// computation corresponding to i-th logical device. If the attribute does not
// exist, the all inputs are placed on logical core 0.
mlir::LogicalResult ExtractInputsForLogicalDevices(
    int num_cores_per_replica, mlir::tf_device::ClusterFuncOp cluster_func,
    mlir::OpBuilder* builder,
    llvm::SmallVectorImpl<llvm::SmallVector<mlir::Value, 4>>* input_list);

  // i-th logical core.
  llvm::SmallVector<llvm::SmallVector<mlir::Value, 4>, 4> input_list;
  builder->setInsertionPoint(*new_parallel_execute);
  auto result = tensorflow::ExtractInputsForLogicalDevices(
      num_cores_per_replica, cluster_func, builder, &input_list);
  if (failed(result)) return failure();

  const bool replicated = tpu_devices.size() != 1;

        << hlosharding.status().message();
    return;
  }
  op->setAttr(shard_str,
              mlir::StringAttr::get(op->getContext(), hlosharding->ToString()));
}

mlir::LogicalResult ExtractInputsForLogicalDevices(
    const int num_cores_per_replica,
    mlir::tf_device::ClusterFuncOp cluster_func, mlir::OpBuilder* builder,
    llvm::SmallVectorImpl<llvm::SmallVector<mlir::Value, 4>>* input_list) {