Add support for binary arithmetic expressions with mixed scalar and vector operands

Bug: tint:376
Change-Id: I2994ff7394efa903050b470a850b41628d5b775c
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/52324
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
This commit is contained in:
Antonio Maiorano 2021-05-27 21:07:56 +00:00 committed by Tint LUCI CQ
parent d5f4ea22f0
commit eaed2b6ce2
10 changed files with 909 additions and 18 deletions

View File

@ -2097,23 +2097,35 @@ bool Resolver::Binary(ast::BinaryExpression* expr) {
SetType(expr, lhs_type);
return true;
}
}
// Binary arithmetic expressions with mixed scalar, vector, and matrix
// operands
if (expr->IsMultiply()) {
// Multiplication of a vector and a scalar
if (lhs_type->Is<F32>() && rhs_vec_elem_type &&
rhs_vec_elem_type->Is<F32>()) {
SetType(expr, rhs_type);
return true;
}
if (lhs_vec_elem_type && lhs_vec_elem_type->Is<F32>() &&
rhs_type->Is<F32>()) {
// Binary arithmetic expressions with mixed scalar and vector operands
if (lhs_vec_elem_type && (lhs_vec_elem_type == rhs_type)) {
if (expr->IsModulo()) {
if (rhs_type->is_integer_scalar()) {
SetType(expr, lhs_type);
return true;
}
} else if (rhs_type->is_numeric_scalar()) {
SetType(expr, lhs_type);
return true;
}
}
if (rhs_vec_elem_type && (rhs_vec_elem_type == lhs_type)) {
if (expr->IsModulo()) {
if (lhs_type->is_integer_scalar()) {
SetType(expr, rhs_type);
return true;
}
} else if (lhs_type->is_numeric_scalar()) {
SetType(expr, rhs_type);
return true;
}
}
}
// Matrix arithmetic
// TODO(amaiorano): matrix-matrix addition and subtraction
if (expr->IsMultiply()) {
auto* lhs_mat = lhs_type->As<Matrix>();
auto* lhs_mat_elem_type = lhs_mat ? lhs_mat->type() : nullptr;
auto* rhs_mat = rhs_type->As<Matrix>();

View File

@ -1298,16 +1298,52 @@ static constexpr Params all_valid_cases[] = {
Params{Op::kDivide, ast_vec3<f32>, ast_vec3<f32>, sem_vec3<sem_f32>},
Params{Op::kModulo, ast_vec3<f32>, ast_vec3<f32>, sem_vec3<sem_f32>},
// Binary arithmetic expressions with mixed scalar, vector, and matrix
// operands
Params{Op::kMultiply, ast_vec3<f32>, ast_f32, sem_vec3<sem_f32>},
Params{Op::kMultiply, ast_f32, ast_vec3<f32>, sem_vec3<sem_f32>},
// Binary arithmetic expressions with mixed scalar and vector operands
Params{Op::kAdd, ast_vec3<i32>, ast_i32, sem_vec3<sem_i32>},
Params{Op::kSubtract, ast_vec3<i32>, ast_i32, sem_vec3<sem_i32>},
Params{Op::kMultiply, ast_vec3<i32>, ast_i32, sem_vec3<sem_i32>},
Params{Op::kDivide, ast_vec3<i32>, ast_i32, sem_vec3<sem_i32>},
Params{Op::kModulo, ast_vec3<i32>, ast_i32, sem_vec3<sem_i32>},
Params{Op::kAdd, ast_i32, ast_vec3<i32>, sem_vec3<sem_i32>},
Params{Op::kSubtract, ast_i32, ast_vec3<i32>, sem_vec3<sem_i32>},
Params{Op::kMultiply, ast_i32, ast_vec3<i32>, sem_vec3<sem_i32>},
Params{Op::kDivide, ast_i32, ast_vec3<i32>, sem_vec3<sem_i32>},
Params{Op::kModulo, ast_i32, ast_vec3<i32>, sem_vec3<sem_i32>},
Params{Op::kAdd, ast_vec3<u32>, ast_u32, sem_vec3<sem_u32>},
Params{Op::kSubtract, ast_vec3<u32>, ast_u32, sem_vec3<sem_u32>},
Params{Op::kMultiply, ast_vec3<u32>, ast_u32, sem_vec3<sem_u32>},
Params{Op::kDivide, ast_vec3<u32>, ast_u32, sem_vec3<sem_u32>},
Params{Op::kModulo, ast_vec3<u32>, ast_u32, sem_vec3<sem_u32>},
Params{Op::kAdd, ast_u32, ast_vec3<u32>, sem_vec3<sem_u32>},
Params{Op::kSubtract, ast_u32, ast_vec3<u32>, sem_vec3<sem_u32>},
Params{Op::kMultiply, ast_u32, ast_vec3<u32>, sem_vec3<sem_u32>},
Params{Op::kDivide, ast_u32, ast_vec3<u32>, sem_vec3<sem_u32>},
Params{Op::kModulo, ast_u32, ast_vec3<u32>, sem_vec3<sem_u32>},
Params{Op::kAdd, ast_vec3<f32>, ast_f32, sem_vec3<sem_f32>},
Params{Op::kSubtract, ast_vec3<f32>, ast_f32, sem_vec3<sem_f32>},
Params{Op::kMultiply, ast_vec3<f32>, ast_f32, sem_vec3<sem_f32>},
Params{Op::kDivide, ast_vec3<f32>, ast_f32, sem_vec3<sem_f32>},
// NOTE: no kModulo for ast_vec3<f32>, ast_f32
// Params{Op::kModulo, ast_vec3<f32>, ast_f32, sem_vec3<sem_f32>},
Params{Op::kAdd, ast_f32, ast_vec3<f32>, sem_vec3<sem_f32>},
Params{Op::kSubtract, ast_f32, ast_vec3<f32>, sem_vec3<sem_f32>},
Params{Op::kMultiply, ast_f32, ast_vec3<f32>, sem_vec3<sem_f32>},
Params{Op::kDivide, ast_f32, ast_vec3<f32>, sem_vec3<sem_f32>},
// NOTE: no kModulo for ast_f32, ast_vec3<f32>
// Params{Op::kModulo, ast_f32, ast_vec3<f32>, sem_vec3<sem_f32>},
// Matrix arithmetic
Params{Op::kMultiply, ast_mat3x3<f32>, ast_f32, sem_mat3x3<sem_f32>},
Params{Op::kMultiply, ast_f32, ast_mat3x3<f32>, sem_mat3x3<sem_f32>},
Params{Op::kMultiply, ast_vec3<f32>, ast_mat3x3<f32>, sem_vec3<sem_f32>},
Params{Op::kMultiply, ast_mat3x3<f32>, ast_vec3<f32>, sem_vec3<sem_f32>},
// TODO(amaiorano): add mat+mat and mat-mat
Params{Op::kMultiply, ast_mat3x3<f32>, ast_mat3x3<f32>,
sem_mat3x3<sem_f32>},

