blob: 11b99714a527797893c0e7794d505b5316efaae9 [file] [log] [blame] [edit]
// Copyright 2020 The Tint Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "src/tint/lang/wgsl/resolver/validator.h"
#include <algorithm>
#include <limits>
#include <utility>
#include "src/tint/lang/core/fluent_types.h"
#include "src/tint/lang/core/type/abstract_numeric.h"
#include "src/tint/lang/core/type/array.h"
#include "src/tint/lang/core/type/atomic.h"
#include "src/tint/lang/core/type/depth_multisampled_texture.h"
#include "src/tint/lang/core/type/depth_texture.h"
#include "src/tint/lang/core/type/multisampled_texture.h"
#include "src/tint/lang/core/type/pointer.h"
#include "src/tint/lang/core/type/reference.h"
#include "src/tint/lang/core/type/sampled_texture.h"
#include "src/tint/lang/core/type/sampler.h"
#include "src/tint/lang/core/type/storage_texture.h"
#include "src/tint/lang/core/type/texture_dimension.h"
#include "src/tint/lang/wgsl/ast/alias.h"
#include "src/tint/lang/wgsl/ast/assignment_statement.h"
#include "src/tint/lang/wgsl/ast/bitcast_expression.h"
#include "src/tint/lang/wgsl/ast/break_statement.h"
#include "src/tint/lang/wgsl/ast/call_statement.h"
#include "src/tint/lang/wgsl/ast/continue_statement.h"
#include "src/tint/lang/wgsl/ast/disable_validation_attribute.h"
#include "src/tint/lang/wgsl/ast/discard_statement.h"
#include "src/tint/lang/wgsl/ast/for_loop_statement.h"
#include "src/tint/lang/wgsl/ast/id_attribute.h"
#include "src/tint/lang/wgsl/ast/if_statement.h"
#include "src/tint/lang/wgsl/ast/index_attribute.h"
#include "src/tint/lang/wgsl/ast/internal_attribute.h"
#include "src/tint/lang/wgsl/ast/interpolate_attribute.h"
#include "src/tint/lang/wgsl/ast/loop_statement.h"
#include "src/tint/lang/wgsl/ast/return_statement.h"
#include "src/tint/lang/wgsl/ast/switch_statement.h"
#include "src/tint/lang/wgsl/ast/traverse_expressions.h"
#include "src/tint/lang/wgsl/ast/unary_op_expression.h"
#include "src/tint/lang/wgsl/ast/variable_decl_statement.h"
#include "src/tint/lang/wgsl/ast/workgroup_attribute.h"
#include "src/tint/lang/wgsl/sem/break_if_statement.h"
#include "src/tint/lang/wgsl/sem/call.h"
#include "src/tint/lang/wgsl/sem/for_loop_statement.h"
#include "src/tint/lang/wgsl/sem/function.h"
#include "src/tint/lang/wgsl/sem/if_statement.h"
#include "src/tint/lang/wgsl/sem/loop_statement.h"
#include "src/tint/lang/wgsl/sem/materialize.h"
#include "src/tint/lang/wgsl/sem/member_accessor_expression.h"
#include "src/tint/lang/wgsl/sem/statement.h"
#include "src/tint/lang/wgsl/sem/struct.h"
#include "src/tint/lang/wgsl/sem/switch_statement.h"
#include "src/tint/lang/wgsl/sem/value_constructor.h"
#include "src/tint/lang/wgsl/sem/value_conversion.h"
#include "src/tint/lang/wgsl/sem/variable.h"
#include "src/tint/lang/wgsl/sem/while_statement.h"
#include "src/tint/utils/containers/map.h"
#include "src/tint/utils/containers/reverse.h"
#include "src/tint/utils/containers/transform.h"
#include "src/tint/utils/macros/defer.h"
#include "src/tint/utils/macros/scoped_assignment.h"
#include "src/tint/utils/math/math.h"
#include "src/tint/utils/text/string.h"
#include "src/tint/utils/text/string_stream.h"
using namespace tint::core::fluent_types; // NOLINT
namespace tint::resolver {
namespace {
constexpr size_t kMaxFunctionParameters = 255;
constexpr size_t kMaxSwitchCaseSelectors = 16383;
bool IsValidStorageTextureDimension(core::type::TextureDimension dim) {
switch (dim) {
case core::type::TextureDimension::k1d:
case core::type::TextureDimension::k2d:
case core::type::TextureDimension::k2dArray:
case core::type::TextureDimension::k3d:
return true;
default:
return false;
}
}
bool IsValidStorageTextureTexelFormat(core::TexelFormat format) {
switch (format) {
case core::TexelFormat::kBgra8Unorm:
case core::TexelFormat::kR32Uint:
case core::TexelFormat::kR32Sint:
case core::TexelFormat::kR32Float:
case core::TexelFormat::kRg32Uint:
case core::TexelFormat::kRg32Sint:
case core::TexelFormat::kRg32Float:
case core::TexelFormat::kRgba8Unorm:
case core::TexelFormat::kRgba8Snorm:
case core::TexelFormat::kRgba8Uint:
case core::TexelFormat::kRgba8Sint:
case core::TexelFormat::kRgba16Uint:
case core::TexelFormat::kRgba16Sint:
case core::TexelFormat::kRgba16Float:
case core::TexelFormat::kRgba32Uint:
case core::TexelFormat::kRgba32Sint:
case core::TexelFormat::kRgba32Float:
return true;
default:
return false;
}
}
// Helper to stringify a pipeline IO attribute.
std::string AttrToStr(const ast::Attribute* attr) {
return Switch(
attr, //
[&](const ast::BuiltinAttribute*) { return "@builtin"; },
[&](const ast::LocationAttribute*) { return "@location"; });
}
template <typename CALLBACK>
void TraverseCallChain(const sem::Function* from, const sem::Function* to, CALLBACK&& callback) {
for (auto* f : from->TransitivelyCalledFunctions()) {
if (f == to) {
callback(f);
return;
}
if (f->TransitivelyCalledFunctions().Contains(to)) {
TraverseCallChain(f, to, callback);
callback(f);
return;
}
}
TINT_ICE() << "TraverseCallChain() 'from' does not transitively call 'to'";
}
} // namespace
Validator::Validator(
ProgramBuilder* builder,
SemHelper& sem,
const core::Extensions& enabled_extensions,
const Hashmap<const core::type::Type*, const Source*, 8>& atomic_composite_info,
Hashset<TypeAndAddressSpace, 8>& valid_type_storage_layouts)
: symbols_(builder->Symbols()),
diagnostics_(builder->Diagnostics()),
sem_(sem),
enabled_extensions_(enabled_extensions),
atomic_composite_info_(atomic_composite_info),
valid_type_storage_layouts_(valid_type_storage_layouts) {
// Set default severities for filterable diagnostic rules.
diagnostic_filters_.Set(core::CoreDiagnosticRule::kDerivativeUniformity,
core::DiagnosticSeverity::kError);
diagnostic_filters_.Set(core::ChromiumDiagnosticRule::kUnreachableCode,
core::DiagnosticSeverity::kWarning);
}
Validator::~Validator() = default;
void Validator::AddError(const std::string& msg, const Source& source) const {
diagnostics_.add_error(diag::System::Resolver, msg, source);
}
void Validator::AddWarning(const std::string& msg, const Source& source) const {
diagnostics_.add_warning(diag::System::Resolver, msg, source);
}
void Validator::AddNote(const std::string& msg, const Source& source) const {
diagnostics_.add_note(diag::System::Resolver, msg, source);
}
bool Validator::AddDiagnostic(core::DiagnosticRule rule,
const std::string& msg,
const Source& source) const {
auto severity = diagnostic_filters_.Get(rule);
if (severity != core::DiagnosticSeverity::kOff) {
diag::Diagnostic d{};
d.severity = ToSeverity(severity);
d.system = diag::System::Resolver;
d.source = source;
d.message = msg;
diagnostics_.add(std::move(d));
if (severity == core::DiagnosticSeverity::kError) {
return false;
}
}
return true;
}
// https://gpuweb.github.io/gpuweb/wgsl/#plain-types-section
bool Validator::IsPlain(const core::type::Type* type) const {
return type->IsAnyOf<core::type::Scalar, core::type::Atomic, core::type::Vector,
core::type::Matrix, core::type::Array, core::type::Struct>();
}
// https://gpuweb.github.io/gpuweb/wgsl/#fixed-footprint-types
bool Validator::IsFixedFootprint(const core::type::Type* type) const {
return Switch(
type, //
[&](const core::type::Vector*) { return true; }, //
[&](const core::type::Matrix*) { return true; }, //
[&](const core::type::Atomic*) { return true; },
[&](const core::type::Array* arr) {
return !arr->Count()->Is<core::type::RuntimeArrayCount>() &&
IsFixedFootprint(arr->ElemType());
},
[&](const core::type::Struct* str) {
for (auto* member : str->Members()) {
if (!IsFixedFootprint(member->Type())) {
return false;
}
}
return true;
},
[&](Default) { return type->Is<core::type::Scalar>(); });
}
// https://gpuweb.github.io/gpuweb/wgsl.html#host-shareable-types
bool Validator::IsHostShareable(const core::type::Type* type) const {
if (type->IsAnyOf<core::type::I32, core::type::U32, core::type::F32, core::type::F16>()) {
return true;
}
return Switch(
type, //
[&](const core::type::Vector* vec) { return IsHostShareable(vec->type()); },
[&](const core::type::Matrix* mat) { return IsHostShareable(mat->type()); },
[&](const core::type::Array* arr) { return IsHostShareable(arr->ElemType()); },
[&](const core::type::Struct* str) {
for (auto* member : str->Members()) {
if (!IsHostShareable(member->Type())) {
return false;
}
}
return true;
},
[&](const core::type::Atomic* atomic) { return IsHostShareable(atomic->Type()); });
}
// https://gpuweb.github.io/gpuweb/wgsl.html#storable-types
bool Validator::IsStorable(const core::type::Type* type) const {
return IsPlain(type) || type->IsAnyOf<core::type::Texture, core::type::Sampler>();
}
const ast::Statement* Validator::ClosestContinuing(bool stop_at_loop,
sem::Statement* current_statement) const {
for (const auto* s = current_statement; s != nullptr; s = s->Parent()) {
if (stop_at_loop && s->Is<sem::LoopStatement>()) {
break;
}
if (s->Is<sem::LoopContinuingBlockStatement>()) {
return s->Declaration();
}
if (auto* f = As<sem::ForLoopStatement>(s->Parent())) {
if (f->Declaration()->continuing == s->Declaration()) {
return s->Declaration();
}
if (stop_at_loop) {
break;
}
}
if (Is<sem::WhileStatement>(s->Parent())) {
if (stop_at_loop) {
break;
}
}
}
return nullptr;
}
bool Validator::Atomic(const ast::TemplatedIdentifier* a, const core::type::Atomic* s) const {
// https://gpuweb.github.io/gpuweb/wgsl/#atomic-types
// T must be either u32 or i32.
if (!s->Type()->IsAnyOf<core::type::U32, core::type::I32>()) {
AddError("atomic only supports i32 or u32 types", a->arguments[0]->source);
return false;
}
return true;
}
bool Validator::Pointer(const ast::TemplatedIdentifier* a, const core::type::Pointer* s) const {
if (s->AddressSpace() == core::AddressSpace::kUndefined) {
AddError("ptr missing address space", a->source);
return false;
}
if (a->arguments.Length() > 2) { // ptr<address-space, type [, access]>
// https://www.w3.org/TR/WGSL/#access-mode-defaults
// When writing a variable declaration or a pointer type in WGSL source:
// * For the storage address space, the access mode is optional, and defaults to read.
