[llvm] [AMDGPU] Recognise bitmask operations as srcmods on select (PR #152119)

[Chris Jackson via llvm-commits]

================
@@ -3036,6 +3036,42 @@ bool AMDGPUDAGToDAGISel::SelectVOP3ModsImpl(SDValue In, SDValue &Src,
     Src = Src.getOperand(0);
   }
 
+  // Convert various sign-bit masks on integers to src mods. Currently disabled
+  // for 16-bit types as the codegen replaces the operand without adding a
+  // srcmod. This is intentionally finding the cases where we are performing
+  // float neg and abs on int types, the goal is not to obtain two's complement
+  // neg or abs. Limit converison to select operands via the nonCanonalizing
+  // pattern.
+  // TODO: Add 16-bit support.
+  if (IsCanonicalizing)
+    return true;
+
+  unsigned Opc = Src->getOpcode();
+  EVT VT = Src.getValueType();
+  if ((Opc != ISD::AND && Opc != ISD::OR && Opc != ISD::XOR) ||
+      (VT != MVT::i32 && VT != MVT::v2i32 && VT != MVT::i64))
----------------
chrisjbris wrote:

I'm sorry but I don't understand why only one bit would be flipped. Wouldn't the src modifier still be placed on the whole operand?
I'm not sure how to resolve this.

https://github.com/llvm/llvm-project/pull/152119