[PATCH] D146776: [llvm] Preliminary fat-lto-objects support

[Arthur Eubanks via Phabricator via llvm-commits]

aeubanks added a comment.

for your use case, is it ok for performance to not be 100% if you go down the non-LTO route? presumably if you're not using LTO you're not caring about squeezing out 100% performance. for example in Chrome-land we've talked about this sort of thing for tests where we don't really care if performance is optimal, so an approximation of -O2 (or whatever) is good enough. if we solidify this as a tradeoff and don't promise 100% compatibility with the default optimization pipeline, I'm much more comfortable with that as long as people don't complain if FatLTO non-LTO performance ever slightly degrades


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D146776/new/

https://reviews.llvm.org/D146776