[llvm] [AMDGPU] Fix canonicalization of truncated values. (PR #83054)

[Harald van Dijk via llvm-commits]

================
@@ -194,7 +194,26 @@ class HasOneUseTernaryOp<SDPatternOperator op> : PatFrag<
   }];
 }
 
-class is_canonicalized<SDPatternOperator op> : PatFrag<
+class is_canonicalized_1<SDPatternOperator op> : PatFrag<
+  (ops node:$src0),
+  (op $src0),
+  [{
+    const SITargetLowering &Lowering =
+              *static_cast<const SITargetLowering *>(getTargetLowering());
+
+    return Lowering.isCanonicalized(*CurDAG, N->getOperand(0));
+   }]> {
+
+  // TODO: Improve the Legalizer for g_build_vector in Global Isel to match this class
+  let GISelPredicateCode = [{
+    const SITargetLowering *TLI = static_cast<const SITargetLowering *>(
+      MF.getSubtarget().getTargetLowering());
+
+    return TLI->isCanonicalized(MI.getOperand(1).getReg(), const_cast<MachineFunction&>(MF));
----------------
hvdijk wrote:

Good question, I can take a look at that later today. This is basically just copied and pasted from the existing `is_canonicalized` form that already uses `const_cast`, but if we can change that, that would be good.

https://github.com/llvm/llvm-project/pull/83054