[llvm] [AMDGPU] Fold fneg into select source modifiers (PR #201426)

[Matt Arsenault via llvm-commits]

================
@@ -1725,6 +1725,106 @@ bb5:                                              ; preds = %bb, %.entry
 }
 
 
+; f32 select(fneg, fneg): both fnegs fold into the cndmask source modifiers.
+define float @select_fneg_fneg_f32(i1 %cond, float %arg0, float %arg1) {
----------------
arsenm wrote:

Are you sure identical tests don't already exist?

https://github.com/llvm/llvm-project/pull/201426