View File

@ -1719,6 +1719,31 @@ uint32_t Builder::GenerateShortCircuitBinaryExpression(
return result_id;
}
uint32_t Builder::GenerateSplat(uint32_t scalar_id, const sem::Type* vec_type) {
// Create a new vector to splat scalar into
auto splat_vector = result_op();
auto* splat_vector_type =
builder_.create<sem::Pointer>(vec_type, ast::StorageClass::kFunction);
push_function_var(
{Operand::Int(GenerateTypeIfNeeded(splat_vector_type)), splat_vector,
Operand::Int(ConvertStorageClass(ast::StorageClass::kFunction)),
Operand::Int(GenerateConstantNullIfNeeded(vec_type))});
// Splat scalar into vector
auto splat_result = result_op();
OperandList ops;
ops.push_back(Operand::Int(GenerateTypeIfNeeded(vec_type)));
ops.push_back(splat_result);
for (size_t i = 0; i < vec_type->As<sem::Vector>()->size(); ++i) {
ops.push_back(Operand::Int(scalar_id));
}
if (!push_function_inst(spv::Op::OpCompositeConstruct, ops)) {
return 0;
}
return splat_result.to_i();
}
uint32_t Builder::GenerateBinaryExpression(ast::BinaryExpression* expr) {
// There is special logic for short circuiting operators.
if (expr->IsLogicalAnd() || expr->IsLogicalOr()) {
@ -1749,6 +1774,33 @@ uint32_t Builder::GenerateBinaryExpression(ast::BinaryExpression* expr) {
// should have been rejected by validation.
auto* lhs_type = TypeOf(expr->lhs())->UnwrapRef();
auto* rhs_type = TypeOf(expr->rhs())->UnwrapRef();
// For vector-scalar arithmetic operations, splat scalar into a vector. We
// skip this for multiply as we can use OpVectorTimesScalar.
const bool is_float_scalar_vector_multiply =
expr->IsMultiply() &&
((lhs_type->is_float_scalar() && rhs_type->is_float_vector()) ||
(lhs_type->is_float_vector() && rhs_type->is_float_scalar()));
if (expr->IsArithmetic() && !is_float_scalar_vector_multiply) {
if (lhs_type->Is<sem::Vector>() && rhs_type->is_numeric_scalar()) {
uint32_t splat_vector_id = GenerateSplat(rhs_id, lhs_type);
if (splat_vector_id == 0) {
return 0;
}
rhs_id = splat_vector_id;
rhs_type = lhs_type;
} else if (lhs_type->is_numeric_scalar() && rhs_type->Is<sem::Vector>()) {
uint32_t splat_vector_id = GenerateSplat(lhs_id, rhs_type);
if (splat_vector_id == 0) {
return 0;
}
lhs_id = splat_vector_id;
lhs_type = rhs_type;
}
}
bool lhs_is_float_or_vec = lhs_type->is_float_scalar_or_vector();
bool lhs_is_unsigned = lhs_type->is_unsigned_scalar_or_vector();

