[llvm-branch-commits] [llvm] [LoongArch] Custom legalize for 256-bit vector trunc (2/2) (PR #170246)

[via llvm-branch-commits]

zhaoqi5 wrote:

> There is another patch merged for solving the same issue, e.g. #202485
> 
> Maybe we could closed this patch series?

Sure. Although the methods are different.

BTW. The current implementation achieves better or equivalent performance than this commit on most tests, but this commit has slightly better results on `trunc_v4i64_to_v4i8` and `xvmsk_ogt_v4f64_concat_poison` tests.

https://github.com/llvm/llvm-project/pull/170246