Avoid clone

Author: hsjts0u

mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp

@@ -990,18 +990,18 @@ tileAndFuseFirstExtractUse(RewriterBase &rewriter, Diagnostic &diag,
   // from the bbArg instead. This allows to reuse the output tensor (instead of
   // creating a new one) of the container when both producer and container write
   // to the same output.
+  bool cloned = false;
   if (LoopLikeOpInterface containerLoop =
-          dyn_cast<LoopLikeOpInterface>(sliceOpToTile->getParentOp())) {
+          dyn_cast<LoopLikeOpInterface>(sliceOpToTile->getParentOp());
+      containerLoop && dyn_cast<DestinationStyleOpInterface>(producerOp)) {
     Operation *clone = rewriter.clone(*producerOp);
+    cloned = true;
     rewriter.modifyOpInPlace(clone, [&]() {
       // Iterate over the outputs of the producer and over the loop bbArgs and
       // check if any bbArg points to the same value as the producer output. In
       // such case, make the producer output point to the bbArg directly.
-      auto dpsInterface = dyn_cast<DestinationStyleOpInterface>(clone);
-      if (!dpsInterface)
-        return;
-
-      for (OpOperand &initOperandPtr : dpsInterface.getDpsInitsMutable()) {
+      for (OpOperand &initOperandPtr :
+           cast<DestinationStyleOpInterface>(clone).getDpsInitsMutable()) {
         Value producerOperand =
             clone->getOperand(initOperandPtr.getOperandNumber());
         for (BlockArgument containerIterArg :
@@ -1063,7 +1063,7 @@ tileAndFuseFirstExtractUse(RewriterBase &rewriter, Diagnostic &diag,
       resultNumber, offsets, sizes);
 
   // Cleanup clone.
-  if (dyn_cast<LoopLikeOpInterface>(containingOp))
+  if (dyn_cast<LoopLikeOpInterface>(containingOp) && cloned)
     rewriter.eraseOp(tileableProducer);
 
   return std::make_tuple(tileAndFuseResult->tiledOps, newContainingOp);