View File

@ -473,6 +473,13 @@ class Builder {
/// @returns true if the vector was successfully generated
bool GenerateVectorType(const sem::Vector* vec, const Operand& result);
/// Generates instructions to splat `scalar_id` into a vector of type
/// `vec_type`
/// @param scalar_id scalar to splat
/// @param vec_type type of vector
/// @returns id of the new vector
uint32_t GenerateSplat(uint32_t scalar_id, const sem::Type* vec_type);
/// Converts AST image format to SPIR-V and pushes an appropriate capability.
/// @param format AST image format type
/// @returns SPIR-V image format type

View File

@ -863,6 +863,222 @@ OpBranch %9
)");
}
enum class Type { f32, i32, u32 };
ast::Expression* MakeVectorExpr(ProgramBuilder* builder, Type type) {
switch (type) {
case Type::f32:
return builder->vec3<ProgramBuilder::f32>(1.f, 1.f, 1.f);
case Type::i32:
return builder->vec3<ProgramBuilder::i32>(1, 1, 1);
case Type::u32:
return builder->vec3<ProgramBuilder::u32>(1u, 1u, 1u);
}
return nullptr;
}
ast::Expression* MakeScalarExpr(ProgramBuilder* builder, Type type) {
switch (type) {
case Type::f32:
return builder->Expr(1.f);
case Type::i32:
return builder->Expr(1);
case Type::u32:
return builder->Expr(1u);
}
return nullptr;
}
std::string OpTypeDecl(Type type) {
switch (type) {
case Type::f32:
return "OpTypeFloat 32";
case Type::i32:
return "OpTypeInt 32 1";
case Type::u32:
return "OpTypeInt 32 0";
}
return {};
}
struct Param {
Type type;
ast::BinaryOp op;
std::string name;
};
using MixedBinaryArithTest = TestParamHelper<Param>;
TEST_P(MixedBinaryArithTest, VectorScalar) {
auto& param = GetParam();
ast::Expression* lhs = MakeVectorExpr(this, param.type);
ast::Expression* rhs = MakeScalarExpr(this, param.type);
std::string op_type_decl = OpTypeDecl(param.type);
auto* expr = create<ast::BinaryExpression>(param.op, lhs, rhs);
WrapInFunction(expr);
spirv::Builder& b = Build();
ASSERT_TRUE(b.Build()) << b.error();
EXPECT_EQ(DumpBuilder(b), R"(OpCapability Shader
OpMemoryModel Logical GLSL450
OpEntryPoint GLCompute %3 "test_function"
OpExecutionMode %3 LocalSize 1 1 1
OpName %3 "test_function"
%2 = OpTypeVoid
%1 = OpTypeFunction %2
%6 = )" + op_type_decl + R"(
%5 = OpTypeVector %6 3
%7 = OpConstant %6 1
%8 = OpConstantComposite %5 %7 %7 %7
%11 = OpTypePointer Function %5
%12 = OpConstantNull %5
%3 = OpFunction %2 None %1
%4 = OpLabel
%10 = OpVariable %11 Function %12
%13 = OpCompositeConstruct %5 %7 %7 %7
%9 = )" + param.name + R"( %5 %8 %13
OpReturn
OpFunctionEnd
)");
Validate(b);
}
TEST_P(MixedBinaryArithTest, ScalarVector) {
auto& param = GetParam();
ast::Expression* lhs = MakeScalarExpr(this, param.type);
ast::Expression* rhs = MakeVectorExpr(this, param.type);
std::string op_type_decl = OpTypeDecl(param.type);
auto* expr = create<ast::BinaryExpression>(param.op, lhs, rhs);
WrapInFunction(expr);
spirv::Builder& b = Build();
ASSERT_TRUE(b.Build()) << b.error();
EXPECT_EQ(DumpBuilder(b), R"(OpCapability Shader
OpMemoryModel Logical GLSL450
OpEntryPoint GLCompute %3 "test_function"
OpExecutionMode %3 LocalSize 1 1 1
OpName %3 "test_function"
%2 = OpTypeVoid
%1 = OpTypeFunction %2
%5 = )" + op_type_decl + R"(
%6 = OpConstant %5 1
%7 = OpTypeVector %5 3
%8 = OpConstantComposite %7 %6 %6 %6
