[PATCH] D134119: [msan] Handle x86_avx_cmp_pd_256 and x86_avx_cmp_ps_256

[Phoebe Wang via Phabricator via llvm-commits]

pengfei added inline comments.


================
Comment at: llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp:3761
-      // generates reasonably looking IR that fails in the backend with "Do not
-      // know how to split the result of this operator!".
       handleVectorComparePackedIntrinsic(I);
----------------
vitalybuka wrote:
> RKSimon wrote:
> > Have you been able to determine whether this has been addressed?
> this error happens only without -mattr=+avx which is expected
> and it's true for many other avx intrinsic here
> 
> So I assume this comment is by mistake here @eugenis 
> 
> 
The FE checked the target attributes before generating such intrinsics, so I think we can always assume we meet the requirments.
In fact, now we generate such intrinsics only for a few cases under constrained FP: https://github.com/llvm/llvm-project/blob/main/clang/lib/CodeGen/CGBuiltin.cpp#L14812-L14866


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D134119/new/

https://reviews.llvm.org/D134119