[pypy-issue] [issue718] string.plit('\x00') takes 300% longer under pypy1.5 w/jit

[Robert Collins]

New submission from Robert Collins <robertc at robertcollins.net>:

under cpython:

11:37 < lifeless>         3    0.000    0.000    0.280    0.093 
journals.py:678(_tokens)


under pypy(1.5, 64-bit build, your official download)
11:38 < lifeless>         3    0.000    0.000    1.161    0.387 
journals.py:678(_tokens)

_tokens is defined as

    def _tokens():
        return content.split('\x00')

content in the timed case is a 74MB bytestring consisting of \x00 separated fields; 
the fields are highly redundant - for example one in 5 in the case that produced 
this is 'add', another one in 5 is 'file', and the rest are 'dir', 'symlink' and 
relative paths, and floats (as ascii). There are 3331176 fields 
(len(content.split('\x00'))==3331176)

----------
effort: ???
messages: 2507
nosy: lifeless, pypy-issue
priority: bug
release: ???
status: unread
title: string.plit('\x00') takes 300% longer under pypy1.5 w/jit

_______________________________________________________
PyPy development tracker <pypy-dev-issue at codespeak.net>
<https://codespeak.net/issue/pypy-dev/issue718>
_______________________________________________________