// index |cluster_func_output_index|. Regions of parallel_execute may
// have different return values depending on output sharding configuration.
mlir::LogicalResult LookupClusterToCoreIndex(
    const mlir::Location& location,
    llvm::SmallVector<llvm::SmallVector<int, 4>, 4> cluster_to_core_index,
    const int core_id, const int cluster_func_output_index,
    int* core_output_index) {