zequanwu added a comment. In D157632#4581219 <https://reviews.llvm.org/D157632#4581219>, @ellis wrote:
> In D157632#4580576 <https://reviews.llvm.org/D157632#4580576>, @zequanwu > wrote: > >> BTW, I noticed something strange with `-pgo-function-entry-coverage` when >> merging via llvm-profdata. > > This is intentional. The two raw profiles individually store blocks as either > covered or not, but when we merge them they get converted to counters and > accumulated. > https://github.com/llvm/llvm-project/blob/6a0feb1503e21432e63d93b44357bad43f8026d1/llvm/lib/ProfileData/InstrProf.cpp#L726 > Then in `PGOUseFunc::populateCoverage()` we annotate blocks as alive if their > counter value is non-zero. > https://github.com/llvm/llvm-project/blob/1aee2434ce4c9b5785cbc8f72cbbbd64f9e85297/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp#L1399 > My logic was that the indexed profile represents these counters as ints, so > we might as well accumulate them. Also, this makes the implementation simpler. Thanks for explanation. That makes sense. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D157632/new/ https://reviews.llvm.org/D157632 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
