Skip to content

[webgpu] allow overloads to Program::AddIndices #24021

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 3 commits into from
Mar 13, 2025
Merged

[webgpu] allow overloads to Program::AddIndices #24021

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5b4ee70
[webgpu] allow overloads to Program::AddIndices
fs-eire Mar 13, 2025
49ddf81
use r-value ref for params of AddIndices() when possible
fs-eire Mar 13, 2025
970c10e
fix r-value ref for AddInputs and AddOutputs
fs-eire Mar 13, 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
2 changes: 1 addition & 1 deletion onnxruntime/contrib_ops/webgpu/bert/flash_attention.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,7 +97,7 @@ Status CopyKVCache(onnxruntime::webgpu::ComputeContext& context, const WebgpuAtt
}
program.AddOutputs({{present_key, ProgramTensorMetadataDependency::Rank, components},
{present_value, ProgramTensorMetadataDependency::Rank, components}})
.AddIndices(valid_present_shape);
.AddIndices(std::move(valid_present_shape));
program.SetDispatchGroupSize(onnxruntime::narrow<uint32_t>(valid_kv_size + 63 / 64))
.SetWorkgroupSize(64)
.CacheHint(has_past, parameters.qkv_format_, parameters.past_present_share_buffer_)
Expand Down
6 changes: 3 additions & 3 deletions onnxruntime/core/providers/webgpu/math/binary_elementwise_ops.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,9 +190,9 @@ Status BinaryElementwise::ComputeInternal(ComputeContext& context) const {
// Mode Vectorize broadcast
// cache hint: "V{a_rank};{b_rank};{output_rank}"
program
.AddIndices(reshaped_output_shape)
.AddIndices(reshaped_lhs_shape)
.AddIndices(reshaped_rhs_shape)
.AddIndices(std::move(reshaped_output_shape))
.AddIndices(std::move(reshaped_lhs_shape))
.AddIndices(std::move(reshaped_rhs_shape))
.CacheHint("V");
} else {
// Mode Broadcast
Expand Down
14 changes: 2 additions & 12 deletions onnxruntime/core/providers/webgpu/program.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -289,7 +289,7 @@ ProgramBase::ProgramBase(std::string_view name, ProgramMetadata&& metadata)
}

ProgramBase& ProgramBase::AddInput(ProgramInput&& input) {
inputs_.emplace_back(input);
inputs_.emplace_back(std::move(input));
return *this;
}

Expand All @@ -299,7 +299,7 @@ ProgramBase& ProgramBase::AddInputs(std::initializer_list<ProgramInput> inputs)
}

ProgramBase& ProgramBase::AddOutput(ProgramOutput&& output) {
outputs_.emplace_back(output);
outputs_.emplace_back(std::move(output));
return *this;
}

Expand All @@ -308,16 +308,6 @@ ProgramBase& ProgramBase::AddOutputs(std::initializer_list<ProgramOutput> output
return *this;
}

ProgramBase& ProgramBase::AddIndices(const TensorShape& shape) {
indices_.emplace_back(shape);
return *this;
}

ProgramBase& ProgramBase::AddIndices(TensorShape&& shape) {
indices_.emplace_back(shape);
return *this;
}

ProgramBase& ProgramBase::SetDispatchGroupSize(uint32_t x) {
return SetDispatchGroupSize(x, 1, 1);
}
Expand Down
8 changes: 5 additions & 3 deletions onnxruntime/core/providers/webgpu/program.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -271,9 +271,11 @@ class ProgramBase {
// add multiple program outputs
ProgramBase& AddOutputs(std::initializer_list<ProgramOutput> outputs);
// add a program variable for indices
ProgramBase& AddIndices(const TensorShape& shape);
// add a program variable for indices
ProgramBase& AddIndices(TensorShape&& shape);
template <typename... Args>
ProgramBase& AddIndices(Args&&... args) {
indices_.emplace_back(std::forward<Args>(args)...);
return *this;
}

// set the size of dispatch groups. Y and Z are 1 if not specified.
ProgramBase& SetDispatchGroupSize(uint32_t x);
Expand Down
2 changes: 1 addition & 1 deletion onnxruntime/core/providers/webgpu/tensor/expand.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,7 +53,7 @@ Status Expand::ComputeInternal(ComputeContext& context) const {
{data_size},
});
if (components_i != components_o) {
program.AddIndices(output_shape);
program.AddIndices(std::move(output_shape));
}
return context.RunProgram(program);
}
Expand Down