::error_type; + } + } - if (ir != NULL) { - if (!ir->type->is_integer()) { - _mesa_glsl_error(& loc, state, "array size must be integer type"); - } else if (!ir->type->is_scalar()) { - _mesa_glsl_error(& loc, state, "array size must be scalar type"); - } else { - ir_constant *const size = ir->constant_expression_value(); - - if (size == NULL) { - _mesa_glsl_error(& loc, state, "array size must be a " - "constant valued expression"); - } else if (size->value.i[0] <= 0) { - _mesa_glsl_error(& loc, state, "array size must be > 0"); - } else { - assert(size->type == ir->type); - length = size->value.u[0]; - - /* If the array size is const (and we've verified that - * it is) then no instructions should have been emitted - * when we converted it to HIR. If they were emitted, - * then either the array size isn't const after all, or - * we are emitting unnecessary instructions. - */ - assert(dummy_instructions.is_empty()); - } - } + for (exec_node *node = array_specifier->array_dimensions.tail_pred; + !node->is_head_sentinel(); node = node->prev) { + unsigned array_size = process_array_size(node, state); + array_type = glsl_type::get_array_instance(array_type, + array_size); } + + if (array_specifier->is_unsized_array) + array_type = glsl_type::get_array_instance(array_type, 0); } - const glsl_type *array_type = glsl_type::get_array_instance(base, length); - return array_type != NULL ? array_type : glsl_type::error_type; + return array_type; } @@ -1841,10 +1905,8 @@ ast_type_specifier::glsl_type(const char **name, type = state->symbols->get_type(this->type_name); *name = this->type_name; - if (this->is_array) { - YYLTYPE loc = this->get_location(); - type = process_array_type(&loc, type, this->array_size, state); - } + YYLTYPE loc = this->get_location(); + type = process_array_type(&loc, type, this->array_specifier, state); return type; } @@ -2593,6 +2655,13 @@ process_initializer(ir_variable *var, ast_declaration *decl, ? "attribute" : "varying"); } + /* If the initializer is an ast_aggregate_initializer, recursively store + * type information from the LHS into it, so that its hir() function can do + * type checking. + */ + if (decl->initializer->oper == ast_aggregate) + _mesa_ast_set_aggregate_type(var->type, decl->initializer); + ir_dereference *const lhs = new(state) ir_dereference_variable(var); ir_rvalue *rhs = decl->initializer->hir(initializer_instructions, state); @@ -2823,8 +2892,7 @@ ast_declarator_list::hir(exec_list *instructions, } foreach_list_typed (ast_declaration, decl, link, &this->declarations) { - assert(!decl->is_array); - assert(decl->array_size == NULL); + assert(decl->array_specifier == NULL); assert(decl->initializer == NULL); ir_variable *const earlier = @@ -2959,14 +3027,8 @@ ast_declarator_list::hir(exec_list *instructions, continue; } - if (decl->is_array) { - var_type = process_array_type(&loc, decl_type, decl->array_size, - state); - if (var_type->is_error()) - continue; - } else { - var_type = decl_type; - } + var_type = process_array_type(&loc, decl_type, decl->array_specifier, + state); var = new(ctx) ir_variable(var_type, decl->identifier, ir_var_auto); @@ -3098,8 +3160,9 @@ ast_declarator_list::hir(exec_list *instructions, * vectors. Vertex shader inputs cannot be arrays or * structures." */ - const glsl_type *check_type = var->type->is_array() - ? var->type->fields.array : var->type; + const glsl_type *check_type = var->type; + while (check_type->is_array()) + check_type = check_type->element_type(); switch (check_type->base_type) { case GLSL_TYPE_FLOAT: @@ -3523,9 +3586,7 @@ ast_parameter_declarator::hir(exec_list *instructions, /* This only handles "vec4 foo[..]". The earlier specifier->glsl_type(...) * call already handled the "vec4[..] foo" case. */ - if (this->is_array) { - type = process_array_type(&loc, type, this->array_size, state); - } + type = process_array_type(&loc, type, this->array_specifier, state); if (!type->is_error() && type->is_unsized_array()) { _mesa_glsl_error(&loc, state, "arrays passed as parameters must have " @@ -4463,7 +4524,7 @@ ast_type_specifier::hir(exec_list *instructions, return NULL; } - if (this->is_array) { + if (this->array_specifier != NULL) { _mesa_glsl_error(&loc, state, "default precision statements do not apply to " "arrays"); @@ -4653,10 +4714,8 @@ ast_process_structure_or_interface_block(exec_list *instructions, "members"); } - if (decl->is_array) { - field_type = process_array_type(&loc, decl_type, decl->array_size, - state); - } + field_type = process_array_type(&loc, decl_type, + decl->array_specifier, state); fields[i].type = field_type; fields[i].name = decl->identifier; fields[i].location = -1; @@ -4882,7 +4941,7 @@ ast_interface_block::hir(exec_list *instructions, _mesa_shader_stage_to_string(state->stage)); } if (this->instance_name == NULL || - strcmp(this->instance_name, "gl_in") != 0 || !this->is_array) { + strcmp(this->instance_name, "gl_in") != 0 || this->array_specifier == NULL) { _mesa_glsl_error(&loc, state, "gl_PerVertex input must be redeclared as " "gl_in[]"); @@ -4984,7 +5043,7 @@ ast_interface_block::hir(exec_list *instructions, * variable (or input block, see interface blocks below) needs to be * declared as an array. */ - if (state->stage == MESA_SHADER_GEOMETRY && !this->is_array && + if (state->stage == MESA_SHADER_GEOMETRY && this->array_specifier == NULL && var_mode == ir_var_shader_in) { _mesa_glsl_error(&loc, state, "geometry shader inputs must be arrays"); } @@ -5018,7 +5077,7 @@ ast_interface_block::hir(exec_list *instructions, ir_variable *var; - if (this->is_array) { + if (this->array_specifier != NULL) { /* Section 4.3.7 (Interface Blocks) of the GLSL 1.50 spec says: * * For uniform blocks declared an array, each individual array @@ -5038,7 +5097,7 @@ ast_interface_block::hir(exec_list *instructions, * interface array size *doesn't* need to be specified is on a * geometry shader input. */ - if (this->array_size == NULL && + if (this->array_specifier->is_unsized_array && (state->stage != MESA_SHADER_GEOMETRY || !this->layout.flags.q.in)) { _mesa_glsl_error(&loc, state, "only geometry shader inputs may be unsized " @@ -5047,7 +5106,7 @@ ast_interface_block::hir(exec_list *instructions, } const glsl_type *block_array_type = - process_array_type(&loc, block_type, this->array_size, state); + process_array_type(&loc, block_type, this->array_specifier, state); var = new(state) ir_variable(block_array_type, this->instance_name, @@ -5079,7 +5138,7 @@ ast_interface_block::hir(exec_list *instructions, /* In order to have an array size, the block must also be declared with * an instane name. */ - assert(!this->is_array); + assert(this->array_specifier == NULL); for (unsigned i = 0; i < num_variables; i++) { ir_variable *var = diff --git a/mesalib/src/glsl/ast_type.cpp b/mesalib/src/glsl/ast_type.cpp index d758bfa1f..637da0dfb 100644 --- a/mesalib/src/glsl/ast_type.cpp +++ b/mesalib/src/glsl/ast_type.cpp @@ -32,14 +32,8 @@ ast_type_specifier::print(void) const printf("%s ", type_name); } - if (is_array) { - printf("[ "); - - if (array_size) { - array_size->print(); - } - - printf("] "); + if (array_specifier) { + array_specifier->print(); } } diff --git a/mesalib/src/glsl/builtin_functions.cpp b/mesalib/src/glsl/builtin_functions.cpp index b1065f48d..ed58ddddb 100755 --- a/mesalib/src/glsl/builtin_functions.cpp +++ b/mesalib/src/glsl/builtin_functions.cpp @@ -576,20 +576,9 @@ private: ir_function_signature *_atomic_op(const char *intrinsic, builtin_available_predicate avail); - ir_function_signature *_min3(builtin_available_predicate avail, - const glsl_type *x_type, - const glsl_type *y_type, - const glsl_type *z_type); - - ir_function_signature *_max3(builtin_available_predicate avail, - const glsl_type *x_type, - const glsl_type *y_type, - const glsl_type *z_type); - - ir_function_signature *_mid3(builtin_available_predicate avail, - const glsl_type *x_type, - const glsl_type *y_type, - const glsl_type *z_type); + B1(min3) + B1(max3) + B1(mid3) #undef B0 #undef B1 @@ -2128,54 +2117,54 @@ builtin_builder::create_builtins() NULL); add_function("min3", - _min3(shader_trinary_minmax, glsl_type::float_type, glsl_type::float_type, glsl_type::float_type), - _min3(shader_trinary_minmax, glsl_type::vec2_type, glsl_type::vec2_type, glsl_type::vec2_type), - _min3(shader_trinary_minmax, glsl_type::vec3_type, glsl_type::vec3_type, glsl_type::vec3_type), - _min3(shader_trinary_minmax, glsl_type::vec4_type, glsl_type::vec4_type, glsl_type::vec4_type), - - _min3(shader_trinary_minmax, glsl_type::int_type, glsl_type::int_type, glsl_type::int_type), - _min3(shader_trinary_minmax, glsl_type::ivec2_type, glsl_type::ivec2_type, glsl_type::ivec2_type), - _min3(shader_trinary_minmax, glsl_type::ivec3_type, glsl_type::ivec3_type, glsl_type::ivec3_type), - _min3(shader_trinary_minmax, glsl_type::ivec4_type, glsl_type::ivec4_type, glsl_type::ivec4_type), - - _min3(shader_trinary_minmax, glsl_type::uint_type, glsl_type::uint_type, glsl_type::uint_type), - _min3(shader_trinary_minmax, glsl_type::uvec2_type, glsl_type::uvec2_type, glsl_type::uvec2_type), - _min3(shader_trinary_minmax, glsl_type::uvec3_type, glsl_type::uvec3_type, glsl_type::uvec3_type), - _min3(shader_trinary_minmax, glsl_type::uvec4_type, glsl_type::uvec4_type, glsl_type::uvec4_type), + _min3(glsl_type::float_type), + _min3(glsl_type::vec2_type), + _min3(glsl_type::vec3_type), + _min3(glsl_type::vec4_type), + + _min3(glsl_type::int_type), + _min3(glsl_type::ivec2_type), + _min3(glsl_type::ivec3_type), + _min3(glsl_type::ivec4_type), + + _min3(glsl_type::uint_type), + _min3(glsl_type::uvec2_type), + _min3(glsl_type::uvec3_type), + _min3(glsl_type::uvec4_type), NULL); add_function("max3", - _max3(shader_trinary_minmax, glsl_type::float_type, glsl_type::float_type, glsl_type::float_type), - _max3(shader_trinary_minmax, glsl_type::vec2_type, glsl_type::vec2_type, glsl_type::vec2_type), - _max3(shader_trinary_minmax, glsl_type::vec3_type, glsl_type::vec3_type, glsl_type::vec3_type), - _max3(shader_trinary_minmax, glsl_type::vec4_type, glsl_type::vec4_type, glsl_type::vec4_type), - - _max3(shader_trinary_minmax, glsl_type::int_type, glsl_type::int_type, glsl_type::int_type), - _max3(shader_trinary_minmax, glsl_type::ivec2_type, glsl_type::ivec2_type, glsl_type::ivec2_type), - _max3(shader_trinary_minmax, glsl_type::ivec3_type, glsl_type::ivec3_type, glsl_type::ivec3_type), - _max3(shader_trinary_minmax, glsl_type::ivec4_type, glsl_type::ivec4_type, glsl_type::ivec4_type), - - _max3(shader_trinary_minmax, glsl_type::uint_type, glsl_type::uint_type, glsl_type::uint_type), - _max3(shader_trinary_minmax, glsl_type::uvec2_type, glsl_type::uvec2_type, glsl_type::uvec2_type), - _max3(shader_trinary_minmax, glsl_type::uvec3_type, glsl_type::uvec3_type, glsl_type::uvec3_type), - _max3(shader_trinary_minmax, glsl_type::uvec4_type, glsl_type::uvec4_type, glsl_type::uvec4_type), + _max3(glsl_type::float_type), + _max3(glsl_type::vec2_type), + _max3(glsl_type::vec3_type), + _max3(glsl_type::vec4_type), + + _max3(glsl_type::int_type), + _max3(glsl_type::ivec2_type), + _max3(glsl_type::ivec3_type), + _max3(glsl_type::ivec4_type), + + _max3(glsl_type::uint_type), + _max3(glsl_type::uvec2_type), + _max3(glsl_type::uvec3_type), + _max3(glsl_type::uvec4_type), NULL); add_function("mid3", - _mid3(shader_trinary_minmax, glsl_type::float_type, glsl_type::float_type, glsl_type::float_type), - _mid3(shader_trinary_minmax, glsl_type::vec2_type, glsl_type::vec2_type, glsl_type::vec2_type), - _mid3(shader_trinary_minmax, glsl_type::vec3_type, glsl_type::vec3_type, glsl_type::vec3_type), - _mid3(shader_trinary_minmax, glsl_type::vec4_type, glsl_type::vec4_type, glsl_type::vec4_type), - - _mid3(shader_trinary_minmax, glsl_type::int_type, glsl_type::int_type, glsl_type::int_type), - _mid3(shader_trinary_minmax, glsl_type::ivec2_type, glsl_type::ivec2_type, glsl_type::ivec2_type), - _mid3(shader_trinary_minmax, glsl_type::ivec3_type, glsl_type::ivec3_type, glsl_type::ivec3_type), - _mid3(shader_trinary_minmax, glsl_type::ivec4_type, glsl_type::ivec4_type, glsl_type::ivec4_type), - - _mid3(shader_trinary_minmax, glsl_type::uint_type, glsl_type::uint_type, glsl_type::uint_type), - _mid3(shader_trinary_minmax, glsl_type::uvec2_type, glsl_type::uvec2_type, glsl_type::uvec2_type), - _mid3(shader_trinary_minmax, glsl_type::uvec3_type, glsl_type::uvec3_type, glsl_type::uvec3_type), - _mid3(shader_trinary_minmax, glsl_type::uvec4_type, glsl_type::uvec4_type, glsl_type::uvec4_type), + _mid3(glsl_type::float_type), + _mid3(glsl_type::vec2_type), + _mid3(glsl_type::vec3_type), + _mid3(glsl_type::vec4_type), + + _mid3(glsl_type::int_type), + _mid3(glsl_type::ivec2_type), + _mid3(glsl_type::ivec3_type), + _mid3(glsl_type::ivec4_type), + + _mid3(glsl_type::uint_type), + _mid3(glsl_type::uvec2_type), + _mid3(glsl_type::uvec3_type), + _mid3(glsl_type::uvec4_type), NULL); #undef F @@ -4064,14 +4053,12 @@ builtin_builder::_atomic_op(const char *intrinsic, } ir_function_signature * -builtin_builder::_min3(builtin_available_predicate avail, - const glsl_type *x_type, const glsl_type *y_type, - const glsl_type *z_type) +builtin_builder::_min3(const glsl_type *type) { - ir_variable *x = in_var(x_type, "x"); - ir_variable *y = in_var(y_type, "y"); - ir_variable *z = in_var(z_type, "z"); - MAKE_SIG(x_type, avail, 3, x, y, z); + ir_variable *x = in_var(type, "x"); + ir_variable *y = in_var(type, "y"); + ir_variable *z = in_var(type, "z"); + MAKE_SIG(type, shader_trinary_minmax, 3, x, y, z); ir_expression *min3 = min2(x, min2(y,z)); body.emit(ret(min3)); @@ -4080,14 +4067,12 @@ builtin_builder::_min3(builtin_available_predicate avail, } ir_function_signature * -builtin_builder::_max3(builtin_available_predicate avail, - const glsl_type *x_type, const glsl_type *y_type, - const glsl_type *z_type) +builtin_builder::_max3(const glsl_type *type) { - ir_variable *x = in_var(x_type, "x"); - ir_variable *y = in_var(y_type, "y"); - ir_variable *z = in_var(z_type, "z"); - MAKE_SIG(x_type, avail, 3, x, y, z); + ir_variable *x = in_var(type, "x"); + ir_variable *y = in_var(type, "y"); + ir_variable *z = in_var(type, "z"); + MAKE_SIG(type, shader_trinary_minmax, 3, x, y, z); ir_expression *max3 = max2(x, max2(y,z)); body.emit(ret(max3)); @@ -4096,14 +4081,12 @@ builtin_builder::_max3(builtin_available_predicate avail, } ir_function_signature * -builtin_builder::_mid3(builtin_available_predicate avail, - const glsl_type *x_type, const glsl_type *y_type, - const glsl_type *z_type) +builtin_builder::_mid3(const glsl_type *type) { - ir_variable *x = in_var(x_type, "x"); - ir_variable *y = in_var(y_type, "y"); - ir_variable *z = in_var(z_type, "z"); - MAKE_SIG(x_type, avail, 3, x, y, z); + ir_variable *x = in_var(type, "x"); + ir_variable *y = in_var(type, "y"); + ir_variable *z = in_var(type, "z"); + MAKE_SIG(type, shader_trinary_minmax, 3, x, y, z); ir_expression *mid3 = max2(min2(x, y), max2(min2(x, z), min2(y, z))); body.emit(ret(mid3)); diff --git a/mesalib/src/glsl/builtin_variables.cpp b/mesalib/src/glsl/builtin_variables.cpp index f630923ed..d6bc3c073 100644 --- a/mesalib/src/glsl/builtin_variables.cpp +++ b/mesalib/src/glsl/builtin_variables.cpp @@ -780,6 +780,8 @@ void builtin_variable_generator::generate_gs_special_vars() { add_output(VARYING_SLOT_LAYER, int_t, "gl_Layer"); + if (state->ARB_viewport_array_enable) + add_output(VARYING_SLOT_VIEWPORT, int_t, "gl_ViewportIndex"); /* Although gl_PrimitiveID appears in tessellation control and tessellation * evaluation shaders, it has a different function there than it has in diff --git a/mesalib/src/glsl/glcpp/glcpp-parse.y b/mesalib/src/glsl/glcpp/glcpp-parse.y index 55c498195..184e5c237 100644 --- a/mesalib/src/glsl/glcpp/glcpp-parse.y +++ b/mesalib/src/glsl/glcpp/glcpp-parse.y @@ -30,7 +30,6 @@ #include "glcpp.h" #include "main/core.h" /* for struct gl_extensions */ -#include "main/mtypes.h" /* for gl_api enum */ static void yyerror (YYLTYPE *locp, glcpp_parser_t *parser, const char *error); @@ -135,7 +134,7 @@ _glcpp_parser_skip_stack_pop (glcpp_parser_t *parser, YYLTYPE *loc); static void _glcpp_parser_handle_version_declaration(glcpp_parser_t *parser, intmax_t version, - const char *ident); + const char *ident, bool explicitly_set); static int glcpp_parser_lex (YYSTYPE *yylval, YYLTYPE *yylloc, glcpp_parser_t *parser); @@ -194,12 +193,15 @@ line: control_line { ralloc_asprintf_rewrite_tail (&parser->output, &parser->output_length, "\n"); } -| HASH_LINE pp_tokens NEWLINE { +| HASH_LINE { + glcpp_parser_resolve_version(parser); + } pp_tokens NEWLINE { + if (parser->skip_stack == NULL || parser->skip_stack->type == SKIP_NO_SKIP) { _glcpp_parser_expand_and_lex_from (parser, - LINE_EXPANDED, $2); + LINE_EXPANDED, $3); } } | text_line { @@ -238,25 +240,35 @@ expanded_line: } ; -control_line: - HASH_DEFINE OBJ_IDENTIFIER replacement_list NEWLINE { - _define_object_macro (parser, & @2, $2, $3); +define: + OBJ_IDENTIFIER replacement_list NEWLINE { + _define_object_macro (parser, & @1, $1, $2); } -| HASH_DEFINE FUNC_IDENTIFIER '(' ')' replacement_list NEWLINE { - _define_function_macro (parser, & @2, $2, NULL, $5); +| FUNC_IDENTIFIER '(' ')' replacement_list NEWLINE { + _define_function_macro (parser, & @1, $1, NULL, $4); } -| HASH_DEFINE FUNC_IDENTIFIER '(' identifier_list ')' replacement_list NEWLINE { - _define_function_macro (parser, & @2, $2, $4, $6); +| FUNC_IDENTIFIER '(' identifier_list ')' replacement_list NEWLINE { + _define_function_macro (parser, & @1, $1, $3, $5); } -| HASH_UNDEF IDENTIFIER NEWLINE { - macro_t *macro = hash_table_find (parser->defines, $2); +; + +control_line: + HASH_DEFINE { + glcpp_parser_resolve_version(parser); + } define +| HASH_UNDEF { + glcpp_parser_resolve_version(parser); + } IDENTIFIER NEWLINE { + macro_t *macro = hash_table_find (parser->defines, $3); if (macro) { - hash_table_remove (parser->defines, $2); + hash_table_remove (parser->defines, $3); ralloc_free (macro); } - ralloc_free ($2); + ralloc_free ($3); } -| HASH_IF conditional_tokens NEWLINE { +| HASH_IF { + glcpp_parser_resolve_version(parser); + } conditional_tokens NEWLINE { /* Be careful to only evaluate the 'if' expression if * we are not skipping. When we are skipping, we * simply push a new 0-valued 'if' onto the skip @@ -268,7 +280,7 @@ control_line: parser->skip_stack->type == SKIP_NO_SKIP) { _glcpp_parser_expand_and_lex_from (parser, - IF_EXPANDED, $2); + IF_EXPANDED, $3); } else { @@ -286,15 +298,19 @@ control_line: } _glcpp_parser_skip_stack_push_if (parser, & @1, 0); } -| HASH_IFDEF IDENTIFIER junk NEWLINE { - macro_t *macro = hash_table_find (parser->defines, $2); - ralloc_free ($2); +| HASH_IFDEF { + glcpp_parser_resolve_version(parser); + } IDENTIFIER junk NEWLINE { + macro_t *macro = hash_table_find (parser->defines, $3); + ralloc_free ($3); _glcpp_parser_skip_stack_push_if (parser, & @1, macro != NULL); } -| HASH_IFNDEF IDENTIFIER junk NEWLINE { - macro_t *macro = hash_table_find (parser->defines, $2); - ralloc_free ($2); - _glcpp_parser_skip_stack_push_if (parser, & @1, macro == NULL); +| HASH_IFNDEF { + glcpp_parser_resolve_version(parser); + } IDENTIFIER junk NEWLINE { + macro_t *macro = hash_table_find (parser->defines, $3); + ralloc_free ($3); + _glcpp_parser_skip_stack_push_if (parser, & @2, macro == NULL); } | HASH_ELIF conditional_tokens NEWLINE { /* Be careful to only evaluate the 'elif' expression @@ -358,12 +374,14 @@ control_line: _glcpp_parser_skip_stack_pop (parser, & @1); } NEWLINE | HASH_VERSION integer_constant