This function is called about 80 times during test startup and most of
the invokes the same compiler command. Avoiding running a new process and
parsing the output into a python dict by caching the output. This
noticeably speeds up running a single lit test: for me this patch changes
the time it takes to run lit with --no-execute for a single test from
~7 seconds to ~5.5 seconds.
Details
Details
- Reviewers
ldionne - Group Reviewers
Restricted Project
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Unit Tests
Unit Tests
Event Timeline
Comment Actions
Thanks for opening these reviews. I proposed a unified approach to solve this problem in https://reviews.llvm.org/D89003, please let me know what you think.