// * For other address spaces, the access mode must not be written.
if (s->AddressSpace() != core::AddressSpace::kStorage) {
AddError("only pointers in <storage> address space may specify an access mode",
a->source);
return false;
}
}
return CheckTypeAccessAddressSpace(s->StoreType(), s->Access(), s->AddressSpace(), tint::Empty,
a->source);
}
bool Validator::StorageTexture(const core::type::StorageTexture* t, const Source& source) const {
switch (t->access()) {
case core::Access::kReadWrite:
if (!enabled_extensions_.Contains(
core::Extension::kChromiumExperimentalReadWriteStorageTexture)) {
AddError(
"read-write storage textures require the "
"chromium_experimental_read_write_storage_texture extension to be enabled",
source);
return false;
}
break;
case core::Access::kWrite:
break;
case core::Access::kUndefined:
AddError("storage texture missing access control", source);
return false;
default:
AddError("storage textures currently only support 'write' access control", source);
return false;
}
if (!IsValidStorageTextureDimension(t->dim())) {
AddError("cube dimensions for storage textures are not supported", source);
return false;
}
if (!IsValidStorageTextureTexelFormat(t->texel_format())) {
AddError(
"image format must be one of the texel formats specified for storage "
"textues in https://gpuweb.github.io/gpuweb/wgsl/#texel-formats",
source);
return false;
}
return true;
}
bool Validator::SampledTexture(const core::type::SampledTexture* t, const Source& source) const {
if (!t->type()->UnwrapRef()->IsAnyOf<core::type::F32, core::type::I32, core::type::U32>()) {
AddError("texture_2d<type>: type must be f32, i32 or u32", source);
return false;
}
return true;
}
bool Validator::MultisampledTexture(const core::type::MultisampledTexture* t,
const Source& source) const {
if (t->dim() != core::type::TextureDimension::k2d) {
AddError("only 2d multisampled textures are supported", source);
return false;
}
if (!t->type()->UnwrapRef()->IsAnyOf<core::type::F32, core::type::I32, core::type::U32>()) {
AddError("texture_multisampled_2d<type>: type must be f32, i32 or u32", source);
return false;
}
return true;
}
bool Validator::Materialize(const core::type::Type* to,
const core::type::Type* from,
const Source& source) const {
if (core::type::Type::ConversionRank(from, to) == core::type::Type::kNoConversion) {
AddError("cannot convert value of type '" + sem_.TypeNameOf(from) + "' to type '" +
sem_.TypeNameOf(to) + "'",
source);
return false;
}
return true;
}
bool Validator::VariableInitializer(const ast::Variable* v,
const core::type::Type* storage_ty,
const sem::ValueExpression* initializer) const {
auto* initializer_ty = initializer->Type();
auto* value_type = initializer_ty->UnwrapRef(); // Implicit load of RHS
// Value type has to match storage type
if (storage_ty != value_type) {
StringStream s;
s << "cannot initialize " << v->Kind() << " of type '" << sem_.TypeNameOf(storage_ty)
<< "' with value of type '" << sem_.TypeNameOf(initializer_ty) << "'";
AddError(s.str(), v->source);
return false;
}
return true;
}
bool Validator::AddressSpaceLayout(const core::type::Type* store_ty,
core::AddressSpace address_space,
Source source) const {
// https://gpuweb.github.io/gpuweb/wgsl/#storage-class-layout-constraints
auto is_uniform_struct_or_array = [address_space](const core::type::Type* ty) {
return address_space == core::AddressSpace::kUniform &&
ty->IsAnyOf<core::type::Array, core::type::Struct>();
};
auto is_uniform_struct = [address_space](const core::type::Type* ty) {
return address_space == core::AddressSpace::kUniform && ty->Is<core::type::Struct>();
};
auto required_alignment_of = [&](const core::type::Type* ty) {
uint32_t actual_align = ty->Align();
uint32_t required_align = actual_align;
if (is_uniform_struct_or_array(ty)) {
required_align = tint::RoundUp(16u, actual_align);
}
return required_align;
};
auto member_name_of = [](const core::type::StructMember* sm) { return sm->Name().Name(); };
// Only validate the [type + address space] once
if (!valid_type_storage_layouts_.Add(TypeAndAddressSpace{store_ty, address_space})) {
return true;
}
if (!core::IsHostShareable(address_space)) {
return true;
}
auto note_usage = [&] {
AddNote("'" + store_ty->FriendlyName() + "' used in address space '" +
tint::ToString(address_space) + "' here",
source);
};
// Among three host-shareable address spaces, f16 is supported in "uniform" and
// "storage" address space, but not "push_constant" address space yet.
if (Is<core::type::F16>(store_ty->DeepestElement()) &&
address_space == core::AddressSpace::kPushConstant) {
AddError("using f16 types in 'push_constant' address space is not implemented yet", source);
return false;
}
if (auto* str = store_ty->As<sem::Struct>()) {
for (size_t i = 0; i < str->Members().Length(); ++i) {
auto* const m = str->Members()[i];
uint32_t required_align = required_alignment_of(m->Type());
// Recurse into the member type.
if (!AddressSpaceLayout(m->Type(), address_space, m->Declaration()->type->source)) {
AddNote("see layout of struct:\n" + str->Layout(), str->Declaration()->source);
note_usage();
return false;
}
// Validate that member is at a valid byte offset
if (m->Offset() % required_align != 0 &&
!enabled_extensions_.Contains(
core::Extension::kChromiumInternalRelaxedUniformLayout)) {
AddError("the offset of a struct member of type '" +
m->Type()->UnwrapRef()->FriendlyName() + "' in address space '" +
tint::ToString(address_space) + "' must be a multiple of " +
std::to_string(required_align) + " bytes, but '" + member_name_of(m) +
"' is currently at offset " + std::to_string(m->Offset()) +
". Consider setting @align(" + std::to_string(required_align) +
") on this member",
m->Declaration()->source);
AddNote("see layout of struct:\n" + str->Layout(), str->Declaration()->source);
if (auto* member_str = m->Type()->As<sem::Struct>()) {
AddNote("and layout of struct member:\n" + member_str->Layout(),
member_str->Declaration()->source);
}
note_usage();
return false;
}
// For uniform buffers, validate that the number of bytes between the previous member of
// type struct and the current is a multiple of 16 bytes.
auto* const prev_member = (i == 0) ? nullptr : str->Members()[i - 1];
if (prev_member && is_uniform_struct(prev_member->Type())) {
const uint32_t prev_to_curr_offset = m->Offset() - prev_member->Offset();
if (prev_to_curr_offset % 16 != 0 &&
!enabled_extensions_.Contains(
core::Extension::kChromiumInternalRelaxedUniformLayout)) {
AddError(
"uniform storage requires that the number of bytes between the start of "
"the previous member of type struct and the current member be a multiple "
"of 16 bytes, but there are currently " +
std::to_string(prev_to_curr_offset) + " bytes between '" +
member_name_of(prev_member) + "' and '" + member_name_of(m) +
"'. Consider setting @align(16) on this member",
m->Declaration()->source);
AddNote("see layout of struct:\n" + str->Layout(), str->Declaration()->source);
auto* prev_member_str = prev_member->Type()->As<sem::Struct>();
AddNote("and layout of previous member struct:\n" + prev_member_str->Layout(),
prev_member_str->Declaration()->source);
note_usage();
return false;
}
}
}
}
// For uniform buffer array members, validate that array elements are aligned to 16 bytes
if (auto* arr = store_ty->As<core::type::Array>()) {
// Recurse into the element type.
// TODO(crbug.com/tint/1388): Ideally we'd pass the source for nested element type here, but
// we can't easily get that from the semantic node. We should consider recursing through the
// AST type nodes instead.
if (!AddressSpaceLayout(arr->ElemType(), address_space, source)) {
return false;
}
if (address_space == core::AddressSpace::kUniform &&
!enabled_extensions_.Contains(core::Extension::kChromiumInternalRelaxedUniformLayout)) {
// We already validated that this array member is itself aligned to 16 bytes above, so
// we only need to validate that stride is a multiple of 16 bytes.
if (arr->Stride() % 16 != 0) {
// Since WGSL has no stride attribute, try to provide a useful hint for how the
// shader author can resolve the issue.
std::string hint;
if (arr->ElemType()->Is<core::type::Scalar>()) {
hint = "Consider using a vector or struct as the element type instead.";
} else if (auto* vec = arr->ElemType()->As<core::type::Vector>();
vec && vec->type()->Size() == 4) {
hint = "Consider using a vec4 instead.";
} else if (arr->ElemType()->Is<sem::Struct>()) {
hint = "Consider using the @size attribute on the last struct member.";
} else {
hint =
"Consider wrapping the element type in a struct and using the @size "
"attribute.";
}
AddError(
"uniform storage requires that array elements are aligned to 16 bytes, but "
"array element of type '" +
arr->ElemType()->FriendlyName() + "' has a stride of " +
std::to_string(arr->Stride()) + " bytes. " + hint,
source);
return false;
}
}
}
return true;
}
bool Validator::LocalVariable(const sem::Variable* local) const {
auto* decl = local->Declaration();
if (IsArrayWithOverrideCount(local->Type())) {
RaiseArrayWithOverrideCountError(decl->type ? decl->type->source
: decl->initializer->source);
return false;
}
return Switch(
decl, //
[&](const ast::Var* var) {
if (IsValidationEnabled(var->attributes,
ast::DisabledValidation::kIgnoreAddressSpace)) {
if (!local->Type()->UnwrapRef()->IsConstructible()) {
AddError("function-scope 'var' must have a constructible type",
var->type ? var->type->source : var->source);
return false;
}
}
return Var(local);
}, //
[&](const ast::Let*) { return Let(local); }, //
[&](const ast::Const*) { return true; }, //
[&](Default) {
TINT_ICE() << "Validator::Variable() called with a unknown variable type: "
<< decl->TypeInfo().name;
return false;
});
}
bool Validator::GlobalVariable(
const sem::GlobalVariable* global,
const Hashmap<OverrideId, const sem::Variable*, 8>& override_ids) const {
auto* decl = global->Declaration();
if (global->AddressSpace() != core::AddressSpace::kWorkgroup &&
IsArrayWithOverrideCount(global->Type())) {
RaiseArrayWithOverrideCountError(decl->type ? decl->type->source
: decl->initializer->source);
return false;
}
bool ok = Switch(
decl, //
[&](const ast::Var* var) {
if (auto* init = global->Initializer();
init && init->Stage() > core::EvaluationStage::kOverride) {
AddError("module-scope 'var' initializer must be a constant or override-expression",
init->Declaration()->source);
return false;
}
if (!var->declared_address_space && !global->Type()->UnwrapRef()->is_handle()) {
AddError(
"module-scope 'var' declarations that are not of texture or sampler types must "
"provide an address space",
decl->source);
return false;
}
return Var(global);
},
[&](const ast::Override*) { return Override(global, override_ids); },
[&](const ast::Const*) { return Const(global); },
[&](Default) {
TINT_ICE() << "Validator::GlobalVariable() called with a unknown variable type: "
<< decl->TypeInfo().name;
return false;
});
if (!ok) {
return false;
}
if (global->AddressSpace() == core::AddressSpace::kFunction) {
AddError("module-scope 'var' must not use address space 'function'", decl->source);
return false;
}
switch (global->AddressSpace()) {
case core::AddressSpace::kUniform:
case core::AddressSpace::kStorage:
case core::AddressSpace::kHandle: {
// https://gpuweb.github.io/gpuweb/wgsl/#resource-interface
// Each resource variable must be declared with both group and binding attributes.
