[mlir][Tensor] Apply ClangTidy performance finding (NFC)

diff --git a/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp b/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
index 9841165..4e95243 100644 (file)
--- a/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
+++ b/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
@@ -3330,7 +3330,7 @@ ShapedType PackOp::inferPackedType(ShapedType sourceType,
                                    ArrayRef<int64_t> innerDimsPos,
                                    ArrayRef<int64_t> outerDimsPerm) {
   SmallVector<int64_t> resultShape = llvm::to_vector(sourceType.getShape());
-  for (auto tiledDim : llvm::enumerate(innerDimsPos)) {
+  for (const auto &tiledDim : llvm::enumerate(innerDimsPos)) {
     if (ShapedType::isDynamic(resultShape[tiledDim.value()]))
       continue;
     if (ShapedType::isDynamic(innerTileSizes[tiledDim.index()])) {