[PATCH] D115497: [Inline] Disable deferred inlining

Jonas Paulsson via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Feb 8 16:27:38 PST 2022


jonpa added a comment.

It seems that the imagick regression on SystemZ per above where inlining of @SetPixelCacheNexusPixels is no longer done involves the memcpy performed directly after entering the function. A 32-byte struct is built by the caller and passed to this function which then only does a memcpy of that struct as a source.

I wonder if perhaps "partial inlining" could be used for this to just inline the memcpy part into caller? It seems this was enabled by default back in 2017, but now that does not appear to be the case. Has this changed with the move to the new pass manager somehow? (I have to enable it with '-mllvm -enable-partial-inlining' and I don't see any target using it...)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D115497/new/

https://reviews.llvm.org/D115497



More information about the llvm-commits mailing list