forked from microsoft/onnxruntime
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[JS/WebGPU] Support Tile operator (microsoft#17123)
### Description As title ### Motivation and Context Improve WebGPU op coverage
- Loading branch information
1 parent
e43de82
commit 471fd27
Showing
8 changed files
with
305 additions
and
2 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,98 @@ | ||
| // Copyright (c) Microsoft Corporation. All rights reserved. | ||
| // Licensed under the MIT License. | ||
|
|
||
| import {DataType} from '../../../wasm-common'; | ||
| import {TensorView} from '../../tensor'; | ||
| import {ShapeUtil} from '../../util'; | ||
| import {ComputeContext, GpuDataType, ProgramInfo, ProgramMetadata} from '../types'; | ||
|
|
||
| import {inputVariable, outputVariable, ShaderHelper} from './common'; | ||
|
|
||
| export const tileProgramMetadata = { | ||
| name: 'Tile', | ||
| inputTypes: [GpuDataType.default] | ||
| }; | ||
|
|
||
| const getRepeats = (repeatsTensorView: TensorView): readonly number[] => | ||
| Array.from(repeatsTensorView.getBigInt64Array(), Number); | ||
|
|
||
|
|
||
| const validateInputs = (inputs: readonly TensorView[]): void => { | ||
| if (!inputs || inputs.length !== 2) { | ||
| throw new Error('Tile requires 2 inputs.'); | ||
| } | ||
|
|
||
| if (inputs[0].dataType !== DataType.float && inputs[0].dataType !== DataType.int32 && | ||
| inputs[0].dataType !== DataType.uint32) { | ||
| throw new Error('Tile only support float, int32, and uint32 data types'); | ||
| } | ||
|
|
||
| if (inputs[1].dataType !== DataType.int64) { | ||
| throw new Error('Tile `repeats` input should be of int64 data type'); | ||
| } | ||
|
|
||
| if (inputs[1].dims.length !== 1) { | ||
| throw new Error('Tile `repeats` input should be 1-D'); | ||
| } | ||
|
|
||
| const repeats: readonly number[] = getRepeats(inputs[1]); | ||
|
|
||
| if (repeats.length !== inputs[0].dims.length) { | ||
| throw new Error('Tile `repeats` input should have same number of elements as rank of input data tensor'); | ||
| } | ||
| }; | ||
|
|
||
| const getOutputShape = (inputShape: readonly number[], repeats: readonly number[]): readonly number[] => { | ||
| const outputShape: number[] = []; | ||
|
|
||
| for (let i = 0; i < inputShape.length; ++i) { | ||
| outputShape.push(inputShape[i] * repeats[i]); | ||
| } | ||
|
|
||
| return outputShape; | ||
| }; | ||
|
|
||
| export const createTileProgramInfo = | ||
| (tileProgramMetadata: ProgramMetadata, inputs: readonly TensorView[]): ProgramInfo => { | ||
| const inputShape = inputs[0].dims; | ||
| const repeats: readonly number[] = getRepeats(inputs[1]); | ||
| const outputShape = getOutputShape(inputShape, repeats); | ||
| const outputSize = ShapeUtil.size(outputShape); | ||
|
|
||
| const dataType = inputs[0].dataType; | ||
| const input = inputVariable('input', dataType, inputShape); | ||
| const output = outputVariable('output', dataType, outputShape); | ||
|
|
||
| const getShaderSource = (shaderHelper: ShaderHelper) => ` | ||
| const inputShape = ${input.indices(...inputShape)}; | ||
| ${shaderHelper.declareVariables(input, output)} | ||
| ${output.impl('offsetToIndices')} | ||
| ${input.impl('indicesToOffset', 'get')} | ||
| ${shaderHelper.mainStart()} | ||
| ${shaderHelper.guardAgainstOutOfBoundsWorkgroupSizes(outputSize)} | ||
| let outputIndices = ${output.offsetToIndices('global_idx')}; | ||
| var inputIndices: ${input.type.indices}; | ||
| for (var i = 0; i < ${inputShape.length}; i++) { | ||
| let inputDimValue = ${output.indicesGet('outputIndices', 'i')} % ${input.indicesGet('inputShape', 'i')}; | ||
| ${input.indicesSet('inputIndices', 'i', 'inputDimValue')} | ||
| } | ||
| ${output.setByOffset('global_idx', input.getByIndices('inputIndices'))} | ||
| }`; | ||
|
|
||
| return { | ||
| ...tileProgramMetadata, | ||
| outputs: [{dims: outputShape, dataType: inputs[0].dataType, gpuDataType: GpuDataType.default}], | ||
| getShaderSource, | ||
| dispatchGroup: () => ({x: Math.ceil(outputSize / 64 /* workgroup size */)}) | ||
| }; | ||
| }; | ||
|
|
||
| export const tile = (context: ComputeContext): void => { | ||
| validateInputs(context.inputs); | ||
| const repeats: readonly number[] = getRepeats(context.inputs[1]); | ||
| const cacheHint = repeats.toString(); | ||
| context.compute( | ||
| {...tileProgramMetadata, cacheHint, get: () => createTileProgramInfo(tileProgramMetadata, context.inputs)}, | ||
