[llvm] [NVPTX] Propagate truncate to operands (PR #98666)

[Artem Belevich via llvm-commits]

================
@@ -5541,6 +5541,53 @@ static SDValue PerformREMCombine(SDNode *N,
   return SDValue();
 }
 
+// truncate (logic_op x, y) --> logic_op (truncate x), (truncate y)
+// This will reduce register pressure.
+static SDValue PerformTruncCombine(SDNode *N,
+                                   TargetLowering::DAGCombinerInfo &DCI) {
+  if (!DCI.isBeforeLegalizeOps())
+    return SDValue();
+
+  SDValue LogicalOp = N->getOperand(0);
+  switch (LogicalOp.getOpcode()) {
+  default:
+    break;
+  case ISD::ADD:
+  case ISD::SUB:
+  case ISD::MUL:
----------------
Artem-B wrote:

Nit. We could push the optimization for `mul` even further.
If the result is truncated to 32-bit, we could convert it to mul.wide of 16-bit operands. Not sure, though, if it's actually going to help on the SASS level, though. This is probably something that ptxas can deal with.


https://github.com/llvm/llvm-project/pull/98666