%11 = OpTypePointer Function %7
%12 = OpConstantNull %7
%3 = OpFunction %2 None %1
%4 = OpLabel
%10 = OpVariable %11 Function %12
%13 = OpCompositeConstruct %7 %6 %6 %6
%9 = )" + param.name + R"( %7 %13 %8
OpReturn
OpFunctionEnd
)");
Validate(b);
}
INSTANTIATE_TEST_SUITE_P(
BuilderTest,
MixedBinaryArithTest,
testing::Values(Param{Type::f32, ast::BinaryOp::kAdd, "OpFAdd"},
Param{Type::f32, ast::BinaryOp::kDivide, "OpFDiv"},
// NOTE: Modulo not allowed on mixed float scalar-vector
// Param{Type::f32, ast::BinaryOp::kModulo, "OpFMod"},
// NOTE: We test f32 multiplies separately as we emit
// OpVectorTimesScalar for this case
// Param{Type::i32, ast::BinaryOp::kMultiply, "OpIMul"},
Param{Type::f32, ast::BinaryOp::kSubtract, "OpFSub"},
Param{Type::i32, ast::BinaryOp::kAdd, "OpIAdd"},
Param{Type::i32, ast::BinaryOp::kDivide, "OpSDiv"},
Param{Type::i32, ast::BinaryOp::kModulo, "OpSMod"},
Param{Type::i32, ast::BinaryOp::kMultiply, "OpIMul"},
Param{Type::i32, ast::BinaryOp::kSubtract, "OpISub"},
Param{Type::u32, ast::BinaryOp::kAdd, "OpIAdd"},
Param{Type::u32, ast::BinaryOp::kDivide, "OpUDiv"},
Param{Type::u32, ast::BinaryOp::kModulo, "OpUMod"},
Param{Type::u32, ast::BinaryOp::kMultiply, "OpIMul"},
Param{Type::u32, ast::BinaryOp::kSubtract, "OpISub"}));
using MixedBinaryArithMultiplyTest = TestParamHelper<Param>;
TEST_P(MixedBinaryArithMultiplyTest, VectorScalar) {
auto& param = GetParam();
ast::Expression* lhs = MakeVectorExpr(this, param.type);
ast::Expression* rhs = MakeScalarExpr(this, param.type);
std::string op_type_decl = OpTypeDecl(param.type);
auto* expr = create<ast::BinaryExpression>(param.op, lhs, rhs);
WrapInFunction(expr);
spirv::Builder& b = Build();
ASSERT_TRUE(b.Build()) << b.error();
EXPECT_EQ(DumpBuilder(b), R"(OpCapability Shader
OpMemoryModel Logical GLSL450
OpEntryPoint GLCompute %3 "test_function"
OpExecutionMode %3 LocalSize 1 1 1
OpName %3 "test_function"
%2 = OpTypeVoid
%1 = OpTypeFunction %2
%6 = )" + op_type_decl + R"(
%5 = OpTypeVector %6 3
%7 = OpConstant %6 1
%8 = OpConstantComposite %5 %7 %7 %7
%3 = OpFunction %2 None %1
%4 = OpLabel
%9 = OpVectorTimesScalar %5 %8 %7
OpReturn
OpFunctionEnd
)");
Validate(b);
}
TEST_P(MixedBinaryArithMultiplyTest, ScalarVector) {
auto& param = GetParam();
ast::Expression* lhs = MakeScalarExpr(this, param.type);
ast::Expression* rhs = MakeVectorExpr(this, param.type);
std::string op_type_decl = OpTypeDecl(param.type);
auto* expr = create<ast::BinaryExpression>(param.op, lhs, rhs);
WrapInFunction(expr);
spirv::Builder& b = Build();
ASSERT_TRUE(b.Build()) << b.error();
EXPECT_EQ(DumpBuilder(b), R"(OpCapability Shader
OpMemoryModel Logical GLSL450
OpEntryPoint GLCompute %3 "test_function"
OpExecutionMode %3 LocalSize 1 1 1
OpName %3 "test_function"
%2 = OpTypeVoid
%1 = OpTypeFunction %2
%5 = )" + op_type_decl + R"(
%6 = OpConstant %5 1
%7 = OpTypeVector %5 3
%8 = OpConstantComposite %7 %6 %6 %6
%3 = OpFunction %2 None %1
%4 = OpLabel
%9 = OpVectorTimesScalar %7 %8 %6
OpReturn
OpFunctionEnd
)");
Validate(b);
}
INSTANTIATE_TEST_SUITE_P(BuilderTest,
MixedBinaryArithMultiplyTest,
testing::Values(Param{
Type::f32, ast::BinaryOp::kMultiply, "OpFMul"}));
} // namespace
} // namespace spirv
} // namespace writer

