[llvm] [PreISelIntrinsicLowering] Lower bounded memcpy/memmove to masked load/store (PR #212710)

[via llvm-commits]

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

----------------
Harishankar14 wrote:

@arsenm  I did try this  by set target-cpu to skylake-avx512 (which has AVX512BW) both via -mcpu and as a direct function attribute, but the pass still declines. The implied features don't seem to reach the TTI query (isLegalMaskedLoad) in this pass  only an explicit target-features attribute makes it fire. Am I missing a step to get target-cpu to expand into features for TTI here? Just an intution, what if we use -mattr instead 

https://github.com/llvm/llvm-project/pull/212710