[PATCH] D147740: [NFC][llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed

[William Junda Huang via Phabricator via llvm-commits]

huangjd added a comment.

Changing the hash algorithm will break backwards compatibility, making previously collected profile invalid, which is a big concern.  Currently MD5 of sampled functions are written to the profile file in ext binary format, and while the profile is being used for compilation, function names in the current module are MD5 hashed to match against the profile. In general, the number of functions in a module is much much smaller than the number of functions in the profile, so hashing them all isn't significant. Note that there's no MD5 hashing while decoding the profile itself (except hash_combine for CS context), and by profiling the reader itself I can confirm that MD5 related function calls are insignificant.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D147740/new/

https://reviews.llvm.org/D147740