View File

@ -0,0 +1,70 @@
fn vector_scalar_f32() {
var v : vec3<f32>;
var s : f32;
var r : vec3<f32>;
r = v + s;
r = v - s;
r = v * s;
r = v / s;
//r = v % s;
}
fn vector_scalar_i32() {
var v : vec3<i32>;
var s : i32;
var r : vec3<i32>;
r = v + s;
r = v - s;
r = v * s;
r = v / s;
r = v % s;
}
fn vector_scalar_u32() {
var v : vec3<u32>;
var s : u32;
var r : vec3<u32>;
r = v + s;
r = v - s;
r = v * s;
r = v / s;
r = v % s;
}
fn scalar_vector_f32() {
var v : vec3<f32>;
var s : f32;
var r : vec3<f32>;
r = s + v;
r = s - v;
r = s * v;
r = s / v;
//r = s % v;
}
fn scalar_vector_i32() {
var v : vec3<i32>;
var s : i32;
var r : vec3<i32>;
r = s + v;
r = s - v;
r = s * v;
r = s / v;
r = s % v;
}
fn scalar_vector_u32() {
var v : vec3<u32>;
var s : u32;
var r : vec3<u32>;
r = s + v;
r = s - v;
r = s * v;
r = s / v;
r = s % v;
}
[[stage(fragment)]]
fn main() -> [[location(0)]] vec4<f32> {
return vec4<f32>(0.0,0.0,0.0,0.0);
}

View File

@ -0,0 +1,73 @@
struct tint_symbol {
float4 value : SV_Target0;
};
void vector_scalar_f32() {
float3 v = float3(0.0f, 0.0f, 0.0f);
float s = 0.0f;
float3 r = float3(0.0f, 0.0f, 0.0f);
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
}
void vector_scalar_i32() {
int3 v = int3(0, 0, 0);
int s = 0;
int3 r = int3(0, 0, 0);
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
r = (v % s);
}
void vector_scalar_u32() {
uint3 v = uint3(0u, 0u, 0u);
uint s = 0u;
uint3 r = uint3(0u, 0u, 0u);
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
r = (v % s);
}
void scalar_vector_f32() {
float3 v = float3(0.0f, 0.0f, 0.0f);
float s = 0.0f;
float3 r = float3(0.0f, 0.0f, 0.0f);
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
}
void scalar_vector_i32() {
int3 v = int3(0, 0, 0);
int s = 0;
int3 r = int3(0, 0, 0);
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
r = (s % v);
}
void scalar_vector_u32() {
uint3 v = uint3(0u, 0u, 0u);
uint s = 0u;
uint3 r = uint3(0u, 0u, 0u);
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
r = (s % v);
}
tint_symbol main() {
const tint_symbol tint_symbol_1 = {float4(0.0f, 0.0f, 0.0f, 0.0f)};
return tint_symbol_1;
}

View File

@ -0,0 +1,75 @@
#include <metal_stdlib>
using namespace metal;
struct tint_symbol_1 {
float4 value [[color(0)]];
};
void vector_scalar_f32() {
float3 v = 0.0f;
float s = 0.0f;
float3 r = 0.0f;
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
}
void vector_scalar_i32() {
int3 v = 0;
int s = 0;
int3 r = 0;
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
r = (v % s);
}
void vector_scalar_u32() {
uint3 v = 0u;
uint s = 0u;
uint3 r = 0u;
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
r = (v % s);
}
void scalar_vector_f32() {
float3 v = 0.0f;
float s = 0.0f;
float3 r = 0.0f;
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
}
void scalar_vector_i32() {
int3 v = 0;
int s = 0;
int3 r = 0;
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
r = (s % v);
}
void scalar_vector_u32() {
uint3 v = 0u;
uint s = 0u;
uint3 r = 0u;
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
r = (s % v);
}
fragment tint_symbol_1 tint_symbol() {
return {float4(0.0f, 0.0f, 0.0f, 0.0f)};
}

