david-xl wrote: > > > . For IR PGO, there is basically no need to do so as the instrumentation > > > and profile-use should be in-sync. For front-end instrumentation, there > > > seem to be some use cases to use out of sync profile: > > > https://reviews.llvm.org/D51240. > > > > > > Thanks for double checking. I noticed the ICP and stale profile tolerance > > discussions when read the Phab history; it's good Phab review history are > > still available nowadays. > > IRPGO profiles could be used along with supplementary sample-pgo profiles. > > I'll probably read relevant code in llvm-profdata to understand how these > > interact in theory mostly for my own curiosity (hopefully no rough edges as > > long as `llvm-profdata` uses the same pgo name format used by latest > > compiler) > > For irpgo with supplementary profiles, this line to build a map ( > > https://github.com/llvm/llvm-project/blob/44dc1e0baae7c4b8a02ba06dcf396d3d452aa873/llvm/tools/llvm-profdata/llvm-profdata.cpp#L982 > > ) needs an update. Will do it together with the test > [update](https://github.com/llvm/llvm-project/pull/74008#discussion_r1421018997) > in this pull request.
A follow up patch to fix the tool is fine too (or may be better). https://github.com/llvm/llvm-project/pull/74008 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
