[PATCH] D44785: Lowering x86 adds/addus/subs/subus intrinsics (llvm part)

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

What happens when one of the inputs is a constant (e.g. saturated increment:: _mm_adds_epi8(X, _mm_set1_epi8(1)))?



================
Comment at: test/CodeGen/X86/avx2-intrinsics-fast-isel.ll:5
 
 ; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/avx2-builtins.c
 
----------------
As noted here, you should be using the same IR as is generated in the clang builtins tests updated in D44786


Repository:
  rL LLVM

https://reviews.llvm.org/D44785