View File

@ -0,0 +1,282 @@
; SPIR-V
; Version: 1.3
; Generator: Google Tint Compiler; 0
; Bound: 200
; Schema: 0
OpCapability Shader
OpMemoryModel Logical GLSL450
OpEntryPoint Fragment %main "main" %tint_symbol_1
OpExecutionMode %main OriginUpperLeft
OpName %tint_symbol_1 "tint_symbol_1"
OpName %vector_scalar_f32 "vector_scalar_f32"
OpName %v "v"
OpName %s "s"
OpName %r "r"
OpName %vector_scalar_i32 "vector_scalar_i32"
OpName %v_0 "v"
OpName %s_0 "s"
OpName %r_0 "r"
OpName %vector_scalar_u32 "vector_scalar_u32"
OpName %v_1 "v"
OpName %s_1 "s"
OpName %r_1 "r"
OpName %scalar_vector_f32 "scalar_vector_f32"
OpName %v_2 "v"
OpName %s_2 "s"
OpName %r_2 "r"
OpName %scalar_vector_i32 "scalar_vector_i32"
OpName %v_3 "v"
OpName %s_3 "s"
OpName %r_3 "r"
OpName %scalar_vector_u32 "scalar_vector_u32"
OpName %v_4 "v"
OpName %s_4 "s"
OpName %r_4 "r"
OpName %tint_symbol_2 "tint_symbol_2"
OpName %tint_symbol "tint_symbol"
OpName %main "main"
OpDecorate %tint_symbol_1 Location 0
%float = OpTypeFloat 32
%v4float = OpTypeVector %float 4
%_ptr_Output_v4float = OpTypePointer Output %v4float
%5 = OpConstantNull %v4float
%tint_symbol_1 = OpVariable %_ptr_Output_v4float Output %5
%void = OpTypeVoid
%6 = OpTypeFunction %void
%v3float = OpTypeVector %float 3
%_ptr_Function_v3float = OpTypePointer Function %v3float
%13 = OpConstantNull %v3float
%_ptr_Function_float = OpTypePointer Function %float
%16 = OpConstantNull %float
%int = OpTypeInt 32 1
%v3int = OpTypeVector %int 3
%_ptr_Function_v3int = OpTypePointer Function %v3int
%42 = OpConstantNull %v3int
%_ptr_Function_int = OpTypePointer Function %int
%45 = OpConstantNull %int
%uint = OpTypeInt 32 0
%v3uint = OpTypeVector %uint 3
%_ptr_Function_v3uint = OpTypePointer Function %v3uint
%78 = OpConstantNull %v3uint
%_ptr_Function_uint = OpTypePointer Function %uint
%81 = OpConstantNull %uint
%191 = OpTypeFunction %void %v4float
%float_0 = OpConstant %float 0
%199 = OpConstantComposite %v4float %float_0 %float_0 %float_0 %float_0
%vector_scalar_f32 = OpFunction %void None %6
%9 = OpLabel
%v = OpVariable %_ptr_Function_v3float Function %13
%s = OpVariable %_ptr_Function_float Function %16
%r = OpVariable %_ptr_Function_v3float Function %13
%21 = OpVariable %_ptr_Function_v3float Function %13
%26 = OpVariable %_ptr_Function_v3float Function %13
%34 = OpVariable %_ptr_Function_v3float Function %13
%18 = OpLoad %v3float %v
%19 = OpLoad %float %s
%22 = OpCompositeConstruct %v3float %19 %19 %19
%20 = OpFAdd %v3float %18 %22
OpStore %r %20
%23 = OpLoad %v3float %v
%24 = OpLoad %float %s
%27 = OpCompositeConstruct %v3float %24 %24 %24
%25 = OpFSub %v3float %23 %27
OpStore %r %25
%28 = OpLoad %v3float %v
%29 = OpLoad %float %s
%30 = OpVectorTimesScalar %v3float %28 %29
OpStore %r %30
%31 = OpLoad %v3float %v
%32 = OpLoad %float %s
%35 = OpCompositeConstruct %v3float %32 %32 %32
%33 = OpFDiv %v3float %31 %35
OpStore %r %33
OpReturn
OpFunctionEnd
%vector_scalar_i32 = OpFunction %void None %6
%37 = OpLabel
%v_0 = OpVariable %_ptr_Function_v3int Function %42
%s_0 = OpVariable %_ptr_Function_int Function %45
%r_0 = OpVariable %_ptr_Function_v3int Function %42
%50 = OpVariable %_ptr_Function_v3int Function %42
%55 = OpVariable %_ptr_Function_v3int Function %42
%60 = OpVariable %_ptr_Function_v3int Function %42
%65 = OpVariable %_ptr_Function_v3int Function %42
%70 = OpVariable %_ptr_Function_v3int Function %42
