diff options
Diffstat (limited to 'src')
26 files changed, 3712 insertions, 35 deletions
diff --git a/src/glsl/Makefile.am b/src/glsl/Makefile.am index 23c6fe8bb6c..7af9a709d5a 100644 --- a/src/glsl/Makefile.am +++ b/src/glsl/Makefile.am @@ -77,7 +77,7 @@ check_PROGRAMS = \ tests/sampler-types-test \ tests/uniform-initializer-test -noinst_PROGRAMS = glsl_compiler +noinst_PROGRAMS = glsl_compiler spirv2nir tests_blob_test_SOURCES = \ tests/blob_test.c @@ -162,6 +162,16 @@ glsl_compiler_LDADD = \ $(top_builddir)/src/libglsl_util.la \ $(PTHREAD_LIBS) +spirv2nir_SOURCES = \ + standalone_scaffolding.cpp \ + standalone_scaffolding.h \ + nir/spirv2nir.c + +spirv2nir_LDADD = \ + libglsl.la \ + $(top_builddir)/src/libglsl_util.la \ + $(PTHREAD_LIBS) + glsl_test_SOURCES = \ standalone_scaffolding.cpp \ tests/common.c \ diff --git a/src/glsl/Makefile.sources b/src/glsl/Makefile.sources index d784a810723..a234ac6f8e2 100644 --- a/src/glsl/Makefile.sources +++ b/src/glsl/Makefile.sources @@ -59,6 +59,7 @@ NIR_FILES = \ nir/nir_remove_dead_variables.c \ nir/nir_search.c \ nir/nir_search.h \ + nir/nir_spirv.h \ nir/nir_split_var_copies.c \ nir/nir_sweep.c \ nir/nir_to_ssa.c \ @@ -68,6 +69,8 @@ NIR_FILES = \ nir/nir_worklist.c \ nir/nir_worklist.h \ nir/nir_types.cpp \ + nir/spirv_to_nir.c \ + nir/spirv_glsl450_to_nir.c \ $(NIR_GENERATED_FILES) # libglsl diff --git a/src/glsl/ast_to_hir.cpp b/src/glsl/ast_to_hir.cpp index f1daee38d7a..8713cd85f0a 100644 --- a/src/glsl/ast_to_hir.cpp +++ b/src/glsl/ast_to_hir.cpp @@ -970,6 +970,7 @@ do_comparison(void *mem_ctx, int operation, ir_rvalue *op0, ir_rvalue *op1) case GLSL_TYPE_SAMPLER: case GLSL_TYPE_IMAGE: case GLSL_TYPE_INTERFACE: + case GLSL_TYPE_FUNCTION: case GLSL_TYPE_ATOMIC_UINT: /* I assume a comparison of a struct containing a sampler just * ignores the sampler present in the type. diff --git a/src/glsl/glsl_types.cpp b/src/glsl/glsl_types.cpp index 9c9b7efcbc7..0d83ee68e42 100644 --- a/src/glsl/glsl_types.cpp +++ b/src/glsl/glsl_types.cpp @@ -32,6 +32,7 @@ mtx_t glsl_type::mutex = _MTX_INITIALIZER_NP; hash_table *glsl_type::array_types = NULL; hash_table *glsl_type::record_types = NULL; hash_table *glsl_type::interface_types = NULL; +hash_table *glsl_type::function_types = NULL; void *glsl_type::mem_ctx = NULL; void @@ -159,6 +160,39 @@ glsl_type::glsl_type(const glsl_struct_field *fields, unsigned num_fields, mtx_unlock(&glsl_type::mutex); } +glsl_type::glsl_type(const glsl_type *return_type, + const glsl_function_param *params, unsigned num_params) : + gl_type(0), + base_type(GLSL_TYPE_FUNCTION), + sampler_dimensionality(0), sampler_shadow(0), sampler_array(0), + sampler_type(0), interface_packing(0), + vector_elements(0), matrix_columns(0), + length(num_params) +{ + unsigned int i; + + mtx_lock(&glsl_type::mutex); + + init_ralloc_type_ctx(); + + this->fields.parameters = rzalloc_array(this->mem_ctx, + glsl_function_param, num_params + 1); + + /* We store the return type as the first parameter */ + this->fields.parameters[0].type = return_type; + this->fields.parameters[0].in = false; + this->fields.parameters[0].out = true; + + /* We store the i'th parameter in slot i+1 */ + for (i = 0; i < length; i++) { + this->fields.parameters[i + 1].type = params[i].type; + this->fields.parameters[i + 1].in = params[i].in; + this->fields.parameters[i + 1].out = params[i].out; + } + + mtx_unlock(&glsl_type::mutex); +} + bool glsl_type::contains_sampler() const @@ -827,6 +861,72 @@ glsl_type::get_interface_instance(const glsl_struct_field *fields, } +static int +function_key_compare(const void *a, const void *b) +{ + const glsl_type *const key1 = (glsl_type *) a; + const glsl_type *const key2 = (glsl_type *) b; + + if (key1->length != key2->length) + return 1; + + return memcmp(key1->fields.parameters, key2->fields.parameters, + (key1->length + 1) * sizeof(*key1->fields.parameters)); +} + + +static unsigned +function_key_hash(const void *a) +{ + const glsl_type *const key = (glsl_type *) a; + char hash_key[128]; + unsigned size = 0; + + size = snprintf(hash_key, sizeof(hash_key), "%08x", key->length); + + for (unsigned i = 0; i < key->length; i++) { + if (size >= sizeof(hash_key)) + break; + + size += snprintf(& hash_key[size], sizeof(hash_key) - size, + "%p", (void *) key->fields.structure[i].type); + } + + return hash_table_string_hash(& hash_key); +} + +const glsl_type * +glsl_type::get_function_instance(const glsl_type *return_type, + const glsl_function_param *params, + unsigned num_params) +{ + const glsl_type key(return_type, params, num_params); + + mtx_lock(&glsl_type::mutex); + + if (function_types == NULL) { + function_types = hash_table_ctor(64, function_key_hash, + function_key_compare); + } + + const glsl_type *t = (glsl_type *) hash_table_find(function_types, &key); + if (t == NULL) { + mtx_unlock(&glsl_type::mutex); + t = new glsl_type(return_type, params, num_params); + mtx_lock(&glsl_type::mutex); + + hash_table_insert(function_types, (void *) t, t); + } + + assert(t->base_type == GLSL_TYPE_FUNCTION); + assert(t->length == num_params); + + mtx_unlock(&glsl_type::mutex); + + return t; +} + + const glsl_type * glsl_type::get_mul_type(const glsl_type *type_a, const glsl_type *type_b) { @@ -955,6 +1055,7 @@ glsl_type::component_slots() const case GLSL_TYPE_IMAGE: return 1; + case GLSL_TYPE_FUNCTION: case GLSL_TYPE_SAMPLER: case GLSL_TYPE_ATOMIC_UINT: case GLSL_TYPE_VOID: @@ -1326,6 +1427,7 @@ glsl_type::count_attribute_slots() const case GLSL_TYPE_ARRAY: return this->length * this->fields.array->count_attribute_slots(); + case GLSL_TYPE_FUNCTION: case GLSL_TYPE_SAMPLER: case GLSL_TYPE_IMAGE: case GLSL_TYPE_ATOMIC_UINT: diff --git a/src/glsl/glsl_types.h b/src/glsl/glsl_types.h index 5645dcd5011..2d4718572af 100644 --- a/src/glsl/glsl_types.h +++ b/src/glsl/glsl_types.h @@ -56,6 +56,7 @@ enum glsl_base_type { GLSL_TYPE_IMAGE, GLSL_TYPE_ATOMIC_UINT, GLSL_TYPE_STRUCT, + GLSL_TYPE_FUNCTION, GLSL_TYPE_INTERFACE, GLSL_TYPE_ARRAY, GLSL_TYPE_VOID, @@ -178,7 +179,7 @@ struct glsl_type { */ union { const struct glsl_type *array; /**< Type of array elements. */ - const struct glsl_type *parameters; /**< Parameters to function. */ + struct glsl_function_param *parameters; /**< Parameters to function. */ struct glsl_struct_field *structure; /**< List of struct fields. */ } fields; @@ -276,6 +277,13 @@ struct glsl_type { const char *block_name); /** + * Get the instance of a function type + */ + static const glsl_type *get_function_instance(const struct glsl_type *return_type, + const glsl_function_param *parameters, + unsigned num_params); + + /** * Get the type resulting from a multiplication of \p type_a * \p type_b */ static const glsl_type *get_mul_type(const glsl_type *type_a, @@ -688,6 +696,10 @@ private: glsl_type(const glsl_struct_field *fields, unsigned num_fields, enum glsl_interface_packing packing, const char *name); + /** Constructor for interface types */ + glsl_type(const glsl_type *return_type, + const glsl_function_param *params, unsigned num_params); + /** Constructor for array types */ glsl_type(const glsl_type *array, unsigned length); @@ -700,6 +712,9 @@ private: /** Hash table containing the known interface types. */ static struct hash_table *interface_types; + /** Hash table containing the known function types. */ + static struct hash_table *function_types; + static int record_key_compare(const void *a, const void *b); static unsigned record_key_hash(const void *key); @@ -727,6 +742,10 @@ private: /*@}*/ }; +#undef DECL_TYPE +#undef STRUCT_TYPE +#endif /* __cplusplus */ + struct glsl_struct_field { const struct glsl_type *type; const char *name; @@ -770,14 +789,17 @@ struct glsl_struct_field { int stream; }; +struct glsl_function_param { + const struct glsl_type *type; + + bool in; + bool out; +}; + static inline unsigned int glsl_align(unsigned int a, unsigned int align) { return (a + align - 1) / align * align; } -#undef DECL_TYPE -#undef STRUCT_TYPE -#endif /* __cplusplus */ - #endif /* GLSL_TYPES_H */ diff --git a/src/glsl/ir_clone.cpp b/src/glsl/ir_clone.cpp index 914e0e4d540..636c143ddc2 100644 --- a/src/glsl/ir_clone.cpp +++ b/src/glsl/ir_clone.cpp @@ -357,6 +357,7 @@ ir_constant::clone(void *mem_ctx, struct hash_table *ht) const return c; } + case GLSL_TYPE_FUNCTION: case GLSL_TYPE_SAMPLER: case GLSL_TYPE_IMAGE: case GLSL_TYPE_ATOMIC_UINT: diff --git a/src/glsl/link_uniform_initializers.cpp b/src/glsl/link_uniform_initializers.cpp index 69073841ea4..60bfc9c15c9 100644 --- a/src/glsl/link_uniform_initializers.cpp +++ b/src/glsl/link_uniform_initializers.cpp @@ -88,6 +88,7 @@ copy_constant_to_storage(union gl_constant_value *storage, case GLSL_TYPE_IMAGE: case GLSL_TYPE_ATOMIC_UINT: case GLSL_TYPE_INTERFACE: + case GLSL_TYPE_FUNCTION: case GLSL_TYPE_VOID: case GLSL_TYPE_ERROR: /* All other types should have already been filtered by other diff --git a/src/glsl/main.cpp b/src/glsl/main.cpp index ccac8399646..fc54ddd7eb1 100644 --- a/src/glsl/main.cpp +++ b/src/glsl/main.cpp @@ -41,12 +41,6 @@ static int glsl_version = 330; -extern "C" void -_mesa_error_no_memory(const char *caller) -{ - fprintf(stderr, "Mesa error: out of memory in %s", caller); -} - static void initialize_context(struct gl_context *ctx, gl_api api) { diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h index 697d37e95ac..61306e9b7e0 100644 --- a/src/glsl/nir/nir.h +++ b/src/glsl/nir/nir.h @@ -782,6 +782,15 @@ NIR_DEFINE_CAST(nir_deref_as_var, nir_deref, nir_deref_var, deref) NIR_DEFINE_CAST(nir_deref_as_array, nir_deref, nir_deref_array, deref) NIR_DEFINE_CAST(nir_deref_as_struct, nir_deref, nir_deref_struct, deref) +/** Returns the tail of a deref chain */ +static inline nir_deref * +nir_deref_tail(nir_deref *deref) +{ + while (deref->child) + deref = deref->child; + return deref; +} + typedef struct { nir_instr instr; diff --git a/src/glsl/nir/nir_lower_io.c b/src/glsl/nir/nir_lower_io.c index 03eed04e1e9..561bebd3a9c 100644 --- a/src/glsl/nir/nir_lower_io.c +++ b/src/glsl/nir/nir_lower_io.c @@ -67,6 +67,7 @@ type_size(const struct glsl_type *type) return 0; case GLSL_TYPE_IMAGE: return 0; + case GLSL_TYPE_FUNCTION: case GLSL_TYPE_VOID: case GLSL_TYPE_ERROR: case GLSL_TYPE_DOUBLE: diff --git a/src/glsl/nir/nir_lower_var_copies.c b/src/glsl/nir/nir_lower_var_copies.c index 21672901f04..98c107aa50e 100644 --- a/src/glsl/nir/nir_lower_var_copies.c +++ b/src/glsl/nir/nir_lower_var_copies.c @@ -53,17 +53,6 @@ deref_next_wildcard_parent(nir_deref *deref) return NULL; } -/* Returns the last deref in the chain. - */ -static nir_deref * -get_deref_tail(nir_deref *deref) -{ - while (deref->child) - deref = deref->child; - - return deref; -} - /* This function recursively walks the given deref chain and replaces the * given copy instruction with an equivalent sequence load/store * operations. @@ -121,8 +110,8 @@ emit_copy_load_store(nir_intrinsic_instr *copy_instr, } else { /* In this case, we have no wildcards anymore, so all we have to do * is just emit the load and store operations. */ - src_tail = get_deref_tail(src_tail); - dest_tail = get_deref_tail(dest_tail); + src_tail = nir_deref_tail(src_tail); + dest_tail = nir_deref_tail(dest_tail); assert(src_tail->type == dest_tail->type); diff --git a/src/glsl/nir/nir_spirv.h b/src/glsl/nir/nir_spirv.h new file mode 100644 index 00000000000..789d30cd672 --- /dev/null +++ b/src/glsl/nir/nir_spirv.h @@ -0,0 +1,38 @@ +/* + * Copyright © 2015 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. + * + * Authors: + * Jason Ekstrand ([email protected]) + * + */ + +#pragma once + +#ifndef _NIR_SPIRV_H_ +#define _NIR_SPIRV_H_ + +#include "nir.h" + +nir_shader *spirv_to_nir(const uint32_t *words, size_t word_count, + const nir_shader_compiler_options *options); + +#endif /* _NIR_SPIRV_H_ */ diff --git a/src/glsl/nir/nir_split_var_copies.c b/src/glsl/nir/nir_split_var_copies.c index fc72c078c77..5c163b59819 100644 --- a/src/glsl/nir/nir_split_var_copies.c +++ b/src/glsl/nir/nir_split_var_copies.c @@ -66,14 +66,6 @@ struct split_var_copies_state { void *dead_ctx; }; -static nir_deref * -get_deref_tail(nir_deref *deref) -{ - while (deref->child != NULL) - deref = deref->child; - return deref; -} - /* Recursively constructs deref chains to split a copy instruction into * multiple (if needed) copy instructions with full-length deref chains. * External callers of this function should pass the tail and head of the @@ -225,8 +217,8 @@ split_var_copies_block(nir_block *block, void *void_state) nir_deref *dest_head = &intrinsic->variables[0]->deref; nir_deref *src_head = &intrinsic->variables[1]->deref; - nir_deref *dest_tail = get_deref_tail(dest_head); - nir_deref *src_tail = get_deref_tail(src_head); + nir_deref *dest_tail = nir_deref_tail(dest_head); + nir_deref *src_tail = nir_deref_tail(src_head); switch (glsl_get_base_type(src_tail->type)) { case GLSL_TYPE_ARRAY: diff --git a/src/glsl/nir/nir_types.cpp b/src/glsl/nir/nir_types.cpp index 62176f508a1..35421506545 100644 --- a/src/glsl/nir/nir_types.cpp +++ b/src/glsl/nir/nir_types.cpp @@ -70,6 +70,18 @@ glsl_get_struct_field(const glsl_type *type, unsigned index) return type->fields.structure[index].type; } +const glsl_type * +glsl_get_function_return_type(const glsl_type *type) +{ + return type->fields.parameters[0].type; +} + +const glsl_function_param * +glsl_get_function_param(const glsl_type *type, unsigned index) +{ + return &type->fields.parameters[index + 1]; +} + const struct glsl_type * glsl_get_column_type(const struct glsl_type *type) { @@ -112,6 +124,20 @@ glsl_get_struct_elem_name(const struct glsl_type *type, unsigned index) return type->fields.structure[index].name; } +glsl_sampler_dim +glsl_get_sampler_dim(const struct glsl_type *type) +{ + assert(glsl_type_is_sampler(type)); + return (glsl_sampler_dim)type->sampler_dimensionality; +} + +glsl_base_type +glsl_get_sampler_result_type(const struct glsl_type *type) +{ + assert(glsl_type_is_sampler(type)); + return (glsl_base_type)type->sampler_type; +} + bool glsl_type_is_void(const glsl_type *type) { @@ -131,11 +157,37 @@ glsl_type_is_scalar(const