webgpu: support BroadcastArgs kernel #7247
Merged
Conversation
This file contains hidden or 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
gyagp
approved these changes
Jan 5, 2023
| if (index < uniforms.size) { | ||
| var s0 = 1.0; | ||
| var s1 = 1.0; | ||
| let indexS0 = index - uniforms.size + uniforms.s0Length; |
There was a problem hiding this comment.
nit, maybe
Suggested change
| let indexS0 = index - uniforms.size + uniforms.s0Length; | |
| let indexS0 = index - uniforms.size + uniforms.s0Size; |
qjia7
reviewed
Jan 5, 2023
| const program = new BroadcastArgsProgram(outputSize); | ||
| const uniformData = | ||
| [{type: 'int32', data: [s0Size]}, {type: 'int32', data: [s1Size]}]; | ||
| return backend.runWebGPUProgram(program, [s0, s1], 'int32', uniformData); |
Contributor
There was a problem hiding this comment.
Can you also add a cpu path if s0 and s1 is on cpu since the inputs are really small?
Neilblaze
approved these changes
Jan 14, 2023
qjia7
approved these changes
Jan 16, 2023
| const {s0, s1} = inputs; | ||
|
|
||
| if (backend.shouldExecuteOnCPU([s0, s1])) { | ||
| const s0BufferInfo = backend.tensorMap.get(s0.dataId); |
Contributor
There was a problem hiding this comment.
nit: s0BufferInfo -> s0TensorInfo
Contributor
Author
|
@gyagp Could you review it again, because I added cpu path after you approved. |
gyagp
reviewed
Jan 16, 2023
| const {s0, s1} = inputs; | ||
|
|
||
| if (backend.shouldExecuteOnCPU([s0, s1])) { | ||
| const s0TensorInfo = backend.tensorMap.get(s0.dataId); |
Contributor
Author
There was a problem hiding this comment.
- cpu backend does not export the CPU impl for other backends.
- cpu backend also calls the tfjs-core common function to implement the feature, it is better for other backends to call the tfjs-core common function directly.
There was a problem hiding this comment.
Per discussion, a similar interface (xxxImplCPU) to be used as CPU fallback is a good approach. However, this may need some changes at CPU backend, so let's refine this in a future PR. LGTM with this one.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
To see the logs from the Cloud Build CI, please join either our discussion or announcement mailing list.
This change is