[llvm] [X86] Narrow vXi32/vXi64 usubsat to vpsubusb/vpsubusw when LHS is known to fit in fewer bits (PR #206592)

[via llvm-commits]

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

----------------
sipher-01 wrote:

I've update the test file to use -mcpu=x86-64, -mcpu=x86-64-v2, -mcpu=x86-64-v3, and -mcpu=x86-64-v4 (via the usual RUN lines + CHECK prefixes, similar to how other vector-codegen tests in this file structure it) instead of the current SSE2/SSE41/AVX1/AVX2/AVX512 split, so we get baseline coverage that maps directly to the microarchitecture levels rather than individual ISA flags.

https://github.com/llvm/llvm-project/pull/206592