[PATCH] D27395: Fixed store operation for a vector of i1.

[Elena Demikhovsky via Phabricator via llvm-commits]

delena added a comment.

All "store", "load", "masked.store" "masked.load" should be compatible between them. At the beginning I thought that the operation should have the same behavior across all X86 targets. But now I see that the code is not target specific and the decision should be common for all targets. I also see that there are a lot of bugs there and the operations are not really tested.
In my opinion, the better and easier approach would be  storing i1 vectors in bytes. I'll also fix AVX-512 and add tests.


Repository:
  rL LLVM

https://reviews.llvm.org/D27395