if (!decl->HasBindingPoint()) {
AddError("resource variables require @group and @binding attributes", decl->source);
return false;
}
break;
}
default: {
auto* binding_attr = ast::GetAttribute<ast::BindingAttribute>(decl->attributes);
auto* group_attr = ast::GetAttribute<ast::GroupAttribute>(decl->attributes);
if (binding_attr || group_attr) {
// https://gpuweb.github.io/gpuweb/wgsl/#attribute-binding
// Must only be applied to a resource variable
AddError("non-resource variables must not have @group or @binding attributes",
decl->source);
return false;
}
}
}
return true;
}
bool Validator::Var(const sem::Variable* v) const {
auto* var = v->Declaration()->As<ast::Var>();
auto* store_ty = v->Type()->UnwrapRef();
if (!IsStorable(store_ty)) {
AddError(sem_.TypeNameOf(store_ty) + " cannot be used as the type of a var", var->source);
return false;
}
if (store_ty->is_handle() && var->declared_address_space) {
// https://gpuweb.github.io/gpuweb/wgsl/#module-scope-variables
// If the store type is a texture type or a sampler type, then the variable declaration must
// not have a address space attribute. The address space will always be handle.
AddError("variables of type '" + sem_.TypeNameOf(store_ty) +
"' must not specifiy an address space",
var->source);
return false;
}
if (var->declared_access) {
// https://www.w3.org/TR/WGSL/#access-mode-defaults
// When writing a variable declaration or a pointer type in WGSL source:
// * For the storage address space, the access mode is optional, and defaults to read.
// * For other address spaces, the access mode must not be written.
if (v->AddressSpace() != core::AddressSpace::kStorage) {
AddError("only variables in <storage> address space may specify an access mode",
var->source);
return false;
}
}
if (var->initializer) {
switch (v->AddressSpace()) {
case core::AddressSpace::kPrivate:
case core::AddressSpace::kFunction:
break; // Allowed an initializer
default:
// https://gpuweb.github.io/gpuweb/wgsl/#var-and-let
// Optionally has an initializer expression, if the variable is in the private or
// function address spaces.
AddError("var of address space '" + tint::ToString(v->AddressSpace()) +
"' cannot have an initializer. var initializers are only supported "
"for the address spaces 'private' and 'function'",
var->source);
return false;
}
}
if (!CheckTypeAccessAddressSpace(v->Type()->UnwrapRef(), v->Access(), v->AddressSpace(),
var->attributes, var->source)) {
return false;
}
if (IsValidationEnabled(var->attributes, ast::DisabledValidation::kIgnoreAddressSpace) &&
(v->AddressSpace() == core::AddressSpace::kIn ||
v->AddressSpace() == core::AddressSpace::kOut)) {
AddError("invalid use of input/output address space", var->source);
return false;
}
return true;
}
bool Validator::Let(const sem::Variable* v) const {
auto* decl = v->Declaration();
auto* storage_ty = v->Type()->UnwrapRef();
if (!(storage_ty->IsConstructible() || storage_ty->Is<core::type::Pointer>())) {
AddError(sem_.TypeNameOf(storage_ty) + " cannot be used as the type of a 'let'",
decl->source);
return false;
}
return true;
}
bool Validator::Override(const sem::GlobalVariable* v,
const Hashmap<OverrideId, const sem::Variable*, 8>& override_ids) const {
auto* decl = v->Declaration();
auto* storage_ty = v->Type()->UnwrapRef();
if (auto* init = v->Initializer(); init && init->Stage() > core::EvaluationStage::kOverride) {
AddError("'override' initializer must be an override-expression",
init->Declaration()->source);
return false;
}
for (auto* attr : decl->attributes) {
if (attr->Is<ast::IdAttribute>()) {
auto id = v->OverrideId();
if (auto var = override_ids.Find(id); var && *var != v) {
AddError("@id values must be unique", attr->source);
AddNote(
"a override with an ID of " + std::to_string(id.value) +
" was previously declared here:",
ast::GetAttribute<ast::IdAttribute>((*var)->Declaration()->attributes)->source);
return false;
}
}
}
if (!storage_ty->Is<core::type::Scalar>()) {
AddError(sem_.TypeNameOf(storage_ty) + " cannot be used as the type of a 'override'",
decl->source);
return false;
}
return true;
}
bool Validator::Const(const sem::Variable*) const {
return true;
}
bool Validator::Parameter(const sem::Variable* var) const {
auto* decl = var->Declaration();
if (IsValidationDisabled(decl->attributes, ast::DisabledValidation::kFunctionParameter)) {
return true;
}
if (auto* ref = var->Type()->As<core::type::Pointer>()) {
if (IsValidationEnabled(decl->attributes, ast::DisabledValidation::kIgnoreAddressSpace)) {
bool ok = false;
auto sc = ref->AddressSpace();
switch (sc) {
case core::AddressSpace::kFunction:
case core::AddressSpace::kPrivate:
ok = true;
break;
case core::AddressSpace::kStorage:
case core::AddressSpace::kUniform:
case core::AddressSpace::kWorkgroup:
ok = enabled_extensions_.Contains(
core::Extension::kChromiumExperimentalFullPtrParameters);
break;
default:
break;
}
if (!ok) {
StringStream ss;
ss << "function parameter of pointer type cannot be in '" << sc
<< "' address space";
AddError(ss.str(), decl->source);
return false;
}
}
}
if (IsPlain(var->Type())) {
if (!var->Type()->IsConstructible()) {
AddError("type of function parameter must be constructible", decl->type->source);
return false;
}
} else if (!var->Type()
->IsAnyOf<core::type::Texture, core::type::Sampler, core::type::Pointer>()) {
AddError("type of function parameter cannot be " + sem_.TypeNameOf(var->Type()),
decl->source);
return false;
}
return true;
}
bool Validator::BuiltinAttribute(const ast::BuiltinAttribute* attr,
const core::type::Type* storage_ty,
ast::PipelineStage stage,
const bool is_input) const {
auto* type = storage_ty->UnwrapRef();
StringStream stage_name;
stage_name << stage;
bool is_stage_mismatch = false;
bool is_output = !is_input;
auto builtin = sem_.Get(attr)->Value();
switch (builtin) {
case core::BuiltinValue::kPosition:
if (stage != ast::PipelineStage::kNone &&
!((is_input && stage == ast::PipelineStage::kFragment) ||
(is_output && stage == ast::PipelineStage::kVertex))) {
is_stage_mismatch = true;
}
if (!(type->is_float_vector() && type->As<core::type::Vector>()->Width() == 4)) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'vec4<f32>'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kGlobalInvocationId:
case core::BuiltinValue::kLocalInvocationId:
case core::BuiltinValue::kNumWorkgroups:
case core::BuiltinValue::kWorkgroupId:
if (stage != ast::PipelineStage::kNone &&
!(stage == ast::PipelineStage::kCompute && is_input)) {
is_stage_mismatch = true;
}
if (!(type->is_unsigned_integer_vector() &&
type->As<core::type::Vector>()->Width() == 3)) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'vec3<u32>'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kFragDepth:
if (stage != ast::PipelineStage::kNone &&
!(stage == ast::PipelineStage::kFragment && !is_input)) {
is_stage_mismatch = true;
}
if (!type->Is<core::type::F32>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'f32'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kFrontFacing:
if (stage != ast::PipelineStage::kNone &&
!(stage == ast::PipelineStage::kFragment && is_input)) {
is_stage_mismatch = true;
}
if (!type->Is<core::type::Bool>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'bool'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kLocalInvocationIndex:
if (stage != ast::PipelineStage::kNone &&
!(stage == ast::PipelineStage::kCompute && is_input)) {
is_stage_mismatch = true;
}
if (!type->Is<core::type::U32>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'u32'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kVertexIndex:
case core::BuiltinValue::kInstanceIndex:
if (stage != ast::PipelineStage::kNone &&
!(stage == ast::PipelineStage::kVertex && is_input)) {
is_stage_mismatch = true;
}
if (!type->Is<core::type::U32>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'u32'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kSampleMask:
if (stage != ast::PipelineStage::kNone && !(stage == ast::PipelineStage::kFragment)) {
is_stage_mismatch = true;
}
if (!type->Is<core::type::U32>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'u32'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kSampleIndex:
if (stage != ast::PipelineStage::kNone &&
!(stage == ast::PipelineStage::kFragment && is_input)) {
is_stage_mismatch = true;
}
if (!type->Is<core::type::U32>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'u32'";
AddError(err.str(), attr->source);
return false;
}
break;
case core::BuiltinValue::kSubgroupInvocationId:
case core::BuiltinValue::kSubgroupSize:
if (!enabled_extensions_.Contains(core::Extension::kChromiumExperimentalSubgroups)) {
StringStream err;
err << "use of @builtin(" << builtin
<< ") attribute requires enabling extension 'chromium_experimental_subgroups'";
AddError(err.str(), attr->source);
return false;
}
if (!type->Is<core::type::U32>()) {
StringStream err;
err << "store type of @builtin(" << builtin << ") must be 'u32'";
AddError(err.str(), attr->source);
return false;
}
if (stage != ast::PipelineStage::kNone && stage != ast::PipelineStage::kCompute) {
StringStream err;
err << "@builtin(" << builtin << ") is only valid as a compute shader input";
AddError(err.str(), attr->source);
return false;
}
break;
default:
break;
}
if (is_stage_mismatch) {
StringStream err;
err << "@builtin(" << builtin << ") cannot be used in "
<< (is_input ? "input of " : "output of ") << stage_name.str() << " pipeline stage";
AddError(err.str(), attr->source);
return false;
}
return true;
}
bool Validator::InterpolateAttribute(const ast::InterpolateAttribute* attr,
const core::type::Type* storage_ty) const {
auto* type = storage_ty->UnwrapRef();
auto i_type = sem_.AsInterpolationType(sem_.Get(attr->type));
if (TINT_UNLIKELY(!i_type)) {
return false;
}
if (type->is_integer_scalar_or_vector() && i_type->Value() != core::InterpolationType::kFlat) {
AddError("interpolation type must be 'flat' for integral user-defined IO types",
attr->source);
return false;
}
if (attr->sampling && i_type->Value() == core::InterpolationType::kFlat) {
AddError("flat interpolation attribute must not have a sampling parameter", attr->source);
return false;
}
return true;
}
bool Validator::Function(const sem::Function* func, ast::PipelineStage stage) const {
auto* decl = func->Declaration();
for (auto* attr : decl->attributes) {
bool ok = Switch(
attr, //
[&](const ast::WorkgroupAttribute*) {
if (decl->PipelineStage() != ast::PipelineStage::kCompute) {
AddError("@workgroup_size is only valid for compute stages", attr->source);
return false;
}
return true;
},
[&](const ast::MustUseAttribute*) {
if (func->ReturnType()->Is<core::type::Void>()) {
AddError("@must_use can only be applied to functions that return a value",
attr->source);
return false;
}
return true;
},
[&](Default) { return true; });
if (!ok) {
return false;
}
}
if (decl->params.Length() > kMaxFunctionParameters) {
AddError("function declares " + std::to_string(decl->params.Length()) +
" parameters, maximum is " + std::to_string(kMaxFunctionParameters),
decl->source);
return false;
}
if (!func->ReturnType()->Is<core::type::Void>()) {
if (!func->ReturnType()->IsConstructible()) {
AddError("function return type must be a constructible type",
decl->return_type->source);
return false;
}
if (decl->body) {
sem::Behaviors behaviors{sem::Behavior::kNext};
if (auto* last = decl->body->Last()) {
behaviors = sem_.Get(last)->Behaviors();
}
if (behaviors.Contains(sem::Behavior::kNext)) {
AddError("missing return at end of function", decl->source);
return false;
}
} else if (TINT_UNLIKELY(IsValidationEnabled(
decl->attributes, ast::DisabledValidation::kFunctionHasNoBody))) {
TINT_ICE() << "Function " << decl->name->symbol.Name() << " has no body";
}
}
if (decl->IsEntryPoint()) {
if (!EntryPoint(func, stage)) {
return false;
}
}
// https://www.w3.org/TR/WGSL/#behaviors-rules
// a function behavior is always one of {}, or {Next}.
