[llvm] [AMDGPU] [DO NOT MERGE] Nonsuccessful Attempt At Using SelectionDAG Hooks for abs i8/i16 (PR #167064)

[Matt Arsenault via llvm-commits]

================
@@ -8139,6 +8157,25 @@ SDValue SITargetLowering::lowerDEBUGTRAP(SDValue Op, SelectionDAG &DAG) const {
   return DAG.getNode(AMDGPUISD::TRAP, SL, MVT::Other, Ops);
 }
 
+// sign-extend and use the 32-bit ABS operation for 16-bit ABS with SGPRs
+SDValue SITargetLowering::lowerABSi16(SDValue Op, SelectionDAG &DAG) const {
+  assert(Op.getOpcode() == ISD::ABS &&
+         "Tried to select abs with non-abs opcode.");
+  assert((Op.getValueType() == MVT::i16 || Op.getValueType() == MVT::i8) &&
+         "Tried to select abs i16 lowering with non-i16 type.");
+
+  // divergent means will not end up using SGPRs
+  if (Op->isDivergent())
+    return SDValue();
+
+  //(abs i16 (i16 op1)) -> (trunc i16 (abs i32 (sext i32 (i16 op1))))
----------------
arsenm wrote:

```suggestion
  // (abs i16 (i16 op1)) -> (trunc i16 (abs i32 (sext i32 (i16 op1))))
```

https://github.com/llvm/llvm-project/pull/167064