[llvm-branch-commits] [llvm] [LoongArch] Try to avoid casts around logical vector ops on lasx (PR #163523)

[via llvm-branch-commits]

https://github.com/wangleiat commented:

This optimization mirrors the corresponding x86 implementation, and the correctness looks sound. Based on the test case changes, it also appears to be a significant improvement.

https://github.com/llvm/llvm-project/pull/163523