[PATCH] D28979: [X86] Add demanded elts support for the inputs to pclmul intrinsic

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

In https://reviews.llvm.org/D28979#652966, @craig.topper wrote:

> Optimize to undef if both inputs are undef.


Should we be optimizing to undef or zero? I don't have an answer but I just got stung for having written muldq/muludq optimizing to undef as generic multiplies become zero. Does clmul count as a multiply?


https://reviews.llvm.org/D28979