NEWLINE { - _glcpp_parser_handle_version_declaration(parser, $2, NULL); + _glcpp_parser_handle_version_declaration(parser, $2, NULL, true); } | HASH_VERSION integer_constant IDENTIFIER NEWLINE { - _glcpp_parser_handle_version_declaration(parser, $2, $3); + _glcpp_parser_handle_version_declaration(parser, $2, $3, true); + } +| HASH NEWLINE { + glcpp_parser_resolve_version(parser); } -| HASH NEWLINE ; integer_constant: @@ -1168,10 +1186,9 @@ static void add_builtin_define(glcpp_parser_t *parser, } glcpp_parser_t * -glcpp_parser_create (const struct gl_extensions *extensions, int api) +glcpp_parser_create (const struct gl_extensions *extensions) { glcpp_parser_t *parser; - int language_version; parser = ralloc (NULL, glcpp_parser_t); @@ -1197,99 +1214,14 @@ glcpp_parser_create (const struct gl_extensions *extensions, int api) parser->info_log_length = 0; parser->error = 0; + parser->extensions = extensions; + parser->version_resolved = false; + parser->has_new_line_number = 0; parser->new_line_number = 1; parser->has_new_source_number = 0; parser->new_source_number = 0; - parser->is_gles = false; - - /* Add pre-defined macros. */ - if (api == API_OPENGLES2) { - parser->is_gles = true; - add_builtin_define(parser, "GL_ES", 1); - - if (extensions != NULL) { - if (extensions->OES_EGL_image_external) - add_builtin_define(parser, "GL_OES_EGL_image_external", 1); - } - } else { - add_builtin_define(parser, "GL_ARB_draw_buffers", 1); - add_builtin_define(parser, "GL_ARB_texture_rectangle", 1); - - if (extensions != NULL) { - if (extensions->EXT_texture_array) { - add_builtin_define(parser, "GL_EXT_texture_array", 1); - } - - if (extensions->ARB_fragment_coord_conventions) - add_builtin_define(parser, "GL_ARB_fragment_coord_conventions", - 1); - - if (extensions->ARB_explicit_attrib_location) - add_builtin_define(parser, "GL_ARB_explicit_attrib_location", 1); - - if (extensions->ARB_shader_texture_lod) - add_builtin_define(parser, "GL_ARB_shader_texture_lod", 1); - - if (extensions->ARB_draw_instanced) - add_builtin_define(parser, "GL_ARB_draw_instanced", 1); - - if (extensions->ARB_conservative_depth) { - add_builtin_define(parser, "GL_AMD_conservative_depth", 1); - add_builtin_define(parser, "GL_ARB_conservative_depth", 1); - } - - if (extensions->ARB_shader_bit_encoding) - add_builtin_define(parser, "GL_ARB_shader_bit_encoding", 1); - - if (extensions->ARB_uniform_buffer_object) - add_builtin_define(parser, "GL_ARB_uniform_buffer_object", 1); - - if (extensions->ARB_texture_cube_map_array) - add_builtin_define(parser, "GL_ARB_texture_cube_map_array", 1); - - if (extensions->ARB_shading_language_packing) - add_builtin_define(parser, "GL_ARB_shading_language_packing", 1); - - if (extensions->ARB_texture_multisample) - add_builtin_define(parser, "GL_ARB_texture_multisample", 1); - - if (extensions->ARB_texture_query_levels) - add_builtin_define(parser, "GL_ARB_texture_query_levels", 1); - - if (extensions->ARB_texture_query_lod) - add_builtin_define(parser, "GL_ARB_texture_query_lod", 1); - - if (extensions->ARB_gpu_shader5) - add_builtin_define(parser, "GL_ARB_gpu_shader5", 1); - - if (extensions->AMD_vertex_shader_layer) - add_builtin_define(parser, "GL_AMD_vertex_shader_layer", 1); - - if (extensions->ARB_shading_language_420pack) - add_builtin_define(parser, "GL_ARB_shading_language_420pack", 1); - - if (extensions->ARB_sample_shading) - add_builtin_define(parser, "GL_ARB_sample_shading", 1); - - if (extensions->EXT_shader_integer_mix) - add_builtin_define(parser, "GL_EXT_shader_integer_mix", 1); - - if (extensions->ARB_texture_gather) - add_builtin_define(parser, "GL_ARB_texture_gather", 1); - - if (extensions->ARB_shader_atomic_counters) - add_builtin_define(parser, "GL_ARB_shader_atomic_counters", 1); - - if (extensions->AMD_shader_trinary_minmax) - add_builtin_define(parser, "GL_AMD_shader_trinary_minmax", 1); - } - } - - language_version = 110; - add_builtin_define(parser, "__VERSION__", language_version); - return parser; } @@ -2087,24 +2019,106 @@ _glcpp_parser_skip_stack_pop (glcpp_parser_t *parser, YYLTYPE *loc) static void _glcpp_parser_handle_version_declaration(glcpp_parser_t *parser, intmax_t version, - const char *es_identifier) + const char *es_identifier, + bool explicitly_set) { - macro_t *macro = hash_table_find (parser->defines, "__VERSION__"); - if (macro) { - hash_table_remove (parser->defines, "__VERSION__"); - ralloc_free (macro); - } + const struct gl_extensions *extensions = parser->extensions; + + parser->version_resolved = true; + add_builtin_define (parser, "__VERSION__", version); - /* If we didn't have a GLES context to begin with, (indicated - * by parser->api), then the version declaration here might - * indicate GLES. */ - if (! parser->is_gles && - (version == 100 || - (es_identifier && (strcmp(es_identifier, "es") == 0)))) - { - parser->is_gles = true; - add_builtin_define (parser, "GL_ES", 1); + parser->is_gles = (version == 100) || + (es_identifier && + (strcmp(es_identifier, "es") == 0)); + + /* Add pre-defined macros. */ + if (parser->is_gles) { + add_builtin_define(parser, "GL_ES", 1); + + if (extensions != NULL) { + if (extensions->OES_EGL_image_external) + add_builtin_define(parser, "GL_OES_EGL_image_external", 1); + } + } else { + add_builtin_define(parser, "GL_ARB_draw_buffers", 1); + add_builtin_define(parser, "GL_ARB_texture_rectangle", 1); + + if (extensions != NULL) { + if (extensions->EXT_texture_array) + add_builtin_define(parser, "GL_EXT_texture_array", 1); + + if (extensions->ARB_arrays_of_arrays) + add_builtin_define(parser, "GL_ARB_arrays_of_arrays", 1); + + if (extensions->ARB_fragment_coord_conventions) + add_builtin_define(parser, "GL_ARB_fragment_coord_conventions", + 1); + + if (extensions->ARB_explicit_attrib_location) + add_builtin_define(parser, "GL_ARB_explicit_attrib_location", 1); + + if (extensions->ARB_shader_texture_lod) + add_builtin_define(parser, "GL_ARB_shader_texture_lod", 1); + + if (extensions->ARB_draw_instanced) + add_builtin_define(parser, "GL_ARB_draw_instanced", 1); + + if (extensions->ARB_conservative_depth) { + add_builtin_define(parser, "GL_AMD_conservative_depth", 1); + add_builtin_define(parser, "GL_ARB_conservative_depth", 1); + } + + if (extensions->ARB_shader_bit_encoding) + add_builtin_define(parser, "GL_ARB_shader_bit_encoding", 1); + + if (extensions->ARB_uniform_buffer_object) + add_builtin_define(parser, "GL_ARB_uniform_buffer_object", 1); + + if (extensions->ARB_texture_cube_map_array) + add_builtin_define(parser, "GL_ARB_texture_cube_map_array", 1); + + if (extensions->ARB_shading_language_packing) + add_builtin_define(parser, "GL_ARB_shading_language_packing", 1); + + if (extensions->ARB_texture_multisample) + add_builtin_define(parser, "GL_ARB_texture_multisample", 1); + + if (extensions->ARB_texture_query_levels) + add_builtin_define(parser, "GL_ARB_texture_query_levels", 1); + + if (extensions->ARB_texture_query_lod) + add_builtin_define(parser, "GL_ARB_texture_query_lod", 1); + + if (extensions->ARB_gpu_shader5) + add_builtin_define(parser, "GL_ARB_gpu_shader5", 1); + + if (extensions->AMD_vertex_shader_layer) + add_builtin_define(parser, "GL_AMD_vertex_shader_layer", 1); + + if (extensions->ARB_shading_language_420pack) + add_builtin_define(parser, "GL_ARB_shading_language_420pack", 1); + + if (extensions->ARB_sample_shading) + add_builtin_define(parser, "GL_ARB_sample_shading", 1); + + if (extensions->ARB_texture_gather) + add_builtin_define(parser, "GL_ARB_texture_gather", 1); + + if (extensions->ARB_shader_atomic_counters) + add_builtin_define(parser, "GL_ARB_shader_atomic_counters", 1); + + if (extensions->AMD_shader_trinary_minmax) + add_builtin_define(parser, "GL_AMD_shader_trinary_minmax", 1); + + if (extensions->ARB_viewport_array) + add_builtin_define(parser, "GL_ARB_viewport_array", 1); + } + } + + if (extensions != NULL) { + if (extensions->EXT_shader_integer_mix) + add_builtin_define(parser, "GL_EXT_shader_integer_mix", 1); } if (version >= 150) @@ -2118,8 +2132,23 @@ _glcpp_parser_handle_version_declaration(glcpp_parser_t *parser, intmax_t versio if (version >= 130 || parser->is_gles) add_builtin_define (parser, "GL_FRAGMENT_PRECISION_HIGH", 1); - ralloc_asprintf_rewrite_tail (&parser->output, &parser->output_length, - "#version %" PRIiMAX "%s%s", version, - es_identifier ? " " : "", - es_identifier ? es_identifier : ""); + if (explicitly_set) { + ralloc_asprintf_rewrite_tail (&parser->output, &parser->output_length, + "#version %" PRIiMAX "%s%s", version, + es_identifier ? " " : "", + es_identifier ? es_identifier : ""); + } +} + +/* GLSL version is no version is explicitly specified. */ +#define IMPLICIT_GLSL_VERSION 110 + +void +glcpp_parser_resolve_version(glcpp_parser_t *parser) +{ + if (parser->version_resolved) + return; + + _glcpp_parser_handle_version_declaration(parser, IMPLICIT_GLSL_VERSION, + NULL, false); } diff --git a/mesalib/src/glsl/glcpp/glcpp.c b/mesalib/src/glsl/glcpp/glcpp.c index 13a32fd7d..b632f1715 100644 --- a/mesalib/src/glsl/glcpp/glcpp.c +++ b/mesalib/src/glsl/glcpp/glcpp.c @@ -109,7 +109,6 @@ load_text_file(void *ctx, const char *filename) static void init_fake_gl_context (struct gl_context *gl_ctx) { - gl_ctx->API = API_OPENGL_COMPAT; gl_ctx->Const.DisableGLSLLineContinuations = false; } diff --git a/mesalib/src/glsl/glcpp/glcpp.h b/mesalib/src/glsl/glcpp/glcpp.h index 85f3fdcd2..4aa200a63 100644 --- a/mesalib/src/glsl/glcpp/glcpp.h +++ b/mesalib/src/glsl/glcpp/glcpp.h @@ -182,6 +182,8 @@ struct glcpp_parser { size_t output_length; size_t info_log_length; int error; + const struct gl_extensions *extensions; + bool version_resolved; bool has_new_line_number; int new_line_number; bool has_new_source_number; @@ -192,7 +194,7 @@ struct glcpp_parser { struct gl_extensions; glcpp_parser_t * -glcpp_parser_create (const struct gl_extensions *extensions, int api); +glcpp_parser_create (const struct gl_extensions *extensions); int glcpp_parser_parse (glcpp_parser_t *parser); @@ -200,6 +202,9 @@ glcpp_parser_parse (glcpp_parser_t *parser); void glcpp_parser_destroy (glcpp_parser_t *parser); +void +glcpp_parser_resolve_version(glcpp_parser_t *parser); + int glcpp_preprocess(void *ralloc_ctx, const char **shader, char **info_log, const struct gl_extensions *extensions, struct gl_context *g_ctx); diff --git a/mesalib/src/glsl/glcpp/pp.c b/mesalib/src/glsl/glcpp/pp.c index 7e1b6c689..637a58f9c 100644 --- a/mesalib/src/glsl/glcpp/pp.c +++ b/mesalib/src/glsl/glcpp/pp.c @@ -139,7 +139,7 @@ glcpp_preprocess(void *ralloc_ctx, const char **shader, char **info_log, const struct gl_extensions *extensions, struct gl_context *gl_ctx) { int errors; - glcpp_parser_t *parser = glcpp_parser_create (extensions, gl_ctx->API); + glcpp_parser_t *parser = glcpp_parser_create (extensions); if (! gl_ctx->Const.DisableGLSLLineContinuations) *shader = remove_line_continuations(parser, *shader); @@ -151,6 +151,8 @@ glcpp_preprocess(void *ralloc_ctx, const char **shader, char **info_log, if (parser->skip_stack) glcpp_error (&parser->skip_stack->loc, parser, "Unterminated #if\n"); + glcpp_parser_resolve_version(parser); + ralloc_strcat(info_log, parser->info_log); ralloc_steal(ralloc_ctx, parser->output); diff --git a/mesalib/src/glsl/glsl_parser.yy b/mesalib/src/glsl/glsl_parser.yy index 52f674b63..5d8dfac02 100644 --- a/mesalib/src/glsl/glsl_parser.yy +++ b/mesalib/src/glsl/glsl_parser.yy @@ -98,6 +98,7 @@ static bool match_layout_qualifier(const char *s1, const char *s2, ast_node *node; ast_type_specifier *type_specifier; + ast_array_specifier *array_specifier; ast_fully_specified_type *fully_specified_type; ast_function *function; ast_parameter_declarator *parameter_declarator; @@ -203,6 +204,7 @@ static bool match_layout_qualifier(const char *s1, const char *s2, %type <type_qualifier> interface_qualifier %type <type_specifier> type_specifier %type <type_specifier> type_specifier_nonarray +%type <array_specifier> array_specifier %type <identifier> basic_type_specifier_nonarray %type <fully_specified_type> fully_specified_type %type <function> function_prototype @@ -881,7 +883,7 @@ parameter_declarator: $$->type->specifier = $1; $$->identifier = $2; } - | type_specifier any_identifier '[' constant_expression ']' + | type_specifier any_identifier array_specifier { void *ctx = state; $$ = new(ctx) ast_parameter_declarator(); @@ -890,8 +892,7 @@ parameter_declarator: $$->type->set_location(yylloc); $$->type->specifier = $1; $$->identifier = $2; - $$->is_array = true; - $$->array_size = $4; + $$->array_specifier = $3; } ; @@ -977,76 +978,42 @@ init_declarator_list: | init_declarator_list ',' any_identifier { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($3, false, NULL, NULL); + ast_declaration *decl = new(ctx) ast_declaration($3, NULL, NULL); decl->set_location(yylloc); $$ = $1; $$->declarations.push_tail(&decl->link); state->symbols->add_variable(new(state) ir_variable(NULL, $3, ir_var_auto)); } - | init_declarator_list ',' any_identifier '[' ']' + | init_declarator_list ',' any_identifier array_specifier { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($3, true, NULL, NULL); + ast_declaration *decl = new(ctx) ast_declaration($3, $4, NULL); decl->set_location(yylloc); $$ = $1; $$->declarations.push_tail(&decl->link); state->symbols->add_variable(new(state) ir_variable(NULL, $3, ir_var_auto)); } - | init_declarator_list ',' any_identifier '[' constant_expression ']' + | init_declarator_list ',' any_identifier array_specifier '=' initializer { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($3, true, $5, NULL); + ast_declaration *decl = new(ctx) ast_declaration($3, $4, $6); decl->set_location(yylloc); $$ = $1; $$->declarations.push_tail(&decl->link); state->symbols->add_variable(new(state) ir_variable(NULL, $3, ir_var_auto)); } - | init_declarator_list ',' any_identifier '[' ']' '=' initializer - { - void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($3, true, NULL, $7); - decl->set_location(yylloc); - - $$ = $1; - $$->declarations.push_tail(&decl->link); - state->symbols->add_variable(new(state) ir_variable(NULL, $3, ir_var_auto)); - if ($7->oper == ast_aggregate) { - ast_aggregate_initializer *ai = (ast_aggregate_initializer *)$7; - ast_type_specifier *type = new(ctx) ast_type_specifier($1->type->specifier, true, NULL); - _mesa_ast_set_aggregate_type(type, ai, state); - } - } - | init_declarator_list ',' any_identifier '[' constant_expression ']' '=' initializer - { - void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($3, true, $5, $8); - decl->set_location(yylloc); - - $$ = $1; - $$->declarations.push_tail(&decl->link); - state->symbols->add_variable(new(state) ir_variable(NULL, $3, ir_var_auto)); - if ($8->oper == ast_aggregate) { - ast_aggregate_initializer *ai = (ast_aggregate_initializer *)$8; - ast_type_specifier *type = new(ctx) ast_type_specifier($1->type->specifier, true, $5); - _mesa_ast_set_aggregate_type(type, ai, state); - } - } | init_declarator_list ',' any_identifier '=' initializer { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($3, false, NULL, $5); + ast_declaration *decl = new(ctx) ast_declaration($3, NULL, $5); decl->set_location(yylloc); $$ = $1; $$->declarations.push_tail(&decl->link); state->symbols->add_variable(new(state) ir_variable(NULL, $3, ir_var_auto)); - if ($5->oper == ast_aggregate) { - ast_aggregate_initializer *ai = (ast_aggregate_initializer *)$5; - _mesa_ast_set_aggregate_type($1->type->specifier, ai, state); - } } ; @@ -1062,74 +1029,43 @@ single_declaration: | fully_specified_type any_identifier { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, false, NULL, NULL); + ast_declaration *decl = new(ctx) ast_declaration($2, NULL, NULL); $$ = new(ctx) ast_declarator_list($1); $$->set_location(yylloc); $$->declarations.push_tail(&decl->link); } - | fully_specified_type any_identifier '[' ']' + | fully_specified_type any_identifier array_specifier { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, true, NULL, NULL); + ast_declaration *decl = new(ctx) ast_declaration($2, $3, NULL); $$ = new(ctx) ast_declarator_list($1); $$->set_location(yylloc); $$->declarations.push_tail(&decl->link); } - | fully_specified_type any_identifier '[' constant_expression ']' + | fully_specified_type any_identifier array_specifier '=' initializer { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, true, $4, NULL); + ast_declaration *decl = new(ctx) ast_declaration($2, $3, $5); $$ = new(ctx) ast_declarator_list($1); $$->set_location(yylloc); $$->declarations.push_tail(&decl->link); } - | fully_specified_type any_identifier '[' ']' '=' initializer - { - void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, true, NULL, $6); - - $$ = new(ctx) ast_declarator_list($1); - $$->set_location(yylloc); - $$->declarations.push_tail(&decl->link); - if ($6->oper == ast_aggregate) { - ast_aggregate_initializer *ai = (ast_aggregate_initializer *)$6; - ast_type_specifier *type = new(ctx) ast_type_specifier($1->specifier, true, NULL); - _mesa_ast_set_aggregate_type(type, ai, state); - } - } - | fully_specified_type any_identifier '[' constant_expression ']' '=' initializer - { - void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, true, $4, $7); - - $$ = new(ctx) ast_declarator_list($1); - $$->set_location(yylloc); - $$->declarations.push_tail(&decl->link); - if ($7->oper == ast_aggregate) { - ast_aggregate_initializer *ai = (ast_aggregate_initializer *)$7; - ast_type_specifier *type = new(ctx) ast_type_specifier($1->specifier, true, $4); - _mesa_ast_set_aggregate_type(type, ai, state); - } - } | fully_specified_type any_identifier '=' initializer { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, false, NULL, $4); + ast_declaration *decl = new(ctx) ast_declaration($2, NULL, $4); $$ = new(ctx) ast_declarator_list($1); $$->set_location(yylloc); $$->declarations.push_tail(&decl->link); - if ($4->oper == ast_aggregate) { - _mesa_ast_set_aggregate_type($1->specifier, $4, state); - } } | INVARIANT variable_identifier // Vertex only. { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, false, NULL, NULL); + ast_declaration *decl = new(ctx) ast_declaration($2, NULL, NULL); $$ = new(ctx) ast_declarator_list(NULL); $$->set_location(yylloc); @@ -1612,19 +1548,51 @@ storage_qualifier: } ; -type_specifier: - type_specifier_nonarray - | type_specifier_nonarray '[' ']' +array_specifier: + '[' ']' + { + void *ctx = state; + $$ = new(ctx) ast_array_specifier(yylloc); + } + | '[' constant_expression ']' + { + void *ctx = state; + $$ = new(ctx) ast_array_specifier(yylloc, $2); + } + | array_specifier '[' ']' { $$ = $1; - $$->is_array = true; - $$->array_size = NULL; + + if (!state->ARB_arrays_of_arrays_enable) { + _mesa_glsl_error(& @1, state, + "GL_ARB_arrays_of_arrays " + "required for defining arrays of arrays"); + } else { + _mesa_glsl_error(& @1, state, + "only the outermost array dimension can " + "be unsized"); + } } - | type_specifier_nonarray '[' constant_expression ']' + | array_specifier '[' constant_expression ']' { $$ = $1; - $$->is_array = true; - $$->array_size = $3; + + if (!state->ARB_arrays_of_arrays_enable) { + _mesa_glsl_error(& @1, state, + "GL_ARB_arrays_of_arrays " + "required for defining arrays of arrays"); + } + + $$->add_dimension($3); + } + ; + +type_specifier: + type_specifier_nonarray + | type_specifier_nonarray array_specifier + { + $$ = $1; + $$->array_specifier = $2; } ; @@ -1804,19 +1772,13 @@ struct_declarator: any_identifier { void *ctx = state; - $$ = new(ctx) ast_declaration($1, false, NULL, NULL); - $$->set_location(yylloc); - } - | any_identifier '[' ']' - { - void *ctx = state; - $$ = new(ctx) ast_declaration($1, true, NULL, NULL); + $$ = new(ctx) ast_declaration($1, NULL, NULL); $$->set_location(yylloc); } - | any_identifier '[' constant_expression ']' + | any_identifier array_specifier { void *ctx = state; - $$ = new(ctx) ast_declaration($1, true, $3, NULL); + $$ = new(ctx) ast_declaration($1, $2, NULL); $$->set_location(yylloc); } ; @@ -1974,7 +1936,7 @@ condition: | fully_specified_type any_identifier '=' initializer { void *ctx = state; - ast_declaration *decl = new(ctx) ast_declaration($2, false, NULL, $4); + ast_declaration *decl = new(ctx) ast_declaration($2, NULL, $4); ast_declarator_list *declarator = new(ctx) ast_declarator_list($1); decl->set_location(yylloc); declarator->set_location(yylloc); @@ -2309,22 +2271,17 @@ instance_name_opt: /* empty */ { $$ = new(state) ast_interface_block(*state->default_uniform_qualifier, - NULL, false, NULL); + NULL, NULL); } | NEW_IDENTIFIER { $$ = new(state) ast_interface_block(*state->default_uniform_qualifier, - $1, false, NULL); - } - | NEW_IDENTIFIER '[' constant_expression ']' - { - $$ = new(state) ast_interface_block(*state->default_uniform_qualifier, - $1, true, $3); + $1, NULL); } - | NEW_IDENTIFIER '[' ']' + | NEW_IDENTIFIER array_specifier { $$ = new(state) ast_interface_block(*state->default_uniform_qualifier, - $1, true, NULL); + $1, $2); } ; diff --git a/mesalib/src/glsl/glsl_parser_extras.cpp b/mesalib/src/glsl/glsl_parser_extras.cpp index 21dc3abd7..87784ed69 100644 --- a/mesalib/src/glsl/glsl_parser_extras.cpp +++ b/mesalib/src/glsl/glsl_parser_extras.cpp @@ -50,7 +50,7 @@ glsl_compute_version_string(void *mem_ctx, bool is_es, unsigned version) static unsigned known_desktop_glsl_versions[] = - { 110, 120, 130, 140, 150, 330, 400, 410, 420, 430 }; + { 110, 120, 130, 140, 150, 330, 400, 410, 420, 430, 440 }; _mesa_glsl_parse_state::_mesa_glsl_parse_state(struct gl_context *_ctx, @@ -291,6 +291,10 @@ _mesa_glsl_parse_state::process_version_directive(YYLTYPE *locp, int version, } } + if (this->es_shader) { + this->ARB_texture_rectangle_enable = false; + } + this->language_version = version; bool supported = false; @@ -484,6 +488,7 @@ struct _mesa_glsl_extension { static const _mesa_glsl_extension _mesa_glsl_supported_extensions[] = { /* API availability */ /* name GL ES supported flag */ + EXT(ARB_arrays_of_arrays, true, false, ARB_arrays_of_arrays), EXT(ARB_conservative_depth, true, false, ARB_conservative_depth), EXT(ARB_draw_buffers, true, false, dummy_true), EXT(ARB_draw_instanced, true, false, ARB_draw_instanced), @@ -513,6 +518,7 @@ static const _mesa_glsl_extension _mesa_glsl_supported_extensions[] = { EXT(ARB_shader_atomic_counters, true, false, ARB_shader_atomic_counters), EXT(ARB_sample_shading, true, false, ARB_sample_shading), EXT(AMD_shader_trinary_minmax, true, false, dummy_true), + EXT(ARB_viewport_array, true, false, ARB_viewport_array), }; #undef EXT @@ -634,25 +640,6 @@ _mesa_glsl_process_extension(const char *name, YYLTYPE *name_locp, /** - * Returns the name of the type of a column of a matrix. E.g., - * - * "mat3" -> "vec3" - * "mat4x2" -> "vec2" - */ -static const char * -_mesa_ast_get_matrix_column_type_name(const char *matrix_type_name) -{ - static const char *vec_name[] = { "vec2", "vec3", "vec4" }; - - /* The number of elements in a row of a matrix is specified by the last - * character of the matrix type name. - */ - long rows = strtol(matrix_type_name + strlen(matrix_type_name) - 1, - NULL, 10); - return vec_name[rows - 2]; -} - -/** * Recurses through <type> and <expr> if <expr> is an aggregate initializer * and sets <expr>'s <constructor_type> field to <type>. Gives later functions * (process_array_constructor, et al) sufficient information to do type @@ -699,37 +686,19 @@ _mesa_ast_get_matrix_column_type_name(const char *matrix_type_name) * doesn't contain sufficient information to determine if the types match. */ void -_mesa_ast_set_aggregate_type(const ast_type_specifier *type, - ast_expression *expr, - _mesa_glsl_parse_state *state) +_mesa_ast_set_aggregate_type(const glsl_type *type, + ast_expression *expr) { - void *ctx = state; ast_aggregate_initializer *ai = (ast_aggregate_initializer *)expr; - ai->constructor_type = (ast_type_specifier *)type; - - bool is_declaration = ai->constructor_type->structure != NULL; - if (!is_declaration) { - /* Look up <type> name in the symbol table to see if it's a struct. */ - const ast_type_specifier *struct_type = - state->symbols->get_type_ast(type->type_name); - ai->constructor_type->structure = - struct_type ? new(ctx) ast_struct_specifier(*struct_type->structure) - : NULL; - } + ai->constructor_type = type; /* If the aggregate is an array, recursively set its elements' types. */ - if (type->is_array) { - /* We want to set the element type which is not an array itself, so make - * a copy of the array type and set its is_array field to false. + if (type->is_array()) { + /* Each array element has the type type->element_type(). * * E.g., if <type> if struct S[2] we want to set each element's type to * struct S. - * - * FINISHME: Update when ARB_array_of_arrays is supported. */ - const ast_type_specifier *non_array_type = - new(ctx) ast_type_specifier(type, false, NULL); - for (exec_node *expr_node = ai->expressions.head; !expr_node->is_tail_sentinel(); expr_node = expr_node->next) { @@ -737,84 +706,33 @@ _mesa_ast_set_aggregate_type(const ast_type_specifier *type, link); if (expr->oper == ast_aggregate) - _mesa_ast_set_aggregate_type(non_array_type, expr, state); + _mesa_ast_set_aggregate_type(type->element_type(), expr); } /* If the aggregate is a struct, recursively set its fields' types. */ - } else if (ai->constructor_type->structure) { - ai->constructor_type->structure->is_declaration = is_declaration; + } else if (type->is_record()) { exec_node *expr_node = ai->expressions.head; - /* Iterate through the struct's fields' declarations. E.g., iterate from - * "float a, b" to "int c" in the struct below. - * - * struct { - * float a, b; - * int c; - * } s; - */ - for (exec_node *decl_list_node = - ai->constructor_type->structure->declarations.head; - !decl_list_node->is_tail_sentinel(); - decl_list_node = decl_list_node->next) { - ast_declarator_list *decl_list = exec_node_data(ast_declarator_list, - decl_list_node, link); - - for (exec_node *decl_node = decl_list->declarations.head; - !decl_node->is_tail_sentinel() && !expr_node->is_tail_sentinel(); - decl_node = decl_node->next, expr_node = expr_node->next) { - ast_declaration *decl = exec_node_data(ast_declaration, decl_node, - link); - ast_expression *expr = exec_node_data(ast_expression, expr_node, - link); - - bool is_array = decl_list->type->specifier->is_array; - ast_expression *array_size = decl_list->type->specifier->array_size; - - /* Recognize variable declarations with the bracketed size attached - * to the type rather than the variable name as arrays. E.g., - * - * float a[2]; - * float[2] b; - * - * are both arrays, but <a>'s array_size is decl->array_size, while - * <b>'s array_size is decl_list->type->specifier->array_size. - */ - if (!is_array) { - /* FINISHME: Update when ARB_array_of_arrays is supported. */ - is_array = decl->is_array; - array_size = decl->array_size; - } - - /* Declaration shadows the <type> parameter. */ - ast_type_specifier *type = - new(ctx) ast_type_specifier(decl_list->type->specifier, - is_array, array_size); + /* Iterate through the struct's fields. */ + for (unsigned i = 0; !expr_node->is_tail_sentinel() && i < type->length; + i++, expr_node = expr_node->next) { + ast_expression *expr = exec_node_data(ast_expression, expr_node, + link); - if (expr->oper == ast_aggregate) - _mesa_ast_set_aggregate_type(type, expr, state); + if (expr->oper == ast_aggregate) { + _mesa_ast_set_aggregate_type(type->fields.structure[i].type, expr); } } - } else { - /* If the aggregate is a matrix, set its columns' types. */ - const char *name; - const glsl_type *const constructor_type = - ai->constructor_type->glsl_type(&name, state); - - if (constructor_type->is_matrix()) { - for (exec_node *expr_node = ai->expressions.head; - !expr_node->is_tail_sentinel(); - expr_node = expr_node->next) { - ast_expression *expr = exec_node_data(ast_expression, expr_node, - link); - - /* Declaration shadows the <type> parameter. */ - ast_type_specifier *type = new(ctx) - ast_type_specifier(_mesa_ast_get_matrix_column_type_name(name)); - - if (expr->oper == ast_aggregate) - _mesa_ast_set_aggregate_type(type, expr, state); - } + /* If the aggregate is a matrix, set its columns' types. */ + } else if (type->is_matrix()) { + for (exec_node *expr_node = ai->expressions.head; + !expr_node->is_tail_sentinel(); + expr_node = expr_node->next) { + ast_expression *expr = exec_node_data(ast_expression, expr_node, + link); + + if (expr->oper == ast_aggregate) + _mesa_ast_set_aggregate_type(type->column_type(), expr); } } } @@ -876,16 +794,10 @@ ast_node::ast_node(void) static void -ast_opt_array_size_print(bool is_array, const ast_expression *array_size) +ast_opt_array_dimensions_print(const ast_array_specifier *array_specifier) { - if (is_array) { - printf("[ "); - - if (array_size) - array_size->print(); - - printf("] "); - } + if (array_specifier) + array_specifier->print(); } @@ -1108,7 +1020,7 @@ ast_parameter_declarator::print(void) const type->print(); if (identifier) printf("%s ", identifier); - ast_opt_array_size_print(is_array, array_size); + ast_opt_array_dimensions_print(array_specifier); } @@ -1124,7 +1036,7 @@ void ast_declaration::print(void) const { printf("%s ", identifier); - ast_opt_array_size_print(is_array, array_size); + ast_opt_array_dimensions_print(array_specifier); if (initializer) { printf("= "); @@ -1133,13 +1045,12 @@ ast_declaration::print(void) const } -ast_declaration::ast_declaration(const char *identifier, bool is_array, - ast_expression *array_size, +ast_declaration::ast_declaration(const char *identifier, + ast_array_specifier *array_specifier, ast_expression *initializer) { this->identifier = identifier; - this->is_array = is_array; - this->array_size = array_size; + this->array_specifier = array_specifier; this->initializer = initializer; } @@ -1560,9 +1471,13 @@ do_common_optimization(exec_list *ir, bool linked, progress = do_copy_propagation(ir) || progress; progress = do_copy_propagation_elements(ir) || progress; - if (options->PreferDP4 && !linked) + if (options->OptimizeForAOS && !linked) progress = opt_flip_matrices(ir) || progress; + if (linked && options->OptimizeForAOS) { + progress = do_vectorize(ir) || progress; + } + if (linked) progress = do_dead_code(ir, uniform_locations_assigned) || progress; else diff --git a/mesalib/src/glsl/glsl_parser_extras.h b/mesalib/src/glsl/glsl_parser_extras.h index 21a27c9ea..8c900e637 100644 --- a/mesalib/src/glsl/glsl_parser_extras.h +++ b/mesalib/src/glsl/glsl_parser_extras.h @@ -298,6 +298,8 @@ struct _mesa_glsl_parse_state { * \name Enable bits for GLSL extensions */ /*@{*/ + bool ARB_arrays_of_arrays_enable; + bool ARB_arrays_of_arrays_warn; bool ARB_draw_buffers_enable; bool ARB_draw_buffers_warn; bool ARB_draw_instanced_enable; @@ -356,6 +358,8 @@ struct _mesa_glsl_parse_state { bool ARB_shader_atomic_counters_warn; bool AMD_shader_trinary_minmax_enable; bool AMD_shader_trinary_minmax_warn; + bool ARB_viewport_array_enable; + bool ARB_viewport_array_warn; /*@}*/ /** Extensions supported by the OpenGL implementation. */ diff --git a/mesalib/src/glsl/glsl_types.cpp b/mesalib/src/glsl/glsl_types.cpp index 12d4ac0ee..1b0b3ef88 100644 --- a/mesalib/src/glsl/glsl_types.cpp +++ b/mesalib/src/glsl/glsl_types.cpp @@ -300,8 +300,20 @@ glsl_type::glsl_type(const glsl_type *array, unsigned length) : if (length == 0) snprintf(n, name_length, "%s[]", array->name); - else - snprintf(n, name_length, "%s[%u]", array->name, length); + else { + /* insert outermost dimensions in the correct spot + * otherwise the dimension order will be backwards + */ + const char *pos = strchr(array->name, '['); + if (pos) { + int idx = pos - array->name; + snprintf(n, idx+1, "%s", array->name); + snprintf(n + idx, name_length - idx, "[%u]%s", + length, array->name + idx); + } else { + snprintf(n, name_length, "%s[%u]", array->name, length); + } + } this->name = n; } @@ -449,6 +461,42 @@ glsl_type::get_array_instance(const glsl_type *base, unsigned array_size) } +bool +glsl_type::record_compare(const glsl_type *b) const +{ + if (this->length != b->length) + return false; + + if (this->interface_packing != b->interface_packing) + return false; + + for (unsigned i = 0; i < this->length; i++) { + if (this->fields.structure[i].type != b->fields.structure[i].type) + return false; + if (strcmp(this->fields.structure[i].name, + b->fields.structure[i].name) != 0) + return false; + if (this->fields.structure[i].row_major + != b->fields.structure[i].row_major) + return false; + if (this->fields.structure[i].location + != b->fields.structure[i].location) + return false; + if (this->fields.structure[i].interpolation + != b->fields.structure[i].interpolation) + return false; + if (this->fields.structure[i].centroid + != b->fields.structure[i].centroid) + return false; + if (this->fields.structure[i].sample + != b->fields.structure[i].sample) + return false; + } + + return true; +} + + int glsl_type::record_key_compare(const void *a, const void *b) { @@ -461,36 +509,7 @@ glsl_type::record_key_compare(const void *a, const void *b) if (strcmp(key1->name, key2->name) != 0) return 1; - if (key1->length != key2->length) - return 1; - - if (key1->interface_packing != key2->interface_packing) - return 1; - - for (unsigned i = 0; i < key1->length; i++) { - if (key1->fields.structure[i].type != key2->fields.structure[i].type) - return 1; - if (strcmp(key1->fields.structure[i].name, - key2->fields.structure[i].name) != 0) - return 1; - if (key1->fields.structure[i].row_major - != key2->fields.structure[i].row_major) - return 1; - if (key1->fields.structure[i].location - != key2->fields.structure[i].location) - return 1; - if (key1->fields.structure[i].interpolation - != key2->fields.structure[i].interpolation) - return 1; - if (key1->fields.structure[i].centroid - != key2->fields.structure[i].centroid) - return 1; - if (key1->fields.structure[i].sample - != key2->fields.structure[i].sample) - return 1; - } - - return 0; + return !key1->record_compare(key2); } diff --git a/mesalib/src/glsl/glsl_types.h b/mesalib/src/glsl/glsl_types.h index fb7c9288d..f88758a9a 100644 --- a/mesalib/src/glsl/glsl_types.h +++ b/mesalib/src/glsl/glsl_types.h @@ -542,6 +542,13 @@ struct glsl_type { */ int sampler_coordinate_components() const; + /** + * Compare a record type against another record type. + * + * This is useful for matching record types declared across shader stages. + */ + bool record_compare(const glsl_type *b) const; + private: /** * ralloc context for all glsl_type allocations diff --git a/mesalib/src/glsl/ir.h b/mesalib/src/glsl/ir.h index 2ae8513a6..19e8383b2 100644 --- a/mesalib/src/glsl/ir.h +++ b/mesalib/src/glsl/ir.h @@ -148,7 +148,7 @@ public: * in particular. No support for other instruction types (assignments, * jumps, calls, etc.) is planned. */ - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); protected: ir_instruction() @@ -1413,7 +1413,7 @@ public: return this; } - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); virtual ir_expression *clone(void *mem_ctx, struct hash_table *ht) const; @@ -1741,7 +1741,7 @@ public: virtual ir_visitor_status accept(ir_hierarchical_visitor *); - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); /** * Return a string representing the ir_texture_opcode. @@ -1847,7 +1847,7 @@ public: virtual ir_visitor_status accept(ir_hierarchical_visitor *); - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); bool is_lvalue() const { @@ -1913,7 +1913,7 @@ public: return this; } - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); /** * Get the variable that is ultimately referenced by an r-value @@ -1973,7 +1973,7 @@ public: return this; } - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); /** * Get the variable that is ultimately referenced by an r-value @@ -2109,7 +2109,7 @@ public: virtual ir_visitor_status accept(ir_hierarchical_visitor *); - virtual bool equals(ir_instruction *ir); + virtual bool equals(ir_instruction *ir, enum ir_node_type ignore = ir_type_unset); /** * Get a particular component of a constant as a specific type diff --git a/mesalib/src/glsl/ir_equals.cpp b/mesalib/src/glsl/ir_equals.cpp index 7cfe1e66b..484530019 100644 --- a/mesalib/src/glsl/ir_equals.cpp +++ b/mesalib/src/glsl/ir_equals.cpp @@ -28,12 +28,12 @@ * can't access a's vtable in that case. */ static bool -possibly_null_equals(ir_instruction *a, ir_instruction *b) +possibly_null_equals(ir_instruction *a, ir_instruction *b, enum ir_node_type ignore) { if (!a || !b) return !a && !b; - return a->equals(b); + return a->equals(b, ignore); } /** @@ -41,13 +41,13 @@ possibly_null_equals(ir_instruction *a, ir_instruction *b) * about. */ bool -ir_instruction::equals(ir_instruction *ir) +ir_instruction::equals(ir_instruction *ir, enum ir_node_type) { return false; } bool -ir_constant::equals(ir_instruction *ir) +ir_constant::equals(ir_instruction *ir, enum ir_node_type ignore) { const ir_constant *other = ir->as_constant(); if (!other) @@ -65,7 +65,7 @@ ir_constant::equals(ir_instruction *ir) } bool -ir_dereference_variable::equals(ir_instruction *ir) +ir_dereference_variable::equals(ir_instruction *ir, enum ir_node_type ignore) { const ir_dereference_variable *other = ir->as_dereference_variable(); if (!other) @@ -75,7 +75,7 @@ ir_dereference_variable::equals(ir_instruction *ir) } bool -ir_dereference_array::equals(ir_instruction *ir) +ir_dereference_array::equals(ir_instruction *ir, enum ir_node_type ignore) { const ir_dereference_array *other = ir->as_dereference_array(); if (!other) @@ -84,17 +84,17 @@ ir_dereference_array::equals(ir_instruction *ir) if (type != other->type) return false; - if (!array->equals(other->array)) + if (!array->equals(other->array, ignore)) return false; - if (!array_index->equals(other->array_index)) + if (!array_index->equals(other->array_index, ignore)) return false; return true; } bool -ir_swizzle::equals(ir_instruction *ir) +ir_swizzle::equals(ir_instruction *ir, enum ir_node_type ignore) { const ir_swizzle *other = ir->as_swizzle(); if (!other) @@ -103,18 +103,20 @@ ir_swizzle::equals(ir_instruction *ir) if (type != other->type) return false; - if (mask.x != other->mask.x || - mask.y != other->mask.y || - mask.z != other->mask.z || - mask.w != other->mask.w) { - return false; + if (ignore != ir_type_swizzle) { + if (mask.x != other->mask.x || + mask.y != other->mask.y || + mask.z != other->mask.z || + mask.w != other->mask.w) { + return false; + } } - return val->equals(other->val); + return val->equals(other->val, ignore); } bool -ir_texture::equals(ir_instruction *ir) +ir_texture::equals(ir_instruction *ir, enum ir_node_type ignore) { const ir_texture *other = ir->as_texture(); if (!other) @@ -126,19 +128,19 @@ ir_texture::equals(ir_instruction *ir) if (op != other->op) return false; - if (!possibly_null_equals(coordinate, other->coordinate)) + if (!possibly_null_equals(coordinate, other->coordinate, ignore)) return false; - if (!possibly_null_equals(projector, other->projector)) + if (!possibly_null_equals(projector, other->projector, ignore)) return