Skip to content

[WebGPU EP] Add GEMM implementation #24023

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
fs-eire merged 10 commits into from
Mar 26, 2025
Merged

[WebGPU EP] Add GEMM implementation #24023

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
a2a7d5f
implement
Mar 14, 2025
b1a43f3
handle A/B
Mar 17, 2025
2198ff9
fix pipeline error
Mar 18, 2025
9438a2b
remove unneccessary variable
Mar 18, 2025
4f31592
optimize
Mar 18, 2025
bc54156
resolve comments
Mar 19, 2025
e29ed33
Merge remote-tracking branch 'origin/main' into user/xiaofeihan/imple…
Mar 19, 2025
996f514
use cachehint
Mar 20, 2025
ff864c6
remove strjoin
Mar 21, 2025
337f0c2
Merge remote-tracking branch 'origin/main' into user/xiaofeihan/imple…
fs-eire Mar 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
233 changes: 233 additions & 0 deletions onnxruntime/core/providers/webgpu/math/gemm.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,233 @@
// Copyright (c) Microsoft Corporation. All rights reserved.
Comment thread Fixed
// Licensed under the MIT License.

#include "core/providers/webgpu/math/gemm.h"

#include "core/providers/webgpu/shader_helper.h"
#include "core/providers/webgpu/webgpu_supported_types.h"

namespace onnxruntime {
namespace webgpu {

#define WEBGPU_GEMM_VERSIONED_KERNEL(start, end) \
ONNX_OPERATOR_VERSIONED_KERNEL_EX( \
Gemm, \
kOnnxDomain, \
start, \
end, \
kWebGpuExecutionProvider, \
(*KernelDefBuilder::Create()) \
.TypeConstraint("T", WebGpuSupportedNumberTypes()), \
Gemm);

#define WEBGPU_GEMM_KERNEL(version) \
ONNX_OPERATOR_KERNEL_EX( \
Gemm, \
kOnnxDomain, \
version, \
kWebGpuExecutionProvider, \
(*KernelDefBuilder::Create()) \
.TypeConstraint("T", WebGpuSupportedNumberTypes()), \
Gemm);

WEBGPU_GEMM_VERSIONED_KERNEL(7, 8)
WEBGPU_GEMM_VERSIONED_KERNEL(9, 10)
WEBGPU_GEMM_VERSIONED_KERNEL(11, 12)
WEBGPU_GEMM_KERNEL(13)

Status GemmProgram::GenerateShaderCode(ShaderHelper& shader) const {
const uint32_t TILE_SIZE = 16;

// Add shared memory arrays
shader.AdditionalImplementation() << "var<workgroup> tile_a: array<array<output_value_t, " << TILE_SIZE << ">, " << TILE_SIZE << ">;\n"
<< "var<workgroup> tile_b: array<array<output_value_t, " << TILE_SIZE << ">, " << TILE_SIZE << ">;\n\n";

const ShaderVariableHelper& output = shader.AddOutput("output", ShaderUsage::UseUniform | ShaderUsage::UseValueTypeAlias);

shader.MainFunctionBody() << " var value = output_value_t(0);\n\n"
<< " let tile_col_start = (workgroup_id.x % uniforms.num_tile_n) * " << TILE_SIZE << "u;\n"
Comment thread
xiaofeihan1 marked this conversation as resolved.
Outdated
<< " let tile_row_start = (workgroup_id.x / uniforms.num_tile_n) * " << TILE_SIZE << "u;\n";

// When K == 0, we don't bind A and B. Because WebGPU doesn't support binding a zero-sized buffer.
if (need_handle_matmul_) {
const ShaderVariableHelper& A = shader.AddInput("A", ShaderUsage::UseUniform);
const ShaderVariableHelper& B = shader.AddInput("B", ShaderUsage::UseUniform);

shader.MainFunctionBody()
<< " let num_tiles = (uniforms.K - 1u) / " << TILE_SIZE << "u + 1u;\n"
<< " var k_start = 0u;\n"
<< " for (var t = 0u; t < num_tiles; t = t + 1u) {\n";

