Investigate MT evaluation performance regression in pipelines
Open
Area: Performance
- Dominant language
- C#
- Stars
- 5.5k
- Forks
- 1.5k
- Avg merge
- 1d 8h
- Merged PRs (30d)
- 141
Description
MT evaluation is significantly slower in pipelines than on Dev Box (about 5x vs about 2x).
Goals:
1. Find the primary cause(s) of the MT evaluation slowdown.
2. Identify actionable fixes.
3. Validate improvements with before/after measurements.
Known so far:
PRE cache was ruled out as the primary cause (it adds lock contention but does not explain the regression).
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.