false; - if (!possibly_null_equals(shadow_comparitor, other->shadow_comparitor)) + if (!possibly_null_equals(shadow_comparitor, other->shadow_comparitor, ignore)) return false; - if (!possibly_null_equals(offset, other->offset)) + if (!possibly_null_equals(offset, other->offset, ignore)) return false; - if (!sampler->equals(other->sampler)) + if (!sampler->equals(other->sampler, ignore)) return false; switch (op) { @@ -147,26 +149,26 @@ ir_texture::equals(ir_instruction *ir) case ir_query_levels: break; case ir_txb: - if (!lod_info.bias->equals(other->lod_info.bias)) + if (!lod_info.bias->equals(other->lod_info.bias, ignore)) return false; break; case ir_txl: case ir_txf: case ir_txs: - if (!lod_info.lod->equals(other->lod_info.lod)) + if (!lod_info.lod->equals(other->lod_info.lod, ignore)) return false; break; case ir_txd: - if (!lod_info.grad.dPdx->equals(other->lod_info.grad.dPdx) || - !lod_info.grad.dPdy->equals(other->lod_info.grad.dPdy)) + if (!lod_info.grad.dPdx->equals(other->lod_info.grad.dPdx, ignore) || + !lod_info.grad.dPdy->equals(other->lod_info.grad.dPdy, ignore)) return false; break; case ir_txf_ms: - if (!lod_info.sample_index->equals(other->lod_info.sample_index)) + if (!lod_info.sample_index->equals(other->lod_info.sample_index, ignore)) return false; break; case ir_tg4: - if (!lod_info.component->equals(other->lod_info.component)) + if (!lod_info.component->equals(other->lod_info.component, ignore)) return false; break; default: @@ -177,7 +179,7 @@ ir_texture::equals(ir_instruction *ir) } bool -ir_expression::equals(ir_instruction *ir) +ir_expression::equals(ir_instruction *ir, enum ir_node_type ignore) { const ir_expression *other = ir->as_expression(); if (!other) @@ -190,7 +192,7 @@ ir_expression::equals(ir_instruction *ir) return false; for (unsigned i = 0; i < get_num_operands(); i++) { - if (!operands[i]->equals(other->operands[i])) + if (!operands[i]->equals(other->operands[i], ignore)) return false; } diff --git a/mesalib/src/glsl/ir_optimization.h b/mesalib/src/glsl/ir_optimization.h index 3ca9f5744..055d65547 100644 --- a/mesalib/src/glsl/ir_optimization.h +++ b/mesalib/src/glsl/ir_optimization.h @@ -98,6 +98,7 @@ bool do_mat_op_to_vec(exec_list *instructions); bool do_noop_swizzle(exec_list *instructions); bool do_structure_splitting(exec_list *instructions); bool do_swizzle_swizzle(exec_list *instructions); +bool do_vectorize(exec_list *instructions); bool do_tree_grafting(exec_list *instructions); bool do_vec_index_to_cond_assign(exec_list *instructions); bool do_vec_index_to_swizzle(exec_list *instructions); diff --git a/mesalib/src/glsl/link_atomics.cpp b/mesalib/src/glsl/link_atomics.cpp index db9c53965..d92cdb117 100644 --- a/mesalib/src/glsl/link_atomics.cpp +++ b/mesalib/src/glsl/link_atomics.cpp @@ -105,9 +105,10 @@ namespace { ir_variable *var = ((ir_instruction *)node)->as_variable(); if (var && var->type->contains_atomic()) { - unsigned id; + unsigned id = 0; bool found = prog->UniformHash->get(id, var->name); assert(found); + (void) found; active_atomic_buffer *buf = &buffers[var->data.binding]; /* If this is the first time the buffer is used, increment diff --git a/mesalib/src/glsl/linker.cpp b/mesalib/src/glsl/linker.cpp index 85a4d3883..93b475497 100644 --- a/mesalib/src/glsl/linker.cpp +++ b/mesalib/src/glsl/linker.cpp @@ -609,6 +609,10 @@ cross_validate_globals(struct gl_shader_program *prog, if (var->type->length != 0) { existing->type = var->type; } + } else if (var->type->is_record() + && existing->type->is_record() + && existing->type->record_compare(var->type)) { + existing->type = var->type; } else { linker_error(prog, "%s `%s' declared as type " "`%s' and type `%s'\n", @@ -1992,19 +1996,14 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog) /* Separate the shaders into groups based on their type. */ - struct gl_shader **vert_shader_list; - unsigned num_vert_shaders = 0; - struct gl_shader **frag_shader_list; - unsigned num_frag_shaders = 0; - struct gl_shader **geom_shader_list; - unsigned num_geom_shaders = 0; - - vert_shader_list = (struct gl_shader **) - calloc(prog->NumShaders, sizeof(struct gl_shader *)); - frag_shader_list = (struct gl_shader **) - calloc(prog->NumShaders, sizeof(struct gl_shader *)); - geom_shader_list = (struct gl_shader **) - calloc(prog->NumShaders, sizeof(struct gl_shader *)); + struct gl_shader **shader_list[MESA_SHADER_STAGES]; + unsigned num_shaders[MESA_SHADER_STAGES]; + + for (int i = 0; i < MESA_SHADER_STAGES; i++) { + shader_list[i] = (struct gl_shader **) + calloc(prog->NumShaders, sizeof(struct gl_shader *)); + num_shaders[i] = 0; + } unsigned min_version = UINT_MAX; unsigned max_version = 0; @@ -2020,20 +2019,9 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog) goto done; } - switch (prog->Shaders[i]->Stage) { - case MESA_SHADER_VERTEX: - vert_shader_list[num_vert_shaders] = prog->Shaders[i]; - num_vert_shaders++; - break; - case MESA_SHADER_FRAGMENT: - frag_shader_list[num_frag_shaders] = prog->Shaders[i]; - num_frag_shaders++; - break; - case MESA_SHADER_GEOMETRY: - geom_shader_list[num_geom_shaders] = prog->Shaders[i]; - num_geom_shaders++; - break; - } + gl_shader_stage shader_type = prog->Shaders[i]->Stage; + shader_list[shader_type][num_shaders[shader_type]] = prog->Shaders[i]; + num_shaders[shader_type]++; } /* In desktop GLSL, different shader versions may be linked together. In @@ -2050,7 +2038,8 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog) /* Geometry shaders have to be linked with vertex shaders. */ - if (num_geom_shaders > 0 && num_vert_shaders == 0) { + if (num_shaders[MESA_SHADER_GEOMETRY] > 0 && + num_shaders[MESA_SHADER_VERTEX] == 0) { linker_error(prog, "Geometry shader must be linked with " "vertex shader\n"); goto done; @@ -2065,55 +2054,39 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog) /* Link all shaders for a particular stage and validate the result. */ - if (num_vert_shaders > 0) { - gl_shader *const sh = - link_intrastage_shaders(mem_ctx, ctx, prog, vert_shader_list, - num_vert_shaders); - - if (!prog->LinkStatus) - goto done; - - validate_vertex_shader_executable(prog, sh); - if (!prog->LinkStatus) - goto done; - prog->LastClipDistanceArraySize = prog->Vert.ClipDistanceArraySize; + for (int stage = 0; stage < MESA_SHADER_STAGES; stage++) { + if (num_shaders[stage] > 0) { + gl_shader *const sh = + link_intrastage_shaders(mem_ctx, ctx, prog, shader_list[stage], + num_shaders[stage]); - _mesa_reference_shader(ctx, &prog->_LinkedShaders[MESA_SHADER_VERTEX], - sh); - } - - if (num_frag_shaders > 0) { - gl_shader *const sh = - link_intrastage_shaders(mem_ctx, ctx, prog, frag_shader_list, - num_frag_shaders); - - if (!prog->LinkStatus) - goto done; + if (!prog->LinkStatus) + goto done; - validate_fragment_shader_executable(prog, sh); - if (!prog->LinkStatus) - goto done; + switch (stage) { + case MESA_SHADER_VERTEX: + validate_vertex_shader_executable(prog, sh); + break; + case MESA_SHADER_GEOMETRY: + validate_geometry_shader_executable(prog, sh); + break; + case MESA_SHADER_FRAGMENT: + validate_fragment_shader_executable(prog, sh); + break; + } + if (!prog->LinkStatus) + goto done; - _mesa_reference_shader(ctx, &prog->_LinkedShaders[MESA_SHADER_FRAGMENT], - sh); + _mesa_reference_shader(ctx, &prog->_LinkedShaders[stage], sh); + } } - if (num_geom_shaders > 0) { - gl_shader *const sh = - link_intrastage_shaders(mem_ctx, ctx, prog, geom_shader_list, - num_geom_shaders); - - if (!prog->LinkStatus) - goto done; - - validate_geometry_shader_executable(prog, sh); - if (!prog->LinkStatus) - goto done; + if (num_shaders[MESA_SHADER_GEOMETRY] > 0) prog->LastClipDistanceArraySize = prog->Geom.ClipDistanceArraySize; - - _mesa_reference_shader(ctx, &prog->_LinkedShaders[MESA_SHADER_GEOMETRY], - sh); - } + else if (num_shaders[MESA_SHADER_VERTEX] > 0) + prog->LastClipDistanceArraySize = prog->Vert.ClipDistanceArraySize; + else + prog->LastClipDistanceArraySize = 0; /* Not used */ /* Here begins the inter-stage linking phase. Some initial validation is * performed, then locations are assigned for uniforms, attributes, and @@ -2371,11 +2344,8 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog) /* FINISHME: Assign fragment shader output locations. */ done: - free(vert_shader_list); - free(frag_shader_list); - free(geom_shader_list); - for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) { + free(shader_list[i]); if (prog->_LinkedShaders[i] == NULL) continue; diff --git a/mesalib/src/glsl/opt_algebraic.cpp b/mesalib/src/glsl/opt_algebraic.cpp index 332f0b77b..d1f6435f4 100644 --- a/mesalib/src/glsl/opt_algebraic.cpp +++ b/mesalib/src/glsl/opt_algebraic.cpp @@ -285,6 +285,58 @@ ir_algebraic_visitor::handle_expression(ir_expression *ir) reassociate_constant(ir, 0, op_const[0], op_expr[1]); if (op_const[1] && !op_const[0]) reassociate_constant(ir, 1, op_const[1], op_expr[0]); + + /* Replace (-x + y) * a + x and commutative variations with lrp(x, y, a). + * + * (-x + y) * a + x + * (x * -a) + (y * a) + x + * x + (x * -a) + (y * a) + * x * (1 - a) + y * a + * lrp(x, y, a) + */ + for (int mul_pos = 0; mul_pos < 2; mul_pos++) { + ir_expression *mul = op_expr[mul_pos]; + + if (!mul || mul->operation != ir_binop_mul) + continue; + + /* Multiply found on one of the operands. Now check for an + * inner addition operation. + */ + for (int inner_add_pos = 0; inner_add_pos < 2; inner_add_pos++) { + ir_expression *inner_add = + mul->operands[inner_add_pos]->as_expression(); + + if (!inner_add || inner_add->operation != ir_binop_add) + continue; + + /* Inner addition found on one of the operands. Now check for + * one of the operands of the inner addition to be the negative + * of x_operand. + */ + for (int neg_pos = 0; neg_pos < 2; neg_pos++) { + ir_expression *neg = + inner_add->operands[neg_pos]->as_expression(); + + if (!neg || neg->operation != ir_unop_neg) + continue; + + ir_rvalue *x_operand = ir->operands[1 - mul_pos]; + + if (!neg->operands[0]->equals(x_operand)) + continue; + + ir_rvalue *y_operand = inner_add->operands[1 - neg_pos]; + ir_rvalue *a_operand = mul->operands[1 - inner_add_pos]; + + if (x_operand->type != y_operand->type || + x_operand->type != a_operand->type) + continue; + + return lrp(x_operand, y_operand, a_operand); + } + } + } break; case ir_binop_sub: diff --git a/mesalib/src/glsl/opt_vectorize.cpp b/mesalib/src/glsl/opt_vectorize.cpp new file mode 100644 index 000000000..9ca811a86 --- /dev/null +++ b/mesalib/src/glsl/opt_vectorize.cpp @@ -0,0 +1,319 @@ +/* + * Copyright © 2013 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 opt_vectorize.cpp + * + * Combines scalar assignments of the same expression (modulo swizzle) to + * multiple channels of the same variable into a single vectorized expression + * and assignment. + * + * Many generated shaders contain scalarized code. That is, they contain + * + * r1.x = log2(v0.x); + * r1.y = log2(v0.y); + * r1.z = log2(v0.z); + * + * rather than + * + * r1.xyz = log2(v0.xyz); + * + * We look for consecutive assignments of the same expression (modulo swizzle) + * to each channel of the same variable. + * + * For instance, we want to convert these three scalar operations + * + * (assign (x) (var_ref r1) (expression float log2 (swiz x (var_ref v0)))) + * (assign (y) (var_ref r1) (expression float log2 (swiz y (var_ref v0)))) + * (assign (z) (var_ref r1) (expression float log2 (swiz z (var_ref v0)))) + * + * into a single vector operation + * + * (assign (xyz) (var_ref r1) (expression vec3 log2 (swiz xyz (var_ref v0)))) + */ + +#include "ir.h" +#include "ir_visitor.h" +#include "ir_optimization.h" +#include "glsl_types.h" +#include "program/prog_instruction.h" + +namespace { + +class ir_vectorize_visitor : public ir_hierarchical_visitor { +public: + void clear() + { + assignment[0] = NULL; + assignment[1] = NULL; + assignment[2] = NULL; + assignment[3] = NULL; + current_assignment = NULL; + last_assignment = NULL; + channels = 0; + has_swizzle = false; + } + + ir_vectorize_visitor() + { + clear(); + progress = false; + } + + virtual ir_visitor_status visit_enter(ir_assignment *); + virtual ir_visitor_status visit_enter(ir_swizzle *); + + virtual ir_visitor_status visit_leave(ir_assignment *); + + void try_vectorize(); + + ir_assignment *assignment[4]; + ir_assignment *current_assignment, *last_assignment; + unsigned channels; + bool has_swizzle; + + bool progress; +}; + +} /* unnamed namespace */ + +/** + * Rewrites the swizzles and types of a right-hand side of an assignment. + * + * From the example above, this function would be called (by visit_tree()) on + * the nodes of the tree (expression float log2 (swiz z (var_ref v0))), + * rewriting it into (expression vec3 log2 (swiz xyz (var_ref v0))). + * + * The function modifies only ir_expressions and ir_swizzles. For expressions + * it sets a new type and swizzles any scalar dereferences into appropriately + * sized vector arguments. For example, if combining + * + * (assign (x) (var_ref r1) (expression float + (swiz x (var_ref v0) (var_ref v1)))) + * (assign (y) (var_ref r1) (expression float + (swiz y (var_ref v0) (var_ref v1)))) + * + * where v1 is a scalar, rewrite_swizzle() would insert a swizzle on + * (var_ref v1) such that the final result was + * + * (assign (xy) (var_ref r1) (expression vec2 + (swiz xy (var_ref v0)) + * (swiz xx (var_ref v1)))) + * + * For swizzles, it sets a new type, and if the variable being swizzled is a + * vector it overwrites the swizzle mask with the ir_swizzle_mask passed as the + * data parameter. If the swizzled variable is scalar, then the swizzle was + * added by an earlier call to rewrite_swizzle() on an expression, so the + * mask should not be modified. + */ +static void +rewrite_swizzle(ir_instruction *ir, void *data) +{ + ir_swizzle_mask *mask = (ir_swizzle_mask *)data; + + switch (ir->ir_type) { + case ir_type_swizzle: { + ir_swizzle *swz = (ir_swizzle *)ir; + if (swz->val->type->is_vector()) { + swz->mask = *mask; + } + swz->type = glsl_type::get_instance(swz->type->base_type, + mask->num_components, 1); + break; + } + case ir_type_expression: { + ir_expression *expr = (ir_expression *)ir; + expr->type = glsl_type::get_instance(expr->type->base_type, + mask->num_components, 1); + for (unsigned i = 0; i < 4; i++) { + if (expr->operands[i]) { + ir_dereference *deref = expr->operands[i]->as_dereference(); + if (deref && deref->type->is_scalar()) { + expr->operands[i] = new(ir) ir_swizzle(deref, 0, 0, 0, 0, + mask->num_components); + } + } + } + break; + } + default: + break; + } +} + +/** + * Attempt to vectorize the previously saved assignments, and clear them from + * consideration. + * + * If the assignments are able to be combined, it modifies in-place the last + * assignment seen to be an equivalent vector form of the scalar assignments. + * It then removes the other now obsolete scalar assignments. + */ +void +ir_vectorize_visitor::try_vectorize() +{ + if (this->last_assignment && this->channels > 1) { + ir_swizzle_mask mask = {0, 1, 2, 3, channels, 0}; + + visit_tree(this->last_assignment->rhs, rewrite_swizzle, &mask); + + this->last_assignment->write_mask = 0; + + for (unsigned i = 0; i < 4; i++) { + if (this->assignment[i]) { + this->last_assignment->write_mask |= 1 << i; + + if (this->assignment[i] != this->last_assignment) { + this->assignment[i]->remove(); + } + } + } + + this->progress = true; + } + clear(); +} + +/** + * Returns whether the write mask is a single channel. + */ +static bool +single_channel_write_mask(unsigned write_mask) +{ + return write_mask != 0 && (write_mask & (write_mask - 1)) == 0; +} + +/** + * Translates single-channeled write mask to single-channeled swizzle. + */ +static unsigned +write_mask_to_swizzle(unsigned write_mask) +{ + switch (write_mask) { + case WRITEMASK_X: return SWIZZLE_X; + case WRITEMASK_Y: return SWIZZLE_Y; + case WRITEMASK_Z: return SWIZZLE_Z; + case WRITEMASK_W: return SWIZZLE_W; + } + assert(!"not reached"); + unreachable(); +} + +/** + * Returns whether a single-channeled write mask matches a swizzle. + */ +static bool +write_mask_matches_swizzle(unsigned write_mask, + const ir_swizzle *swz) +{ + return ((write_mask == WRITEMASK_X && swz->mask.x == SWIZZLE_X) || + (write_mask == WRITEMASK_Y && swz->mask.x == SWIZZLE_Y) || + (write_mask == WRITEMASK_Z && swz->mask.x == SWIZZLE_Z) || + (write_mask == WRITEMASK_W && swz->mask.x == SWIZZLE_W)); +} + +/** + * Upon entering an ir_assignment, attempt to vectorize the currently tracked + * assignments if the current assignment is not suitable. Keep a pointer to + * the current assignment. + */ +ir_visitor_status +ir_vectorize_visitor::visit_enter(ir_assignment *ir) +{ + ir_dereference *lhs = this->last_assignment != NULL ? + this->last_assignment->lhs : NULL; + ir_rvalue *rhs = this->last_assignment != NULL ? + this->last_assignment->rhs : NULL; + + if (ir->condition || + this->channels >= 4 || + !single_channel_write_mask(ir->write_mask) || + (lhs && !ir->lhs->equals(lhs)) || + (rhs && !ir->rhs->equals(rhs, ir_type_swizzle))) { + try_vectorize(); + } + + this->current_assignment = ir; + + return visit_continue; +} + +/** + * Upon entering an ir_swizzle, set ::has_swizzle if we're visiting from an + * ir_assignment (i.e., that ::current_assignment is set) and the swizzle mask + * matches the current assignment's write mask. + * + * If the write mask doesn't match the swizzle mask, remove the current + * assignment from further consideration. + */ +ir_visitor_status +ir_vectorize_visitor::visit_enter(ir_swizzle *ir) +{ + if (this->current_assignment) { + if (write_mask_matches_swizzle(this->current_assignment->write_mask, ir)) { + this->has_swizzle = true; + } else { + this->current_assignment = NULL; + } + } + return visit_continue; +} + +/** + * Upon leaving an ir_assignment, save a pointer to it in ::assignment[] if + * the swizzle mask(s) found were appropriate. Also save a pointer in + * ::last_assignment so that we can compare future assignments with it. + * + * Finally, clear ::current_assignment and ::has_swizzle. + */ +ir_visitor_status +ir_vectorize_visitor::visit_leave(ir_assignment *ir) +{ + if (this->has_swizzle && this->current_assignment) { + assert(this->current_assignment == ir); + + unsigned channel = write_mask_to_swizzle(this->current_assignment->write_mask); + this->assignment[channel] = ir; + this->channels++; + + this->last_assignment = this->current_assignment; + } + this->current_assignment = NULL; + this->has_swizzle = false; + return visit_continue; +} + +/** + * Combines scalar assignments of the same expression (modulo swizzle) to + * multiple channels of the same variable into a single vectorized expression + * and assignment. + */ +bool +do_vectorize(exec_list *instructions) +{ + ir_vectorize_visitor v; + + v.run(instructions); + + /* Try to vectorize the last assignments seen. */ + v.try_vectorize(); + + return v.progress; +} diff --git a/mesalib/src/glsl/standalone_scaffolding.cpp b/mesalib/src/glsl/standalone_scaffolding.cpp index 257d2e7a9..91794719b 100644 --- a/mesalib/src/glsl/standalone_scaffolding.cpp +++ b/mesalib/src/glsl/standalone_scaffolding.cpp @@ -110,6 +110,7 @@ void initialize_context_to_defaults(struct gl_context *ctx, gl_api api) ctx->Extensions.ARB_texture_query_levels = true; ctx->Extensions.ARB_texture_query_lod = true; ctx->Extensions.ARB_uniform_buffer_object = true; + ctx->Extensions.ARB_viewport_array = true; ctx->Extensions.OES_EGL_image_external = true; ctx->Extensions.OES_standard_derivatives = true; |