if (TINT_UNLIKELY(func->Behaviors() != sem::Behaviors{} &&
func->Behaviors() != sem::Behavior::kNext)) {
auto name = decl->name->symbol.Name();
TINT_ICE() << "function '" << name << "' behaviors are: " << func->Behaviors();
}
return true;
}
bool Validator::EntryPoint(const sem::Function* func, ast::PipelineStage stage) const {
auto* decl = func->Declaration();
// Use a lambda to validate the entry point attributes for a type.
// Persistent state is used to track which builtins and locations have already been seen, in
// order to catch conflicts.
// TODO(jrprice): This state could be stored in sem::Function instead, and then passed to
// sem::Function since it would be useful there too.
Hashset<core::BuiltinValue, 4> builtins;
Hashset<std::pair<uint32_t, uint32_t>, 8> locationsAndIndexes;
enum class ParamOrRetType {
kParameter,
kReturnType,
};
// Inner lambda that is applied to a type and all of its members.
auto validate_entry_point_attributes_inner = [&](VectorRef<const ast::Attribute*> attrs,
const core::type::Type* ty, Source source,
ParamOrRetType param_or_ret,
bool is_struct_member,
std::optional<uint32_t> location,
std::optional<uint32_t> index) {
// Scan attributes for pipeline IO attributes.
// Check for overlap with attributes that have been seen previously.
const ast::Attribute* pipeline_io_attribute = nullptr;
const ast::LocationAttribute* location_attribute = nullptr;
const ast::IndexAttribute* index_attribute = nullptr;
const ast::InterpolateAttribute* interpolate_attribute = nullptr;
const ast::InvariantAttribute* invariant_attribute = nullptr;
for (auto* attr : attrs) {
auto is_invalid_compute_shader_attribute = false;
if (auto* builtin_attr = attr->As<ast::BuiltinAttribute>()) {
auto builtin = sem_.Get(builtin_attr)->Value();
if (pipeline_io_attribute) {
AddError("multiple entry point IO attributes", attr->source);
AddNote("previously consumed " + AttrToStr(pipeline_io_attribute),
pipeline_io_attribute->source);
return false;
}
pipeline_io_attribute = attr;
if (builtins.Contains(builtin)) {
StringStream err;
err << "@builtin(" << builtin << ") appears multiple times as pipeline "
<< (param_or_ret == ParamOrRetType::kParameter ? "input" : "output");
AddError(err.str(), decl->source);
return false;
}
if (!BuiltinAttribute(builtin_attr, ty, stage,
/* is_input */ param_or_ret == ParamOrRetType::kParameter)) {
return false;
}
builtins.Add(builtin);
} else if (auto* loc_attr = attr->As<ast::LocationAttribute>()) {
location_attribute = loc_attr;
if (pipeline_io_attribute) {
AddError("multiple entry point IO attributes", attr->source);
AddNote("previously consumed " + AttrToStr(pipeline_io_attribute),
pipeline_io_attribute->source);
return false;
}
pipeline_io_attribute = attr;
bool is_input = param_or_ret == ParamOrRetType::kParameter;
if (TINT_UNLIKELY(!location.has_value())) {
TINT_ICE() << "Location has no value";
return false;
}
if (!LocationAttribute(loc_attr, ty, stage, source, is_input)) {
return false;
}
} else if (auto* index_attr = attr->As<ast::IndexAttribute>()) {
index_attribute = index_attr;
return IndexAttribute(index_attr, stage);
} else if (auto* interpolate = attr->As<ast::InterpolateAttribute>()) {
if (decl->PipelineStage() == ast::PipelineStage::kCompute) {
is_invalid_compute_shader_attribute = true;
} else if (!InterpolateAttribute(interpolate, ty)) {
return false;
}
interpolate_attribute = interpolate;
} else if (auto* invariant = attr->As<ast::InvariantAttribute>()) {
if (decl->PipelineStage() == ast::PipelineStage::kCompute) {
is_invalid_compute_shader_attribute = true;
}
invariant_attribute = invariant;
}
if (is_invalid_compute_shader_attribute) {
std::string input_or_output =
param_or_ret == ParamOrRetType::kParameter ? "inputs" : "output";
AddError("@" + attr->Name() + " is not valid for compute shader " + input_or_output,
attr->source);
return false;
}
}
if (IsValidationEnabled(attrs, ast::DisabledValidation::kEntryPointParameter)) {
if (is_struct_member && ty->Is<core::type::Struct>()) {
AddError("nested structures cannot be used for entry point IO", source);
return false;
}
if (!ty->Is<core::type::Struct>() && !pipeline_io_attribute) {
std::string err = "missing entry point IO attribute";
if (!is_struct_member) {
err += (param_or_ret == ParamOrRetType::kParameter ? " on parameter"
: " on return type");
}
AddError(err, source);
return false;
}
if (pipeline_io_attribute && pipeline_io_attribute->Is<ast::LocationAttribute>()) {
if (ty->is_integer_scalar_or_vector() && !interpolate_attribute) {
if (decl->PipelineStage() == ast::PipelineStage::kVertex &&
param_or_ret == ParamOrRetType::kReturnType) {
AddError(
"integral user-defined vertex outputs must have a flat interpolation "
"attribute",
source);
return false;
}
if (decl->PipelineStage() == ast::PipelineStage::kFragment &&
param_or_ret == ParamOrRetType::kParameter) {
AddError(
"integral user-defined fragment inputs must have a flat interpolation "
"attribute",
source);
return false;
}
}
}
if (index_attribute) {
if (Is<ast::LocationAttribute>(pipeline_io_attribute)) {
AddError("index attribute must only be used with @location",
index_attribute->source);
return false;
}
// Because HLSL specifies dual source blending targets with SV_Target0 and 1, we
// should restrict targets with index attributes to location 0 for easy translation
// in the backend writers.
if (location.value() != 0) {
AddError("index attribute must only be used with @location(0)",
index_attribute->source);
return false;
}
}
if (location_attribute) {
uint32_t idx = 0xffffffff;
if (index_attribute) {
idx = index.value();
}
std::pair<uint32_t, uint32_t> locationAndIndex(location.value(), idx);
if (!locationsAndIndexes.Add(locationAndIndex)) {
StringStream err;
if (!index_attribute) {
err << "@location(" << location.value() << ") appears multiple times";
} else {
err << "@location(" << location.value() << ") @index(" << index.value()
<< ") appears multiple times";
}
AddError(err.str(), location_attribute->source);
return false;
}
}
if (interpolate_attribute) {
if (!pipeline_io_attribute ||
!pipeline_io_attribute->Is<ast::LocationAttribute>()) {
AddError("interpolate attribute must only be used with @location",
interpolate_attribute->source);
return false;
}
}
if (invariant_attribute) {
bool has_position = false;
if (pipeline_io_attribute) {
if (auto* builtin_attr = pipeline_io_attribute->As<ast::BuiltinAttribute>()) {
auto builtin = sem_.Get(builtin_attr)->Value();
has_position = (builtin == core::BuiltinValue::kPosition);
}
}
if (!has_position) {
AddError("invariant attribute must only be applied to a position builtin",
invariant_attribute->source);
return false;
}
}
}
return true;
};
// Outer lambda for validating the entry point attributes for a type.
auto validate_entry_point_attributes = [&](VectorRef<const ast::Attribute*> attrs,
const core::type::Type* ty, Source source,
ParamOrRetType param_or_ret,
std::optional<uint32_t> location,
std::optional<uint32_t> index) {
if (!validate_entry_point_attributes_inner(attrs, ty, source, param_or_ret,
/*is_struct_member*/ false, location, index)) {
return false;
}
if (auto* str = ty->As<sem::Struct>()) {
for (auto* member : str->Members()) {
if (!validate_entry_point_attributes_inner(
member->Declaration()->attributes, member->Type(),
member->Declaration()->source, param_or_ret,
/*is_struct_member*/ true, member->Attributes().location,
member->Attributes().index)) {
AddNote("while analyzing entry point '" + decl->name->symbol.Name() + "'",
decl->source);
return false;
}
}
}
return true;
};
for (auto* param : func->Parameters()) {
auto* param_decl = param->Declaration();
if (!validate_entry_point_attributes(param_decl->attributes, param->Type(),
param_decl->source, ParamOrRetType::kParameter,
param->Location(), std::nullopt)) {
return false;
}
}
// Clear IO sets after parameter validation. Builtin and location attributes in return types
// should be validated independently from those used in parameters.
builtins.Clear();
locationsAndIndexes.Clear();
if (!func->ReturnType()->Is<core::type::Void>()) {
if (!validate_entry_point_attributes(decl->return_type_attributes, func->ReturnType(),
decl->source, ParamOrRetType::kReturnType,
func->ReturnLocation(), func->ReturnIndex())) {
return false;
}
}
if (decl->PipelineStage() == ast::PipelineStage::kVertex &&
!builtins.Contains(core::BuiltinValue::kPosition)) {
// Check module-scope variables, as the SPIR-V sanitizer generates these.
