[profiler] update CUDA runtime kernel identification logic #157890
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/157890
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit b0b748d with merge base ee72338 ( UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@pytorchbot label "release notes: profiler" |
torch/profiler/_utils.py
Outdated
|
|
||
| name = str(getattr(e, "name", e)).lower() | ||
| # Exclude launcher and memory events | ||
| exclude_patterns = ["cudalaunch", "cudamem"] |
There was a problem hiding this comment.
Not sure if this is necessary. cudalaunch will already have DeviceType.CPU. I think we should just check if kernel is in the name...
There was a problem hiding this comment.
Thanks, I misunderstood it due to my lack of background. Then how about keeping the current structure and removing #TODO? (cc @davidchencsl )
There was a problem hiding this comment.
I think until we can procedurally determine it is a kernel besides looking at the name we should keep the todo
There was a problem hiding this comment.
@sraikund16 Thanks for your review. For focusing on excluding memory kernel, I updated for more edge-cases in this commit. Could you take a look at this PR?
mikaylagawarecki
added
the
triaged
|
@pytorchbot rebase |
|
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
Use cases: - "mem" : "cudaMemcpy", "cudaMemset" - "alloc" : "cudaMalloc", "cudaMallocManaged" - "free" : "cudaFree"
|
Successfully rebased |
pytorchmergebot
force-pushed
the
profiler_kernel
branch
from
1f85809 to
b0b748d
Compare
|
@pytorchbot merge |
|
Pull workflow has not been scheduled for the PR yet. It could be because author doesn't have permissions to run those or skip-checks keywords were added to PR/commits, aborting merge. Please get/give approval for the workflows and/or remove skip ci decorators before next merge attempt. If you think this is a mistake, please contact PyTorch Dev Infra. |
|
@pytorchmergebot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Update CUDA kernel detection to exclude memory API calls References: - https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__MEMORY.html - https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__EXECUTION.html Pull Request resolved: #157890 Approved by: https://github.com/sraikund16
5 participants
Update CUDA kernel detection to exclude memory API calls
References: