[llvm] [AMDGPU] Legalize strict fp_extend from bf16 (PR #215477)

[Matt Arsenault via llvm-commits]

================
@@ -5006,10 +5006,13 @@ SDValue SITargetLowering::lowerFP_EXTEND(SDValue Op, SelectionDAG &DAG) const {
       DAG.getNode(ISD::BITCAST, SL, SrcVT.changeTypeToInteger(), Src);
 
   EVT DstVT = Op.getValueType();
+  SDValue Result = DAG.getNode(ISD::BF16_TO_FP, SL, DstVT, BitCast);
+  // bf16 -> f32/f64 extension is exact and cannot raise an FP exception, so
+  // the strict chain can pass through unchanged.
----------------
arsenm wrote:

It may raise an exception if the input is snan 

https://github.com/llvm/llvm-project/pull/215477