// Fill workgroup shared memory
if (transA_ && transB_) {
shader.MainFunctionBody() << " var col = tile_row_start + local_id.x;\n"
<< " var row = k_start + local_id.y;\n"
<< " if (col < uniforms.M && row < uniforms.K) {\n"
<< " tile_a[local_id.y][local_id.x] = " << A.GetByOffset("row * uniforms.M + col") << ";\n"
<< " } else {\n"
<< " tile_a[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n\n"
<< " col = k_start + local_id.x;\n"
<< " row = tile_col_start + local_id.y;\n"
<< " if (col < uniforms.K && row < uniforms.N) {\n"
<< " tile_b[local_id.y][local_id.x] = " << B.GetByOffset("row * uniforms.K + col") << ";\n"
<< " } else {\n"
<< " tile_b[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n";
} else if (transA_ && !transB_) {
shader.MainFunctionBody() << " var col = tile_row_start + local_id.x;\n"
<< " var row = k_start + local_id.y;\n"
<< " if (col < uniforms.M && row < uniforms.K) {\n"
<< " tile_a[local_id.y][local_id.x] = " << A.GetByOffset("row * uniforms.M + col") << ";\n"
<< " } else {\n"
<< " tile_a[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n\n"
<< " col = tile_col_start + local_id.x;\n"
<< " row = k_start + local_id.y;\n"
<< " if (col < uniforms.N && row < uniforms.K) {\n"
<< " tile_b[local_id.y][local_id.x] = " << B.GetByOffset("row * uniforms.N + col") << ";\n"
<< " } else {\n"
<< " tile_b[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n";
} else if (!transA_ && transB_) {
shader.MainFunctionBody() << " var col = k_start + local_id.x;\n"
<< " var row = tile_row_start + local_id.y;\n"
<< " if (col < uniforms.K && row < uniforms.M) {\n"
<< " tile_a[local_id.y][local_id.x] = " << A.GetByOffset("row * uniforms.K + col") << ";\n"
<< " } else {\n"
<< " tile_a[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n\n"
<< " col = k_start + local_id.x;\n"
<< " row = tile_col_start + local_id.y;\n"
<< " if (col < uniforms.K && row < uniforms.N) {\n"
<< " tile_b[local_id.y][local_id.x] = " << B.GetByOffset("row * uniforms.K + col") << ";\n"
<< " } else {\n"
<< " tile_b[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n";
} else {
shader.MainFunctionBody() << " var col = k_start + local_id.x;\n"
<< " var row = tile_row_start + local_id.y;\n"
<< " if (col < uniforms.K && row < uniforms.M) {\n"
<< " tile_a[local_id.y][local_id.x] = " << A.GetByOffset("row * uniforms.K + col") << ";\n"
<< " } else {\n"
<< " tile_a[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n\n"
<< " col = tile_col_start + local_id.x;\n"
<< " row = k_start + local_id.y;\n"
<< " if (col < uniforms.N && row < uniforms.K) {\n"
<< " tile_b[local_id.y][local_id.x] = " << B.GetByOffset("row * uniforms.N + col") << ";\n"
<< " } else {\n"
<< " tile_b[local_id.y][local_id.x] = output_value_t(0);\n"
<< " }\n";
}

shader.MainFunctionBody() << " k_start = k_start + " << TILE_SIZE << "u;\n"
<< " workgroupBarrier();\n\n"
<< " for (var k = 0u; k < " << TILE_SIZE << "u; k = k + 1u) {\n";

if (transA_ && transB_) {
shader.MainFunctionBody() << " value = value + tile_a[k][local_id.y] * tile_b[local_id.x][k];\n";
} else if (transA_ && !transB_) {
shader.MainFunctionBody() << " value = value + tile_a[k][local_id.y] * tile_b[k][local_id.x];\n";
} else if (!transA_ && transB_) {
shader.MainFunctionBody() << " value = value + tile_a[local_id.y][k] * tile_b[local_id.x][k];\n";
} else {
shader.MainFunctionBody() << " value = value + tile_a[local_id.y][k] * tile_b[k][local_id.x];\n";
}

shader.MainFunctionBody() << " }\n"
<< " workgroupBarrier();\n"
<< " }\n\n";
}

// Calculate Alpha
if (alpha_) {
Comment thread
xiaofeihan1 marked this conversation as resolved.
shader.MainFunctionBody() << " value = value * output_value_t(uniforms.alpha);\n";
}

shader.MainFunctionBody() << " let m = tile_row_start + local_id.y;\n"
<< " let n = tile_col_start + local_id.x;\n";

// Calculate Bias
if (need_handle_bias_) {
const ShaderVariableHelper& C = shader.AddInput("C", ShaderUsage::UseUniform);
shader.MainFunctionBody() << " value = value + output_value_t(uniforms.beta) * "
<< C.GetByOffset(C.BroadcastedIndicesToOffset("vec2(m, n)", output)) << ";\n";
}

// Write output
shader.MainFunctionBody() << " if (m < uniforms.M && n < uniforms.N) {\n"
<< " " << output.SetByOffset("m * uniforms.N + n", "value") << "\n"
<< " }\n";

return Status::OK();
}

Status Gemm::ComputeInternal(ComputeContext& context) const {
const auto* A = context.Input<Tensor>(0);
const auto* B = context.Input<Tensor>(1);
const auto* C = context.Input<Tensor>(2);

if (A == nullptr || B == nullptr) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT, "Gemm requires input tensors A and B.");
}

const auto& A_shape = A->Shape();
const auto& B_shape = B->Shape();

if (A_shape.NumDimensions() != 2 || B_shape.NumDimensions() != 2) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT, "Input tensors A and B must be 2 dimensional.");
}

uint32_t M = gsl::narrow<uint32_t>(transA_ ? A_shape[1] : A_shape[0]);
uint32_t K = gsl::narrow<uint32_t>(transA_ ? A_shape[0] : A_shape[1]);
uint32_t N = gsl::narrow<uint32_t>(transB_ ? B_shape[0] : B_shape[1]);

if ((transA_ ? A_shape[0] : A_shape[1]) != (transB_ ? B_shape[1] : B_shape[0])) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT, "Inner dimensions of A and B must match.");
}

std::vector<int64_t> output_dims{M, N};
auto* Y = context.Output(0, output_dims);
int64_t output_size = Y->Shape().Size();

if (output_size == 0) {
return Status::OK();
}

// WebGPU doesn't support binding a zero-sized buffer, so we need to check if K == 0
bool need_handle_matmul = K != 0;
Comment thread
xiaofeihan1 marked this conversation as resolved.
Outdated
bool need_handle_bias = C && beta_;

