#pypy on 2021-03-16 — irc logs at freenode.irclog.whitequark.org

06:07 <mattip> own tests of module/zipimport are really slow

06:08 <bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/138 [default]

06:08 <bbot2> Started: http://buildbot.pypy.org/builders/own-win-x86-64/builds/215 [default]

06:08 <mattip> on aarch64 they timeout, on x86_64 they take ~20 minutes

06:09 <fijal> wow pypy is fast

06:11 <fijal> I'm analyzing a calendar and it's 8s (pypy) vs 30s (cpy)

06:11 <fijal> cold JIT

06:13 <cfbolz> nice

06:20 <fijal> that's 3.7 btw

06:20 <fijal> but it's parsing, which is not exactly pypy forte

06:20 <fijal> (I think it's predominantly parsing)

06:32 <cfbolz> mattip: I just realized that json speed was already mentioned in an issue last summer: https://foss.heptapod.net/pypy/pypy/-/issues/3238

06:40 <mattip> heh

06:41 <mattip> someone on stackoverflow was complaining about Decimal being slow

06:42 <mattip> it turns out we have a cffi- wrapper for libmpdec, but do not ship it by default

06:43 <mattip> so most users will use the pure-python _pydecimal

06:45 <cfbolz> mattip: wait no

06:45 <cfbolz> are you sure it's that way round?

06:45 <cfbolz> iirc the cffi one is slower than the pure python one

06:45 <cfbolz> and py3 ships the former, py2 the latter

06:47 <mattip> on py3.7, _decimal_build.py is commented out of lib_pypy/pypy_tools/build_cffi_imports.py

06:47 <cfbolz> ok, good

06:47 <cfbolz> because the cffi one is slower ;-)

06:49 <mattip> ahh. right, issue 3024

07:06 <mattip> ok. answered

07:49 <cfbolz> mattip: is there a benchmark?

07:50 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/138 [default]

07:50 <bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/139 [py3.7]

08:14 <bbot2> Success: http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/3218 [default]

08:14 <cfbolz> Found it

08:14 <cfbolz> Yeah well

08:14 <bbot2> Failure: http://buildbot.pypy.org/builders/own-win-x86-64/builds/215 [default]

08:14 <bbot2> Started: http://buildbot.pypy.org/builders/own-win-x86-64/builds/216 [py3.7]

08:14 <cfbolz> Not much to say, I fear

08:39 <cfbolz> mattip: I wonder whether his reproducer shows the real problem

08:39 <cfbolz> but yes, our re should be good at the problem anyway ;-)

08:41 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-linux-aarch64/builds/899 [py3.7]

08:46 <mattip> here is the regex reproducer as a script, with the "a in b" alternative

08:46 <mattip> https://paste.ubuntu.com/p/BHD9HmV6cn/

08:47 <cfbolz> mattip: you will be happy to learn that I suspect that it's the sphinx problem :-P

08:48 <mattip> :)

08:48 <mattip> ahh, got it

08:52 <cfbolz> it's a long chain of bridges to promote the current regex :-(

09:01 <cfbolz> mattip: there's a solution, but I am not sure whether it will make other code worse

09:03 <cfbolz> mattip: can you maybe file an issue about re?

10:00 <bbot2> Started: http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/3219 [py3.7]

10:29 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/139 [py3.7]

11:32 <cfbolz> mattip: I did it now

11:35 <bbot2> Failure: http://buildbot.pypy.org/builders/own-win-x86-64/builds/216 [py3.7]

11:39 <bbot2> Success: http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/3219 [py3.7]

13:08 <mattip> cfbolz: thanks

13:16 <bbot2> Started: http://buildbot.pypy.org/builders/own-win-x86-32/builds/2601 [mattip: 32-bit windows builds, py3.7]

13:16 <bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5763 [mattip: 32-bit windows builds, py3.7]

14:11 <bbot2> Failure: http://buildbot.pypy.org/builders/own-win-x86-32/builds/2601 [mattip: 32-bit windows builds, py3.7]

16:33 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5763 [mattip: 32-bit windows builds, py3.7]

16:54 <bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/7542 [ronan: force build, hpy-update-to-eb07982b6]

16:54 <bbot2> Started: http://buildbot.pypy.org/builders/own-linux-x86-64/builds/8600 [ronan: force build, hpy-update-to-eb07982b6]

18:07 <bbot2> Failure: http://buildbot.pypy.org/builders/own-linux-x86-64/builds/8600 [ronan: force build, hpy-update-to-eb07982b6]

18:22 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/7542 [ronan: force build, hpy-update-to-eb07982b6]

18:29 <bbot2> Started: http://buildbot.pypy.org/builders/own-win-x86-32/builds/2602 [mattip: 332-bit windows builds]

18:29 <bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5764 [mattip: 332-bit windows builds]

19:04 <bbot2> Failure: http://buildbot.pypy.org/builders/own-win-x86-32/builds/2602 [mattip: 332-bit windows builds]

21:28 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5764 [mattip: 332-bit windows builds]