Skip to content
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

Jump to bottom

[webgpu native] Add transpose shared #22098

Open
wants to merge 1 commit into
base: fs-eire/webgpu-ep
Choose a base branch
from
+90 −20
Open

[webgpu native] Add transpose shared #22098

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
102 changes: 85 additions & 17 deletions onnxruntime/core/providers/webgpu/tensor/transpose.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,11 +47,11 @@ ONNX_OPERATOR_KERNEL_EX(
.TypeConstraint("T", WebGpuSupportedNumberTypes()),
Transpose);

const std::string AppendPermFunction(gsl::span<const size_t> perm) {
const std::string AppendPermFunction(gsl::span<const int64_t> perm) {
std::ostringstream ss;
ss.imbue(std::locale::classic());
ss << "fn perm(i: y_indices_t)->x_indices_t {\n"
" var a: x_indices_t;\n";
ss << "fn perm(i: output_indices_t)->a_indices_t {\n"
" var a: a_indices_t;\n";
for (auto i = 0; i < perm.size(); ++i) {
ss << " a[" << perm[i] << "] = i[" << i << "];\n";
}
Expand All @@ -60,21 +60,62 @@ const std::string AppendPermFunction(gsl::span<const size_t> perm) {
return ss.str();
}

auto SqueezeShape(const gsl::span<const int64_t>& shape, const gsl::span<const size_t>& adjusted_perm, InlinedVector<int64_t>& new_shape, InlinedVector<int64_t>& new_perm) {
for (auto i = 0; i < shape.size(); ++i) {
if (shape[i] != 1) {
new_shape.push_back(shape[i]);
}
if (shape[adjusted_perm[i]] != 1) {
new_perm.push_back(adjusted_perm[i]);
}
}
};

Status TransposeProgram::GenerateShaderCode(ShaderHelper& shader) const {
const auto& input = shader.AddInput("x", ShaderUsage::UseUniform | ShaderUsage::UseIndicesTypeAlias);
const auto& output = shader.AddOutput("y", ShaderUsage::UseUniform | ShaderUsage::UseIndicesTypeAlias);
shader.AppendImplementation(AppendPermFunction(this->perm_));
shader.SetMainFunctionBody(shader.GuardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size"),
" let indices = ", output.OffsetToIndices("global_idx"),
";\n"
" let x_indices = perm(indices); \n"
" ",
output.SetByOffset("global_idx", input.GetByIndices("x_indices")));
const auto& input = shader.AddInput("a", ShaderUsage::UseUniform | ShaderUsage::UseIndicesTypeAlias);
const auto& output = shader.AddOutput("output", ShaderUsage::UseUniform | ShaderUsage::UseIndicesTypeAlias | ShaderUsage::UseValueTypeAlias);

if (use_shared_) {
const auto tile_size = std::to_string(tile_size_);
shader.AppendImplementation("var<workgroup> tile : array<array<output_value_t, " + tile_size + " + 1>, " + tile_size + ">;\n");
shader.SetMainFunctionBody(
" let stride = (uniforms.output_shape[1] - 1) / " + tile_size +
" + 1;\n"
" let workgroup_id_x = workgroup_idx % stride;\n"
" let workgroup_id_y = workgroup_idx / stride;\n"
" let input_col = workgroup_id_y * " +
tile_size +
"u + local_id.x;\n"
" let input_row = workgroup_id_x * " +
tile_size +
"u + local_id.y;\n"
" if (input_row < uniforms.a_shape[0] && input_col < uniforms.a_shape[1]) {\n"
" tile[local_id.y][local_id.x] = " +
input.GetByIndices("a_indices_t(input_row, input_col)") +
";\n"
" }\n"
" workgroupBarrier();\n"
" let output_col = workgroup_id_x * " +
tile_size +
"u + local_id.x;\n"
" let output_row = workgroup_id_y * " +
tile_size +
"u + local_id.y;\n"
" if (output_row < uniforms.output_shape[0] && output_col < uniforms.output_shape[1]) {\n " +
output.SetByIndices("output_indices_t(output_row, output_col)", "tile[local_id.x][local_id.y]") + "\n }");
} else {
shader.AppendImplementation(AppendPermFunction(this->perm_));
shader.SetMainFunctionBody(shader.GuardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size"),
" let indices = ", output.OffsetToIndices("global_idx"),
";\n"
" let x_indices = perm(indices);\n",
" ",
output.SetByOffset("global_idx", input.GetByIndices("x_indices")));
}
return Status::OK();
}

