>>106962091
actually tried it and the end-pointer and span version is 60% faster, not slower