[llvm] [AMDGPU] Recognise bitmask operations as srcmods (PR #149110)

[Shilei Tian via llvm-commits]

================
@@ -3036,6 +3036,41 @@ bool AMDGPUDAGToDAGISel::SelectVOP3ModsImpl(SDValue In, SDValue &Src,
     Src = Src.getOperand(0);
   }
 
+  // Convert various sign-bit masks 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.
+  // TODO: Add 16-bit support.
+  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))
+    return true;
+
+  ConstantSDNode *CRHS = dyn_cast<ConstantSDNode>(Src->getOperand(1));
+  if (!CRHS)
+    return true;
+
+  // Recognise (xor a, 0x80000000) as NEG SrcMod.
+  if (Opc == ISD::XOR && CRHS->getAPIntValue().isSignMask()) {
+    Mods |= SISrcMods::NEG;
+    Src = Src.getOperand(0);
+
----------------
shiltian wrote:

unnecessary empty line

https://github.com/llvm/llvm-project/pull/149110