[llvm] [AMDGPU] Use `v_cvt_pk_*` instructions for i16_f32 saturated conversions (PR #202680)

[Igor Wodiany via llvm-commits]

================
@@ -724,6 +724,21 @@ def S_MAX_U32 : SOP2_32 <"s_max_u32",
 } // End isCommutable = 1
 } // End Defs = [SCC]
 
+// f32 -> i16 saturated conversions. Required because those conversions
----------------
IgWod wrote:

I'm double checking it. We probably should wait with this patch until we have an answer in case we need to remove s_cvt cases - this would actually simplify GISel part a bit.

https://github.com/llvm/llvm-project/pull/202680