%47 = OpLoad %v3int %v_0
%48 = OpLoad %int %s_0
%51 = OpCompositeConstruct %v3int %48 %48 %48
%49 = OpIAdd %v3int %47 %51
OpStore %r_0 %49
%52 = OpLoad %v3int %v_0
%53 = OpLoad %int %s_0
%56 = OpCompositeConstruct %v3int %53 %53 %53
%54 = OpISub %v3int %52 %56
OpStore %r_0 %54
%57 = OpLoad %v3int %v_0
%58 = OpLoad %int %s_0
%61 = OpCompositeConstruct %v3int %58 %58 %58
%59 = OpIMul %v3int %57 %61
OpStore %r_0 %59
%62 = OpLoad %v3int %v_0
%63 = OpLoad %int %s_0
%66 = OpCompositeConstruct %v3int %63 %63 %63
%64 = OpSDiv %v3int %62 %66
OpStore %r_0 %64
%67 = OpLoad %v3int %v_0
%68 = OpLoad %int %s_0
%71 = OpCompositeConstruct %v3int %68 %68 %68
%69 = OpSMod %v3int %67 %71
OpStore %r_0 %69
OpReturn
OpFunctionEnd
%vector_scalar_u32 = OpFunction %void None %6
%73 = OpLabel
%v_1 = OpVariable %_ptr_Function_v3uint Function %78
%s_1 = OpVariable %_ptr_Function_uint Function %81
%r_1 = OpVariable %_ptr_Function_v3uint Function %78
%86 = OpVariable %_ptr_Function_v3uint Function %78
%91 = OpVariable %_ptr_Function_v3uint Function %78
%96 = OpVariable %_ptr_Function_v3uint Function %78
%101 = OpVariable %_ptr_Function_v3uint Function %78
%106 = OpVariable %_ptr_Function_v3uint Function %78
%83 = OpLoad %v3uint %v_1
%84 = OpLoad %uint %s_1
%87 = OpCompositeConstruct %v3uint %84 %84 %84
%85 = OpIAdd %v3uint %83 %87
OpStore %r_1 %85
%88 = OpLoad %v3uint %v_1
%89 = OpLoad %uint %s_1
%92 = OpCompositeConstruct %v3uint %89 %89 %89
%90 = OpISub %v3uint %88 %92
OpStore %r_1 %90
%93 = OpLoad %v3uint %v_1
%94 = OpLoad %uint %s_1
%97 = OpCompositeConstruct %v3uint %94 %94 %94
%95 = OpIMul %v3uint %93 %97
OpStore %r_1 %95
%98 = OpLoad %v3uint %v_1
%99 = OpLoad %uint %s_1
%102 = OpCompositeConstruct %v3uint %99 %99 %99
%100 = OpUDiv %v3uint %98 %102
OpStore %r_1 %100
%103 = OpLoad %v3uint %v_1
%104 = OpLoad %uint %s_1
%107 = OpCompositeConstruct %v3uint %104 %104 %104
%105 = OpUMod %v3uint %103 %107
OpStore %r_1 %105
OpReturn
OpFunctionEnd
%scalar_vector_f32 = OpFunction %void None %6
%109 = OpLabel
%v_2 = OpVariable %_ptr_Function_v3float Function %13
%s_2 = OpVariable %_ptr_Function_float Function %16
%r_2 = OpVariable %_ptr_Function_v3float Function %13
%116 = OpVariable %_ptr_Function_v3float Function %13
%121 = OpVariable %_ptr_Function_v3float Function %13
%129 = OpVariable %_ptr_Function_v3float Function %13
%113 = OpLoad %float %s_2
%114 = OpLoad %v3float %v_2
%117 = OpCompositeConstruct %v3float %113 %113 %113
%115 = OpFAdd %v3float %117 %114
OpStore %r_2 %115
%118 = OpLoad %float %s_2
%119 = OpLoad %v3float %v_2
%122 = OpCompositeConstruct %v3float %118 %118 %118
%120 = OpFSub %v3float %122 %119
OpStore %r_2 %120
%123 = OpLoad %float %s_2
%124 = OpLoad %v3float %v_2
%125 = OpVectorTimesScalar %v3float %124 %123
OpStore %r_2 %125
%126 = OpLoad %float %s_2
%127 = OpLoad %v3float %v_2
%130 = OpCompositeConstruct %v3float %126 %126 %126
%128 = OpFDiv %v3float %130 %127
OpStore %r_2 %128
OpReturn
OpFunctionEnd
%scalar_vector_i32 = OpFunction %void None %6
%132 = OpLabel
%v_3 = OpVariable %_ptr_Function_v3int Function %42
%s_3 = OpVariable %_ptr_Function_int Function %45
%r_3 = OpVariable %_ptr_Function_v3int Function %42
%139 = OpVariable %_ptr_Function_v3int Function %42
%144 = OpVariable %_ptr_Function_v3int Function %42
%149 = OpVariable %_ptr_Function_v3int Function %42
%154 = OpVariable %_ptr_Function_v3int Function %42
