[llvm] [LoongArch] Custom lowering for 128-bit vector integer shifts (PR #171097)

[via llvm-commits]

================
@@ -9,8 +9,10 @@ declare <16 x i8> @llvm.bitreverse.v16i8(<16 x i8>)
 define <16 x i8> @test_bitreverse_v16i8(<16 x i8> %a) nounwind {
 ; LA32-LABEL: test_bitreverse_v16i8:
 ; LA32:       # %bb.0:
-; LA32-NEXT:    vslli.b $vr1, $vr0, 4
+; LA32-NEXT:    vandi.b $vr1, $vr0, 15
----------------
heiher wrote:

It appears that some optimizations are missing, which will probably be resolved in a subsequent patch.

https://github.com/llvm/llvm-project/pull/171097