struct glsl_type *type) } bool +glsl_type_is_vector_or_scalar(const struct glsl_type *type) +{ + return type->is_vector() || type->is_scalar(); +} + +bool glsl_type_is_matrix(const struct glsl_type *type) { return type->is_matrix(); } +bool +glsl_type_is_sampler(const struct glsl_type *type) +{ + return type->is_sampler(); +} + +bool +glsl_sampler_type_is_shadow(const struct glsl_type *type) +{ + assert(glsl_type_is_sampler(type)); + return type->sampler_shadow; +} + +bool +glsl_sampler_type_is_array(const struct glsl_type *type) +{ + assert(glsl_type_is_sampler(type)); + return type->sampler_array; +} + const glsl_type * glsl_void_type(void) { @@ -149,13 +201,72 @@ glsl_float_type(void) } const glsl_type * +glsl_int_type(void) +{ + return glsl_type::int_type; +} + +const glsl_type * +glsl_uint_type(void) +{ + return glsl_type::uint_type; +} + +const glsl_type * +glsl_bool_type(void) +{ + return glsl_type::bool_type; +} + +const glsl_type * glsl_vec4_type(void) { return glsl_type::vec4_type; } const glsl_type * +glsl_scalar_type(enum glsl_base_type base_type) +{ + return glsl_type::get_instance(base_type, 1, 1); +} + +const glsl_type * +glsl_vector_type(enum glsl_base_type base_type, unsigned components) +{ + assert(components > 1 && components <= 4); + return glsl_type::get_instance(base_type, components, 1); +} + +const glsl_type * +glsl_matrix_type(enum glsl_base_type base_type, unsigned rows, unsigned columns) +{ + assert(rows > 1 && rows <= 4 && columns > 1 && columns <= 4); + return glsl_type::get_instance(base_type, rows, columns); +} + +const glsl_type * glsl_array_type(const glsl_type *base, unsigned elements) { return glsl_type::get_array_instance(base, elements); } + +const glsl_type * +glsl_struct_type(const glsl_struct_field *fields, + unsigned num_fields, const char *name) +{ + return glsl_type::get_record_instance(fields, num_fields, name); +} + +const struct glsl_type * +glsl_sampler_type(enum glsl_sampler_dim dim, bool is_shadow, bool is_array, + enum glsl_base_type base_type) +{ + return glsl_type::get_sampler_instance(dim, is_shadow, is_array, base_type); +} + +const glsl_type * +glsl_function_type(const glsl_type *return_type, + const glsl_function_param *params, unsigned num_params) +{ + return glsl_type::get_function_instance(return_type, params, num_params); +} diff --git a/src/glsl/nir/nir_types.h b/src/glsl/nir/nir_types.h index 276d4ad6234..ceb131c9f47 100644 --- a/src/glsl/nir/nir_types.h +++ b/src/glsl/nir/nir_types.h @@ -49,6 +49,12 @@ const struct glsl_type *glsl_get_array_element(const struct glsl_type *type); const struct glsl_type *glsl_get_column_type(const struct glsl_type *type); +const struct glsl_type * +glsl_get_function_return_type(const struct glsl_type *type); + +const struct glsl_function_param * +glsl_get_function_param(const struct glsl_type *type, unsigned index); + enum glsl_base_type glsl_get_base_type(const struct glsl_type *type); unsigned glsl_get_vector_elements(const struct glsl_type *type); @@ -62,17 +68,40 @@ unsigned glsl_get_length(const struct glsl_type *type); const char *glsl_get_struct_elem_name(const struct glsl_type *type, unsigned index); +enum glsl_sampler_dim glsl_get_sampler_dim(const struct glsl_type *type); +enum glsl_base_type glsl_get_sampler_result_type(const struct glsl_type *type); bool glsl_type_is_void(const struct glsl_type *type); bool glsl_type_is_vector(const struct glsl_type *type); bool glsl_type_is_scalar(const struct glsl_type *type); +bool glsl_type_is_vector_or_scalar(const struct glsl_type *type); bool glsl_type_is_matrix(const struct glsl_type *type); +bool glsl_type_is_sampler(const struct glsl_type *type); +bool glsl_sampler_type_is_shadow(const struct glsl_type *type); +bool glsl_sampler_type_is_array(const struct glsl_type *type); const struct glsl_type *glsl_void_type(void); const struct glsl_type *glsl_float_type(void); +const struct glsl_type *glsl_int_type(void); +const struct glsl_type *glsl_uint_type(void); +const struct glsl_type *glsl_bool_type(void); + const struct glsl_type *glsl_vec4_type(void); +const struct glsl_type *glsl_scalar_type(enum glsl_base_type base_type); +const struct glsl_type *glsl_vector_type(enum glsl_base_type base_type, + unsigned components); +const struct glsl_type *glsl_matrix_type(enum glsl_base_type base_type, + unsigned rows, unsigned columns); const struct glsl_type *glsl_array_type(const struct glsl_type *base, unsigned elements); +const struct glsl_type *glsl_struct_type(const struct glsl_struct_field *fields, + unsigned num_fields, const char *name); +const struct glsl_type *glsl_sampler_type(enum glsl_sampler_dim dim, + bool is_shadow, bool is_array, + enum glsl_base_type base_type); +const struct glsl_type * glsl_function_type(const struct glsl_type *return_type, + const struct glsl_function_param *params, + unsigned num_params); #ifdef __cplusplus } diff --git a/src/glsl/nir/spirv.h b/src/glsl/nir/spirv.h new file mode 100644 index 00000000000..da717ecd342 --- /dev/null +++ b/src/glsl/nir/spirv.h @@ -0,0 +1,1304 @@ +/* +** Copyright (c) 2015 The Khronos Group Inc. +** +** Permission is hereby granted, free of charge, to any person obtaining a copy +** of this software and/or associated documentation files (the "Materials"), +** to deal in the Materials without restriction, including without limitation +** the rights to use, copy, modify, merge, publish, distribute, sublicense, +** and/or sell copies of the Materials, and to permit persons to whom the +** Materials are furnished to do so, subject to the following conditions: +** +** The above copyright notice and this permission notice shall be included in +** all copies or substantial portions of the Materials. +** +** MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS KHRONOS +** STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS SPECIFICATIONS AND +** HEADER INFORMATION ARE LOCATED AT https://www.khronos.org/registry/ +** +** THE MATERIALS ARE 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 MATERIALS OR THE USE OR OTHER DEALINGS +** IN THE MATERIALS. +*/ + +/* +** This header is automatically generated by the same tool that creates +** the Binary Section of the SPIR-V specification. +*/ + +/* +** Specification revision 30. +** Enumeration tokens for SPIR-V, in three styles: C, C++, generic. +** - C++ will have the tokens in the "spv" name space, with no prefix. +** - C will have tokens with as "Spv" prefix. +** +** Some tokens act like mask values, which can be OR'd together, +** while others are mutually exclusive. The mask-like ones have +** "Mask" in their name, and a parallel enum that has the shift +** amount (1 << x) for each corresponding enumerant. +*/ + +#ifndef spirv_H +#define spirv_H + +#ifdef __cplusplus + +namespace spv { + +static const int MagicNumber = 0x07230203; +static const int Version = 99; + +typedef unsigned int Id; + +static const unsigned int OpCodeMask = 0xFFFF; +static const unsigned int WordCountShift = 16; + +enum SourceLanguage { + SourceLanguageUnknown = 0, + SourceLanguageESSL = 1, + SourceLanguageGLSL = 2, + SourceLanguageOpenCL = 3, +}; + +enum ExecutionModel { + ExecutionModelVertex = 0, + ExecutionModelTessellationControl = 1, + ExecutionModelTessellationEvaluation = 2, + ExecutionModelGeometry = 3, + ExecutionModelFragment = 4, + ExecutionModelGLCompute = 5, + ExecutionModelKernel = 6, +}; + +enum AddressingModel { + AddressingModelLogical = 0, + AddressingModelPhysical32 = 1, + AddressingModelPhysical64 = 2, +}; + +enum MemoryModel { + MemoryModelSimple = 0, + MemoryModelGLSL450 = 1, + MemoryModelOpenCL12 = 2, + MemoryModelOpenCL20 = 3, + MemoryModelOpenCL21 = 4, +}; + +enum ExecutionMode { + ExecutionModeInvocations = 0, + ExecutionModeSpacingEqual = 1, + ExecutionModeSpacingFractionalEven = 2, + ExecutionModeSpacingFractionalOdd = 3, + ExecutionModeVertexOrderCw = 4, + ExecutionModeVertexOrderCcw = 5, + ExecutionModePixelCenterInteger = 6, + ExecutionModeOriginUpperLeft = 7, + ExecutionModeEarlyFragmentTests = 8, + ExecutionModePointMode = 9, + ExecutionModeXfb = 10, + ExecutionModeDepthReplacing = 11, + ExecutionModeDepthAny = 12, + ExecutionModeDepthGreater = 13, + ExecutionModeDepthLess = 14, + ExecutionModeDepthUnchanged = 15, + ExecutionModeLocalSize = 16, + ExecutionModeLocalSizeHint = 17, + ExecutionModeInputPoints = 18, + ExecutionModeInputLines = 19, + ExecutionModeInputLinesAdjacency = 20, + ExecutionModeInputTriangles = 21, + ExecutionModeInputTrianglesAdjacency = 22, + ExecutionModeInputQuads = 23, + ExecutionModeInputIsolines = 24, + ExecutionModeOutputVertices = 25, + ExecutionModeOutputPoints = 26, + ExecutionModeOutputLineStrip = 27, + ExecutionModeOutputTriangleStrip = 28, + ExecutionModeVecTypeHint = 29, + ExecutionModeContractionOff = 30, +}; + +enum StorageClass { + StorageClassUniformConstant = 0, + StorageClassInput = 1, + StorageClassUniform = 2, + StorageClassOutput = 3, + StorageClassWorkgroupLocal = 4, + StorageClassWorkgroupGlobal = 5, + StorageClassPrivateGlobal = 6, + StorageClassFunction = 7, + StorageClassGeneric = 8, + StorageClassPrivate = 9, + StorageClassAtomicCounter = 10, +}; + +enum Dim { + Dim1D = 0, + Dim2D = 1, + Dim3D = 2, + DimCube = 3, + DimRect = 4, + DimBuffer = 5, +}; + +enum SamplerAddressingMode { + SamplerAddressingModeNone = 0, + SamplerAddressingModeClampToEdge = 1, + SamplerAddressingModeClamp = 2, + SamplerAddressingModeRepeat = 3, + SamplerAddressingModeRepeatMirrored = 4, +}; + +enum SamplerFilterMode { + SamplerFilterModeNearest = 0, + SamplerFilterModeLinear = 1, +}; + +enum FPFastMathModeShift { + FPFastMathModeNotNaNShift = 0, + FPFastMathModeNotInfShift = 1, + FPFastMathModeNSZShift = 2, + FPFastMathModeAllowRecipShift = 3, + FPFastMathModeFastShift = 4, +}; + +enum FPFastMathModeMask { + FPFastMathModeMaskNone = 0, + FPFastMathModeNotNaNMask = 0x00000001, + FPFastMathModeNotInfMask = 0x00000002, + FPFastMathModeNSZMask = 0x00000004, + FPFastMathModeAllowRecipMask = 0x00000008, + FPFastMathModeFastMask = 0x00000010, +}; + +enum FPRoundingMode { + FPRoundingModeRTE = 0, + FPRoundingModeRTZ = 1, + FPRoundingModeRTP = 2, + FPRoundingModeRTN = 3, +}; + +enum LinkageType { + LinkageTypeExport = 0, + LinkageTypeImport = 1, +}; + +enum AccessQualifier { + AccessQualifierReadOnly = 0, + AccessQualifierWriteOnly = 1, + AccessQualifierReadWrite = 2, +}; + +enum FunctionParameterAttribute { + FunctionParameterAttributeZext = 0, + FunctionParameterAttributeSext = 1, + FunctionParameterAttributeByVal = 2, + FunctionParameterAttributeSret = 3, + FunctionParameterAttributeNoAlias = 4, + FunctionParameterAttributeNoCapture = 5, + FunctionParameterAttributeSVM = 6, + FunctionParameterAttributeNoWrite = 7, + FunctionParameterAttributeNoReadWrite = 8, +}; + +enum Decoration { + DecorationPrecisionLow = 0, + DecorationPrecisionMedium = 1, + DecorationPrecisionHigh = 2, + DecorationBlock = 3, + DecorationBufferBlock = 4, + DecorationRowMajor = 5, + DecorationColMajor = 6, + DecorationGLSLShared = 7, + DecorationGLSLStd140 = 8, + DecorationGLSLStd430 = 9, + DecorationGLSLPacked = 10, + DecorationSmooth = 11, + DecorationNoperspective = 12, + DecorationFlat = 13, + DecorationPatch = 14, + DecorationCentroid = 15, + DecorationSample = 16, + DecorationInvariant = 17, + DecorationRestrict = 18, + DecorationAliased = 19, + DecorationVolatile = 20, + DecorationConstant = 21, + DecorationCoherent = 22, + DecorationNonwritable = 23, + DecorationNonreadable = 24, + DecorationUniform = 25, + DecorationNoStaticUse = 26, + DecorationCPacked = 27, + DecorationSaturatedConversion = 28, + DecorationStream = 29, + DecorationLocation = 30, + DecorationComponent = 31, + DecorationIndex = 32, + DecorationBinding = 33, + DecorationDescriptorSet = 34, + DecorationOffset = 35, + DecorationAlignment = 36, + DecorationXfbBuffer = 37, + DecorationStride = 38, + DecorationBuiltIn = 39, + DecorationFuncParamAttr = 40, + DecorationFPRoundingMode = 41, + DecorationFPFastMathMode = 42, + DecorationLinkageAttributes = 43, + DecorationSpecId = 44, +}; + +enum BuiltIn { + BuiltInPosition = 0, + BuiltInPointSize = 1, + BuiltInClipVertex = 2, + BuiltInClipDistance = 3, + BuiltInCullDistance = 4, + BuiltInVertexId = 5, + BuiltInInstanceId = 6, + BuiltInPrimitiveId = 7, + BuiltInInvocationId = 8, + BuiltInLayer = 9, + BuiltInViewportIndex = 10, + BuiltInTessLevelOuter = 11, + BuiltInTessLevelInner = 12, + BuiltInTessCoord = 13, + BuiltInPatchVertices = 14, + BuiltInFragCoord = 15, + BuiltInPointCoord = 16, + BuiltInFrontFacing = 17, + BuiltInSampleId = 18, + BuiltInSamplePosition = 19, + BuiltInSampleMask = 20, + BuiltInFragColor = 21, + BuiltInFragDepth = 22, + BuiltInHelperInvocation = 23, + BuiltInNumWorkgroups = 24, + BuiltInWorkgroupSize = 25, + BuiltInWorkgroupId = 26, + BuiltInLocalInvocationId = 27, + BuiltInGlobalInvocationId = 28, + BuiltInLocalInvocationIndex = 29, + BuiltInWorkDim = 30, + BuiltInGlobalSize = 31, + BuiltInEnqueuedWorkgroupSize = 32, + BuiltInGlobalOffset = 33, + BuiltInGlobalLinearId = 34, + BuiltInWorkgroupLinearId = 35, + BuiltInSubgroupSize = 36, + BuiltInSubgroupMaxSize = 37, + BuiltInNumSubgroups = 38, + BuiltInNumEnqueuedSubgroups = 39, + BuiltInSubgroupId = 40, + BuiltInSubgroupLocalInvocationId = 41, +}; + +enum SelectionControlShift { + SelectionControlFlattenShift = 0, + SelectionControlDontFlattenShift = 1, +}; + +enum SelectionControlMask { + SelectionControlMaskNone = 0, + SelectionControlFlattenMask = 0x00000001, + SelectionControlDontFlattenMask = 0x00000002, +}; + +enum LoopControlShift { + LoopControlUnrollShift = 0, + LoopControlDontUnrollShift = 1, +}; + +enum LoopControlMask { + LoopControlMaskNone = 0, + LoopControlUnrollMask = 0x00000001, + LoopControlDontUnrollMask = 0x00000002, +}; + +enum FunctionControlShift { + FunctionControlInlineShift = 0, + FunctionControlDontInlineShift = 1, + FunctionControlPureShift = 2, + FunctionControlConstShift = 3, +}; + +enum FunctionControlMask { + FunctionControlMaskNone = 0, + FunctionControlInlineMask = 0x00000001, + FunctionControlDontInlineMask = 0x00000002, + FunctionControlPureMask = 0x00000004, + FunctionControlConstMask = 0x00000008, +}; + +enum MemorySemanticsShift { + MemorySemanticsRelaxedShift = 0, + MemorySemanticsSequentiallyConsistentShift = 1, + MemorySemanticsAcquireShift = 2, + MemorySemanticsReleaseShift = 3, + MemorySemanticsUniformMemoryShift = 4, + MemorySemanticsSubgroupMemoryShift = 5, + MemorySemanticsWorkgroupLocalMemoryShift = 6, + MemorySemanticsWorkgroupGlobalMemoryShift = 7, + MemorySemanticsAtomicCounterMemoryShift = 8, + MemorySemanticsImageMemoryShift = 9, +}; + +enum MemorySemanticsMask { + MemorySemanticsMaskNone = 0, + MemorySemanticsRelaxedMask = 0x00000001, + MemorySemanticsSequentiallyConsistentMask = 0x00000002, + MemorySemanticsAcquireMask = 0x00000004, + MemorySemanticsReleaseMask = 0x00000008, + MemorySemanticsUniformMemoryMask = 0x00000010, + MemorySemanticsSubgroupMemoryMask = 0x00000020, + MemorySemanticsWorkgroupLocalMemoryMask = 0x00000040, + MemorySemanticsWorkgroupGlobalMemoryMask = 0x00000080, + MemorySemanticsAtomicCounterMemoryMask = 0x00000100, + MemorySemanticsImageMemoryMask = 0x00000200, +}; + +enum MemoryAccessShift { + MemoryAccessVolatileShift = 0, + MemoryAccessAlignedShift = 1, +}; + +enum MemoryAccessMask { + MemoryAccessMaskNone = 0, + MemoryAccessVolatileMask = 0x00000001, + MemoryAccessAlignedMask = 0x00000002, +}; + +enum ExecutionScope { + ExecutionScopeCrossDevice = 0, + ExecutionScopeDevice = 1, + ExecutionScopeWorkgroup = 2, + ExecutionScopeSubgroup = 3, +}; + +enum GroupOperation { + GroupOperationReduce = 0, + GroupOperationInclusiveScan = 1, + GroupOperationExclusiveScan = 2, +}; + +enum KernelEnqueueFlags { + KernelEnqueueFlagsNoWait = 0, + KernelEnqueueFlagsWaitKernel = 1, + KernelEnqueueFlagsWaitWorkGroup = 2, +}; + +enum KernelProfilingInfoShift { + KernelProfilingInfoCmdExecTimeShift = 0, +}; + +enum KernelProfilingInfoMask { + KernelProfilingInfoMaskNone = 0, + KernelProfilingInfoCmdExecTimeMask = 0x00000001, +}; + +enum Op { + OpNop = 0, + OpSource = 1, + OpSourceExtension = 2, + OpExtension = 3, + OpExtInstImport = 4, + OpMemoryModel = 5, + OpEntryPoint = 6, + OpExecutionMode = 7, + OpTypeVoid = 8, + OpTypeBool = 9, + OpTypeInt = 10, + OpTypeFloat = 11, + OpTypeVector = 12, + OpTypeMatrix = 13, + OpTypeSampler = 14, + OpTypeFilter = 15, + OpTypeArray = 16, + OpTypeRuntimeArray = 17, + OpTypeStruct = 18, + OpTypeOpaque = 19, + OpTypePointer = 20, + OpTypeFunction = 21, + OpTypeEvent = 22, + OpTypeDeviceEvent = 23, + OpTypeReserveId = 24, + OpTypeQueue = 25, + OpTypePipe = 26, + OpConstantTrue = 27, + OpConstantFalse = 28, + OpConstant = 29, + OpConstantComposite = 30, + OpConstantSampler = 31, + OpConstantNullPointer = 32, + OpConstantNullObject = 33, + OpSpecConstantTrue = 34, + OpSpecConstantFalse = 35, + OpSpecConstant = 36, + OpSpecConstantComposite = 37, + OpVariable = 38, + OpVariableArray = 39, + OpFunction = 40, + OpFunctionParameter = 41, + OpFunctionEnd = 42, + OpFunctionCall = 43, + OpExtInst = 44, + OpUndef = 45, + OpLoad = 46, + OpStore = 47, + OpPhi = 48, + OpDecorationGroup = 49, + OpDecorate = 50, + OpMemberDecorate = 51, + OpGroupDecorate = 52, + OpGroupMemberDecorate = 53, + OpName = 54, + OpMemberName = 55, + OpString = 56, + OpLine = 57, + OpVectorExtractDynamic = 58, + OpVectorInsertDynamic = 59, + OpVectorShuffle = 60, + OpCompositeConstruct = 61, + OpCompositeExtract = 62, + OpCompositeInsert = 63, + OpCopyObject = 64, + OpCopyMemory = 65, + OpCopyMemorySized = 66, + OpSampler = 67, + OpTextureSample = 68, + OpTextureSampleDref = 69, + OpTextureSampleLod = 70, + OpTextureSampleProj = 71, + OpTextureSampleGrad = 72, + OpTextureSampleOffset = 73, + OpTextureSampleProjLod = 74, + OpTextureSampleProjGrad = 75, + OpTextureSampleLodOffset = 76, + OpTextureSampleProjOffset = 77, + OpTextureSampleGradOffset = 78, + OpTextureSampleProjLodOffset = 79, + OpTextureSampleProjGradOffset = 80, + OpTextureFetchTexelLod = 81, + OpTextureFetchTexelOffset = 82, + OpTextureFetchSample = 83, + OpTextureFetchTexel = 84, + OpTextureGather = 85, + OpTextureGatherOffset = 86, + OpTextureGatherOffsets = 87, + OpTextureQuerySizeLod = 88, + OpTextureQuerySize = 89, + OpTextureQueryLod = 90, + OpTextureQueryLevels = 91, + OpTextureQuerySamples = 92, + OpAccessChain = 93, + OpInBoundsAccessChain = 94, + OpSNegate = 95, + OpFNegate = 96, + OpNot = 97, + OpAny = 98, + OpAll = 99, + OpConvertFToU = 100, + OpConvertFToS = 101, + OpConvertSToF = 102, + OpConvertUToF = 103, + OpUConvert = 104, + OpSConvert = 105, + OpFConvert = 106, + OpConvertPtrToU = 107, + OpConvertUToPtr = 108, + OpPtrCastToGeneric = 109, + OpGenericCastToPtr = 110, + OpBitcast = 111, + OpTranspose = 112, + OpIsNan = 113, + OpIsInf = 114, + OpIsFinite = 115, + OpIsNormal = 116, + OpSignBitSet = 117, + OpLessOrGreater = 118, + OpOrdered = 119, + OpUnordered = 120, + OpArrayLength = 121, + OpIAdd = 122, + OpFAdd = 123, + OpISub = 124, + OpFSub = 125, + OpIMul = 126, + OpFMul = 127, + OpUDiv = 128, + OpSDiv = 129, + OpFDiv = 130, + OpUMod = 131, + OpSRem = 132, + OpSMod = 133, + OpFRem = 134, + OpFMod = 135, + OpVectorTimesScalar = 136, + OpMatrixTimesScalar = 137, + OpVectorTimesMatrix = 138, + OpMatrixTimesVector = 139, + OpMatrixTimesMatrix = 140, + OpOuterProduct = 141, + OpDot = 142, + OpShiftRightLogical = 143, + OpShiftRightArithmetic = 144, + OpShiftLeftLogical = 145, + OpLogicalOr = 146, + OpLogicalXor = 147, + OpLogicalAnd = 148, + OpBitwiseOr = 149, + OpBitwiseXor = 150, + OpBitwiseAnd = 151, + OpSelect = 152, + OpIEqual = 153, + OpFOrdEqual = 154, + OpFUnordEqual = 155, + OpINotEqual = 156, + OpFOrdNotEqual = 157, + OpFUnordNotEqual = 158, + OpULessThan = 159, + OpSLessThan = 160, + OpFOrdLessThan = 161, + OpFUnordLessThan = 162, + OpUGreaterThan = 163, + OpSGreaterThan = 164, + OpFOrdGreaterThan = 165, + OpFUnordGreaterThan = 166, + OpULessThanEqual = 167, + OpSLessThanEqual = 168, + OpFOrdLessThanEqual = 169, + OpFUnordLessThanEqual = 170, + OpUGreaterThanEqual = 171, + OpSGreaterThanEqual = 172, + OpFOrdGreaterThanEqual = 173, + OpFUnordGreaterThanEqual = 174, + OpDPdx = 175, + OpDPdy = 176, + OpFwidth = 177, + OpDPdxFine = 178, + OpDPdyFine = 179, + OpFwidthFine = 180, + OpDPdxCoarse = 181, + OpDPdyCoarse = 182, + OpFwidthCoarse = 183, + OpEmitVertex = 184, + OpEndPrimitive = 185, + OpEmitStreamVertex = 186, + OpEndStreamPrimitive = 187, + OpControlBarrier = 188, + OpMemoryBarrier = 189, + OpImagePointer = 190, + OpAtomicInit = 191, + OpAtomicLoad = 192, + OpAtomicStore = 193, + OpAtomicExchange = 194, + OpAtomicCompareExchange = 195, + OpAtomicCompareExchangeWeak = 196, + OpAtomicIIncrement = 197, + OpAtomicIDecrement = 198, + OpAtomicIAdd = 199, + OpAtomicISub = 200, + OpAtomicUMin = 201, + OpAtomicUMax = 202, + OpAtomicAnd = 203, + OpAtomicOr = 204, + OpAtomicXor = 205, + OpLoopMerge = 206, + OpSelectionMerge = 207, + OpLabel = 208, + OpBranch = 209, + OpBranchConditional = 210, + OpSwitch = 211, + OpKill = 212, + OpReturn = 213, + OpReturnValue = 214, + OpUnreachable = 215, + OpLifetimeStart = 216, + OpLifetimeStop = 217, + OpCompileFlag = 218, + OpAsyncGroupCopy = 219, + OpWaitGroupEvents = 220, + OpGroupAll = 221, + OpGroupAny = 222, + OpGroupBroadcast = 223, + OpGroupIAdd = 224, + OpGroupFAdd = 225, + OpGroupFMin = 226, + OpGroupUMin = 227, + OpGroupSMin = 228, + OpGroupFMax = 229, + OpGroupUMax = 230, + OpGroupSMax = 231, + OpGenericCastToPtrExplicit = 232, + OpGenericPtrMemSemantics = 233, + OpReadPipe = 234, + OpWritePipe = 235, + OpReservedReadPipe = 236, + OpReservedWritePipe = 237, + OpReserveReadPipePackets = 238, + OpReserveWritePipePackets = 239, + OpCommitReadPipe = 240, + OpCommitWritePipe = 241, + OpIsValidReserveId = 242, + OpGetNumPipePackets = 243, + OpGetMaxPipePackets = 244, + OpGroupReserveReadPipePackets = 245, + OpGroupReserveWritePipePackets = 246, + OpGroupCommitReadPipe = 247, + OpGroupCommitWritePipe = 248, + OpEnqueueMarker = 249, + OpEnqueueKernel = 250, + OpGetKernelNDrangeSubGroupCount = 251, + OpGetKernelNDrangeMaxSubGroupSize = 252, + OpGetKernelWorkGroupSize = 253, + OpGetKernelPreferredWorkGroupSizeMultiple = 254, + OpRetainEvent = 255, + OpReleaseEvent = 256, + OpCreateUserEvent = 257, + OpIsValidEvent = 258, + OpSetUserEventStatus = 259, + OpCaptureEventProfilingInfo = 260, + OpGetDefaultQueue = 261, + OpBuildNDRange = 262, + OpSatConvertSToU = 263, + OpSatConvertUToS = 264, + OpAtomicIMin = 265, + OpAtomicIMax = 266, +}; + +}; // end namespace spv + +#endif // #ifdef __cplusplus + + +#ifndef __cplusplus + +static const int SpvMagicNumber = 0x07230203; +static const int SpvVersion = 99; + +typedef unsigned int SpvId; + +static const unsigned int SpvOpCodeMask = 0xFFFF; +static const unsigned int SpvWordCountShift = 16; + +typedef enum SpvSourceLanguage_ { + SpvSourceLanguageUnknown = 0, + SpvSourceLanguageESSL = 1, + SpvSourceLanguageGLSL = 2, + SpvSourceLanguageOpenCL = 3, +} SpvSourceLanguage; + +typedef enum SpvExecutionModel_ { + SpvExecutionModelVertex = 0, + SpvExecutionModelTessellationControl = 1, + SpvExecutionModelTessellationEvaluation = 2, + SpvExecutionModelGeometry = 3, + SpvExecutionModelFragment = 4, + SpvExecutionModelGLCompute = 5, + SpvExecutionModelKernel = 6, +} SpvExecutionModel; + +typedef enum SpvAddressingModel_ { + SpvAddressingModelLogical = 0, + SpvAddressingModelPhysical32 = 1, + SpvAddressingModelPhysical64 = 2, +} SpvAddressingModel; + +typedef enum SpvMemoryModel_ { + SpvMemoryModelSimple = 0, + SpvMemoryModelGLSL450 = 1, + SpvMemoryModelOpenCL12 = 2, + SpvMemoryModelOpenCL20 = 3, + SpvMemoryModelOpenCL21 = 4, +} SpvMemoryModel; + +typedef enum SpvExecutionMode_ { + SpvExecutionModeInvocations = 0, + SpvExecutionModeSpacingEqual = 1, + SpvExecutionModeSpacingFractionalEven = 2, + SpvExecutionModeSpacingFractionalOdd = 3, + SpvExecutionModeVertexOrderCw = 4, + SpvExecutionModeVertexOrderCcw = 5, + SpvExecutionModePixelCenterInteger = 6, + SpvExecutionModeOriginUpperLeft = 7, + SpvExecutionModeEarlyFragmentTests = 8, + SpvExecutionModePointMode = 9, + SpvExecutionModeXfb = 10, + SpvExecutionModeDepthReplacing = 11, + SpvExecutionModeDepthAny = 12, + SpvExecutionModeDepthGreater = 13, + SpvExecutionModeDepthLess = 14, + SpvExecutionModeDepthUnchanged = 15, + SpvExecutionModeLocalSize = 16, + SpvExecutionModeLocalSizeHint = 17, + SpvExecutionModeInputPoints = 18, + SpvExecutionModeInputLines = 19, + SpvExecutionModeInputLinesAdjacency = 20, + SpvExecutionModeInputTriangles = 21, + SpvExecutionModeInputTrianglesAdjacency = 22, + SpvExecutionModeInputQuads = 23, + SpvExecutionModeInputIsolines = 24, + SpvExecutionModeOutputVertices = 25, + SpvExecutionModeOutputPoints = 26, + SpvExecutionModeOutputLineStrip = 27, + SpvExecutionModeOutputTriangleStrip = 28, + SpvExecutionModeVecTypeHint = 29, + SpvExecutionModeContractionOff = 30, +} SpvExecutionMode; + +typedef enum SpvStorageClass_ { + SpvStorageClassUniformConstant = 0, + SpvStorageClassInput = 1, + SpvStorageClassUniform = 2, + SpvStorageClassOutput = 3, + SpvStorageClassWorkgroupLocal = 4, + SpvStorageClassWorkgroupGlobal = 5, + SpvStorageClassPrivateGlobal = 6, + SpvStorageClassFunction = 7, + SpvStorageClassGeneric = 8, + SpvStorageClassPrivate = 9, + SpvStorageClassAtomicCounter = 10, +} SpvStorageClass; + +typedef enum SpvDim_ { + SpvDim1D = 0, + SpvDim2D = 1, + SpvDim3D = 2, + SpvDimCube = 3, + SpvDimRect = 4, + SpvDimBuffer = 5, +} SpvDim; + +typedef enum SpvSamplerAddressingMode_ { + SpvSamplerAddressingModeNone = 0, + SpvSamplerAddressingModeClampToEdge = 1, + SpvSamplerAddressingModeClamp = 2, + SpvSamplerAddressingModeRepeat = 3, + SpvSamplerAddressingModeRepeatMirrored = 4, +} SpvSamplerAddressingMode; + +typedef enum SpvSamplerFilterMode_ { + SpvSamplerFilterModeNearest = 0, + SpvSamplerFilterModeLinear = 1, +} SpvSamplerFilterMode; + +typedef enum SpvFPFastMathModeShift_ { + SpvFPFastMathModeNotNaNShift = 0, + SpvFPFastMathModeNotInfShift = 1, + SpvFPFastMathModeNSZShift = 2, + SpvFPFastMathModeAllowRecipShift = 3, + SpvFPFastMathModeFastShift = 4, +} SpvFPFastMathModeShift; + +typedef enum SpvFPFastMathModeMask_ { + SpvFPFastMathModeMaskNone = 0, + SpvFPFastMathModeNotNaNMask = 0x00000001, + SpvFPFastMathModeNotInfMask = 0x00000002, + SpvFPFastMathModeNSZMask = 0x00000004, + SpvFPFastMathModeAllowRecipMask = 0x00000008, + SpvFPFastMathModeFastMask = 0x00000010, +} SpvFPFastMathModeMask; + +typedef enum SpvFPRoundingMode_ { + SpvFPRoundingModeRTE = 0, + SpvFPRoundingModeRTZ = 1, + SpvFPRoundingModeRTP = 2, + SpvFPRoundingModeRTN = 3, +} SpvFPRoundingMode; + +typedef enum SpvLinkageType_ { + SpvLinkageTypeExport = 0, + SpvLinkageTypeImport = 1, +} SpvLinkageType; + +typedef enum SpvAccessQualifier_ { + SpvAccessQualifierReadOnly = 0, + SpvAccessQualifierWriteOnly = 1, + SpvAccessQualifierReadWrite = 2, +} SpvAccessQualifier; + +typedef enum SpvFunctionParameterAttribute_ { + SpvFunctionParameterAttributeZext = 0, + SpvFunctionParameterAttributeSext = 1, + SpvFunctionParameterAttributeByVal = 2, + SpvFunctionParameterAttributeSret = 3, + SpvFunctionParameterAttributeNoAlias = 4, + SpvFunctionParameterAttributeNoCapture = 5, + SpvFunctionParameterAttributeSVM = 6, + SpvFunctionParameterAttributeNoWrite = 7, + SpvFunctionParameterAttributeNoReadWrite = 8, +} SpvFunctionParameterAttribute; + +typedef enum SpvDecoration_ { + SpvDecorationPrecisionLow = 0, + SpvDecorationPrecisionMedium = 1, + SpvDecorationPrecisionHigh = 2, + SpvDecorationBlock = 3, + SpvDecorationBufferBlock = 4, + SpvDecorationRowMajor = 5, + SpvDecorationColMajor = 6, + SpvDecorationGLSLShared = 7, + SpvDecorationGLSLStd140 = 8, + SpvDecorationGLSLStd430 = 9, + SpvDecorationGLSLPacked = 10, + SpvDecorationSmooth = 11, + SpvDecorationNoperspective = 12, + SpvDecorationFlat = 13, + SpvDecorationPatch = 14, + SpvDecorationCentroid = 15, + SpvDecorationSample = 16, + SpvDecorationInvariant = 17, + SpvDecorationRestrict = 18, + SpvDecorationAliased = 19, + SpvDecorationVolatile = 20, + SpvDecorationConstant = 21, + SpvDecorationCoherent = 22, + SpvDecorationNonwritable = 23, + SpvDecorationNonreadable = 24, + SpvDecorationUniform = 25, + SpvDecorationNoStaticUse = 26, + SpvDecorationCPacked = 27, + SpvDecorationSaturatedConversion = 28, + SpvDecorationStream = 29, + SpvDecorationLocation = 30, + SpvDecorationComponent = 31, + SpvDecorationIndex = 32, + SpvDecorationBinding = 33, + SpvDecorationDescriptorSet = 34, + SpvDecorationOffset = 35, + SpvDecorationAlignment = 36, + SpvDecorationXfbBuffer = 37, + SpvDecorationStride = 38, + SpvDecorationBuiltIn = 39, + SpvDecorationFuncParamAttr = 40, + SpvDecorationFPRoundingMode = 41, + SpvDecorationFPFastMathMode = 42, + SpvDecorationLinkageAttributes = 43, + SpvDecorationSpecId = 44, +} SpvDecoration; + +typedef enum SpvBuiltIn_ { + SpvBuiltInPosition = 0, + SpvBuiltInPointSize = 1, + SpvBuiltInClipVertex = 2, + SpvBuiltInClipDistance = 3, + SpvBuiltInCullDistance = 4, + SpvBuiltInVertexId = 5, + SpvBuiltInInstanceId = 6, + SpvBuiltInPrimitiveId = 7, + SpvBuiltInInvocationId = 8, + SpvBuiltInLayer = 9, + SpvBuiltInViewportIndex = 10, + SpvBuiltInTessLevelOuter = 11, + SpvBuiltInTessLevelInner = 12, + SpvBuiltInTessCoord = 13, + SpvBuiltInPatchVertices = 14, + SpvBuiltInFragCoord = 15, + SpvBuiltInPointCoord = 16, + SpvBuiltInFrontFacing = 17, + SpvBuiltInSampleId = 18, + SpvBuiltInSamplePosition = 19, + SpvBuiltInSampleMask = 20, + SpvBuiltInFragColor = 21, + SpvBuiltInFragDepth = 22, + SpvBuiltInHelperInvocation = 23, + SpvBuiltInNumWorkgroups = 24, + SpvBuiltInWorkgroupSize = 25, + SpvBuiltInWorkgroupId = 26, + SpvBuiltInLocalInvocationId = 27, + SpvBuiltInGlobalInvocationId = 28, + SpvBuiltInLocalInvocationIndex = 29, + SpvBuiltInWorkDim = 30, + SpvBuiltInGlobalSize = 31, + SpvBuiltInEnqueuedWorkgroupSize = 32, + SpvBuiltInGlobalOffset = 33, + SpvBuiltInGlobalLinearId = 34, + SpvBuiltInWorkgroupLinearId = 35, + SpvBuiltInSubgroupSize = 36, + SpvBuiltInSubgroupMaxSize = 37, + SpvBuiltInNumSubgroups = 38, + SpvBuiltInNumEnqueuedSubgroups = 39, + SpvBuiltInSubgroupId = 40, + SpvBuiltInSubgroupLocalInvocationId = 41, +} SpvBuiltIn; + +typedef enum SpvSelectionControlShift_ { + SpvSelectionControlFlattenShift = 0, + SpvSelectionControlDontFlattenShift = 1, +} SpvSelectionControlShift; + +typedef enum SpvSelectionControlMask_ { + SpvSelectionControlMaskNone = 0, + SpvSelectionControlFlattenMask = 0x00000001, + SpvSelectionControlDontFlattenMask = 0x00000002, +} SpvSelectionControlMask; + +typedef enum SpvLoopControlShift_ { + SpvLoopControlUnrollShift = 0, + SpvLoopControlDontUnrollShift = 1, +} SpvLoopControlShift; + +typedef enum SpvLoopControlMask_ { + SpvLoopControlMaskNone = 0, + SpvLoopControlUnrollMask = 0x00000001, + SpvLoopControlDontUnrollMask = 0x00000002, +} SpvLoopControlMask; + +typedef enum SpvFunctionControlShift_ { + SpvFunctionControlInlineShift = 0, + SpvFunctionControlDontInlineShift = 1, + SpvFunctionControlPureShift = 2, + SpvFunctionControlConstShift = 3, +} SpvFunctionControlShift; + +typedef enum SpvFunctionControlMask_ { + SpvFunctionControlMaskNone = 0, + SpvFunctionControlInlineMask = 0x00000001, + SpvFunctionControlDontInlineMask = 0x00000002, + SpvFunctionControlPureMask = 0x00000004, + SpvFunctionControlConstMask = 0x00000008, +} SpvFunctionControlMask; + +typedef enum SpvMemorySemanticsShift_ { + SpvMemorySemanticsRelaxedShift = 0, + SpvMemorySemanticsSequentiallyConsistentShift = 1, + SpvMemorySemanticsAcquireShift = 2, + SpvMemorySemanticsReleaseShift = 3, + SpvMemorySemanticsUniformMemoryShift = 4, + SpvMemorySemanticsSubgroupMemoryShift = 5, + SpvMemorySemanticsWorkgroupLocalMemoryShift = 6, + SpvMemorySemanticsWorkgroupGlobalMemoryShift = 7, + SpvMemorySemanticsAtomicCounterMemoryShift = 8, + SpvMemorySemanticsImageMemoryShift = 9, +} SpvMemorySemanticsShift; + +typedef enum SpvMemorySemanticsMask_ { + SpvMemorySemanticsMaskNone = 0, + SpvMemorySemanticsRelaxedMask = 0x00000001, + SpvMemorySemanticsSequentiallyConsistentMask = 0x00000002, + SpvMemorySemanticsAcquireMask = 0x00000004, + SpvMemorySemanticsReleaseMask = 0x00000008, + SpvMemorySemanticsUniformMemoryMask = 0x00000010, + SpvMemorySemanticsSubgroupMemoryMask = 0x00000020, + SpvMemorySemanticsWorkgroupLocalMemoryMask = 0x00000040, + SpvMemorySemanticsWorkgroupGlobalMemoryMask = 0x00000080, + SpvMemorySemanticsAtomicCounterMemoryMask = 0x00000100, + SpvMemorySemanticsImageMemoryMask = 0x00000200, +} SpvMemorySemanticsMask; + +typedef enum SpvMemoryAccessShift_ { + SpvMemoryAccessVolatileShift = 0, + SpvMemoryAccessAlignedShift = 1, +} SpvMemoryAccessShift; + +typedef enum SpvMemoryAccessMask_ { + SpvMemoryAccessMaskNone = 0, + SpvMemoryAccessVolatileMask = 0x00000001, + SpvMemoryAccessAlignedMask = 0x00000002, +} SpvMemoryAccessMask; + +typedef enum SpvExecutionScope_ { + SpvExecutionScopeCrossDevice = 0, + SpvExecutionScopeDevice = 1, + SpvExecutionScopeWorkgroup = 2, + SpvExecutionScopeSubgroup = 3, +} SpvExecutionScope; + +typedef enum SpvGroupOperation_ { + SpvGroupOperationReduce = 0, + SpvGroupOperationInclusiveScan = 1, + SpvGroupOperationExclusiveScan = 2, +} SpvGroupOperation; + +typedef enum SpvKernelEnqueueFlags_ { + SpvKernelEnqueueFlagsNoWait = 0, + SpvKernelEnqueueFlagsWaitKernel = 1, + SpvKernelEnqueueFlagsWaitWorkGroup = 2, +} SpvKernelEnqueueFlags; + +typedef enum SpvKernelProfilingInfoShift_ { + SpvKernelProfilingInfoCmdExecTimeShift = 0, +} SpvKernelProfilingInfoShift; + +typedef enum SpvKernelProfilingInfoMask_ { + SpvKernelProfilingInfoMaskNone = 0, + SpvKernelProfilingInfoCmdExecTimeMask = 0x00000001, +} SpvKernelProfilingInfoMask; + +typedef enum SpvOp_ { + SpvOpNop = 0, + SpvOpSource = 1, + SpvOpSourceExtension = 2, + SpvOpExtension = 3, + SpvOpExtInstImport = 4, + SpvOpMemoryModel = 5, + SpvOpEntryPoint = 6, + SpvOpExecutionMode = 7, + SpvOpTypeVoid = 8, + SpvOpTypeBool = 9, + SpvOpTypeInt = 10, + SpvOpTypeFloat = 11, + SpvOpTypeVector = 12, + SpvOpTypeMatrix = 13, + SpvOpTypeSampler = 14, + SpvOpTypeFilter = 15, + SpvOpTypeArray = 16, + SpvOpTypeRuntimeArray = 17, + SpvOpTypeStruct = 18, + SpvOpTypeOpaque = 19, + SpvOpTypePointer = 20, + SpvOpTypeFunction = 21, + SpvOpTypeEvent = 22, + SpvOpTypeDeviceEvent = 23, + SpvOpTypeReserveId = 24, + SpvOpTypeQueue = 25, + SpvOpTypePipe = 26, + SpvOpConstantTrue = 27, + SpvOpConstantFalse = 28, + SpvOpConstant = 29, + SpvOpConstantComposite = 30, + SpvOpConstantSampler = 31, + SpvOpConstantNullPointer = 32, + SpvOpConstantNullObject = 33, + SpvOpSpecConstantTrue = 34, + SpvOpSpecConstantFalse = 35, + SpvOpSpecConstant = 36, + SpvOpSpecConstantComposite = 37, + SpvOpVariable = 38, + SpvOpVariableArray = 39, + SpvOpFunction = 40, + SpvOpFunctionParameter = 41, + SpvOpFunctionEnd = 42, + SpvOpFunctionCall = 43, + SpvOpExtInst = 44, + SpvOpUndef = 45, + SpvOpLoad = 46, + SpvOpStore = 47, + SpvOpPhi = 48, + SpvOpDecorationGroup = 49, + SpvOpDecorate = 50, + SpvOpMemberDecorate = 51, + SpvOpGroupDecorate = 52, + SpvOpGroupMemberDecorate = 53, + SpvOpName = 54, + SpvOpMemberName = 55, + SpvOpString = 56, + SpvOpLine = 57, + SpvOpVectorExtractDynamic = 58, + SpvOpVectorInsertDynamic = 59, + SpvOpVectorShuffle = 60, + SpvOpCompositeConstruct = 61, + SpvOpCompositeExtract = 62, + SpvOpCompositeInsert = 63, + SpvOpCopyObject = 64, + SpvOpCopyMemory = 65, + SpvOpCopyMemorySized = 66, + SpvOpSampler = 67, + SpvOpTextureSample = 68, + SpvOpTextureSampleDref = 69, + SpvOpTextureSampleLod = 70, + SpvOpTextureSampleProj = 71, + SpvOpTextureSampleGrad = 72, + SpvOpTextureSampleOffset = 73, + SpvOpTextureSampleProjLod = 74, + SpvOpTextureSampleProjGrad = 75, + SpvOpTextureSampleLodOffset = 76, + SpvOpTextureSampleProjOffset = 77, + SpvOpTextureSampleGradOffset = 78, + SpvOpTextureSampleProjLodOffset = 79, + SpvOpTextureSampleProjGradOffset = 80, + SpvOpTextureFetchTexelLod = 81, + SpvOpTextureFetchTexelOffset = 82, + SpvOpTextureFetchSample = 83, + SpvOpTextureFetchTexel = 84, + SpvOpTextureGather = 85, + SpvOpTextureGatherOffset = 86, + SpvOpTextureGatherOffsets = 87, + SpvOpTextureQuerySizeLod = 88, + SpvOpTextureQuerySize = 89, + SpvOpTextureQueryLod = 90, + SpvOpTextureQueryLevels = 91, + SpvOpTextureQuerySamples = 92, + SpvOpAccessChain = 93, + SpvOpInBoundsAccessChain = 94, + SpvOpSNegate = 95, + SpvOpFNegate = 96, + SpvOpNot = 97, + SpvOpAny = 98, + SpvOpAll = 99, + SpvOpConvertFToU = 100, + SpvOpConvertFToS = 101, + SpvOpConvertSToF = 102, + SpvOpConvertUToF = 103, + SpvOpUConvert = 104, + SpvOpSConvert = 105, + SpvOpFConvert = 106, + SpvOpConvertPtrToU = 107, + SpvOpConvertUToPtr = 108, + SpvOpPtrCastToGeneric = 109, + SpvOpGenericCastToPtr = 110, + SpvOpBitcast = 111, + SpvOpTranspose = 112, + SpvOpIsNan = 113, + SpvOpIsInf = 114, + SpvOpIsFinite = 115, + SpvOpIsNormal = 116, + SpvOpSignBitSet = 117, + SpvOpLessOrGreater = 118, + SpvOpOrdered = 119, + SpvOpUnordered = 120, + SpvOpArrayLength = 121, + SpvOpIAdd = 122, + SpvOpFAdd = 123, + SpvOpISub = 124, + SpvOpFSub = 125, + SpvOpIMul = 126, + SpvOpFMul = 127, + SpvOpUDiv = 128, + SpvOpSDiv = 129, + SpvOpFDiv = 130, + SpvOpUMod = 131, + SpvOpSRem = 132, + SpvOpSMod = 133, + SpvOpFRem = 134, + SpvOpFMod = 135, + SpvOpVectorTimesScalar = 136, + SpvOpMatrixTimesScalar = 137, + SpvOpVectorTimesMatrix = 138, + SpvOpMatrixTimesVector = 139, + SpvOpMatrixTimesMatrix = 140, + SpvOpOuterProduct = 141, + SpvOpDot = 142, + SpvOpShiftRightLogical = 143, + SpvOpShiftRightArithmetic = 144, + SpvOpShiftLeftLogical = 145, + SpvOpLogicalOr = 146, + SpvOpLogicalXor = 147, + SpvOpLogicalAnd = 148, + SpvOpBitwiseOr = 149, + SpvOpBitwiseXor = 150, + SpvOpBitwiseAnd = 151, + SpvOpSelect = 152, + SpvOpIEqual = 153, + SpvOpFOrdEqual = 154, + SpvOpFUnordEqual = 155, + SpvOpINotEqual = 156, + SpvOpFOrdNotEqual = 157, + SpvOpFUnordNotEqual = 158, + SpvOpULessThan = 159, + SpvOpSLessThan = 160, + SpvOpFOrdLessThan = 161, + SpvOpFUnordLessThan = 162, + SpvOpUGreaterThan = 163, + SpvOpSGreaterThan = 164, + SpvOpFOrdGreaterThan = 165, + SpvOpFUnordGreaterThan = 166, + SpvOpULessThanEqual = 167, + SpvOpSLessThanEqual = 168, + SpvOpFOrdLessThanEqual = 169, + SpvOpFUnordLessThanEqual = 170, + SpvOpUGreaterThanEqual = 171, + SpvOpSGreaterThanEqual = 172, + SpvOpFOrdGreaterThanEqual = 173, + SpvOpFUnordGreaterThanEqual = 174, + SpvOpDPdx = 175, + SpvOpDPdy = 176, + SpvOpFwidth = 177, + SpvOpDPdxFine = 178, + SpvOpDPdyFine = 179, + SpvOpFwidthFine = 180, + SpvOpDPdxCoarse = 181, + SpvOpDPdyCoarse = 182, + SpvOpFwidthCoarse = 183, + SpvOpEmitVertex = 184, + SpvOpEndPrimitive = 185, + SpvOpEmitStreamVertex = 186, + SpvOpEndStreamPrimitive = 187, + SpvOpControlBarrier = 188, + SpvOpMemoryBarrier = 189, + SpvOpImagePointer = 190, + SpvOpAtomicInit = 191, + SpvOpAtomicLoad = 192, + SpvOpAtomicStore = 193, + SpvOpAtomicExchange = 194, + SpvOpAtomicCompareExchange = 195, + SpvOpAtomicCompareExchangeWeak = 196, + SpvOpAtomicIIncrement = 197, + SpvOpAtomicIDecrement = 198, + SpvOpAtomicIAdd = 199, + SpvOpAtomicISub = 200, + SpvOpAtomicUMin = 201, + SpvOpAtomicUMax = 202, + SpvOpAtomicAnd = 203, + SpvOpAtomicOr = 204, + SpvOpAtomicXor = 205, + SpvOpLoopMerge = 206, + SpvOpSelectionMerge = 207, + SpvOpLabel = 208, + SpvOpBranch = 209, + SpvOpBranchConditional = 210, + SpvOpSwitch = 211, + SpvOpKill = 212, + SpvOpReturn = 213, + SpvOpReturnValue = 214, + SpvOpUnreachable = 215, + SpvOpLifetimeStart = 216, + SpvOpLifetimeStop = 217, + SpvOpCompileFlag = 218, + SpvOpAsyncGroupCopy = 219, + SpvOpWaitGroupEvents = 220, + SpvOpGroupAll = 221, + SpvOpGroupAny = 222, + SpvOpGroupBroadcast = 223, + SpvOpGroupIAdd = 224, + SpvOpGroupFAdd = 225, + SpvOpGroupFMin = 226, + SpvOpGroupUMin = 227, + SpvOpGroupSMin = 228, + SpvOpGroupFMax = 229, + SpvOpGroupUMax = 230, + SpvOpGroupSMax = 231, + SpvOpGenericCastToPtrExplicit = 232, + SpvOpGenericPtrMemSemantics = 233, + SpvOpReadPipe = 234, + SpvOpWritePipe = 235, + SpvOpReservedReadPipe = 236, + SpvOpReservedWritePipe = 237, + SpvOpReserveReadPipePackets = 238, + SpvOpReserveWritePipePackets = 239, + SpvOpCommitReadPipe = 240, + SpvOpCommitWritePipe = 241, + SpvOpIsValidReserveId = 242, + SpvOpGetNumPipePackets = 243, + SpvOpGetMaxPipePackets = 244, + SpvOpGroupReserveReadPipePackets = 245, + SpvOpGroupReserveWritePipePackets = 246, + SpvOpGroupCommitReadPipe = 247, + SpvOpGroupCommitWritePipe = 248, + SpvOpEnqueueMarker = 249, + SpvOpEnqueueKernel = 250, + SpvOpGetKernelNDrangeSubGroupCount = 251, + SpvOpGetKernelNDrangeMaxSubGroupSize = 252, + SpvOpGetKernelWorkGroupSize = 253, + SpvOpGetKernelPreferredWorkGroupSizeMultiple = 254, + SpvOpRetainEvent = 255, + SpvOpReleaseEvent = 256, + SpvOpCreateUserEvent = 257, + SpvOpIsValidEvent = 258, + SpvOpSetUserEventStatus = 259, + SpvOpCaptureEventProfilingInfo = 260, + SpvOpGetDefaultQueue = 261, + SpvOpBuildNDRange = 262, + SpvOpSatConvertSToU = 263, + SpvOpSatConvertUToS = 264, + SpvOpAtomicIMin = 265, + SpvOpAtomicIMax = 266, +} SpvOp; + +#endif // #ifndef __cplusplus + +#endif // #ifndef spirv_H diff --git a/src/glsl/nir/spirv2nir.c b/src/glsl/nir/spirv2nir.c new file mode 100644 index 00000000000..0eed23fbc3f --- /dev/null +++ b/src/glsl/nir/spirv2nir.c @@ -0,0 +1,54 @@ +/* + * Copyright © 2015 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. + * + * Authors: + * Jason Ekstrand ([email protected]) + * + */ + +/* + * A simple executable that opens a SPIR-V shader, converts it to NIR, and + * dumps out the result. This should be useful for testing the + * spirv_to_nir code. + */ + +#include "nir_spirv.h" + +#include <sys/mman.h> +#include <sys/types.h> +#include <fcntl.h> +#include <unistd.h> + +int main(int argc, char **argv) +{ + int fd = open(argv[1], O_RDONLY); + off_t len = lseek(fd, 0, SEEK_END); + + assert(len % 4 == 0); + size_t word_count = len / 4; + + const void *map = mmap(NULL, len, PROT_READ, MAP_PRIVATE, fd, 0); + assert(map != NULL); + + nir_shader *shader = spirv_to_nir(map, word_count, NULL); + nir_print_shader(shader, stderr); +} diff --git a/src/glsl/nir/spirv_glsl450_to_nir.c b/src/glsl/nir/spirv_glsl450_to_nir.c new file mode 100644 index 00000000000..3b9d0940aad --- /dev/null +++ b/src/glsl/nir/spirv_glsl450_to_nir.c @@ -0,0 +1,284 @@ +/* + * Copyright © 2015 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. + * + * Authors: + * Jason Ekstrand ([email