Serhiy Storchaka <storch...@gmail.com> added the comment: As you can see, the unpatched code does not depend on the alignment. With patches aligned data (which constitute the vast majority, if not all) decoded much faster and non-aligned data decoded sometimes slightly slower. Time of decoding 2-10-bytes practically does not depend on the string length, most of the time (0.1 ms) occupy the overhead of function calls and objects creation and destruction. But even in this case, the patches show a steady increase in performance. When the overhead costs are reduced advantage becomes stronger. For short strings the second patch is better the first patch, as expected.
---------- Added file: http://bugs.python.org/file25051/decode_ascii_compare.ods _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue14419> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
