[Mlir-commits] [mlir] [Mlir] decompose generic by unfolding projected permutation crash fix (PR #122449)

[Adam Siemieniuk]

================
@@ -159,6 +159,15 @@ LogicalResult DecomposeProjectedPermutation::matchAndRewrite(
     auto map = op.getMatchingIndexingMap(&opOperand);
     if (!map.isProjectedPermutation(false))
       return failure();
+
+    // If we have any inputs that aren't memref or ranked tensor types, reject
+    // the pattern.
+    if (!dyn_cast<ShapedType>(opOperand.get().getType()))
+      return op->emitError("Expected operand #")
----------------
adam-smnk wrote:

Thanks 👍
For the test, it is fine to have an invalid/negative examples and add `CHECK` statements to verify nothing has changed as expected.

Actually, could you add your test to the other `decompose-generic-by-unfolding-projected-permutation.mlir` file and just follow the existing convention.

https://github.com/llvm/llvm-project/pull/122449