Add benchmark for CUDA Graph / HIP Graph trace processing performance
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 52/100
- Issue type
- Feature
- Clarity
- Mostly clear
- Activity status
- Quiet
- Tech stack
- python
- Domain
- performance
Research direction
The issue names no files, tests, or benchmark entry point. Start by locating the existing trace-processing benchmarks and the path used to process trace artifacts, then identify how benchmark traces are stored and timed. Done means adding one graph-heavy trace, a wall-clock benchmark, and a comparable non-graph baseline with reported results.
Written by the indexing model from the issue text.
Description
Summary
Add a benchmark that measures TraceLens trace-processing time on a trace containing CUDA Graph / HIP Graph dispatches. Customers have reported slowdowns when processing graph-heavy traces, and we have no benchmark to track this.
Motivation
- Customers report noticeable slowdown when loading/processing traces that contain CUDA/HIP graph captures.
- Without a targeted benchmark we can't quantify the impact or catch regressions.
Proposed Work
- Collect or synthesize a representative trace — one trace where a large chunk of the kernels are dispatched via CUDA Graph / HIP Graph.
- Add a benchmark target — measure end-to-end trace processing wall-clock time for that trace.
- Compare against a non-graph baseline — run the same benchmark on a similarly-sized trace without graphs so we can see the delta.
Acceptance Criteria
- One graph-heavy trace is available as a benchmark artifact.
- Benchmark measures and reports wall-clock processing time.
- Results are comparable to a non-graph baseline of similar kernel count.
- Dominant language
- Python
- Stars
- 101
- Forks
- 34
- Avg merge
- 2d 1h
- Merged PRs (30d)
- 78
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from AMD-AGI/TraceLens
-
Difficulty 2/5 1-3 hours Newbie friendliness 80/100
Maintainers usually reply within 1 day
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
2 graph capture handling bugsPossibly taken @gabeweisz claimed this 1 day ago. Open
Difficulty 4/5 3-5 days Newbie friendliness 25/100
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 35/100
Maintainers usually reply within 1 day
All issues in AMD-AGI/TraceLens
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 83/100
Maintainers usually reply within 1 day
-
comp/desktop P3 type/feature
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
NousResearch/hermes-agent#154285 ·
Maintainers usually reply within 1 day
-
docs: qualify copied and custom kubectl debug policy notesPossibly taken @nym01 claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
kubescape/cel-admission-library#166 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
SACGF/variantgrid#2804 · 1 comment ·
Maintainers usually reply within 1 day
-
[BUG] HTTP lifecycle dispatch treats a NO_PROXY-only environment as corporate egress and skips the resolve-and-pinned sessionPossibly taken @bodapatisaikrishna claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
Maintainers usually reply within 1 day