Pipeline stall metrics break after a few seconds
- Dominant language
- C++
- Stars
- 3.3k
- Forks
- 528
- Avg merge
- 1h 40m
- Merged PRs (30d)
- 2
Description
Built latest source, after 1-5 seconds `pcm` starts showing
Observed with w7-2595X CPU on Linux kernel 5.15
First noticed on pcm 202502 version from Ubuntu 25.10, this version is more stable but breaks the same way when I manipulate CPU frequency by changing AVX offsets forth and back.
Edit: version 202502 breaks exactly the same, seems that it works only until frequency changes. I tested first under load so frequency was pretty stable, and the only change was induced by my tweaking of AVX offsets. Without load it breaks pretty fast if CPU frequency is not fixed.
Contributor guide
Research direction
The issue provides no file or test entry point. First reproduce pcm on Linux while changing CPU frequency or AVX offsets, then trace the pipeline-stall metric handling; done means the metrics continue reporting correctly across those frequency changes.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp, linux
- Domain
- performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 38/100