protected]) + * + */ + +#include "spirv_to_nir_private.h" + +enum GLSL450Entrypoint { + Round = 0, + RoundEven = 1, + Trunc = 2, + Abs = 3, + Sign = 4, + Floor = 5, + Ceil = 6, + Fract = 7, + + Radians = 8, + Degrees = 9, + Sin = 10, + Cos = 11, + Tan = 12, + Asin = 13, + Acos = 14, + Atan = 15, + Sinh = 16, + Cosh = 17, + Tanh = 18, + Asinh = 19, + Acosh = 20, + Atanh = 21, + Atan2 = 22, + + Pow = 23, + Exp = 24, + Log = 25, + Exp2 = 26, + Log2 = 27, + Sqrt = 28, + InverseSqrt = 29, + + Determinant = 30, + MatrixInverse = 31, + + Modf = 32, // second argument needs the OpVariable = , not an OpLoad + Min = 33, + Max = 34, + Clamp = 35, + Mix = 36, + Step = 37, + SmoothStep = 38, + + FloatBitsToInt = 39, + FloatBitsToUint = 40, + IntBitsToFloat = 41, + UintBitsToFloat = 42, + + Fma = 43, + Frexp = 44, + Ldexp = 45, + + PackSnorm4x8 = 46, + PackUnorm4x8 = 47, + PackSnorm2x16 = 48, + PackUnorm2x16 = 49, + PackHalf2x16 = 50, + PackDouble2x32 = 51, + UnpackSnorm2x16 = 52, + UnpackUnorm2x16 = 53, + UnpackHalf2x16 = 54, + UnpackSnorm4x8 = 55, + UnpackUnorm4x8 = 56, + UnpackDouble2x32 = 57, + + Length = 58, + Distance = 59, + Cross = 60, + Normalize = 61, + Ftransform = 62, + FaceForward = 63, + Reflect = 64, + Refract = 65, + + UaddCarry = 66, + UsubBorrow = 67, + UmulExtended = 68, + ImulExtended = 69, + BitfieldExtract = 70, + BitfieldInsert = 71, + BitfieldReverse = 72, + BitCount = 73, + FindLSB = 74, + FindMSB = 75, + + InterpolateAtCentroid = 76, + InterpolateAtSample = 77, + InterpolateAtOffset = 78, + + Count +}; + +static nir_ssa_def* +build_length(nir_builder *b, nir_ssa_def *vec) +{ + switch (vec->num_components) { + case 1: return nir_fsqrt(b, nir_fmul(b, vec, vec)); + case 2: return nir_fsqrt(b, nir_fdot2(b, vec, vec)); + case 3: return nir_fsqrt(b, nir_fdot3(b, vec, vec)); + case 4: return nir_fsqrt(b, nir_fdot4(b, vec, vec)); + default: + unreachable("Invalid number of components"); + } +} + +static void +handle_glsl450_alu(struct vtn_builder *b, enum GLSL450Entrypoint entrypoint, + const uint32_t *w, unsigned count) +{ + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa); + val->type = vtn_value(b, w[1], vtn_value_type_type)->type; + + /* Collect the various SSA sources */ + unsigned num_inputs = count - 5; + nir_ssa_def *src[3]; + for (unsigned i = 0; i < num_inputs; i++) + src[i] = vtn_ssa_value(b, w[i + 5]); + + nir_op op; + switch (entrypoint) { + case Round: op = nir_op_fround_even; break; /* TODO */ + case RoundEven: op = nir_op_fround_even; break; + case Trunc: op = nir_op_ftrunc; break; + case Abs: op = nir_op_fabs; break; + case Sign: op = nir_op_fsign; break; + case Floor: op = nir_op_ffloor; break; + case Ceil: op = nir_op_fceil; break; + case Fract: op = nir_op_ffract; break; + case Radians: + val->ssa = nir_fmul(&b->nb, src[0], nir_imm_float(&b->nb, 0.01745329251)); + return; + case Degrees: + val->ssa = nir_fmul(&b->nb, src[0], nir_imm_float(&b->nb, 57.2957795131)); + return; + case Sin: op = nir_op_fsin; break; + case Cos: op = nir_op_fcos; break; + case Tan: + val->ssa = nir_fdiv(&b->nb, nir_fsin(&b->nb, src[0]), + nir_fcos(&b->nb, src[0])); + return; + case Pow: op = nir_op_fpow; break; + case Exp2: op = nir_op_fexp2; break; + case Log2: op = nir_op_flog2; break; + case Sqrt: op = nir_op_fsqrt; break; + case InverseSqrt: op = nir_op_frsq; break; + + case Modf: op = nir_op_fmod; break; + case Min: op = nir_op_fmin; break; + case Max: op = nir_op_fmax; break; + case Mix: op = nir_op_flrp; break; + case Step: + val->ssa = nir_sge(&b->nb, src[1], src[0]); + return; + + case FloatBitsToInt: + case FloatBitsToUint: + case IntBitsToFloat: + case UintBitsToFloat: + /* Probably going to be removed from the final version of the spec. */ + val->ssa = src[0]; + return; + + case Fma: op = nir_op_ffma; break; + case Ldexp: op = nir_op_ldexp; break; + + /* Packing/Unpacking functions */ + case PackSnorm4x8: op = nir_op_pack_snorm_4x8; break; + case PackUnorm4x8: op = nir_op_pack_unorm_4x8; break; + case PackSnorm2x16: op = nir_op_pack_snorm_2x16; break; + case PackUnorm2x16: op = nir_op_pack_unorm_2x16; break; + case PackHalf2x16: op = nir_op_pack_half_2x16; break; + case UnpackSnorm4x8: op = nir_op_unpack_snorm_4x8; break; + case UnpackUnorm4x8: op = nir_op_unpack_unorm_4x8; break; + case UnpackSnorm2x16: op = nir_op_unpack_snorm_2x16; break; + case UnpackUnorm2x16: op = nir_op_unpack_unorm_2x16; break; + case UnpackHalf2x16: op = nir_op_unpack_half_2x16; break; + + case Length: + val->ssa = build_length(&b->nb, src[0]); + return; + case Distance: + val->ssa = build_length(&b->nb, nir_fsub(&b->nb, src[0], src[1])); + return; + case Normalize: + val->ssa = nir_fdiv(&b->nb, src[0], build_length(&b->nb, src[0])); + return; + + case UaddCarry: op = nir_op_uadd_carry; break; + case UsubBorrow: op = nir_op_usub_borrow; break; + case BitfieldExtract: op = nir_op_ubitfield_extract; break; /* TODO */ + case BitfieldInsert: op = nir_op_bitfield_insert; break; + case BitfieldReverse: op = nir_op_bitfield_reverse; break; + case BitCount: op = nir_op_bit_count; break; + case FindLSB: op = nir_op_find_lsb; break; + case FindMSB: op = nir_op_ufind_msb; break; /* TODO */ + + case Exp: + case Log: + case Clamp: + case Asin: + case Acos: + case Atan: + case Atan2: + case Sinh: + case Cosh: + case Tanh: + case Asinh: + case Acosh: + case Atanh: + case SmoothStep: + case Frexp: + case PackDouble2x32: + case UnpackDouble2x32: + case Cross: + case Ftransform: + case FaceForward: + case Reflect: + case Refract: + case UmulExtended: + case ImulExtended: + default: + unreachable("Unhandled opcode"); + } + + nir_alu_instr *instr = nir_alu_instr_create(b->shader, op); + nir_ssa_dest_init(&instr->instr, &instr->dest.dest, + glsl_get_vector_elements(val->type), val->name); + val->ssa = &instr->dest.dest.ssa; + + for (unsigned i = 0; i < nir_op_infos[op].num_inputs; i++) + instr->src[i].src = nir_src_for_ssa(src[i]); + + nir_builder_instr_insert(&b->nb, &instr->instr); +} + +bool +vtn_handle_glsl450_instruction(struct vtn_builder *b, uint32_t ext_opcode, + const uint32_t *words, unsigned count) +{ + switch ((enum GLSL450Entrypoint)ext_opcode) { + case Determinant: + case MatrixInverse: + case InterpolateAtCentroid: + case InterpolateAtSample: + case InterpolateAtOffset: + unreachable("Unhandled opcode"); + + default: + handle_glsl450_alu(b, (enum GLSL450Entrypoint)ext_opcode, words, count); + } + + return true; +} diff --git a/src/glsl/nir/spirv_to_nir.c b/src/glsl/nir/spirv_to_nir.c new file mode 100644 index 00000000000..1fc1b8bc5dc --- /dev/null +++ b/src/glsl/nir/spirv_to_nir.c @@ -0,0 +1,1572 @@ +/* + * Copyright © 2015 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. + * + * Authors: + * Jason Ekstrand ([email protected]) + * + */ + +#include "spirv_to_nir_private.h" +#include "nir_vla.h" + +nir_ssa_def * +vtn_ssa_value(struct vtn_builder *b, uint32_t value_id) +{ + struct vtn_value *val = vtn_untyped_value(b, value_id); + switch (val->value_type) { + case vtn_value_type_constant: { + assert(glsl_type_is_vector_or_scalar(val->type)); + unsigned num_components = glsl_get_vector_elements(val->type); + nir_load_const_instr *load = + nir_load_const_instr_create(b->shader, num_components); + + for (unsigned i = 0; i < num_components; i++) + load->value.u[0] = val->constant->value.u[0]; + + nir_builder_instr_insert(&b->nb, &load->instr); + return &load->def; + } + + case vtn_value_type_ssa: + return val->ssa; + default: + unreachable("Invalid type for an SSA value"); + } +} + +static char * +vtn_string_literal(struct vtn_builder *b, const uint32_t *words, + unsigned word_count) +{ + return ralloc_strndup(b, (char *)words, word_count * sizeof(*words)); +} + +static const uint32_t * +vtn_foreach_instruction(struct vtn_builder *b, const uint32_t *start, + const uint32_t *end, vtn_instruction_handler handler) +{ + const uint32_t *w = start; + while (w < end) { + SpvOp opcode = w[0] & SpvOpCodeMask; + unsigned count = w[0] >> SpvWordCountShift; + assert(count >= 1 && w + count <= end); + + if (!handler(b, opcode, w, count)) + return w; + + w += count; + } + assert(w == end); + return w; +} + +static void +vtn_handle_extension(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + switch (opcode) { + case SpvOpExtInstImport: { + struct vtn_value *val = vtn_push_value(b, w[1], vtn_value_type_extension); + if (strcmp((const char *)&w[2], "GLSL.std.450") == 0) { + val->ext_handler = vtn_handle_glsl450_instruction; + } else { + assert(!"Unsupported extension"); + } + break; + } + + case SpvOpExtInst: { + struct vtn_value *val = vtn_value(b, w[3], vtn_value_type_extension); + bool handled = val->ext_handler(b, w[4], w, count); + (void)handled; + assert(handled); + break; + } + + default: + unreachable("Unhandled opcode"); + } +} + +static void +_foreach_decoration_helper(struct vtn_builder *b, + struct vtn_value *base_value, + struct vtn_value *value, + vtn_decoration_foreach_cb cb, void *data) +{ + for (struct vtn_decoration *dec = value->decoration; dec; dec = dec->next) { + if (dec->group) { + assert(dec->group->value_type == vtn_value_type_decoration_group); + _foreach_decoration_helper(b, base_value, dec->group, cb, data); + } else { + cb(b, base_value, dec, data); + } + } +} + +/** Iterates (recursively if needed) over all of the decorations on a value + * + * This function iterates over all of the decorations applied to a given + * value. If it encounters a decoration group, it recurses into the group + * and iterates over all of those decorations as well. + */ +void +vtn_foreach_decoration(struct vtn_builder *b, struct vtn_value *value, + vtn_decoration_foreach_cb cb, void *data) +{ + _foreach_decoration_helper(b, value, value, cb, data); +} + +static void +vtn_handle_decoration(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + switch (opcode) { + case SpvOpDecorationGroup: + vtn_push_value(b, w[1], vtn_value_type_undef); + break; + + case SpvOpDecorate: { + struct vtn_value *val = &b->values[w[1]]; + + struct vtn_decoration *dec = rzalloc(b, struct vtn_decoration); + dec->decoration = w[2]; + dec->literals = &w[3]; + + /* Link into the list */ + dec->next = val->decoration; + val->decoration = dec; + break; + } + + case SpvOpGroupDecorate: { + struct vtn_value *group = &b->values[w[1]]; + assert(group->value_type == vtn_value_type_decoration_group); + + for (unsigned i = 2; i < count; i++) { + struct vtn_value *val = &b->values[w[i]]; + struct vtn_decoration *dec = rzalloc(b, struct vtn_decoration); + dec->group = group; + + /* Link into the list */ + dec->next = val->decoration; + val->decoration = dec; + } + break; + } + + case SpvOpGroupMemberDecorate: + assert(!"Bad instruction. Khronos Bug #13513"); + break; + + default: + unreachable("Unhandled opcode"); + } +} + +static const struct glsl_type * +vtn_handle_type(struct vtn_builder *b, SpvOp opcode, + const uint32_t *args, unsigned count) +{ + switch (opcode) { + case SpvOpTypeVoid: + return glsl_void_type(); + case SpvOpTypeBool: + return glsl_bool_type(); + case SpvOpTypeInt: + return glsl_int_type(); + case SpvOpTypeFloat: + return glsl_float_type(); + + case SpvOpTypeVector: { + const struct glsl_type *base = + vtn_value(b, args[0], vtn_value_type_type)->type; + unsigned elems = args[1]; + + assert(glsl_type_is_scalar(base)); + return glsl_vector_type(glsl_get_base_type(base), elems); + } + + case SpvOpTypeMatrix: { + const struct glsl_type *base = + vtn_value(b, args[0], vtn_value_type_type)->type; + unsigned columns = args[1]; + + assert(glsl_type_is_vector(base)); + return glsl_matrix_type(glsl_get_base_type(base), + glsl_get_vector_elements(base), + columns); + } + + case SpvOpTypeArray: + return glsl_array_type(b->values[args[0]].type, args[1]); + + case SpvOpTypeStruct: { + NIR_VLA(struct glsl_struct_field, fields, count); + for (unsigned i = 0; i < count; i++) { + /* TODO: Handle decorators */ + fields[i].type = vtn_value(b, args[i], vtn_value_type_type)->type; + fields[i].name = ralloc_asprintf(b, "field%d", i); + fields[i].location = -1; + fields[i].interpolation = 0; + fields[i].centroid = 0; + fields[i].sample = 0; + fields[i].matrix_layout = 2; + fields[i].stream = -1; + } + return glsl_struct_type(fields, count, "struct"); + } + + case SpvOpTypeFunction: { + const struct glsl_type *return_type = b->values[args[0]].type; + NIR_VLA(struct glsl_function_param, params, count - 1); + for (unsigned i = 1; i < count; i++) { + params[i - 1].type = vtn_value(b, args[i], vtn_value_type_type)->type; + + /* FIXME: */ + params[i - 1].in = true; + params[i - 1].out = true; + } + return glsl_function_type(return_type, params, count - 1); + } + + case SpvOpTypePointer: + /* FIXME: For now, we'll just do the really lame thing and return + * the same type. The validator should ensure that the proper number + * of dereferences happen + */ + return vtn_value(b, args[1], vtn_value_type_type)->type; + + case SpvOpTypeSampler: { + const struct glsl_type *sampled_type = + vtn_value(b, args[0], vtn_value_type_type)->type; + + assert(glsl_type_is_vector_or_scalar(sampled_type)); + + enum glsl_sampler_dim dim; + switch ((SpvDim)args[1]) { + case SpvDim1D: dim = GLSL_SAMPLER_DIM_1D; break; + case SpvDim2D: dim = GLSL_SAMPLER_DIM_2D; break; + case SpvDim3D: dim = GLSL_SAMPLER_DIM_3D; break; + case SpvDimCube: dim = GLSL_SAMPLER_DIM_CUBE; break; + case SpvDimRect: dim = GLSL_SAMPLER_DIM_RECT; break; + case SpvDimBuffer: dim = GLSL_SAMPLER_DIM_BUF; break; + default: + unreachable("Invalid SPIR-V Sampler dimension"); + } + + /* TODO: Handle the various texture image/filter options */ + (void)args[2]; + + bool is_array = args[3]; + bool is_shadow = args[4]; + + assert(args[5] == 0 && "FIXME: Handl multi-sampled textures"); + + return glsl_sampler_type(dim, is_shadow, is_array, + glsl_get_base_type(sampled_type)); + } + + case SpvOpTypeRuntimeArray: + case SpvOpTypeOpaque: + case SpvOpTypeEvent: + case SpvOpTypeDeviceEvent: + case SpvOpTypeReserveId: + case SpvOpTypeQueue: + case SpvOpTypePipe: + default: + unreachable("Unhandled opcode"); + } +} + +static void +vtn_handle_constant(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_constant); + val->type = vtn_value(b, w[1], vtn_value_type_type)->type; + val->constant = ralloc(b, nir_constant); + switch (opcode) { + case SpvOpConstantTrue: + assert(val->type == glsl_bool_type()); + val->constant->value.u[0] = NIR_TRUE; + break; + case SpvOpConstantFalse: + assert(val->type == glsl_bool_type()); + val->constant->value.u[0] = NIR_FALSE; + break; + case SpvOpConstant: + assert(glsl_type_is_scalar(val->type)); + val->constant->value.u[0] = w[3]; + break; + case SpvOpConstantComposite: { + unsigned elem_count = count - 3; + nir_constant **elems = ralloc_array(b, nir_constant *, elem_count); + for (unsigned i = 0; i < elem_count; i++) + elems[i] = vtn_value(b, w[i + 3], vtn_value_type_constant)->constant; + + switch (glsl_get_base_type(val->type)) { + case GLSL_TYPE_UINT: + case GLSL_TYPE_INT: + case GLSL_TYPE_FLOAT: + case GLSL_TYPE_BOOL: + if (glsl_type_is_matrix(val->type)) { + unsigned rows = glsl_get_vector_elements(val->type); + assert(glsl_get_matrix_columns(val->type) == elem_count); + for (unsigned i = 0; i < elem_count; i++) + for (unsigned j = 0; j < rows; j++) + val->constant->value.u[rows * i + j] = elems[i]->value.u[j]; + } else { + assert(glsl_type_is_vector(val->type)); + assert(glsl_get_vector_elements(val->type) == elem_count); + for (unsigned i = 0; i < elem_count; i++) + val->constant->value.u[i] = elems[i]->value.u[0]; + } + ralloc_free(elems); + break; + + case GLSL_TYPE_STRUCT: + case GLSL_TYPE_ARRAY: + ralloc_steal(val->constant, elems); + val->constant->elements = elems; + break; + + default: + unreachable("Unsupported type for constants"); + } + break; + } + + default: + unreachable("Unhandled opcode"); + } +} + +static void +var_decoration_cb(struct vtn_builder *b, struct vtn_value *val, + const struct vtn_decoration *dec, void *void_var) +{ + assert(val->value_type == vtn_value_type_deref); + assert(val->deref->deref.child == NULL); + assert(val->deref->var == void_var); + + nir_variable *var = void_var; + switch (dec->decoration) { + case SpvDecorationPrecisionLow: + case SpvDecorationPrecisionMedium: + case SpvDecorationPrecisionHigh: + break; /* FIXME: Do nothing with these for now. */ + case SpvDecorationSmooth: + var->data.interpolation = INTERP_QUALIFIER_SMOOTH; + break; + case SpvDecorationNoperspective: + var->data.interpolation = INTERP_QUALIFIER_NOPERSPECTIVE; + break; + case SpvDecorationFlat: + var->data.interpolation = INTERP_QUALIFIER_FLAT; + break; + case SpvDecorationCentroid: + var->data.centroid = true; + break; + case SpvDecorationSample: + var->data.sample = true; + break; + case SpvDecorationInvariant: + var->data.invariant = true; + break; + case SpvDecorationConstant: + assert(var->constant_initializer != NULL); + var->data.read_only = true; + break; + case SpvDecorationNonwritable: + var->data.read_only = true; + break; + case SpvDecorationLocation: + var->data.explicit_location = true; + var->data.location = dec->literals[0]; + break; + case SpvDecorationComponent: + var->data.location_frac = dec->literals[0]; + break; + case SpvDecorationIndex: + var->data.explicit_index = true; + var->data.index = dec->literals[0]; + break; + case SpvDecorationBinding: + var->data.explicit_binding = true; + var->data.binding = dec->literals[0]; + break; + case SpvDecorationBlock: + case SpvDecorationBufferBlock: + case SpvDecorationRowMajor: + case SpvDecorationColMajor: + case SpvDecorationGLSLShared: + case SpvDecorationGLSLStd140: + case SpvDecorationGLSLStd430: + case SpvDecorationGLSLPacked: + case SpvDecorationPatch: + case SpvDecorationRestrict: + case SpvDecorationAliased: + case SpvDecorationVolatile: + case SpvDecorationCoherent: + case SpvDecorationNonreadable: + case SpvDecorationUniform: + /* This is really nice but we have no use for it right now. */ + case SpvDecorationNoStaticUse: + case SpvDecorationCPacked: + case SpvDecorationSaturatedConversion: + case SpvDecorationStream: + case SpvDecorationDescriptorSet: + case SpvDecorationOffset: + case SpvDecorationAlignment: + case SpvDecorationXfbBuffer: + case SpvDecorationStride: + case SpvDecorationBuiltIn: + case SpvDecorationFuncParamAttr: + case SpvDecorationFPRoundingMode: + case SpvDecorationFPFastMathMode: + case SpvDecorationLinkageAttributes: + case SpvDecorationSpecId: + default: + unreachable("Unhandled variable decoration"); + } +} + +static void +vtn_handle_variables(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + switch (opcode) { + case SpvOpVariable: { + const struct glsl_type *type = + vtn_value(b, w[1], vtn_value_type_type)->type; + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_deref); + + nir_variable *var = ralloc(b->shader, nir_variable); + + var->type = type; + var->name = ralloc_strdup(var, val->name); + + switch ((SpvStorageClass)w[3]) { + case SpvStorageClassUniformConstant: + var->data.mode = nir_var_uniform; + var->data.read_only = true; + break; + case SpvStorageClassInput: + var->data.mode = nir_var_shader_in; + var->data.read_only = true; + break; + case SpvStorageClassOutput: + var->data.mode = nir_var_shader_out; + break; + case SpvStorageClassPrivateGlobal: + var->data.mode = nir_var_global; + break; + case SpvStorageClassFunction: + var->data.mode = nir_var_local; + break; + case SpvStorageClassUniform: + case SpvStorageClassWorkgroupLocal: + case SpvStorageClassWorkgroupGlobal: + case SpvStorageClassGeneric: + case SpvStorageClassPrivate: + case SpvStorageClassAtomicCounter: + default: + unreachable("Unhandled variable storage class"); + } + + if (count > 4) { + assert(count == 5); + var->constant_initializer = + vtn_value(b, w[4], vtn_value_type_constant)->constant; + } + + if (var->data.mode == nir_var_local) { + exec_list_push_tail(&b->impl->locals, &var->node); + } else { + exec_list_push_tail(&b->shader->globals, &var->node); + } + + val->deref = nir_deref_var_create(b->shader, var); + + vtn_foreach_decoration(b, val, var_decoration_cb, var); + break; + } + + case SpvOpAccessChain: + case SpvOpInBoundsAccessChain: { + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_deref); + nir_deref_var *base = vtn_value(b, w[3], vtn_value_type_deref)->deref; + val->deref = nir_deref_as_var(nir_copy_deref(b, &base->deref)); + + nir_deref *tail = &val->deref->deref; + while (tail->child) + tail = tail->child; + + for (unsigned i = 0; i < count - 4; i++) { + assert(w[i + 4] < b->value_id_bound); + struct vtn_value *idx_val = &b->values[w[i + 4]]; + + enum glsl_base_type base_type = glsl_get_base_type(tail->type); + switch (base_type) { + case GLSL_TYPE_UINT: + case GLSL_TYPE_INT: + case GLSL_TYPE_FLOAT: + case GLSL_TYPE_DOUBLE: + case GLSL_TYPE_BOOL: + case GLSL_TYPE_ARRAY: { + nir_deref_array *deref_arr = nir_deref_array_create(b); + if (base_type == GLSL_TYPE_ARRAY) { + deref_arr->deref.type = glsl_get_array_element(tail->type); + } else if (glsl_type_is_matrix(tail->type)) { + deref_arr->deref.type = glsl_get_column_type(tail->type); + } else { + assert(glsl_type_is_vector(tail->type)); + deref_arr->deref.type = glsl_scalar_type(base_type); + } + + if (idx_val->value_type == vtn_value_type_constant) { + unsigned idx = idx_val->constant->value.u[0]; + deref_arr->deref_array_type = nir_deref_array_type_direct; + deref_arr->base_offset = idx; + } else { + assert(idx_val->value_type == vtn_value_type_ssa); + deref_arr->deref_array_type = nir_deref_array_type_indirect; + deref_arr->base_offset = 0; + deref_arr->indirect = nir_src_for_ssa(vtn_ssa_value(b, w[1])); + } + tail->child = &deref_arr->deref; + break; + } + + case GLSL_TYPE_STRUCT: { + assert(idx_val->value_type == vtn_value_type_constant); + unsigned idx = idx_val->constant->value.u[0]; + nir_deref_struct *deref_struct = nir_deref_struct_create(b, idx); + deref_struct->deref.type = glsl_get_struct_field(tail->type, idx); + tail->child = &deref_struct->deref; + break; + } + default: + unreachable("Invalid type for deref"); + } + tail = tail->child; + } + break; + } + + case SpvOpCopyMemory: { + nir_deref_var *dest = vtn_value(b, w[1], vtn_value_type_deref)->deref; + nir_deref_var *src = vtn_value(b, w[2], vtn_value_type_deref)->deref; + + nir_intrinsic_instr *copy = + nir_intrinsic_instr_create(b->shader, nir_intrinsic_copy_var); + copy->variables[0] = nir_deref_as_var(nir_copy_deref(copy, &dest->deref)); + copy->variables[1] = nir_deref_as_var(nir_copy_deref(copy, &src->deref)); + + nir_builder_instr_insert(&b->nb, ©->instr); + break; + } + + case SpvOpLoad: { + nir_deref_var *src = vtn_value(b, w[3], vtn_value_type_deref)->deref; + const struct glsl_type *src_type = nir_deref_tail(&src->deref)->type; + + if (glsl_get_base_type(src_type) == GLSL_TYPE_SAMPLER) { + vtn_push_value(b, w[2], vtn_value_type_deref)->deref = src; + return; + } + + assert(glsl_type_is_vector_or_scalar(src_type)); + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa); + + nir_intrinsic_instr *load = + nir_intrinsic_instr_create(b->shader, nir_intrinsic_load_var); + load->variables[0] = nir_deref_as_var(nir_copy_deref(load, &src->deref)); + load->num_components = glsl_get_vector_elements(src_type); + nir_ssa_dest_init(&load->instr, &load->dest, load->num_components, + val->name); + + nir_builder_instr_insert(&b->nb, &load->instr); + val->type = src_type; + + if (src->var->data.mode == nir_var_uniform && + glsl_get_base_type(src_type) == GLSL_TYPE_BOOL) { + /* Uniform boolean loads need to be fixed up since they're defined + * to be zero/nonzero rather than NIR_FALSE/NIR_TRUE. + */ + val->ssa = nir_ine(&b->nb, &load->dest.ssa, nir_imm_int(&b->nb, 0)); + } else { + val->ssa = &load->dest.ssa; + } + break; + } + + case SpvOpStore: { + nir_deref_var *dest = vtn_value(b, w[1], vtn_value_type_deref)->deref; + const struct glsl_type *dest_type = nir_deref_tail(&dest->deref)->type; + struct vtn_value *src_val = vtn_untyped_value(b, w[2]); + if (src_val->value_type == vtn_value_type_ssa) { + assert(glsl_type_is_vector_or_scalar(dest_type)); + nir_intrinsic_instr *store = + nir_intrinsic_instr_create(b->shader, nir_intrinsic_store_var); + store->src[0] = nir_src_for_ssa(src_val->ssa); + store->variables[0] = nir_deref_as_var(nir_copy_deref(store, &dest->deref)); + store->num_components = glsl_get_vector_elements(dest_type); + + nir_builder_instr_insert(&b->nb, &store->instr); + } else { + assert(src_val->value_type == vtn_value_type_constant); + + nir_variable *const_tmp = rzalloc(b->shader, nir_variable); + const_tmp->type = dest_type; + const_tmp->name = "const_temp"; + const_tmp->data.mode = nir_var_local; + const_tmp->data.read_only = true; + exec_list_push_tail(&b->impl->locals, &const_tmp->node); + + nir_intrinsic_instr *copy = + nir_intrinsic_instr_create(b->shader, nir_intrinsic_copy_var); + copy->variables[0] = nir_deref_as_var(nir_copy_deref(copy, &dest->deref)); + copy->variables[1] = nir_deref_var_create(copy, const_tmp); + + nir_builder_instr_insert(&b->nb, ©->instr); + } + break; + } + + case SpvOpVariableArray: + case SpvOpCopyMemorySized: + case SpvOpArrayLength: + case SpvOpImagePointer: + default: + unreachable("Unhandled opcode"); + } +} + +static void +vtn_handle_function_call(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + unreachable("Unhandled opcode"); +} + +static nir_tex_src +vtn_tex_src(struct vtn_builder *b, unsigned index, nir_tex_src_type type) +{ + nir_tex_src src; + src.src = nir_src_for_ssa(vtn_value(b, index, vtn_value_type_ssa)->ssa); + src.src_type = type; + return src; +} + +static void +vtn_handle_texture(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa); + nir_deref_var *sampler = vtn_value(b, w[3], vtn_value_type_deref)->deref; + + nir_tex_src srcs[8]; /* 8 should be enough */ + nir_tex_src *p = srcs; + + unsigned coord_components = 0; + switch (opcode) { + case SpvOpTextureSample: + case SpvOpTextureSampleDref: + case SpvOpTextureSampleLod: + case SpvOpTextureSampleProj: + case SpvOpTextureSampleGrad: + case SpvOpTextureSampleOffset: + case SpvOpTextureSampleProjLod: + case SpvOpTextureSampleProjGrad: + case SpvOpTextureSampleLodOffset: + case SpvOpTextureSampleProjOffset: + case SpvOpTextureSampleGradOffset: + case SpvOpTextureSampleProjLodOffset: + case SpvOpTextureSampleProjGradOffset: + case SpvOpTextureFetchTexelLod: + case SpvOpTextureFetchTexelOffset: + case SpvOpTextureFetchSample: + case SpvOpTextureFetchTexel: + case SpvOpTextureGather: + case SpvOpTextureGatherOffset: + case SpvOpTextureGatherOffsets: + case SpvOpTextureQueryLod: { + /* All these types have the coordinate as their first real argument */ + struct vtn_value *coord = vtn_value(b, w[4], vtn_value_type_ssa); + coord_components = glsl_get_vector_elements(coord->type); + p->src = nir_src_for_ssa(coord->ssa); + p->src_type = nir_tex_src_coord; + p++; + break; + } + default: + break; + } + + nir_texop texop; + switch (opcode) { + case SpvOpTextureSample: + texop = nir_texop_tex; + + if (count == 6) { + texop = nir_texop_txb; + *p++ = vtn_tex_src(b, w[5], nir_tex_src_bias); + } + break; + + case SpvOpTextureSampleDref: + case SpvOpTextureSampleLod: + case SpvOpTextureSampleProj: + case SpvOpTextureSampleGrad: + case SpvOpTextureSampleOffset: + case SpvOpTextureSampleProjLod: + case SpvOpTextureSampleProjGrad: + case SpvOpTextureSampleLodOffset: + case SpvOpTextureSampleProjOffset: + case SpvOpTextureSampleGradOffset: + case SpvOpTextureSampleProjLodOffset: + case SpvOpTextureSampleProjGradOffset: + case SpvOpTextureFetchTexelLod: + case SpvOpTextureFetchTexelOffset: + case SpvOpTextureFetchSample: + case SpvOpTextureFetchTexel: + case SpvOpTextureGather: + case SpvOpTextureGatherOffset: + case SpvOpTextureGatherOffsets: + case SpvOpTextureQuerySizeLod: + case SpvOpTextureQuerySize: + case SpvOpTextureQueryLod: + case SpvOpTextureQueryLevels: + case SpvOpTextureQuerySamples: + default: + unreachable("Unhandled opcode"); + } + + nir_tex_instr *instr = nir_tex_instr_create(b->shader, p - srcs); + + const struct glsl_type *sampler_type = nir_deref_tail(&sampler->deref)->type; + instr->sampler_dim = glsl_get_sampler_dim(sampler_type); + + switch (glsl_get_sampler_result_type(sampler_type)) { + case GLSL_TYPE_FLOAT: instr->dest_type = nir_type_float; break; + case GLSL_TYPE_INT: instr->dest_type = nir_type_int; break; + case GLSL_TYPE_UINT: instr->dest_type = nir_type_unsigned; break; + case GLSL_TYPE_BOOL: instr->dest_type = nir_type_bool; break; + default: + unreachable("Invalid base type for sampler result"); + } + + instr->op = texop; + memcpy(instr->src, srcs, instr->num_srcs * sizeof(*instr->src)); + instr->coord_components = coord_components; + instr->is_array = glsl_sampler_type_is_array(sampler_type); + instr->is_shadow = glsl_sampler_type_is_shadow(sampler_type); + + instr->sampler = sampler; + + nir_ssa_dest_init(&instr->instr, &instr->dest, 4, NULL); + val->ssa = &instr->dest.ssa; + + nir_builder_instr_insert(&b->nb, &instr->instr); +} + +static void +vtn_handle_matrix_alu(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + unreachable("Matrix math not handled"); +} + +static void +vtn_handle_alu(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa); + val->type = vtn_value(b, w[1], vtn_value_type_type)->type; + + /* Collect the various SSA sources */ + unsigned num_inputs = count - 3; + nir_ssa_def *src[4]; + for (unsigned i = 0; i < num_inputs; i++) + src[i] = vtn_ssa_value(b, w[i + 3]); + + /* Indicates that the first two arguments should be swapped. This is + * used for implementing greater-than and less-than-or-equal. + */ + bool swap = false; + + nir_op op; + switch (opcode) { + /* Basic ALU operations */ + case SpvOpSNegate: op = nir_op_ineg; break; + case SpvOpFNegate: op = nir_op_fneg; break; + case SpvOpNot: op = nir_op_inot; break; + + case SpvOpAny: + switch (src[0]->num_components) { + case 1: op = nir_op_imov; break; + case 2: op = nir_op_bany2; break; + case 3: op = nir_op_bany3; break; + case 4: op = nir_op_bany4; break; + } + break; + + case SpvOpAll: + switch (src[0]->num_components) { + case 1: op = nir_op_imov; break; + case 2: op = nir_op_ball2; break; + case 3: op = nir_op_ball3; break; + case 4: op = nir_op_ball4; break; + } + break; + + case SpvOpIAdd: op = nir_op_iadd; break; + case SpvOpFAdd: op = nir_op_fadd; break; + case SpvOpISub: op = nir_op_isub; break; + case SpvOpFSub: op = nir_op_fsub; break; + case SpvOpIMul: op = nir_op_imul; break; + case SpvOpFMul: op = nir_op_fmul; break; + case SpvOpUDiv: op = nir_op_udiv; break; + case SpvOpSDiv: op = nir_op_idiv; break; + case SpvOpFDiv: op = nir_op_fdiv; break; + case SpvOpUMod: op = nir_op_umod; break; + case SpvOpSMod: op = nir_op_umod; break; /* FIXME? */ + case SpvOpFMod: op = nir_op_fmod; break; + + case SpvOpDot: + assert(src[0]->num_components == src[1]->num_components); + switch (src[0]->num_components) { + case 1: op = nir_op_fmul; break; + case 2: op = nir_op_fdot2; break; + case 3: op = nir_op_fdot3; break; + case 4: op = nir_op_fdot4; break; + } + break; + + case SpvOpShiftRightLogical: op = nir_op_ushr; break; + case SpvOpShiftRightArithmetic: op = nir_op_ishr; break; + case SpvOpShiftLeftLogical: op = nir_op_ishl; break; + case SpvOpLogicalOr: op = nir_op_ior; break; + case SpvOpLogicalXor: op = nir_op_ixor; break; + case SpvOpLogicalAnd: op = nir_op_iand; break; + case SpvOpBitwiseOr: op = nir_op_ior; break; + case SpvOpBitwiseXor: op = nir_op_ixor; break; + case SpvOpBitwiseAnd: op = nir_op_iand; break; + case SpvOpSelect: op = nir_op_bcsel; break; + case SpvOpIEqual: op = nir_op_ieq; break; + + /* Comparisons: (TODO: How do we want to handled ordered/unordered?) */ + case SpvOpFOrdEqual: op = nir_op_feq; break; + case SpvOpFUnordEqual: op = nir_op_feq; break; + case SpvOpINotEqual: op = nir_op_ine; break; + case SpvOpFOrdNotEqual: op = nir_op_fne; break; + case SpvOpFUnordNotEqual: op = nir_op_fne; break; + case SpvOpULessThan: op = nir_op_ult; break; + case SpvOpSLessThan: op = nir_op_ilt; break; + case SpvOpFOrdLessThan: op = nir_op_flt; break; + case SpvOpFUnordLessThan: op = nir_op_flt; break; + case SpvOpUGreaterThan: op = nir_op_ult; swap = true; break; + case SpvOpSGreaterThan: op = nir_op_ilt; swap = true; break; + case SpvOpFOrdGreaterThan: op = nir_op_flt; swap = true; break; + case SpvOpFUnordGreaterThan: op = nir_op_flt; swap = true; break; + case SpvOpULessThanEqual: op = nir_op_uge; swap = true; break; + case SpvOpSLessThanEqual: op = nir_op_ige; swap = true; break; + case SpvOpFOrdLessThanEqual: op = nir_op_fge; swap = true; break; + case SpvOpFUnordLessThanEqual: op = nir_op_fge; swap = true; break; + case SpvOpUGreaterThanEqual: op = nir_op_uge; break; + case SpvOpSGreaterThanEqual: op = nir_op_ige; break; + case SpvOpFOrdGreaterThanEqual: op = nir_op_fge; break; + case SpvOpFUnordGreaterThanEqual:op = nir_op_fge; break; + + /* Conversions: */ + case SpvOpConvertFToU: op = nir_op_f2u; break; + case SpvOpConvertFToS: op = nir_op_f2i; break; + case SpvOpConvertSToF: op = nir_op_i2f; break; + case SpvOpConvertUToF: op = nir_op_u2f; break; + case SpvOpBitcast: op = nir_op_imov; break; + case SpvOpUConvert: + case SpvOpSConvert: + op = nir_op_imov; /* TODO: NIR is 32-bit only; these are no-ops. */ + break; + case SpvOpFConvert: + op = nir_op_fmov; + break; + + /* Derivatives: */ + case SpvOpDPdx: op = nir_op_fddx; break; + case SpvOpDPdy: op = nir_op_fddy; break; + case SpvOpDPdxFine: op = nir_op_fddx_fine; break; + case SpvOpDPdyFine: op = nir_op_fddy_fine; break; + case SpvOpDPdxCoarse: op = nir_op_fddx_coarse; break; + case SpvOpDPdyCoarse: op = nir_op_fddy_coarse; break; + case SpvOpFwidth: + val->ssa = nir_fadd(&b->nb, + nir_fabs(&b->nb, nir_fddx(&b->nb, src[0])), + nir_fabs(&b->nb, nir_fddx(&b->nb, src[1]))); + return; + case SpvOpFwidthFine: + val->ssa = nir_fadd(&b->nb, + nir_fabs(&b->nb, nir_fddx_fine(&b->nb, src[0])), + nir_fabs(&b->nb, nir_fddx_fine(&b->nb, src[1]))); + return; + case SpvOpFwidthCoarse: + val->ssa = nir_fadd(&b->nb, + nir_fabs(&b->nb, nir_fddx_coarse(&b->nb, src[0])), + nir_fabs(&b->nb, nir_fddx_coarse(&b->nb, src[1]))); + return; + + case SpvOpVectorTimesScalar: + /* The builder will take care of splatting for us. */ + val->ssa = nir_fmul(&b->nb, src[0], src[1]); + return; + + case SpvOpSRem: + case SpvOpFRem: + unreachable("No NIR equivalent"); + + case SpvOpIsNan: + case SpvOpIsInf: + case SpvOpIsFinite: + case SpvOpIsNormal: + case SpvOpSignBitSet: + case SpvOpLessOrGreater: + case SpvOpOrdered: + case SpvOpUnordered: + default: + unreachable("Unhandled opcode"); + } + + if (swap) { + nir_ssa_def *tmp = src[0]; + src[0] = src[1]; + src[1] = tmp; + } + + nir_alu_instr *instr = nir_alu_instr_create(b->shader, op); + nir_ssa_dest_init(&instr->instr, &instr->dest.dest, + glsl_get_vector_elements(val->type), val->name); + val->ssa = &instr->dest.dest.ssa; + + for (unsigned i = 0; i < nir_op_infos[op].num_inputs; i++) + instr->src[i].src = nir_src_for_ssa(src[i]); + + nir_builder_instr_insert(&b->nb, &instr->instr); +} + +static bool +vtn_handle_preamble_instruction(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + switch (opcode) { + case SpvOpSource: + case SpvOpSourceExtension: + case SpvOpCompileFlag: + case SpvOpExtension: + /* Unhandled, but these are for debug so that's ok. */ + break; + + case SpvOpExtInstImport: + vtn_handle_extension(b, opcode, w, count); + break; + + case SpvOpMemoryModel: + assert(w[1] == SpvAddressingModelLogical); + assert(w[2] == SpvMemoryModelGLSL450); + break; + + case SpvOpEntryPoint: + assert(b->entry_point == NULL); + b->entry_point = &b->values[w[2]]; + b->execution_model = w[1]; + break; + + case SpvOpExecutionMode: + unreachable("Execution modes not yet implemented"); + break; + + case SpvOpString: + vtn_push_value(b, w[1], vtn_value_type_string)->str = + vtn_string_literal(b, &w[2], count - 2); + break; + + case SpvOpName: + b->values[w[1]].name = vtn_string_literal(b, &w[2], count - 2); + break; + + case SpvOpMemberName: + /* TODO */ + break; + + case SpvOpLine: + break; /* Ignored for now */ + + case SpvOpDecorationGroup: + case SpvOpDecorate: + case SpvOpMemberDecorate: + case SpvOpGroupDecorate: + case SpvOpGroupMemberDecorate: + vtn_handle_decoration(b, opcode, w, count); + break; + + case SpvOpTypeVoid: + case SpvOpTypeBool: + case SpvOpTypeInt: + case SpvOpTypeFloat: + case SpvOpTypeVector: + case SpvOpTypeMatrix: + case SpvOpTypeSampler: + case SpvOpTypeArray: + case SpvOpTypeRuntimeArray: + case SpvOpTypeStruct: + case SpvOpTypeOpaque: + case SpvOpTypePointer: + case SpvOpTypeFunction: + case SpvOpTypeEvent: + case SpvOpTypeDeviceEvent: + case SpvOpTypeReserveId: + case SpvOpTypeQueue: + case SpvOpTypePipe: + vtn_push_value(b, w[1], vtn_value_type_type)->type = + vtn_handle_type(b, opcode, &w[2], count - 2); + break; + + case SpvOpConstantTrue: + case SpvOpConstantFalse: + case SpvOpConstant: + case SpvOpConstantComposite: + case SpvOpConstantSampler: + case SpvOpConstantNullPointer: + case SpvOpConstantNullObject: + case SpvOpSpecConstantTrue: + case SpvOpSpecConstantFalse: + case SpvOpSpecConstant: + case SpvOpSpecConstantComposite: + vtn_handle_constant(b, opcode, w, count); + break; + + case SpvOpVariable: + vtn_handle_variables(b, opcode, w, count); + break; + + default: + return false; /* End of preamble */ + } + + return true; +} + +static bool +vtn_handle_first_cfg_pass_instruction(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + switch (opcode) { + case SpvOpFunction: { + assert(b->func == NULL); + b->func = rzalloc(b, struct vtn_function); + + const struct glsl_type *result_type = + vtn_value(b, w[1], vtn_value_type_type)->type; + struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_function); + const struct glsl_type *func_type = + vtn_value(b, w[4], vtn_value_type_type)->type; + + assert(glsl_get_function_return_type(func_type) == result_type); + + nir_function *func = + nir_function_create(b->shader, ralloc_strdup(b->shader, val->name)); + + nir_function_overload *overload = nir_function_overload_create(func); + overload->num_params = glsl_get_length(func_type); + overload->params = ralloc_array(overload, nir_parameter, + overload->num_params); + for (unsigned i = 0; i < overload->num_params; i++) { + const struct glsl_function_param *param = + glsl_get_function_param(func_type, i); + overload->params[i].type = param->type; + if (param->in) { + if (param->out) { + overload->params[i].param_type = nir_parameter_inout; + } else { + overload->params[i].param_type = nir_parameter_in; + } + } else { + if (param->out) { + overload->params[i].param_type = nir_parameter_out; + } else { + assert(!"Parameter is neither in nor out"); + } + } + } + b->func->overload = overload; + break; + } + + case SpvOpFunctionEnd: + b->func = NULL; + break; + + case SpvOpFunctionParameter: + break; /* Does nothing */ + + case SpvOpLabel: { + assert(b->block == NULL); + b->block = rzalloc(b, struct vtn_block); + b->block->label = w; + vtn_push_value(b, w[1], vtn_value_type_block)->block = b->block; + + if (b->func->start_block == NULL) { + /* This is the first block encountered for this function. In this + * case, we set the start block and add it to the list of + * implemented functions that we'll walk later. + */ + b->func->start_block = b->block; + exec_list_push_tail(&b->functions, &b->func->node); + } + break; + } + + case SpvOpBranch: + case SpvOpBranchConditional: + case SpvOpSwitch: + case SpvOpKill: + case SpvOpReturn: + case SpvOpReturnValue: + case SpvOpUnreachable: + assert(b->block); + b->block->branch = w; + b->block = NULL; + break; + + case SpvOpSelectionMerge: + case SpvOpLoopMerge: + assert(b->block && b->block->merge_op == SpvOpNop); + b->block->merge_op = opcode; + b->block->merge_block_id = w[1]; + break; + + default: + /* Continue on as per normal */ + return true; + } + + return true; +} + +static bool +vtn_handle_body_instruction(struct vtn_builder *b, SpvOp opcode, + const uint32_t *w, unsigned count) +{ + switch (opcode) { + case SpvOpLabel: { + struct vtn_block *block = vtn_value(b, w[1], vtn_value_type_block)->block; + assert(block->block == NULL); + + struct exec_node *list_tail = exec_list_get_tail(b->nb.cf_node_list); + nir_cf_node *tail_node = exec_node_data(nir_cf_node, list_tail, node); + assert(tail_node->type == nir_cf_node_block); + block->block = nir_cf_node_as_block(tail_node); + break; + } + + case SpvOpLoopMerge: + case SpvOpSelectionMerge: + /* This is handled by cfg pre-pass and walk_blocks */ + break; + + case SpvOpUndef: + vtn_push_value(b, w[2], vtn_value_type_undef); + break; + + case SpvOpExtInst: + vtn_handle_extension(b, opcode, w, count); + break; + + case SpvOpVariable: + case SpvOpVariableArray: + case SpvOpLoad: + case SpvOpStore: + case SpvOpCopyMemory: + case SpvOpCopyMemorySized: + case SpvOpAccessChain: + case SpvOpInBoundsAccessChain: + case SpvOpArrayLength: + case SpvOpImagePointer: + vtn_handle_variables(b, opcode, w, count); + break; + + case SpvOpFunctionCall: + vtn_handle_function_call(b, opcode, w, count); + break; + + case SpvOpTextureSample: + case SpvOpTextureSampleDref: + case SpvOpTextureSampleLod: + case SpvOpTextureSampleProj: + case SpvOpTextureSampleGrad: + case SpvOpTextureSampleOffset: + case SpvOpTextureSampleProjLod: + case SpvOpTextureSampleProjGrad: + case SpvOpTextureSampleLodOffset: + case SpvOpTextureSampleProjOffset: + case SpvOpTextureSampleGradOffset: + case SpvOpTextureSampleProjLodOffset: + case SpvOpTextureSampleProjGradOffset: + case SpvOpTextureFetchTexelLod: + case SpvOpTextureFetchTexelOffset: + case SpvOpTextureFetchSample: + case SpvOpTextureFetchTexel: + case SpvOpTextureGather: + case SpvOpTextureGatherOffset: + case SpvOpTextureGatherOffsets: + case SpvOpTextureQuerySizeLod: + case SpvOpTextureQuerySize: + case SpvOpTextureQueryLod: + case SpvOpTextureQueryLevels: + case SpvOpTextureQuerySamples: + vtn_handle_texture(b, opcode, w, count); + break; + + case SpvOpSNegate: + case SpvOpFNegate: + case SpvOpNot: + case SpvOpAny: + case SpvOpAll: + case SpvOpConvertFToU: + case SpvOpConvertFToS: + case SpvOpConvertSToF: + case SpvOpConvertUToF: + case SpvOpUConvert: + case SpvOpSConvert: + case SpvOpFConvert: + case SpvOpConvertPtrToU: + case SpvOpConvertUToPtr: + case SpvOpPtrCastToGeneric: + case SpvOpGenericCastToPtr: + case SpvOpBitcast: + case SpvOpIsNan: + case SpvOpIsInf: + case SpvOpIsFinite: + case SpvOpIsNormal: + case SpvOpSignBitSet: + case SpvOpLessOrGreater: + case SpvOpOrdered: + case SpvOpUnordered: + case SpvOpIAdd: + case SpvOpFAdd: + case SpvOpISub: + case SpvOpFSub: + case SpvOpIMul: + case SpvOpFMul: + case SpvOpUDiv: + case SpvOpSDiv: + case SpvOpFDiv: + case SpvOpUMod: + case SpvOpSRem: + case SpvOpSMod: + case SpvOpFRem: + case SpvOpFMod: + case SpvOpVectorTimesScalar: + case SpvOpDot: + case SpvOpShiftRightLogical: + case SpvOpShiftRightArithmetic: + case SpvOpShiftLeftLogical: + case SpvOpLogicalOr: + case SpvOpLogicalXor: + case SpvOpLogicalAnd: + case SpvOpBitwiseOr: + case SpvOpBitwiseXor: + case SpvOpBitwiseAnd: + case SpvOpSelect: + case SpvOpIEqual: + case SpvOpFOrdEqual: + case SpvOpFUnordEqual: + case SpvOpINotEqual: + case SpvOpFOrdNotEqual: + case SpvOpFUnordNotEqual: + case SpvOpULessThan: + case SpvOpSLessThan: + case SpvOpFOrdLessThan: + case SpvOpFUnordLessThan: + case SpvOpUGreaterThan: + case SpvOpSGreaterThan: + case SpvOpFOrdGreaterThan: + case SpvOpFUnordGreaterThan: + case SpvOpULessThanEqual: + case SpvOpSLessThanEqual: + case SpvOpFOrdLessThanEqual: + case SpvOpFUnordLessThanEqual: + case SpvOpUGreaterThanEqual: + case SpvOpSGreaterThanEqual: + case SpvOpFOrdGreaterThanEqual: + case SpvOpFUnordGreaterThanEqual: + case SpvOpDPdx: + case SpvOpDPdy: + case SpvOpFwidth: + case SpvOpDPdxFine: + case SpvOpDPdyFine: + case SpvOpFwidthFine: + case SpvOpDPdxCoarse: + case SpvOpDPdyCoarse: + case SpvOpFwidthCoarse: + vtn_handle_alu(b, opcode, w, count); + break; + + case SpvOpTranspose: + case SpvOpOuterProduct: + case SpvOpMatrixTimesScalar: + case SpvOpVectorTimesMatrix: + case SpvOpMatrixTimesVector: + case SpvOpMatrixTimesMatrix: + vtn_handle_matrix_alu(b, opcode, w, count); + break; + + default: + unreachable("Unhandled opcode"); + } + + return true; +} + +static void +vtn_walk_blocks(struct vtn_builder *b, struct vtn_block *start, + struct vtn_block *break_block, struct vtn_block *cont_block, + struct vtn_block *end_block) +{ + struct vtn_block *block = start; + while (block != end_block) { + if (block->block != NULL) { + /* We've already visited this block once before so this is a + * back-edge. Back-edges are only allowed to point to a loop + * merge. + */ + assert(block == cont_block); + return; + } + + if (block->merge_op == SpvOpLoopMerge) { + /* This is the jump into a loop. */ + cont_block = block; + break_block = vtn_value(b, block->merge_block_id, + vtn_value_type_block)->block; + + nir_loop *loop = nir_loop_create(b->shader); + nir_cf_node_insert_end(b->nb.cf_node_list, &loop->cf_node); + + struct exec_list *old_list = b->nb.cf_node_list; + + /* Reset the merge_op to prerevent infinite recursion */ + block->merge_op = SpvOpNop; + + nir_builder_insert_after_cf_list(&b->nb, &loop->body); + vtn_walk_blocks(b, block, break_block, cont_block, NULL); + + nir_builder_insert_after_cf_list(&b->nb, old_list); + block = break_block; + continue; + } + + const uint32_t *w = block->branch; + SpvOp branch_op = w[0] & SpvOpCodeMask; + + b->block = block; + vtn_foreach_instruction(b, block->label, block->branch, + vtn_handle_body_instruction); + + switch (branch_op) { + case SpvOpBranch: { + struct vtn_block *branch_block = + vtn_value(b, w[1], vtn_value_type_block)->block; + + if (branch_block == break_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_break); + nir_builder_instr_insert(&b->nb, &jump->instr); + + return; + } else if (branch_block == cont_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_continue); + nir_builder_instr_insert(&b->nb, &jump->instr); + + return; + } else if (branch_block == end_block) { + return; + } else { + block = branch_block; + continue; + } + } + + case SpvOpBranchConditional: { + /* Gather up the branch blocks */ + struct vtn_block *then_block = + vtn_value(b, w[2], vtn_value_type_block)->block; + struct vtn_block *else_block = + vtn_value(b, w[3], vtn_value_type_block)->block; + + nir_if *if_stmt = nir_if_create(b->shader); + if_stmt->condition = nir_src_for_ssa(vtn_ssa_value(b, w[1])); + nir_cf_node_insert_end(b->nb.cf_node_list, &if_stmt->cf_node); + + if (then_block == break_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_break); + nir_instr_insert_after_cf_list(&if_stmt->then_list, + &jump->instr); + block = else_block; + } else if (else_block == break_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_break); + nir_instr_insert_after_cf_list(&if_stmt->else_list, + &jump->instr); + block = then_block; + } else if (then_block == cont_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_continue); + nir_instr_insert_after_cf_list(&if_stmt->then_list, + &jump->instr); + block = else_block; + } else if (else_block == cont_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_continue); + nir_instr_insert_after_cf_list(&if_stmt->else_list, + &jump->instr); + block = then_block; + } else { + /* Conventional if statement */ + assert(block->merge_op == SpvOpSelectionMerge); + struct vtn_block *merge_block = + vtn_value(b, block->merge_block_id, vtn_value_type_block)->block; + + struct exec_list *old_list = b->nb.cf_node_list; + + nir_builder_insert_after_cf_list(&b->nb, &if_stmt->then_list); + vtn_walk_blocks(b, then_block, break_block, cont_block, merge_block); + + nir_builder_insert_after_cf_list(&b->nb, &if_stmt->else_list); + vtn_walk_blocks(b, else_block, break_block, cont_block, merge_block); + + nir_builder_insert_after_cf_list(&b->nb, old_list); + block = merge_block; + continue; + } + + /* If we got here then we inserted a predicated break or continue + * above and we need to handle the other case. We already set + * `block` above to indicate what block to visit after the + * predicated break. + */ + + /* It's possible that the other branch is also a break/continue. + * If it is, we handle that here. + */ + if (block == break_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_break); + nir_builder_instr_insert(&b->nb, &jump->instr); + + return; + } else if (block == cont_block) { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_continue); + nir_builder_instr_insert(&b->nb, &jump->instr); + + return; + } + + /* If we got here then there was a predicated break/continue but + * the other half of the if has stuff in it. `block` was already + * set above so there is nothing left for us to do. + */ + continue; + } + + case SpvOpReturn: { + nir_jump_instr *jump = nir_jump_instr_create(b->shader, + nir_jump_return); + nir_builder_instr_insert(&b->nb, &jump->instr); + return; + } + + case SpvOpKill: { + nir_intrinsic_instr *discard = + nir_intrinsic_instr_create(b->shader, nir_intrinsic_discard); + nir_builder_instr_insert(&b->nb, &discard->instr); + return; + } + + case SpvOpSwitch: + case SpvOpReturnValue: + case SpvOpUnreachable: + default: + unreachable("Unhandled opcode"); + } + } +} + +nir_shader * +spirv_to_nir(const uint32_t *words, size_t word_count, + const nir_shader_compiler_options *options) +{ + const uint32_t *word_end = words + word_count; + + /* Handle the SPIR-V header (first 4 dwords) */ + assert(word_count > 5); + + assert(words[0] == SpvMagicNumber); + assert(words[1] == 99); + /* words[2] == generator magic */ + unsigned value_id_bound = words[3]; + assert(words[4] == 0); + + words+= 5; + + nir_shader *shader = nir_shader_create(NULL, options); + + /* Initialize the stn_builder object */ + struct vtn_builder *b = rzalloc(NULL, struct vtn_builder); + b->shader = shader; + b->value_id_bound = value_id_bound; + b->values = ralloc_array(b, struct vtn_value, value_id_bound); + exec_list_make_empty(&b->functions); + + /* Handle all the preamble instructions */ + words = vtn_foreach_instruction(b, words, word_end, + vtn_handle_preamble_instruction); + + /* Do a very quick CFG analysis pass */ + vtn_foreach_instruction(b, words, word_end, + vtn_handle_first_cfg_pass_instruction); + + foreach_list_typed(struct vtn_function, func, node, &b->functions) { + b->impl = nir_function_impl_create(func->overload); + nir_builder_init(&b->nb, b->impl); + nir_builder_insert_after_cf_list(&b->nb, &b->impl->body); + vtn_walk_blocks(b, func->start_block, NULL, NULL, NULL); + } + + ralloc_free(b); + + return shader; +} diff --git a/src/glsl/nir/spirv_to_nir_private.h b/src/glsl/nir/spirv_to_nir_private.h new file mode 100644 index 00000000000..d2b364bdfeb --- /dev/null +++ b/src/glsl/nir/spirv_to_nir_private.h @@ -0,0 +1,148 @@ +/* + * Copyright © 2015 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. + * + * Authors: + * Jason Ekstrand ([email protected]) + * + */ + +#include "nir_spirv.h" +#include "nir_builder.h" +#include "spirv.h" + +struct vtn_builder; +struct vtn_decoration; + +enum vtn_value_type { + vtn_value_type_invalid = 0, + vtn_value_type_undef, + vtn_value_type_string, + vtn_value_type_decoration_group, + vtn_value_type_type, + vtn_value_type_constant, + vtn_value_type_deref, + vtn_value_type_function, + vtn_value_type_block, + vtn_value_type_ssa, + vtn_value_type_extension, +}; + +struct vtn_block { + /* Merge opcode if this block contains a merge; SpvOpNop otherwise. */ + SpvOp merge_op; + uint32_t merge_block_id; + const uint32_t *label; + const uint32_t *branch; + nir_block *block; +}; + +struct vtn_function { + struct exec_node node; + + nir_function_overload *overload; + struct vtn_block *start_block; +}; + +typedef bool (*vtn_instruction_handler)(struct vtn_builder *, uint32_t, + const uint32_t *, unsigned); + +struct vtn_value { + enum vtn_value_type value_type; + const char *name; + struct vtn_decoration *decoration; + const struct glsl_type *type; + union { + void *ptr; + char *str; + nir_constant *constant; + nir_deref_var *deref; + struct vtn_function *func; + struct vtn_block *block; + nir_ssa_def *ssa; + vtn_instruction_handler ext_handler; + }; +}; + +struct vtn_decoration { + struct vtn_decoration *next; + const uint32_t *literals; + struct vtn_value *group; + SpvDecoration decoration; +}; + +struct vtn_builder { + nir_builder nb; + + nir_shader *shader; + nir_function_impl *impl; + struct vtn_block *block; + + unsigned value_id_bound; + struct vtn_value *values; + + SpvExecutionModel execution_model; + struct vtn_value *entry_point; + + struct vtn_function *func; + struct exec_list functions; +}; + +static inline struct vtn_value * +vtn_push_value(struct vtn_builder *b, uint32_t value_id, + enum vtn_value_type value_type) +{ + assert(value_id < b->value_id_bound); + assert(b->values[value_id].value_type == vtn_value_type_invalid); + + b->values[value_id].value_type = value_type; + + return &b->values[value_id]; +} + +static inline struct vtn_value * +vtn_untyped_value(struct vtn_builder *b, uint32_t value_id) +{ + assert(value_id < b->value_id_bound); + return &b->values[value_id]; +} + +static inline struct vtn_value * +vtn_value(struct vtn_builder *b, uint32_t value_id, + enum vtn_value_type value_type) +{ + struct vtn_value *val = vtn_untyped_value(b, value_id); + assert(val->value_type == value_type); + return val; +} + +nir_ssa_def *vtn_ssa_value(struct vtn_builder *b, uint32_t value_id); + +typedef void (*vtn_decoration_foreach_cb)(struct vtn_builder *, + struct vtn_value *, + const struct vtn_decoration *, + void *); + +void vtn_foreach_decoration(struct vtn_builder *b, struct vtn_value *value, + vtn_decoration_foreach_cb cb, void *data); + +bool vtn_handle_glsl450_instruction(struct vtn_builder *b, uint32_t ext_opcode, + const uint32_t *words, unsigned count); diff --git a/src/glsl/standalone_scaffolding.cpp b/src/glsl/standalone_scaffolding.cpp index a109c4e92d2..6e1ecec3235 100644 --- a/src/glsl/standalone_scaffolding.cpp +++ b/src/glsl/standalone_scaffolding.cpp @@ -34,6 +34,12 @@ #include <string.h> #include "util/ralloc.h" +extern "C" void +_mesa_error_no_memory(const char *caller) +{ + fprintf(stderr, "Mesa error: out of memory in %s", caller); +} + void _mesa_warning(struct gl_context *ctx, const char *fmt, ...) { diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp index 8a9bbdbae52..5ce1dfc6633 100644 --- a/src/mesa/drivers/dri/i965/brw_fs.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs.cpp @@ -671,6 +671,7 @@ fs_visitor::type_size(const struct glsl_type *type) case GLSL_TYPE_ERROR: case GLSL_TYPE_INTERFACE: case GLSL_TYPE_DOUBLE: + case GLSL_TYPE_FUNCTION: unreachable("not reached"); } diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp index 975f5f6b2c9..d6bb1178f7c 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp @@ -1354,6 +1354,7 @@ fs_visitor::emit_assignment_writes(fs_reg &l, fs_reg &r, case GLSL_TYPE_VOID: case GLSL_TYPE_ERROR: case GLSL_TYPE_INTERFACE: + case GLSL_TYPE_FUNCTION: unreachable("not reached"); } } diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp index c1fd859fef5..ebfb49acf8d 100644 --- a/src/mesa/drivers/dri/i965/brw_shader.cpp +++ b/src/mesa/drivers/dri/i965/brw_shader.cpp @@ -351,6 +351,7 @@ brw_type_for_base_type(const struct glsl_type *type) case GLSL_TYPE_ERROR: case GLSL_TYPE_INTERFACE: case GLSL_TYPE_DOUBLE: + case GLSL_TYPE_FUNCTION: unreachable("not reached"); } diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp index 5a60fe43bf8..e51c140c0f2 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp @@ -615,6 +615,7 @@ type_size(const struct glsl_type *type) case GLSL_TYPE_DOUBLE: case GLSL_TYPE_ERROR: case GLSL_TYPE_INTERFACE: + case GLSL_TYPE_FUNCTION: unreachable("not reached"); } diff --git a/src/mesa/program/ir_to_mesa.cpp b/src/mesa/program/ir_to_mesa.cpp index 3dcb53702a5..fceed712bdb 100644 --- a/src/mesa/program/ir_to_mesa.cpp +++ b/src/mesa/program/ir_to_mesa.cpp @@ -541,6 +541,7 @@ type_size(const struct glsl_type *type) case GLSL_TYPE_VOID: case GLSL_TYPE_ERROR: case GLSL_TYPE_INTERFACE: + case GLSL_TYPE_FUNCTION: assert(!"Invalid type in type_size"); break; } @@ -2448,6 +2449,7 @@ _mesa_associate_uniform_storage(struct gl_context *ctx, case GLSL_TYPE_STRUCT: case GLSL_TYPE_ERROR: case GLSL_TYPE_INTERFACE: + case GLSL_TYPE_FUNCTION: assert(!"Should not get here."); break; } |