[clang] [compiler-rt] [llvm] [X86] AVX10_V2_AUX Implementation (PR #206888)

[via llvm-commits]

ganeshgit wrote:

@RKSimon Request your review and approval. For narrowing converts where we cannot have select (128/256) we still use _mask version for unmasked (with -1), merge masking and zero masking. For 512, we use select. This follows existing AVX 10.2 convert intrinsics. So, I request your approval for merging this. Currently merging is blocked for your "requested changes"

https://github.com/llvm/llvm-project/pull/206888