bool found = false;
for (auto* global : func->TransitivelyReferencedGlobals()) {
if (auto* builtin_attr =
ast::GetAttribute<ast::BuiltinAttribute>(global->Declaration()->attributes)) {
auto builtin = sem_.Get(builtin_attr)->Value();
if (builtin == core::BuiltinValue::kPosition) {
found = true;
break;
}
}
}
if (!found) {
AddError("a vertex shader must include the 'position' builtin in its return type",
decl->source);
return false;
}
}
if (decl->PipelineStage() == ast::PipelineStage::kCompute) {
if (!ast::HasAttribute<ast::WorkgroupAttribute>(decl->attributes)) {
AddError("a compute shader must include 'workgroup_size' in its attributes",
decl->source);
return false;
}
}
// Validate there are no resource variable binding collisions
Hashmap<BindingPoint, const ast::Variable*, 8> binding_points;
for (auto* global : func->TransitivelyReferencedGlobals()) {
auto* var_decl = global->Declaration()->As<ast::Var>();
if (!var_decl) {
continue;
}
auto bp = global->BindingPoint();
if (!bp) {
continue;
}
if (auto added = binding_points.Add(*bp, var_decl);
!added &&
IsValidationEnabled(decl->attributes,
ast::DisabledValidation::kBindingPointCollision) &&
IsValidationEnabled((*added.value)->attributes,
ast::DisabledValidation::kBindingPointCollision)) {
// https://gpuweb.github.io/gpuweb/wgsl/#resource-interface
// Bindings must not alias within a shader stage: two different variables in the
// resource interface of a given shader must not have the same group and binding values,
// when considered as a pair of values.
auto func_name = decl->name->symbol.Name();
AddError(
"entry point '" + func_name +
"' references multiple variables that use the same resource binding @group(" +
std::to_string(bp->group) + "), @binding(" + std::to_string(bp->binding) + ")",
var_decl->source);
AddNote("first resource binding usage declared here", (*added.value)->source);
return false;
}
}
return true;
}
bool Validator::EvaluationStage(const sem::ValueExpression* expr,
core::EvaluationStage latest_stage,
std::string_view constraint) const {
if (expr->Stage() == core::EvaluationStage::kNotEvaluated) {
return true;
}
if (expr->Stage() > latest_stage) {
auto stage_name = [](core::EvaluationStage stage) -> std::string {
switch (stage) {
case core::EvaluationStage::kRuntime:
return "a runtime-expression";
case core::EvaluationStage::kOverride:
return "an override-expression";
case core::EvaluationStage::kConstant:
return "a const-expression";
case core::EvaluationStage::kNotEvaluated:
return "an unevaluated expression";
}
return "<unknown>";
};
AddError(std::string(constraint) + " requires " + stage_name(latest_stage) +
", but expression is " + stage_name(expr->Stage()),
expr->Declaration()->source);
if (auto* stmt = expr->Stmt()) {
if (auto* decl = As<ast::VariableDeclStatement>(stmt->Declaration())) {
if (decl->variable->Is<ast::Const>()) {
AddNote("consider changing 'const' to 'let'", decl->source);
}
}
}
return false;
}
return true;
}
bool Validator::Statements(VectorRef<const ast::Statement*> stmts) const {
for (auto* stmt : stmts) {
if (!sem_.Get(stmt)->IsReachable()) {
if (!AddDiagnostic(core::ChromiumDiagnosticRule::kUnreachableCode,
"code is unreachable", stmt->source)) {
return false;
}
break;
}
}
return true;
}
bool Validator::Bitcast(const ast::BitcastExpression* cast, const core::type::Type* to) const {
auto* from = sem_.TypeOf(cast->expr)->UnwrapRef();
if (!from->is_numeric_scalar_or_vector()) {
AddError("'" + sem_.TypeNameOf(from) + "' cannot be bitcast", cast->expr->source);
return false;
}
if (!to->is_numeric_scalar_or_vector()) {
AddError("cannot bitcast to '" + sem_.TypeNameOf(to) + "'", cast->type->source);
return false;
}
// Only bitcasts between scalar/vector types of the same bit width are allowed.
if (from->Size() != to->Size()) {
AddError(
"cannot bitcast from '" + sem_.TypeNameOf(from) + "' to '" + sem_.TypeNameOf(to) + "'",
cast->source);
return false;
}
return true;
}
bool Validator::BreakStatement(const sem::Statement* stmt,
sem::Statement* current_statement) const {
if (!stmt->FindFirstParent<sem::LoopBlockStatement, sem::CaseStatement>()) {
AddError("break statement must be in a loop or switch case", stmt->Declaration()->source);
return false;
}
if (ClosestContinuing(/*stop_at_loop*/ true, current_statement) != nullptr) {
AddError(
"`break` must not be used to exit from a continuing block. Use `break-if` instead.",
stmt->Declaration()->source);
return false;
}
return true;
}
bool Validator::ContinueStatement(const sem::Statement* stmt,
sem::Statement* current_statement) const {
if (auto* continuing = ClosestContinuing(/*stop_at_loop*/ true, current_statement)) {
AddError("continuing blocks must not contain a continue statement",
stmt->Declaration()->source);
if (continuing != stmt->Declaration() && continuing != stmt->Parent()->Declaration()) {
AddNote("see continuing block here", continuing->source);
}
return false;
}
if (!stmt->FindFirstParent<sem::LoopBlockStatement>()) {
AddError("continue statement must be in a loop", stmt->Declaration()->source);
return false;
}
return true;
}
bool Validator::Call(const sem::Call* call, sem::Statement* current_statement) const {
if (!call->Target()->MustUse()) {
return true;
}
auto* expr = call->Declaration();
bool is_call_stmt =
current_statement && Is<ast::CallStatement>(current_statement->Declaration(),
[&](auto* stmt) { return stmt->expr == expr; });
if (is_call_stmt) {
// Call target is annotated with @must_use, but was used as a call statement.
Switch(
call->Target(), //
[&](const sem::Function* fn) {
AddError("ignoring return value of function '" +
fn->Declaration()->name->symbol.Name() + "' annotated with @must_use",
call->Declaration()->source);
sem_.NoteDeclarationSource(fn->Declaration());
},
[&](const sem::Builtin* b) {
AddError("ignoring return value of builtin '" + tint::ToString(b->Type()) + "'",
call->Declaration()->source);
},
[&](const sem::ValueConversion*) {
AddError("value conversion evaluated but not used", call->Declaration()->source);
},
[&](const sem::ValueConstructor*) {
AddError("value constructor evaluated but not used", call->Declaration()->source);
},
[&](Default) {
AddError("return value of call not used", call->Declaration()->source);
});
return false;
}
return true;
}
bool Validator::LoopStatement(const sem::LoopStatement* stmt) const {
if (stmt->Behaviors().Empty()) {
AddError("loop does not exit", stmt->Declaration()->source.Begin());
return false;
}
return true;
}
bool Validator::ForLoopStatement(const sem::ForLoopStatement* stmt) const {
if (stmt->Behaviors().Empty()) {
AddError("for-loop does not exit", stmt->Declaration()->source.Begin());
return false;
}
if (auto* cond = stmt->Condition()) {
auto* cond_ty = cond->Type()->UnwrapRef();
if (!cond_ty->Is<core::type::Bool>()) {
AddError("for-loop condition must be bool, got " + sem_.TypeNameOf(cond_ty),
stmt->Condition()->Declaration()->source);
return false;
}
}
return true;
}
bool Validator::WhileStatement(const sem::WhileStatement* stmt) const {
if (stmt->Behaviors().Empty()) {
AddError("while does not exit", stmt->Declaration()->source.Begin());
return false;
}
if (auto* cond = stmt->Condition()) {
auto* cond_ty = cond->Type()->UnwrapRef();
if (!cond_ty->Is<core::type::Bool>()) {
AddError("while condition must be bool, got " + sem_.TypeNameOf(cond_ty),
stmt->Condition()->Declaration()->source);
return false;
}
}
return true;
}
bool Validator::BreakIfStatement(const sem::BreakIfStatement* stmt,
sem::Statement* current_statement) const {
auto* cond_ty = stmt->Condition()->Type()->UnwrapRef();
if (!cond_ty->Is<core::type::Bool>()) {
AddError("break-if statement condition must be bool, got " + sem_.TypeNameOf(cond_ty),
stmt->Condition()->Declaration()->source);
return false;
}
for (const auto* s = current_statement; s != nullptr; s = s->Parent()) {
if (s->Is<sem::LoopStatement>()) {
break;
}
if (auto* continuing = s->As<sem::LoopContinuingBlockStatement>()) {
if (continuing->Declaration()->statements.Back() != stmt->Declaration()) {
AddError("break-if must be the last statement in a continuing block",
stmt->Declaration()->source);
AddNote("see continuing block here", s->Declaration()->source);
return false;
}
return true;
}
}
AddError("break-if must be in a continuing block", stmt->Declaration()->source);
return false;
}
bool Validator::IfStatement(const sem::IfStatement* stmt) const {
auto* cond_ty = stmt->Condition()->Type()->UnwrapRef();
if (!cond_ty->Is<core::type::Bool>()) {
AddError("if statement condition must be bool, got " + sem_.TypeNameOf(cond_ty),
stmt->Condition()->Declaration()->source);
return false;
}
return true;
}
bool Validator::BuiltinCall(const sem::Call* call) const {
if (call->Type()->Is<core::type::Void>()) {
bool is_call_statement = false;
// Some built-in call are not owned by a statement, e.g. a built-in called in global
// variable declaration. Calling no-return-value built-in in these context is invalid as
// well.
if (auto* call_stmt = call->Stmt()) {
if (auto* call_stmt_ast = As<ast::CallStatement>(call_stmt->Declaration())) {
if (call_stmt_ast->expr == call->Declaration()) {
is_call_statement = true;
}
}
}
if (!is_call_statement) {
// https://gpuweb.github.io/gpuweb/wgsl/#function-call-expr
// If the called function does not return a value, a function call statement should be
// used instead.
