Enable CUDA sync event correlation on CUDA 12.9+ - #1552
Open
benghaem wants to merge 1 commit into
Open
Conversation
|
@benghaem has exported this pull request. If you are a Meta employee, you can view the originating Diff in D117757961. |
Summary: ## Summary CUDA 12.9 decoupled `CUPTI_ACTIVITY_KIND_CUDA_EVENT` from `CUPTI_ACTIVITY_KIND_SYNCHRONIZATION`. Kineto continued enabling only synchronization when sync events were requested, so CUDA 13 traces lacked event-record inputs and left `wait_on_stream` and `wait_on_cuda_event_record_corr_id` unresolved. ## Changes Update `CuptiActivityApi.cpp` to enable and disable `CUPTI_ACTIVITY_KIND_CUDA_EVENT` alongside synchronization activity collection on CUDA 12.9 and newer. The CUDA 13 device-timestamp opt-in remains unchanged, and the implementation is mirrored between fbcode and xplat. Add an end-to-end profiler test that creates a cross-stream wait and verifies the exported trace contains resolved source-stream and event-correlation metadata. The test runs whenever Kineto is available without CUDA-version or ROCm gating, uses a backend-neutral name, and validates the metadata contract without depending on backend-specific category or event display strings. Differential Revision: D117757961
benghaem
force-pushed
the
export-D117757961
branch
from
August 27, 2026 22:58
2c12dff to
15e5b87
Compare
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary:
Summary
CUDA 12.9 decoupled
CUPTI_ACTIVITY_KIND_CUDA_EVENTfromCUPTI_ACTIVITY_KIND_SYNCHRONIZATION. Kineto continued enabling only synchronization when sync events were requested, so CUDA 13 traces lacked event-record inputs and leftwait_on_streamandwait_on_cuda_event_record_corr_idunresolved.Changes
Update
CuptiActivityApi.cppto enable and disableCUPTI_ACTIVITY_KIND_CUDA_EVENTalongside synchronization activity collection on CUDA 12.9 and newer. The CUDA 13 device-timestamp opt-in remains unchanged, and the implementation is mirrored between fbcode and xplat.Add an end-to-end profiler test that creates a cross-stream wait and verifies the exported trace contains resolved source-stream and event-correlation metadata. The test runs whenever Kineto is available without CUDA-version or ROCm gating, uses a backend-neutral name, and validates the metadata contract without depending on backend-specific category or event display strings.
Differential Revision: D117757961