[Mlir-commits] [mlir] [mlir][linalg] Bail out tensor.cast pack/unpack fold on unprovable tile sizes (PR #188000)

[Renato Golin]

================
@@ -6476,16 +6485,20 @@ struct FoldTensorCastUnPackOp : public OpRewritePattern<UnPackOp> {
     Value sourceTensor = newOperands[0];
 
     // Get the updated mixed-tile-sizes attribute.
-    SmallVector<OpFoldResult> newMixedTileSizes = getNewMixedTileSizes(
-        rewriter, sourceTensor.getType(), op.getMixedTiles());
+    FailureOr<SmallVector<OpFoldResult>> newMixedTileSizes =
----------------
rengolin wrote:

```suggestion
    auto newMixedTileSizes =
```

https://github.com/llvm/llvm-project/pull/188000