Status Transpose::ComputeInternal(ComputeContext& context) const {
// TODO: there is an optimized version of transpose to port.
const auto* input_tensor = context.Input(0);
const TensorShape& input_shape = input_tensor->Shape();
int32_t rank = gsl::narrow_cast<int32_t>(input_shape.NumDimensions());
Expand All @@ -86,16 +127,43 @@ Status Transpose::ComputeInternal(ComputeContext& context) const {
TensorShape output_shape(output_dims);
auto* output_tensor = context.Output(0, output_shape);

InlinedVector<int64_t> new_shape{};
InlinedVector<int64_t> new_perm{};
SqueezeShape(input_shape.GetDims(), *p_perm, new_shape, new_perm);
const auto channels_last = new_perm == InlinedVector<int64_t>({2, 3, 1});
const auto channels_first = new_perm == InlinedVector<int64_t>({3, 1, 2});
const auto use_shared = (new_shape.size() == 2 && new_perm[0] > new_perm[1]) || channels_last || channels_first;
auto new_input_shape = use_shared ? new_shape : input_shape;
auto new_output_shape = output_dims;
if (use_shared) {
new_input_shape = channels_last
? InlinedVector<int64_t>({new_shape[0], new_shape[1] * new_shape[2]})
: channels_first
? InlinedVector<int64_t>({new_shape[0] * new_shape[1], new_shape[2]})
: new_shape;
new_output_shape = InlinedVector<int64_t>({new_input_shape[1], new_input_shape[0]});
}

uint32_t output_size = gsl::narrow_cast<int32_t>(input_tensor->Shape().Size());
TransposeProgram program{*p_perm};
const auto tile_size = 16;
TransposeProgram program{*p_perm, use_shared, tile_size};
if (use_shared) {
program.SetWorkgroupSize(tile_size, tile_size, 1);
}

program
.CacheHint(absl::StrJoin(*p_perm, "-"))
.AddInputs({{input_tensor, ProgramTensorMetadataDependency::TypeAndRank}})
.AddOutputs({output_tensor})
.SetDispatchGroupSize((output_size + WORKGROUP_SIZE - 1) / WORKGROUP_SIZE)
.AddInputs({{input_tensor, ProgramTensorMetadataDependency::TypeAndRank, new_input_shape, 1}})
.AddOutputs({{output_tensor, ProgramTensorMetadataDependency::TypeAndRank, new_output_shape, 1}})
.SetDispatchGroupSize(static_cast<uint32_t>((new_output_shape[1] + tile_size - 1) / tile_size),
static_cast<uint32_t>(((new_output_shape[0] + tile_size - 1) / tile_size)))
.AddUniformVariables({
{static_cast<uint32_t>(output_size)},
});

use_shared ? program.SetDispatchGroupSize(static_cast<uint32_t>((new_output_shape[1] + tile_size - 1) / tile_size),
static_cast<uint32_t>(((new_output_shape[0] + tile_size - 1) / tile_size)))
: program.SetDispatchGroupSize((output_size + WORKGROUP_SIZE - 1) / WORKGROUP_SIZE);
return context.RunProgram(program);
}

Expand Down
8 changes: 5 additions & 3 deletions onnxruntime/core/providers/webgpu/tensor/transpose.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,16 +13,18 @@ namespace webgpu {

class TransposeProgram final : public Program<TransposeProgram> {
public:
TransposeProgram(const gsl::span<const size_t>& permutations)
: Program{"Transpose"}, perm_(permutations.begin(), permutations.end()) {
TransposeProgram(const gsl::span<const size_t>& permutations, bool use_shared, const int tile_size)
: Program{"Transpose"}, perm_(permutations.begin(), permutations.end()), use_shared_(use_shared), tile_size_(tile_size) {
}

Status GenerateShaderCode(ShaderHelper& sh) const override;

WEBGPU_PROGRAM_DEFINE_UNIFORM_VARIABLES({"output_size", ProgramUniformVariableDataType::Uint32});

private:
InlinedVector<size_t> perm_;
InlinedVector<int64_t> perm_;
const bool use_shared_;
const uint32_t tile_size_;
};

class Transpose final : public WebGpuKernel, public TransposeBase {
Expand Down