[llvm] [AMDGPU] Correctly insert s_nops for dst forwarding hazard (PR #100276)

[Pravin Jagtap via llvm-commits]

================
@@ -2904,6 +2905,15 @@ def getVCMPXOpFromVCMP : InstrMapping {
   let ValueCols = [["1"]];
 }
 
+def FP8DstByteSelTable : GenericTable {
+  let FilterClass = "VOP3_Pseudo";
----------------
pravinjagtap wrote:

I think, its better to filter it on encoded variants of the instruction. For example, V_CVT_SR_FP8_F32 is also defined for _gfx12_ which uses _VOP3_CVT_SR_F8_ByteSel_Profile_ but _HasFP8DstByteSel_ is _false_ there that makes _HasFP8DstByteSel_ inconsistent. I am not sure whether this hazard needs to be resolved for gfx12 or not (I assume NO) 

https://github.com/llvm/llvm-project/pull/100276