[llvm] [AMDGPU] selecting v_sat_pk instruction, version 2 (PR #123297)

[via llvm-commits]

================
@@ -16,6 +16,10 @@ declare i16 @llvm.smax.i16(i16, i16)
 
 declare <2 x i16> @llvm.smin.v2i16(<2 x i16>, <2 x i16>)
 declare <2 x i16> @llvm.smax.v2i16(<2 x i16>, <2 x i16>)
+declare <4 x i16> @llvm.smin.v4i16(<4 x i16>, <4 x i16>)
----------------
Shoreshen wrote:

Hi @arsenm , the case is added~

But currently the v2i16 case will not use saturation pack, only even number of elements are supported. And for all MVT, only v3i16 is odd vector.

https://github.com/llvm/llvm-project/pull/123297