[llvm] [X86] combineSelect - `fold select(c,vtrunc(x),0)` => `X86ISD::MTRUNC(x,c)` (PR #212865)

[Simon Pilgrim via llvm-commits]

================

----------------
RKSimon wrote:

don't use specific cpu targets - use -mattr or -mcpu=x86-64-v4 if you really need to (put with avx512 folds its often good to have a non-VLX RUN as well if you can)

Although tbh I'd try to add this to avx512-trunc.ll (or a related existing file)

https://github.com/llvm/llvm-project/pull/212865