cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end ) | use cffi for calling C | if a pep adds a mere 25-30 [C-API] functions or so, it's a drop in the ocean (cough) - Armin
bionicpup64_2042 has quit [Ping timeout: 246 seconds]
bionicpup64_2042 has joined #pypy
bionicpup64_2042 has quit [Ping timeout: 245 seconds]
<kenaan>
antocuni hpy abdeda0fd3c6 /pypy/: some progress in fixing test_ztranslation, which includes improving the fake objspace
<kenaan>
antocuni hpy b1e1f0479f6b /pypy/module/hpy_universal/_vendored/include/universal/hpy.h: update_vendored to e481b58; this change was made in the hpy-ctypespace branch but needed to be ported to pyhandle/hp...
<kenaan>
antocuni hpy 987c3b5d033e /pypy/module/hpy_universal/apiset.py: give a better name to this helper, it makes it much easier to debug translation errors
<kenaan>
antocuni hpy 121bac5f857c /pypy/: progress towards making test_ztranslation working again: these changes are needed to that functions inside hpy_unive...
<kenaan>
antocuni hpy 4063b6b8b3f3 /pypy/module/hpy_universal/: fix the last bit of translation
<kenaan>
antocuni hpy 4aa8318c0c07 /pypy/objspace/fake/objspace.py: finally fix hpy_universa:test_ztranslation: we need to ensure that the annotator sees some concrete implementation o...
<kenaan>
antocuni hpy e042625501a0 /pypy/: add an option to show Pdb+ in ztranslation tests, so that it is easier to debug translation errors
xcm is now known as Guest72826
Guest72826 has quit [Killed (karatkievich.freenode.net (Nickname regained by services))]
<antocuni>
they are not very scientific: I ran them on my laptop so there are issues about thermal throttling, high variance, etc
<Alex_Gaynor>
But it seems definitely in the same ballpark, and maybe a bit faster.
<Alex_Gaynor>
and pypy is a bit slower than cpython for some reason
<Alex_Gaynor>
antocuni: what is hpy vs. hpy_universal? Universal you can build once and use with both pypy and cpython?
<antocuni>
I think I ran cpython first (although I pasted it later), so it might be that my cpu was less throttled
<antocuni>
Alex_Gaynor: yes
<Alex_Gaynor>
antocuni: Very cool.
<antocuni>
you have two sets of hpy headers: one is cpython-specific and basically translates hpy API into C-API using macros and static inline functions
<Alex_Gaynor>
antocuni: I guess that means various macros are really function calls or so?
<antocuni>
one is "universal" and builds a dll which doesn't link to any external symbol at all. All the API functions are called through a function pointer in the "context" object which you pass around
<antocuni>
Alex_Gaynor: yes
<antocuni>
in particular, in universal mode every incref/decref become function calls to hpy_dup/hpy_close
<antocuni>
so I am surprised that the performance is not so bad on cpython :)
tsaka__ has joined #pypy
<antocuni>
FWIW, another run of the benchmark (this time running hpy_universal before json): as I said, very high variance: http://paste.openstack.org/show/786999/
marky1991 has joined #pypy
hsaliak has quit [Ping timeout: 265 seconds]
hsaliak has joined #pypy
<antocuni>
so, benchmarks again: this time I ran them on bencher4 and there is no significative variance between different runs, so they are probably more meaningful: http://paste.openstack.org/show/787004/
<antocuni>
I'm a bit surprised that ujson_hpy_universal is ~23% slower on pypy than cpython. I am not really sure where the slowdown could come from
<Alex_Gaynor>
antocuni: is somebody working on hyp from cpython, so theoretically it can become the main cpython API?
marky1991 has quit [Ping timeout: 246 seconds]
marky1991 has joined #pypy
_whitelogger_ has joined #pypy
mwhudson has quit [Disconnected by services]
mwhudson_ is now known as mwhudson
mattip_ has joined #pypy
Viech has quit [*.net *.split]
marvin_ has quit [*.net *.split]
altendky has quit [*.net *.split]
_whitelogger has quit [*.net *.split]
dansan has quit [*.net *.split]
mattip has quit [*.net *.split]
Taggnostr has quit [*.net *.split]
wleslie has quit [*.net *.split]
simpson has quit [*.net *.split]
kirma has quit [*.net *.split]
michelp has quit [*.net *.split]
phlebas has quit [*.net *.split]
panni_ has quit [*.net *.split]
phlebas_ is now known as phlebas
panni__ has joined #pypy
michelp_ is now known as michelp
altendky_ is now known as altendky
dansan_ is now known as dansan
marvin_ has joined #pypy
simpson has joined #pypy
<tos9>
antocuni: (I have nothing useful to say besides you are great)