[ExecuTorch][WebGPU] Restore SDPA per-dispatch kernel_name labels#20551
Conversation
The register-tile change (pytorch#20507) rewrote the `update_cache`, QK (`sdpa_compute_attn_weights`), `sdpa_softmax`, and AV (`sdpa_compute_out`) `build_dispatch` call sites and dropped the per-dispatch `kernel_name` labels originally added in pytorch#20167. With the labels gone, `WEBGPU_TIMESTAMP_QUERY` profiling can no longer attribute on-GPU time to the attention stage that produced it (every dispatch reports as the default "dispatch"). This re-threads `kernel_name` through `build_dispatch` (defaulted to `""`, so all other callers are unaffected) into the existing `WebGPUDispatch::kernel_name` field that `WebGPUQueryPool` already reads, and re-applies the four SDPA stage labels. No behavior change when profiling is off; the production `execute()` path is byte-identical.
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20551
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 32 PendingAs of commit ad00635 with merge base b919db7 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
psiddh
left a comment
There was a problem hiding this comment.
Approving this to unbloc the diff train
[ExecuTorch][WebGPU] Restore SDPA per-dispatch kernel_name labels
@diff-train-skip-merge