[Repost] PyPy 1.8 50% slower than PyPy 1.7 on my test case
Sorry to repost, but does anyone have an idea about what I could do to track down the source of the slowdown? Should I run with some trace to try to compare how different parts behave? How could I do that? Thanks, Laurent. Hi and first of all, thanks for that great project. Now to my "problem". I'm doing some puzzle-solving, constraint processing with Python and on my particular program, PyPy 1.8 showed a 50% increase in running time over PyPy 1.7. I'm not doing anything fancy (no numpy, itertools, etc.). The program is here: https://github.com/slowfrog/hexiom To reproduce: - fetch hexiom2.py and level36.txt from github - run 'pypy hexiom2.py -sfirst level36.txt' On my machine (Windows 7) the timings are the following: Python 2.6.5/Windows 32 3m35s Python 2.7.1 (930f0bc4125a, Nov 27 2011, 11:58:57) [PyPy 1.7.0 with MSC v.1500 32 bit] 31s Python 2.7.2 (0e28b379d8b3, Feb 09 2012, 18:31:47) [PyPy 1.8.0 with MSC v.1500 32 bit] 48s I'm using the default options. Do you have any idea what could be causing that? Thanks, Laurent.
On Fri, Feb 17, 2012 at 8:23 PM, Laurent Vaucher <laurentvaucher@gmail.com> wrote:
Hi laurent. Good starting points: * run stuff with cProfile and compare * run stuff with valgrind and compare * compare traces * run PYPYLOG=log pypy <your program> and then analyze output using pypy/tool/logparser.py I'm sorry we did not get back to you, but we're quite busy. Thanks! fijal
Hi Laurent, On Fri, Feb 17, 2012 at 19:23, Laurent Vaucher <laurentvaucher@gmail.com> wrote:
Sorry to repost, but does anyone have an idea about what I could do to track down the source of the slowdown?
We're busy, as Fijal said, but your issue is not forgotten. I have already put it there: https://bugs.pypy.org/issue1051 A bientôt, Armin.
Hi Laurent, Follow-up on the issue with your observed slow-down, https://bugs.pypy.org/issue1051 : It seems to be faster again on a pypy trunk nowadays than it was on pypy 1.8; actually it is faster now than with pypy 1.7. Can you report if you're experiencing the same results on Windows? Do you mind if we steal hexiom2.py and level36.txt and put them in https://bitbucket.org/pypy/benchmarks, so that the program will be run nightly and we can spot speed regressions directly? Thanks! Armin.
That's great news! Thanks! I don't mind at all your using my test files. "I hereby grant you all the rights you need to do that, as I am the sole author of this code." I'm coding for fun, not to assert any copyright, patent or anything. I'll try the latest trunk version and tell you how it goes. Thanks again. Laurent. On Sun, Jun 3, 2012 at 3:46 PM, Armin Rigo <arigo@tunes.org> wrote:
Well, with the latest Windows nightly bundle, it's now 10-20% faster than 1.7 and almost twice as fast as 1.8, depending on the duration of the test. I don't know what you did, but you did it damn well! Bravo. Laurent. On Mon, Jun 4, 2012 at 7:02 AM, Laurent Vaucher <laurentvaucher@gmail.com>wrote:
On Fri, Feb 17, 2012 at 8:23 PM, Laurent Vaucher <laurentvaucher@gmail.com> wrote:
Hi laurent. Good starting points: * run stuff with cProfile and compare * run stuff with valgrind and compare * compare traces * run PYPYLOG=log pypy <your program> and then analyze output using pypy/tool/logparser.py I'm sorry we did not get back to you, but we're quite busy. Thanks! fijal
Hi Laurent, On Fri, Feb 17, 2012 at 19:23, Laurent Vaucher <laurentvaucher@gmail.com> wrote:
Sorry to repost, but does anyone have an idea about what I could do to track down the source of the slowdown?
We're busy, as Fijal said, but your issue is not forgotten. I have already put it there: https://bugs.pypy.org/issue1051 A bientôt, Armin.
Hi Laurent, Follow-up on the issue with your observed slow-down, https://bugs.pypy.org/issue1051 : It seems to be faster again on a pypy trunk nowadays than it was on pypy 1.8; actually it is faster now than with pypy 1.7. Can you report if you're experiencing the same results on Windows? Do you mind if we steal hexiom2.py and level36.txt and put them in https://bitbucket.org/pypy/benchmarks, so that the program will be run nightly and we can spot speed regressions directly? Thanks! Armin.
That's great news! Thanks! I don't mind at all your using my test files. "I hereby grant you all the rights you need to do that, as I am the sole author of this code." I'm coding for fun, not to assert any copyright, patent or anything. I'll try the latest trunk version and tell you how it goes. Thanks again. Laurent. On Sun, Jun 3, 2012 at 3:46 PM, Armin Rigo <arigo@tunes.org> wrote:
Well, with the latest Windows nightly bundle, it's now 10-20% faster than 1.7 and almost twice as fast as 1.8, depending on the duration of the test. I don't know what you did, but you did it damn well! Bravo. Laurent. On Mon, Jun 4, 2012 at 7:02 AM, Laurent Vaucher <laurentvaucher@gmail.com>wrote:
participants (3)
-
Armin Rigo -
Laurent Vaucher -
Maciej Fijalkowski