[PATCH] D60261: Change semantics of fadd/fmul vector reductions.

[Sander de Smalen via Phabricator via llvm-commits]

sdesmalen marked 2 inline comments as done.
sdesmalen added a comment.

Thanks for the review!



================
Comment at: include/llvm/CodeGen/BasicTTIImpl.h:1266
+          /*IsPairwiseForm=*/false); // FIXME: This should be set to
+                                     // 'FMF.allowReassoc()'
     case Intrinsic::experimental_vector_reduce_smax:
----------------
nikic wrote:
> Regardless of IsPairwiseForm, this will compute an unordered reduction cost, just for two different reduction strategies. Just passing FMF.allowReassoc() here wouldn't be meaningful. We'd need a separate flag to indicate ordered reductions.
I've updated the FIXME comment to reflect this before committing the patch.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D60261/new/

https://reviews.llvm.org/D60261