auto* builtin = call->Target()->As<sem::Builtin>();
auto name = tint::ToString(builtin->Type());
AddError("builtin '" + name + "' does not return a value", call->Declaration()->source);
return false;
}
}
return true;
}
bool Validator::TextureBuiltinFunction(const sem::Call* call) const {
auto* builtin = call->Target()->As<sem::Builtin>();
if (!builtin) {
return false;
}
std::string func_name = builtin->str();
auto& signature = builtin->Signature();
auto check_arg_is_constexpr = [&](core::ParameterUsage usage, int min, int max) {
auto signed_index = signature.IndexOf(usage);
if (signed_index < 0) {
return true;
}
auto index = static_cast<size_t>(signed_index);
std::string name{core::ToString(usage)};
auto* arg = call->Arguments()[index];
if (auto values = arg->ConstantValue()) {
if (auto* vector = values->Type()->As<core::type::Vector>()) {
for (size_t i = 0; i < vector->Width(); i++) {
auto value = values->Index(i)->ValueAs<AInt>();
if (value < min || value > max) {
AddError("each component of the " + name + " argument must be at least " +
std::to_string(min) + " and at most " + std::to_string(max) +
". " + name + " component " + std::to_string(i) + " is " +
std::to_string(value),
arg->Declaration()->source);
return false;
}
}
} else {
auto value = values->ValueAs<AInt>();
if (value < min || value > max) {
AddError("the " + name + " argument must be at least " + std::to_string(min) +
" and at most " + std::to_string(max) + ". " + name + " is " +
std::to_string(value),
arg->Declaration()->source);
return false;
}
}
return true;
}
AddError("the " + name + " argument must be a const-expression",
arg->Declaration()->source);
return false;
};
return check_arg_is_constexpr(core::ParameterUsage::kOffset, -8, 7) &&
check_arg_is_constexpr(core::ParameterUsage::kComponent, 0, 3);
}
bool Validator::WorkgroupUniformLoad(const sem::Call* call) const {
auto* builtin = call->Target()->As<sem::Builtin>();
if (!builtin) {
return false;
}
TINT_ASSERT(call->Arguments().Length() > 0);
auto* arg = call->Arguments()[0];
auto* ptr = arg->Type()->As<core::type::Pointer>();
TINT_ASSERT(ptr != nullptr);
auto* ty = ptr->StoreType();
if (ty->Is<core::type::Atomic>() || atomic_composite_info_.Contains(ty)) {
AddError(
"workgroupUniformLoad must not be called with an argument that contains an atomic type",
arg->Declaration()->source);
return false;
}
return true;
}
bool Validator::RequiredExtensionForBuiltinFunction(const sem::Call* call) const {
const auto* builtin = call->Target()->As<sem::Builtin>();
if (!builtin) {
return true;
}
const auto extension = builtin->RequiredExtension();
if (extension == core::Extension::kUndefined) {
return true;
}
if (!enabled_extensions_.Contains(extension)) {
AddError("cannot call built-in function '" + std::string(builtin->str()) +
"' without extension " + tint::ToString(extension),
call->Declaration()->source);
return false;
}
return true;
}
bool Validator::CheckF16Enabled(const Source& source) const {
// Validate if f16 type is allowed.
if (!enabled_extensions_.Contains(core::Extension::kF16)) {
AddError("f16 type used without 'f16' extension enabled", source);
return false;
}
return true;
}
bool Validator::FunctionCall(const sem::Call* call, sem::Statement* current_statement) const {
auto* decl = call->Declaration();
auto* target = call->Target()->As<sem::Function>();
auto sym = target->Declaration()->name->symbol;
auto name = sym.Name();
if (!current_statement) { // Function call at module-scope.
AddError("functions cannot be called at module-scope", decl->source);
return false;
}
if (target->Declaration()->IsEntryPoint()) {
// https://www.w3.org/TR/WGSL/#function-restriction
// An entry point must never be the target of a function call.
AddError("entry point functions cannot be the target of a function call", decl->source);
return false;
}
if (decl->args.Length() != target->Parameters().Length()) {
bool more = decl->args.Length() > target->Parameters().Length();
AddError("too " + (more ? std::string("many") : std::string("few")) +
" arguments in call to '" + name + "', expected " +
std::to_string(target->Parameters().Length()) + ", got " +
std::to_string(call->Arguments().Length()),
decl->source);
return false;
}
for (size_t i = 0; i < call->Arguments().Length(); ++i) {
const sem::Variable* param = target->Parameters()[i];
const ast::Expression* arg_expr = decl->args[i];
auto* param_type = param->Type();
auto* arg_type = sem_.TypeOf(arg_expr)->UnwrapRef();
if (param_type != arg_type) {
AddError("type mismatch for argument " + std::to_string(i + 1) + " in call to '" +
name + "', expected '" + sem_.TypeNameOf(param_type) + "', got '" +
sem_.TypeNameOf(arg_type) + "'",
arg_expr->source);
return false;
}
if (param_type->Is<core::type::Pointer>() &&
!enabled_extensions_.Contains(
core::Extension::kChromiumExperimentalFullPtrParameters)) {
// https://gpuweb.github.io/gpuweb/wgsl/#function-restriction
// Each argument of pointer type to a user-defined function must have the same memory
// view as its root identifier.
// We can validate this by just comparing the store type of the argument with that of
// its root identifier, as these will match iff the memory view is the same.
auto* arg_store_type = arg_type->As<core::type::Pointer>()->StoreType();
auto* root = call->Arguments()[i]->RootIdentifier();
auto* root_ptr_ty = root->Type()->As<core::type::Pointer>();
auto* root_ref_ty = root->Type()->As<core::type::Reference>();
TINT_ASSERT(root_ptr_ty || root_ref_ty);
const core::type::Type* root_store_type;
if (root_ptr_ty) {
root_store_type = root_ptr_ty->StoreType();
} else {
root_store_type = root_ref_ty->StoreType();
}
if (root_store_type != arg_store_type &&
IsValidationEnabled(param->Declaration()->attributes,
ast::DisabledValidation::kIgnoreInvalidPointerArgument)) {
AddError(
"arguments of pointer type must not point to a subset of the originating "
"variable",
arg_expr->source);
return false;
}
}
}
if (call->Type()->Is<core::type::Void>()) {
bool is_call_statement = false;
if (auto* call_stmt = As<ast::CallStatement>(call->Stmt()->Declaration())) {
if (call_stmt->expr == call->Declaration()) {
is_call_statement = true;
}
}
if (!is_call_statement) {
// https://gpuweb.github.io/gpuweb/wgsl/#function-call-expr
// If the called function does not return a value, a function call
// statement should be used instead.
AddError("function '" + name + "' does not return a value", decl->source);
return false;
}
}
return true;
}
bool Validator::StructureInitializer(const ast::CallExpression* ctor,
const core::type::Struct* struct_type) const {
if (!struct_type->IsConstructible()) {
AddError("structure constructor has non-constructible type", ctor->source);
return false;
}
if (ctor->args.Length() > 0) {
if (ctor->args.Length() != struct_type->Members().Length()) {
std::string fm = ctor->args.Length() < struct_type->Members().Length() ? "few" : "many";
AddError("structure constructor has too " + fm + " inputs: expected " +
std::to_string(struct_type->Members().Length()) + ", found " +
std::to_string(ctor->args.Length()),
ctor->source);
return false;
}
for (auto* member : struct_type->Members()) {
auto* value = ctor->args[member->Index()];
auto* value_ty = sem_.TypeOf(value);
if (member->Type() != value_ty->UnwrapRef()) {
AddError(
"type in structure constructor does not match struct member type: expected '" +
sem_.TypeNameOf(member->Type()) + "', found '" + sem_.TypeNameOf(value_ty) +
"'",
value->source);
return false;
}
}
}
return true;
}
bool Validator::ArrayConstructor(const ast::CallExpression* ctor,
const core::type::Array* array_type) const {
auto& values = ctor->args;
auto* elem_ty = array_type->ElemType();
for (auto* value : values) {
auto* value_ty = sem_.TypeOf(value)->UnwrapRef();
if (core::type::Type::ConversionRank(value_ty, elem_ty) ==
core::type::Type::kNoConversion) {
AddError("'" + sem_.TypeNameOf(value_ty) +
"' cannot be used to construct an array of '" + sem_.TypeNameOf(elem_ty) +
"'",
value->source);
return false;
}
}
auto* c = array_type->Count();
if (c->Is<core::type::RuntimeArrayCount>()) {
AddError("cannot construct a runtime-sized array", ctor->source);
return false;
}
if (c->IsAnyOf<sem::NamedOverrideArrayCount, sem::UnnamedOverrideArrayCount>()) {
AddError("cannot construct an array that has an override-expression count", ctor->source);
return false;
}
if (!elem_ty->IsConstructible()) {
AddError("array constructor has non-constructible element type", ctor->source);
return false;
}
if (TINT_UNLIKELY(!c->Is<core::type::ConstantArrayCount>())) {
TINT_ICE() << "Invalid ArrayCount found";
return false;
}
const auto count = c->As<core::type::ConstantArrayCount>()->value;
if (!values.IsEmpty() && (values.Length() != count)) {
std::string fm = values.Length() < count ? "few" : "many";
AddError("array constructor has too " + fm + " elements: expected " +
std::to_string(count) + ", found " + std::to_string(values.Length()),
ctor->source);
return false;
}
return true;
}
bool Validator::Vector(const core::type::Type* el_ty, const Source& source) const {
if (!el_ty->Is<core::type::Scalar>()) {
AddError("vector element type must be 'bool', 'f32', 'f16', 'i32' or 'u32'", source);
return false;
}
return true;
}
bool Validator::Matrix(const core::type::Type* el_ty, const Source& source) const {
if (!el_ty->is_float_scalar()) {
AddError("matrix element type must be 'f32' or 'f16'", source);
return false;
}
return true;
}
bool Validator::PipelineStages(VectorRef<sem::Function*> entry_points) const {
auto backtrace = [&](const sem::Function* func, const sem::Function* entry_point) {
if (func != entry_point) {
TraverseCallChain(entry_point, func, [&](const sem::Function* f) {
AddNote("called by function '" + f->Declaration()->name->symbol.Name() + "'",
f->Declaration()->source);
});
AddNote(
"called by entry point '" + entry_point->Declaration()->name->symbol.Name() + "'",
entry_point->Declaration()->source);
}
};
auto check_workgroup_storage = [&](const sem::Function* func,
const sem::Function* entry_point) {
auto stage = entry_point->Declaration()->PipelineStage();
if (stage != ast::PipelineStage::kCompute) {
for (auto* var : func->DirectlyReferencedGlobals()) {
if (var->AddressSpace() == core::AddressSpace::kWorkgroup) {
StringStream stage_name;
stage_name << stage;
for (auto* user : var->Users()) {
if (func == user->Stmt()->Function()) {
AddError("workgroup memory cannot be used by " + stage_name.str() +
" pipeline stage",
user->Declaration()->source);
break;
}
}
AddNote("variable is declared here", var->Declaration()->source);
backtrace(func, entry_point);
return false;
}
}
}
return true;
};
auto check_builtin_calls = [&](const sem::Function* func, const sem::Function* entry_point) {
auto stage = entry_point->Declaration()->PipelineStage();
for (auto* builtin : func->DirectlyCalledBuiltins()) {
if (!builtin->SupportedStages().Contains(stage)) {
auto* call = func->FindDirectCallTo(builtin);
StringStream err;
err << "built-in cannot be used by " << stage << " pipeline stage";
AddError(err.str(),
call ? call->Declaration()->source : func->Declaration()->source);
backtrace(func, entry_point);
return false;
}
}
return true;
};
auto check_no_discards = [&](const sem::Function* func, const sem::Function* entry_point) {
if (auto* discard = func->DiscardStatement()) {
auto stage = entry_point->Declaration()->PipelineStage();
StringStream err;
err << "discard statement cannot be used in " << stage << " pipeline stage";
AddError(err.str(), discard->Declaration()->source);
backtrace(func, entry_point);
return false;
}
return true;
};
auto check_func = [&](const sem::Function* func, const sem::Function* entry_point) {
if (!check_workgroup_storage(func, entry_point)) {
return false;
}
if (!check_builtin_calls(func, entry_point)) {
return false;
}
if (entry_point->Declaration()->PipelineStage() != ast::PipelineStage::kFragment) {
if (!check_no_discards(func, entry_point)) {
return false;
}
}
return true;
};
for (auto* entry_point : entry_points) {
if (!check_func(entry_point, entry_point)) {
return false;
}
for (auto* func : entry_point->TransitivelyCalledFunctions()) {
if (!check_func(func, entry_point)) {
return false;
}
}
}
return true;
}
bool Validator::PushConstants(VectorRef<sem::Function*> entry_points) const {
for (auto* entry_point : entry_points) {
// State checked and modified by check_push_constant so that it remembers previously seen
// push_constant variables for an entry-point.