GemmProgram program{transA_, transB_, alpha_, need_handle_bias, need_handle_matmul};

if (need_handle_matmul) {
program.AddInputs({{A, ProgramTensorMetadataDependency::Type},
{B, ProgramTensorMetadataDependency::Type}});
}

if (need_handle_bias) {
program.AddInput({C, ProgramTensorMetadataDependency::Rank});
}

const uint32_t TILE_SIZE = 16;
const uint32_t num_tile_n = (N + TILE_SIZE - 1) / TILE_SIZE;
const uint32_t num_tile_m = (M + TILE_SIZE - 1) / TILE_SIZE;

program.AddOutputs({{Y, ProgramTensorMetadataDependency::Type}})
.SetDispatchGroupSize(num_tile_n * num_tile_m)
.SetWorkgroupSize(TILE_SIZE, TILE_SIZE)
.AddUniformVariables({
{static_cast<uint32_t>(num_tile_n)}, // num_tile_n
{static_cast<uint32_t>(M)}, // M
{static_cast<uint32_t>(N)}, // N
{static_cast<uint32_t>(K)}, // K
{alpha_}, // alpha
{beta_} // beta
});

return context.RunProgram(program);
}

} // namespace webgpu
} // namespace onnxruntime
66 changes: 66 additions & 0 deletions onnxruntime/core/providers/webgpu/math/gemm.h
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,66 @@
// Copyright (c) Microsoft Corporation. All rights reserved.
// Licensed under the MIT License.

#pragma once

#include "core/providers/webgpu/webgpu_kernel.h"
#include "core/providers/webgpu/shader_helper.h"
#include "core/providers/webgpu/program.h"

namespace onnxruntime {
namespace webgpu {

class GemmProgram final : public Program<GemmProgram> {
public:
GemmProgram(bool transA, bool transB, float alpha, bool need_handle_bias, bool need_handle_matmul)
: Program{"Gemm"},
transA_{transA},
transB_{transB},
alpha_{alpha},
need_handle_bias_{need_handle_bias},
need_handle_matmul_{need_handle_matmul} {}

Status GenerateShaderCode(ShaderHelper& sh) const override;

WEBGPU_PROGRAM_DEFINE_UNIFORM_VARIABLES(
{"num_tile_n", ProgramUniformVariableDataType::Uint32},
{"M", ProgramUniformVariableDataType::Uint32},
{"N", ProgramUniformVariableDataType::Uint32},
{"K", ProgramUniformVariableDataType::Uint32},
{"alpha", ProgramUniformVariableDataType::Float32},
{"beta", ProgramUniformVariableDataType::Float32});

private:
bool transA_;
bool transB_;
float alpha_;
bool need_handle_bias_;
bool need_handle_matmul_;
};

class Gemm final : public WebGpuKernel {
public:
Gemm(const OpKernelInfo& info) : WebGpuKernel(info) {
int64_t transA_temp;
info.GetAttrOrDefault("transA", &transA_temp, static_cast<int64_t>(0));
transA_ = transA_temp != 0;

int64_t transB_temp;
info.GetAttrOrDefault("transB", &transB_temp, static_cast<int64_t>(0));
transB_ = transB_temp != 0;

info.GetAttrOrDefault("alpha", &alpha_, 1.0f);
info.GetAttrOrDefault("beta", &beta_, 1.0f);
}

Status ComputeInternal(ComputeContext& context) const override;

private:
bool transA_;
bool transB_;
float alpha_;
float beta_;
};

} // namespace webgpu
} // namespace onnxruntime
8 changes: 4 additions & 4 deletions onnxruntime/core/providers/webgpu/webgpu_execution_provider.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -610,10 +610,10 @@ std::unique_ptr<KernelRegistry> RegisterKernels() {
// BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 1, GlobalMaxPool)>,
// BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kMSInternalNHWCDomain, 1, GlobalMaxPool)>,

// BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 7, 8, Gemm)>,
// BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 9, 10, Gemm)>,
// BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 11, 12, Gemm)>,
// BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 13, Gemm)>,
BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 7, 8, Gemm)>,
BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 9, 10, Gemm)>,
BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 11, 12, Gemm)>,
BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 13, Gemm)>,
// BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 1, 12, MatMul)>,
// BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 13, MatMul)>,

Expand Down