[PATCH] D131618: [WIP][Do NOT review] LLD related changes for -ffat-lto-objects support

Arda Unal via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon Aug 29 10:58:48 PDT 2022


arda marked an inline comment as done.
arda added inline comments.


================
Comment at: llvm/lib/Bitcode/Writer/EmbedBitcodePass.cpp:26
+
+PreservedAnalyses EmbedBitcodePass::run(Module &M, ModuleAnalysisManager &AM) {
+  if (M.getGlobalVariable("llvm.embedded.module", true))
----------------
arda wrote:
> sfertile wrote:
> > From the discourse discussion:
> > 1) it was suggested that we remove the existing -fembed-bitcode functionality as Apple has stop supporting it.
> > 2) mentioned that MLGO uses the option to embed the bitcode at various points in the pipeline depending on if its using LTO our not.
> > 
> > Do we want the pass to be a bit more generic and be able to specify the global to use for embedding, and the section name to use as arguments? That way MLGO can keep using the section name it uses now . It also helps consuming tools to disambiguate between bitcode embedded for lto purpose from bitcode embedded for other purposes.
> I should have an answer in a couple of days. I will keep you updated.
We have talked to folks at MLGO and reached a decision that their use case is sufficiently different from ours that it should be handled separately. More detailed discussion here:

[[ https://discourse.llvm.org/t/rfc-ffat-lto-objects-support/63977/15?u=arda | [RFC] -ffat-lto-objects support on Discours ]]


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D131618/new/

https://reviews.llvm.org/D131618



More information about the cfe-commits mailing list