%159 = OpVariable %_ptr_Function_v3int Function %42
%136 = OpLoad %int %s_3
%137 = OpLoad %v3int %v_3
%140 = OpCompositeConstruct %v3int %136 %136 %136
%138 = OpIAdd %v3int %140 %137
OpStore %r_3 %138
%141 = OpLoad %int %s_3
%142 = OpLoad %v3int %v_3
%145 = OpCompositeConstruct %v3int %141 %141 %141
%143 = OpISub %v3int %145 %142
OpStore %r_3 %143
%146 = OpLoad %int %s_3
%147 = OpLoad %v3int %v_3
%150 = OpCompositeConstruct %v3int %146 %146 %146
%148 = OpIMul %v3int %150 %147
OpStore %r_3 %148
%151 = OpLoad %int %s_3
%152 = OpLoad %v3int %v_3
%155 = OpCompositeConstruct %v3int %151 %151 %151
%153 = OpSDiv %v3int %155 %152
OpStore %r_3 %153
%156 = OpLoad %int %s_3
%157 = OpLoad %v3int %v_3
%160 = OpCompositeConstruct %v3int %156 %156 %156
%158 = OpSMod %v3int %160 %157
OpStore %r_3 %158
OpReturn
OpFunctionEnd
%scalar_vector_u32 = OpFunction %void None %6
%162 = OpLabel
%v_4 = OpVariable %_ptr_Function_v3uint Function %78
%s_4 = OpVariable %_ptr_Function_uint Function %81
%r_4 = OpVariable %_ptr_Function_v3uint Function %78
%169 = OpVariable %_ptr_Function_v3uint Function %78
%174 = OpVariable %_ptr_Function_v3uint Function %78
%179 = OpVariable %_ptr_Function_v3uint Function %78
%184 = OpVariable %_ptr_Function_v3uint Function %78
%189 = OpVariable %_ptr_Function_v3uint Function %78
%166 = OpLoad %uint %s_4
%167 = OpLoad %v3uint %v_4
%170 = OpCompositeConstruct %v3uint %166 %166 %166
%168 = OpIAdd %v3uint %170 %167
OpStore %r_4 %168
%171 = OpLoad %uint %s_4
%172 = OpLoad %v3uint %v_4
%175 = OpCompositeConstruct %v3uint %171 %171 %171
%173 = OpISub %v3uint %175 %172
OpStore %r_4 %173
%176 = OpLoad %uint %s_4
%177 = OpLoad %v3uint %v_4
%180 = OpCompositeConstruct %v3uint %176 %176 %176
%178 = OpIMul %v3uint %180 %177
OpStore %r_4 %178
%181 = OpLoad %uint %s_4
%182 = OpLoad %v3uint %v_4
%185 = OpCompositeConstruct %v3uint %181 %181 %181
%183 = OpUDiv %v3uint %185 %182
OpStore %r_4 %183
%186 = OpLoad %uint %s_4
%187 = OpLoad %v3uint %v_4
%190 = OpCompositeConstruct %v3uint %186 %186 %186
%188 = OpUMod %v3uint %190 %187
OpStore %r_4 %188
OpReturn
OpFunctionEnd
%tint_symbol_2 = OpFunction %void None %191
%tint_symbol = OpFunctionParameter %v4float
%194 = OpLabel
OpStore %tint_symbol_1 %tint_symbol
OpReturn
OpFunctionEnd
%main = OpFunction %void None %6
%196 = OpLabel
%197 = OpFunctionCall %void %tint_symbol_2 %199
OpReturn
OpFunctionEnd

View File

@ -0,0 +1,68 @@
fn vector_scalar_f32() {
var v : vec3<f32>;
var s : f32;
var r : vec3<f32>;
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
}
fn vector_scalar_i32() {
var v : vec3<i32>;
var s : i32;
var r : vec3<i32>;
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
r = (v % s);
}
fn vector_scalar_u32() {
var v : vec3<u32>;
var s : u32;
var r : vec3<u32>;
r = (v + s);
r = (v - s);
r = (v * s);
r = (v / s);
r = (v % s);
}
fn scalar_vector_f32() {
var v : vec3<f32>;
var s : f32;
var r : vec3<f32>;
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
}
fn scalar_vector_i32() {
var v : vec3<i32>;
var s : i32;
var r : vec3<i32>;
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
r = (s % v);
}
fn scalar_vector_u32() {
var v : vec3<u32>;
var s : u32;
var r : vec3<u32>;
r = (s + v);
r = (s - v);
r = (s * v);
r = (s / v);
r = (s % v);
}
[[stage(fragment)]]
fn main() -> [[location(0)]] vec4<f32> {
return vec4<f32>(0.0, 0.0, 0.0, 0.0);
}