const sem::Variable* push_constant_var = nullptr;
const sem::Function* push_constant_func = nullptr;
auto check_push_constant = [&](const sem::Function* func, const sem::Function* ep) {
for (auto* var : func->DirectlyReferencedGlobals()) {
if (var->AddressSpace() != core::AddressSpace::kPushConstant ||
var == push_constant_var) {
continue;
}
if (push_constant_var == nullptr) {
push_constant_var = var;
push_constant_func = func;
continue;
}
AddError("entry point '" + ep->Declaration()->name->symbol.Name() +
"' uses two different 'push_constant' variables.",
ep->Declaration()->source);
AddNote("first 'push_constant' variable declaration is here",
var->Declaration()->source);
if (func != ep) {
TraverseCallChain(ep, func, [&](const sem::Function* f) {
AddNote(
"called by function '" + f->Declaration()->name->symbol.Name() + "'",
f->Declaration()->source);
});
AddNote(
"called by entry point '" + ep->Declaration()->name->symbol.Name() + "'",
ep->Declaration()->source);
}
AddNote("second 'push_constant' variable declaration is here",
push_constant_var->Declaration()->source);
if (push_constant_func != ep) {
TraverseCallChain(ep, push_constant_func, [&](const sem::Function* f) {
AddNote(
"called by function '" + f->Declaration()->name->symbol.Name() + "'",
f->Declaration()->source);
});
AddNote(
"called by entry point '" + ep->Declaration()->name->symbol.Name() + "'",
ep->Declaration()->source);
}
return false;
}
return true;
};
if (!check_push_constant(entry_point, entry_point)) {
return false;
}
for (auto* func : entry_point->TransitivelyCalledFunctions()) {
if (!check_push_constant(func, entry_point)) {
return false;
}
}
}
return true;
}
bool Validator::Array(const core::type::Array* arr, const Source& el_source) const {
auto* el_ty = arr->ElemType();
if (!IsPlain(el_ty)) {
AddError(sem_.TypeNameOf(el_ty) + " cannot be used as an element type of an array",
el_source);
return false;
}
if (!IsFixedFootprint(el_ty)) {
AddError("an array element type cannot contain a runtime-sized array", el_source);
return false;
}
if (IsArrayWithOverrideCount(el_ty)) {
RaiseArrayWithOverrideCountError(el_source);
return false;
}
return true;
}
bool Validator::ArrayStrideAttribute(const ast::StrideAttribute* attr,
uint32_t el_size,
uint32_t el_align) const {
auto stride = attr->stride;
bool is_valid_stride = (stride >= el_size) && (stride >= el_align) && (stride % el_align == 0);
if (!is_valid_stride) {
// https://gpuweb.github.io/gpuweb/wgsl/#array-layout-rules
// Arrays decorated with the stride attribute must have a stride that is
// at least the size of the element type, and be a multiple of the
// element type's alignment value.
AddError(
"arrays decorated with the stride attribute must have a stride that is at least the "
"size of the element type, and be a multiple of the element type's alignment value",
attr->source);
return false;
}
return true;
}
bool Validator::Alias(const ast::Alias*) const {
return true;
}
bool Validator::Structure(const sem::Struct* str, ast::PipelineStage stage) const {
if (str->Members().IsEmpty()) {
AddError("structures must have at least one member", str->Declaration()->source);
return false;
}
auto has_index = false;
Hashset<std::pair<uint32_t, uint32_t>, 8> locationsAndIndexes;
for (auto* member : str->Members()) {
if (auto* r = member->Type()->As<core::type::Array>()) {
if (r->Count()->Is<core::type::RuntimeArrayCount>()) {
if (member != str->Members().Back()) {
AddError("runtime arrays may only appear as the last member of a struct",
member->Declaration()->source);
return false;
}
}
if (IsArrayWithOverrideCount(member->Type())) {
RaiseArrayWithOverrideCountError(member->Declaration()->type->source);
return false;
}
} else if (!IsFixedFootprint(member->Type())) {
AddError(
"a struct that contains a runtime array cannot be nested inside another struct",
member->Declaration()->source);
return false;
}
auto has_position = false;
const ast::IndexAttribute* index_attribute = nullptr;
const ast::LocationAttribute* location_attribute = nullptr;
const ast::InvariantAttribute* invariant_attribute = nullptr;
const ast::InterpolateAttribute* interpolate_attribute = nullptr;
for (auto* attr : member->Declaration()->attributes) {
bool ok = Switch(
attr, //
[&](const ast::InvariantAttribute* invariant) {
invariant_attribute = invariant;
return true;
},
[&](const ast::LocationAttribute* location) {
location_attribute = location;
TINT_ASSERT(member->Attributes().location.has_value());
if (!LocationAttribute(location, member->Type(), stage,
member->Declaration()->source)) {
return false;
}
return true;
},
[&](const ast::IndexAttribute* index) {
index_attribute = index;
has_index = true;
return IndexAttribute(index, stage);
},
[&](const ast::BuiltinAttribute* builtin_attr) {
if (!BuiltinAttribute(builtin_attr, member->Type(), stage,
/* is_input */ false)) {
return false;
}
auto builtin = sem_.Get(builtin_attr)->Value();
if (builtin == core::BuiltinValue::kPosition) {
has_position = true;
}
return true;
},
[&](const ast::InterpolateAttribute* interpolate) {
interpolate_attribute = interpolate;
if (!InterpolateAttribute(interpolate, member->Type())) {
return false;
}
return true;
},
[&](const ast::StructMemberSizeAttribute*) {
if (!member->Type()->HasCreationFixedFootprint()) {
AddError(
"@size can only be applied to members where the member's type size "
"can be fully determined at shader creation time",
attr->source);
return false;
}
return true;
},
[&](Default) { return true; });
if (!ok) {
return false;
}
}
if (invariant_attribute && !has_position) {
AddError("invariant attribute must only be applied to a position builtin",
invariant_attribute->source);
return false;
}
if (index_attribute) {
if (!location_attribute) {
AddError("index attribute must only be used with @location",
index_attribute->source);
return false;
}
// Because HLSL specifies dual source blending targets with SV_Target0 and 1, we should
// restrict targets with index attributes to location 0 for easy translation in the
// backend writers.
if (member->Attributes().location.value() != 0) {
AddError("index attribute must only be used with @location(0)",
index_attribute->source);
return false;
}
}
if (interpolate_attribute && !location_attribute) {
AddError("interpolate attribute must only be used with @location",
interpolate_attribute->source);
return false;
}
// Ensure all locations and index pairs are unique
if (location_attribute) {
uint32_t index = 0xffffffff;
if (index_attribute) {
index = member->Attributes().index.value();
}
uint32_t location = member->Attributes().location.value();
if (has_index && location != 0) {
StringStream err;
err << "Multiple render targets are not allowed when using dual source blending. "
"The output @location("
<< location << ") is not allowed as a render target.";
AddError(err.str(), location_attribute->source);
return false;
}
std::pair<uint32_t, uint32_t> locationAndIndex(location, index);
if (!locationsAndIndexes.Add(locationAndIndex)) {
StringStream err;
if (!index_attribute) {
err << "@location(" << location << ") appears multiple times";
AddError(err.str(), location_attribute->source);
} else {
err << "@location(" << location << ") @index(" << index
<< ") appears multiple times";
AddError(err.str(), index_attribute->source);
}
return false;
}
}
}
return true;
}
bool Validator::LocationAttribute(const ast::LocationAttribute* loc_attr,
const core::type::Type* type,
ast::PipelineStage stage,
const Source& source,
const bool is_input) const {
std::string inputs_or_output = is_input ? "inputs" : "output";
if (stage == ast::PipelineStage::kCompute) {
AddError("@" + loc_attr->Name() + " is not valid for compute shader " + inputs_or_output,
loc_attr->source);
return false;
}
if (!type->is_numeric_scalar_or_vector()) {
std::string invalid_type = sem_.TypeNameOf(type);
AddError("cannot apply @location to declaration of type '" + invalid_type + "'", source);
AddNote(
"@location must only be applied to declarations of numeric scalar or numeric vector "
"type",
loc_attr->source);
return false;
}
return true;
}
bool Validator::IndexAttribute(const ast::IndexAttribute* index_attr,
ast::PipelineStage stage) const {
if (!enabled_extensions_.Contains(core::Extension::kChromiumInternalDualSourceBlending)) {
AddError(
"use of '@index' attribute requires enabling extension "
"'chromium_internal_dual_source_blending'",
index_attr->source);
return false;
}
if (stage == ast::PipelineStage::kCompute) {
AddError("@" + index_attr->Name() + " is not valid for compute shader output",
index_attr->source);
return false;
}
if (stage == ast::PipelineStage::kVertex) {
AddError("@" + index_attr->Name() + " is not valid for vertex shader output",
index_attr->source);
return false;
}
return true;
}
bool Validator::Return(const ast::ReturnStatement* ret,
const core::type::Type* func_type,
const core::type::Type* ret_type,
sem::Statement* current_statement) const {
if (func_type->UnwrapRef() != ret_type) {
AddError("return statement type must match its function return type, returned '" +
sem_.TypeNameOf(ret_type) + "', expected '" + sem_.TypeNameOf(func_type) + "'",
ret->source);
return false;
}
auto* sem = sem_.Get(ret);
if (auto* continuing = ClosestContinuing(/*stop_at_loop*/ false, current_statement)) {
AddError("continuing blocks must not contain a return statement", ret->source);
if (continuing != sem->Declaration() && continuing != sem->Parent()->Declaration()) {
AddNote("see continuing block here", continuing->source);
}
return false;
}
return true;
}
bool Validator::SwitchStatement(const ast::SwitchStatement* s) {
if (s->body.Length() > kMaxSwitchCaseSelectors) {
AddError("switch statement has " + std::to_string(s->body.Length()) +
" case selectors, max is " + std::to_string(kMaxSwitchCaseSelectors),
s->source);
return false;
}
auto* cond_ty = sem_.TypeOf(s->condition);
if (!cond_ty->is_integer_scalar()) {
AddError("switch statement selector expression must be of a scalar integer type",
s->condition->source);
return false;
}
const sem::CaseSelector* default_selector = nullptr;
Hashmap<int64_t, Source, 4> selectors;
for (auto* case_stmt : s->body) {
auto* case_sem = sem_.Get<sem::CaseStatement>(case_stmt);
for (auto* selector : case_sem->Selectors()) {
if (selector->IsDefault()) {
if (default_selector != nullptr) {
// More than one default clause
AddError("switch statement must have exactly one default clause",
selector->Declaration()->source);
AddNote("previous default case", default_selector->Declaration()->source);
return false;
}
default_selector = selector;
continue;
}
auto* decl_ty = selector->Value()->Type();
if (cond_ty != decl_ty) {
AddError(
"the case selector values must have the same type as the selector expression.",
selector->Declaration()->source);
return false;
}
auto value = selector->Value()->ValueAs<u32>();
if (auto added = selectors.Add(value, selector->Declaration()->source); !added) {
AddError("duplicate switch case '" +
(decl_ty->IsAnyOf<core::type::I32, core::type::AbstractNumeric>()
? std::to_string(i32(value))
: std::to_string(value)) +
"'",
selector->Declaration()->source);
AddNote("previous case declared here", *added.value);
return false;
}
}
}
if (default_selector == nullptr) {
// No default clause
AddError("switch statement must have a default clause", s->source);
return false;
}
return true;
}
bool Validator::Assignment(const ast::Statement* a, const core::type::Type* rhs_ty) const {
const ast::Expression* lhs;
const ast::Expression* rhs;
if (auto* assign = a->As<ast::AssignmentStatement>()) {
lhs = assign->lhs;
rhs = assign->rhs;
} else if (auto* compound = a->As<ast::CompoundAssignmentStatement>()) {
lhs = compound->lhs;
rhs = compound->rhs;
} else {
TINT_ICE() << "invalid assignment statement";
return false;
}
if (lhs->Is<ast::PhonyExpression>()) {
// https://www.w3.org/TR/WGSL/#phony-assignment-section
auto* ty = rhs_ty->UnwrapRef();
if (!ty->IsConstructible() &&
!ty->IsAnyOf<core::type::Pointer, core::type::Texture, core::type::Sampler,
core::type::AbstractNumeric>()) {
AddError("cannot assign '" + sem_.TypeNameOf(rhs_ty) +
"' to '_'. '_' can only be assigned a constructible, pointer, texture or "
"sampler type",
rhs->source);
return false;
}
return true; // RHS can be anything.
