[PATCH] D13644: AVX-512 bit shuffle in 32-bit mode - fixed a bug

[Elena Demikhovsky via llvm-commits]

delena added a comment.

I considered this approach, but when we deal with mask, we want to create UNDEF node for -1.
So the common helper may be useful, but not in this specific case.

Another possibility is to create a BUILD_VECTOR node from mask array and then rebuild it for 32-bit mode in a common helper.
Seems like a double work for 32.

- Elena


Repository:
  rL LLVM

http://reviews.llvm.org/D13644