| {inputs: [0]}); | ||
| }; |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,147 @@ | ||
| [ | ||
| { | ||
| "name": "Tile 2D - float32", | ||
| "operator": "Tile", | ||
| "attributes": [], | ||
| "cases": [ | ||
| { | ||
| "name": "T[1,2]", | ||
| "inputs": [ | ||
| { | ||
| "data": [1, 2, 3, 4], | ||
| "dims": [2, 2], | ||
| "type": "float32" | ||
| }, | ||
| { | ||
| "data": [1, 2], | ||
| "dims": [2], | ||
| "type": "int64" | ||
| } | ||
| ], | ||
| "outputs": [ | ||
| { | ||
| "data": [1, 2, 1, 2, 3, 4, 3, 4], | ||
| "dims": [2, 4], | ||
| "type": "float32" | ||
| } | ||
| ] | ||
| } | ||
| ] | ||
| }, | ||
| { | ||
| "name": "Tile 2D - int32", | ||
| "operator": "Tile", | ||
| "attributes": [], | ||
| "cases": [ | ||
| { | ||
| "name": "T[1,2]", | ||
| "inputs": [ | ||
| { | ||
| "data": [1, 2, 3, 4], | ||
| "dims": [2, 2], | ||
| "type": "int32" | ||
| }, | ||
| { | ||
| "data": [1, 2], | ||
| "dims": [2], | ||
| "type": "int64" | ||
| } | ||
| ], | ||
| "outputs": [ | ||
| { | ||
| "data": [1, 2, 1, 2, 3, 4, 3, 4], | ||
| "dims": [2, 4], | ||
| "type": "int32" | ||
| } | ||
| ] | ||
| } | ||
| ] | ||
| }, | ||
| { | ||
| "name": "Tile 2D - uint32", | ||
| "operator": "Tile", | ||
| "attributes": [], | ||
| "cases": [ | ||
| { | ||
| "name": "T[1,2]", | ||
| "inputs": [ | ||
| { | ||
| "data": [1, 2, 3, 4], | ||
| "dims": [2, 2], | ||
| "type": "uint32" | ||
| }, | ||
| { | ||
| "data": [2, 1], | ||
| "dims": [2], | ||
| "type": "int64" | ||
| } | ||
| ], | ||
| "outputs": [ | ||
| { | ||
| "data": [1, 2, 3, 4, 1, 2, 3, 4], | ||
| "dims": [4, 2], | ||
| "type": "uint32" | ||
| } | ||
| ] | ||
| } | ||
| ] | ||
| }, | ||
| { | ||
| "name": "Tile 1D - float32", | ||
| "operator": "Tile", | ||
| "attributes": [], | ||
| "cases": [ | ||
| { | ||
| "name": "T[2]", | ||
| "inputs": [ | ||
| { | ||
| "data": [1, 2], | ||
| "dims": [2], | ||
| "type": "float32" | ||
| }, | ||
| { | ||
| "data": [4], | ||
| "dims": [1], | ||
| "type": "int64" | ||
| } | ||
| ], | ||
| "outputs": [ | ||
| { | ||
| "data": [1, 2, 1, 2, 1, 2, 1, 2], | ||
| "dims": [8], | ||
| "type": "float32" | ||
| } | ||
| ] | ||
| } | ||
| ] | ||
| }, | ||
| { | ||
| "name": "Tile 2D all dims", | ||
| "operator": "Tile", | ||
| "attributes": [], | ||
| "cases": [ | ||
| { | ||
| "name": "T[2]", | ||
| "inputs": [ | ||
| { | ||
| "data": [0, 1, 2, 3], | ||
| "dims": [2, 2], | ||
| "type": "float32" | ||
| }, | ||
| { | ||
| "data": [2, 2], | ||
| "dims": [2], | ||
| "type": "int64" | ||
| } | ||
| ], | ||
| "outputs": [ | ||
| { | ||
| "data": [0, 1, 0, 1, 2, 3, 2, 3, 0, 1, 0, 1, 2, 3, 2, 3], | ||
| "dims": [4, 4], | ||
| "type": "float32" | ||
| } | ||
| ] | ||
| } | ||
| ] | ||
| } | ||
| ] |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,36 @@ | ||
| // Copyright (c) Microsoft Corporation. All rights reserved. | ||
| // Licensed under the MIT License. | ||
|
|
||
| #include "core/providers/js/js_kernel.h" | ||
| #include "tile.h" | ||
|
|
||
| namespace onnxruntime { | ||
| namespace js { | ||
| ONNX_OPERATOR_VERSIONED_KERNEL_EX( | ||
| Tile, | ||
| kOnnxDomain, | ||
| 6, | ||
| 12, | ||
| kJsExecutionProvider, | ||
| KernelDefBuilder() | ||
| .TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(), | ||
| DataTypeImpl::GetTensorType<int32_t>(), | ||
| DataTypeImpl::GetTensorType<uint32_t>()}) | ||
| .TypeConstraint("T1", DataTypeImpl::GetTensorType<int64_t>()) | ||
| .InputMemoryType(OrtMemTypeCPU, 1), | ||
| Tile); | ||
|
|
||
| ONNX_OPERATOR_KERNEL_EX( | ||
| Tile, | ||
| kOnnxDomain, | ||
| 13, | ||
| kJsExecutionProvider, | ||
| KernelDefBuilder() | ||
| .TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(), | ||
| DataTypeImpl::GetTensorType<int32_t>(), | ||
| DataTypeImpl::GetTensorType<uint32_t>()}) | ||
| .TypeConstraint("T1", DataTypeImpl::GetTensorType<int64_t>()) | ||
| .InputMemoryType(OrtMemTypeCPU, 1), | ||
| Tile); | ||
| } // namespace js | ||
| } // namespace onnxruntime |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,14 @@ | ||
| // Copyright (c) Microsoft Corporation. All rights reserved. | ||
| // Licensed under the MIT License. | ||
|
|
||
| #pragma once | ||
|
|
||
| #include "core/providers/js/js_kernel.h" | ||
|
|
||
| namespace onnxruntime { | ||
| namespace js { | ||
|
|
||
| JSEP_KERNEL_IMPL(Tile, Tile); | ||
|
|
||
| } // namespace js | ||
| } // namespace onnxruntime |