}
// https://gpuweb.github.io/gpuweb/wgsl/#assignment-statement
auto const* lhs_sem = sem_.GetVal(lhs);
auto const* lhs_ty = lhs_sem->Type();
auto* lhs_ref = lhs_ty->As<core::type::Reference>();
if (!lhs_ref) {
// LHS is not a reference, so it has no storage.
AddError("cannot assign to " + sem_.Describe(lhs_sem), lhs->source);
auto* expr = lhs;
while (expr) {
expr = Switch(
expr, //
[&](const ast::AccessorExpression* e) { return e->object; },
[&](const ast::IdentifierExpression* i) {
if (auto user = sem_.Get<sem::VariableUser>(i)) {
Switch(
user->Variable()->Declaration(), //
[&](const ast::Let* v) {
AddNote("'let' variables are immutable",
user->Declaration()->source);
sem_.NoteDeclarationSource(v);
},
[&](const ast::Const* v) {
AddNote("'const' variables are immutable",
user->Declaration()->source);
sem_.NoteDeclarationSource(v);
},
[&](const ast::Override* v) {
AddNote("'override' variables are immutable",
user->Declaration()->source);
sem_.NoteDeclarationSource(v);
},
[&](const ast::Parameter* v) {
AddNote("parameters are immutable", user->Declaration()->source);
sem_.NoteDeclarationSource(v);
});
}
return nullptr;
});
}
return false;
}
auto* storage_ty = lhs_ref->StoreType();
auto* value_type = rhs_ty->UnwrapRef(); // Implicit load of RHS
// Value type has to match storage type
if (storage_ty != value_type) {
AddError(
"cannot assign '" + sem_.TypeNameOf(rhs_ty) + "' to '" + sem_.TypeNameOf(lhs_ty) + "'",
a->source);
return false;
}
if (!storage_ty->IsConstructible()) {
AddError("storage type of assignment must be constructible", a->source);
return false;
}
if (lhs_ref->Access() == core::Access::kRead) {
AddError("cannot store into a read-only type '" + sem_.RawTypeNameOf(lhs_ty) + "'",
a->source);
return false;
}
return true;
}
bool Validator::IncrementDecrementStatement(const ast::IncrementDecrementStatement* inc) const {
const ast::Expression* lhs = inc->lhs;
// https://gpuweb.github.io/gpuweb/wgsl/#increment-decrement
if (auto* var_user = sem_.Get<sem::VariableUser>(lhs)) {
auto* v = var_user->Variable()->Declaration();
const char* err = Switch(
v, //
[&](const ast::Parameter*) { return "cannot modify function parameter"; },
[&](const ast::Let*) { return "cannot modify 'let'"; },
[&](const ast::Override*) { return "cannot modify 'override'"; });
if (err) {
AddError(err, lhs->source);
AddNote("'" + v->name->symbol.Name() + "' is declared here:", v->source);
return false;
}
}
auto const* lhs_ty = sem_.TypeOf(lhs);
auto* lhs_ref = lhs_ty->As<core::type::Reference>();
if (!lhs_ref) {
// LHS is not a reference, so it has no storage.
AddError("cannot modify value of type '" + sem_.TypeNameOf(lhs_ty) + "'", lhs->source);
return false;
}
if (!lhs_ref->StoreType()->is_integer_scalar()) {
const std::string kind = inc->increment ? "increment" : "decrement";
AddError(kind + " statement can only be applied to an integer scalar", lhs->source);
return false;
}
if (lhs_ref->Access() == core::Access::kRead) {
AddError("cannot modify read-only type '" + sem_.RawTypeNameOf(lhs_ty) + "'", inc->source);
return false;
}
return true;
}
bool Validator::NoDuplicateAttributes(VectorRef<const ast::Attribute*> attributes) const {
Hashmap<const tint::TypeInfo*, Source, 8> seen;
tint::Vector<const ast::DiagnosticControl*, 8> diagnostic_controls;
for (auto* d : attributes) {
if (auto* diag = d->As<ast::DiagnosticAttribute>()) {
// Allow duplicate diagnostic attributes, and check for conflicts later.
diagnostic_controls.Push(&diag->control);
} else {
auto added = seen.Add(&d->TypeInfo(), d->source);
if (!added && !d->Is<ast::InternalAttribute>()) {
AddError("duplicate " + d->Name() + " attribute", d->source);
AddNote("first attribute declared here", *added.value);
return false;
}
}
}
return DiagnosticControls(diagnostic_controls, "attribute");
}
bool Validator::DiagnosticControls(VectorRef<const ast::DiagnosticControl*> controls,
const char* use) const {
// Make sure that no two diagnostic controls conflict.
// They conflict if the rule name is the same and the severity is different.
Hashmap<std::pair<Symbol, Symbol>, const ast::DiagnosticControl*, 8> diagnostics;
for (auto* dc : controls) {
auto category = dc->rule_name->category ? dc->rule_name->category->symbol : Symbol();
auto name = dc->rule_name->name->symbol;
auto diag_added = diagnostics.Add(std::make_pair(category, name), dc);
if (!diag_added && (*diag_added.value)->severity != dc->severity) {
{
StringStream ss;
ss << "conflicting diagnostic " << use;
AddError(ss.str(), dc->rule_name->source);
}
{
StringStream ss;
ss << "severity of '" << dc->rule_name->String() << "' set to '" << dc->severity
<< "' here";
AddNote(ss.str(), (*diag_added.value)->rule_name->source);
}
return false;
}
}
return true;
}
bool Validator::IsValidationDisabled(VectorRef<const ast::Attribute*> attributes,
ast::DisabledValidation validation) const {
for (auto* attribute : attributes) {
if (auto* dv = attribute->As<ast::DisableValidationAttribute>()) {
if (dv->validation == validation) {
return true;
}
}
}
return false;
}
bool Validator::IsValidationEnabled(VectorRef<const ast::Attribute*> attributes,
ast::DisabledValidation validation) const {
return !IsValidationDisabled(attributes, validation);
}
bool Validator::IsArrayWithOverrideCount(const core::type::Type* ty) const {
if (auto* arr = ty->UnwrapRef()->As<core::type::Array>()) {
if (arr->Count()->IsAnyOf<sem::NamedOverrideArrayCount, sem::UnnamedOverrideArrayCount>()) {
return true;
}
}
return false;
}
void Validator::RaiseArrayWithOverrideCountError(const Source& source) const {
AddError(
"array with an 'override' element count can only be used as the store type of a "
"'var<workgroup>'",
source);
}
std::string Validator::VectorPretty(uint32_t size, const core::type::Type* element_type) const {
core::type::Vector vec_type(element_type, size);
return vec_type.FriendlyName();
}
bool Validator::CheckTypeAccessAddressSpace(const core::type::Type* store_ty,
core::Access access,
core::AddressSpace address_space,
VectorRef<const tint::ast::Attribute*> attributes,
const Source& source) const {
if (!AddressSpaceLayout(store_ty, address_space, source)) {
return false;
}
if (address_space == core::AddressSpace::kPushConstant &&
!enabled_extensions_.Contains(core::Extension::kChromiumExperimentalPushConstant) &&
IsValidationEnabled(attributes, ast::DisabledValidation::kIgnoreAddressSpace)) {
AddError(
"use of variable address space 'push_constant' requires enabling extension "
"'chromium_experimental_push_constant'",
source);
return false;
}
if (address_space == core::AddressSpace::kStorage && access == core::Access::kWrite) {
// The access mode for the storage address space can only be 'read' or 'read_write'.
AddError("access mode 'write' is not valid for the 'storage' address space", source);
return false;
}
auto atomic_error = [&]() -> const char* {
if (address_space != core::AddressSpace::kStorage &&
address_space != core::AddressSpace::kWorkgroup) {
return "atomic variables must have <storage> or <workgroup> address space";
}
if (address_space == core::AddressSpace::kStorage && access != core::Access::kReadWrite) {
return "atomic variables in <storage> address space must have read_write access "
"mode";
}
return nullptr;
};
auto check_sub_atomics = [&] {
if (auto atomic_use = atomic_composite_info_.Get(store_ty)) {
if (auto* err = atomic_error()) {
AddError(err, source);
AddNote("atomic sub-type of '" + sem_.TypeNameOf(store_ty) + "' is declared here",
**atomic_use);
return false;
}
}
return true;
};
return Switch(
store_ty, //
[&](const core::type::Atomic*) {
if (auto* err = atomic_error()) {
AddError(err, source);
return false;
}
return true;
},
[&](const core::type::Struct*) { return check_sub_atomics(); }, //
[&](const core::type::Array*) { return check_sub_atomics(); }, //
[&](Default) { return true; });
}
} // namespace tint::resolver