Split up and refactor CLANG_BOLT_INSTRUMENT into support for
perf no-LBR and perf with LBR profiling modes.
Details
Details
- Reviewers
beanz phosek jobnoorman - Group Reviewers
Restricted Project
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
| clang/utils/perf-training/bolt.lit.cfg | ||
|---|---|---|
| 12 | This name would ideally be passed through the generated lit.site.cfg rather than hardcoding it here. | |
| clang/utils/perf-training/perf-helper.py | ||
| 75 | This could be omitted. | |
| 85 | Can we remove this? | |
| 88–97 | Why not assign to perf_args directly? | |
| 115 | I'd call it llvm-bolt for simplicity. | |
| 118–120 | I'd consider providing both --lbr and --no-lbr options. | |
| 123–126 | Why not assign to p2b_args directly? | |
| clang/utils/perf-training/perf-helper.py | ||
|---|---|---|
| 75 | I think it's best to keep it as optional argument, not positional. | |
Comment Actions
Make the name of BOLT-instrumented Clang binary (CLANG_BOLT_INSTRUMENTED)
a user-settable cache variable
This name would ideally be passed through the generated lit.site.cfg rather than hardcoding it here.