cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://botbot.me/freenode/pypy/ ) | use cffi for calling C | "nothing compares to the timeshifter, my personal polar expedition in software" - pedronis
<kenaan_>
arigo cpyext-avoid-roundtrip d519061c7cf0 /rpython/rtyper/lltypesystem/: (antocuni, arigo) Aaaaah. Found that this simple new test in test_ll2ctypes doesn't actually pass....
<kenaan_>
arigo cpyext-avoid-roundtrip 087ef45836c2 /pypy/module/cpyext/: (antocuni, arigo) Move subtype_dealloc to C. Lots of messes that might not be necessary any more t...
<kenaan_>
arigo cpyext-avoid-roundtrip 5593c8b50dc4 /rpython/rtyper/lltypesystem/: (antocuni, arigo) More.
<kenaan_>
arigo cpyext-avoid-roundtrip 05c413f37f2c /pypy/module/cpyext/: (antocuni, arigo) Simplification again, using the fixes in ll2ctypes.
<jantje>
hi! I've downloaded the pre-built python2.7 compatbible pypy 5.8.0, and I've quicky inserted it as a 'drop in' replacement in a software build project, but first measurements show that pypy is executing 50% slower than the actual python2.7. So I think I must be doing something wrong.
<jantje>
a lot of python processes are spawned, so some are quite short actually, but others run for a couple of minutes so that is also where I would expect it to be faster.
ronan has joined #pypy
antocuni has joined #pypy
<jantje>
anything I can check/debug ?
forgottenone has joined #pypy
rokujyouhitoma has joined #pypy
<kenaan_>
antocuni cpyext-avoid-roundtrip 63bf91dc13e5 /pypy/module/cpyext/: (antocuni, arigo): move these two api functions to C
rokujyouhitoma has quit [Ping timeout: 248 seconds]
<fijal>
jantje: hi, are you summing up the total?
<fijal>
jantje: what is your program doing?
<fijal>
jantje: I would suggest having a quick look at vmprof
<fijal>
the_drow: nope, I think we gonna announce there is no real interest
<the_drow>
:(
<fijal>
tos9: not a regression
<the_drow>
Why not open it to the public like STM?
<tos9>
fijal: cool -- is boto a particular bad offender? It just uses pyopenssl no?
<fijal>
it uses ssl module
<fijal>
but also it uses a connection pool
<fijal>
and *also* each connection from connection pool takes ~1M to load the cert file
<fijal>
the_drow: how much money do you think we raised for STM?
<the_drow>
40k or so?
<the_drow>
there were 2 calls for donations
<fijal>
ok, so SFC takes 10%
<fijal>
that leaves us with 35k or so (note that most of the donations were from corpos btw)
<fijal>
most in terms of $ value
<fijal>
which is this weird number that ends up working in a way that we have to do something at a steeply discounted rate
<fijal>
which is ok, but then you need to do something else to fund activities that are not fundable
yuyichao has quit [Ping timeout: 260 seconds]
<the_drow>
I'm not claiming you should work on this at a discounted rate. I think that given enough time, people will contribute enough money for you guys to work on it at your normal rate
realitix has quit [Ping timeout: 264 seconds]
yuyichao has joined #pypy
exarkun has quit [Ping timeout: 258 seconds]
exarkun has joined #pypy
<kenaan_>
antocuni cpyext-avoid-roundtrip 98f6749ae4ed /pypy/module/cpyext/: (antocuni, arigo): port more functions to C, and mark some others which we would like to port
rokujyouhitoma has joined #pypy
antocuni has quit [Ping timeout: 240 seconds]
<jantje>
fijal: ok, thanks, will do.
rokujyouhitoma has quit [Ping timeout: 260 seconds]
<fijal>
the_drow: that's blatantly not the case
<fijal>
the_drow: none of our campaigns ever reached the goal
<fijal>
and expectations are sky high
arigato has joined #pypy
oberstet has quit [Ping timeout: 240 seconds]
rokujyouhitoma has joined #pypy
rokujyouhitoma has quit [Ping timeout: 255 seconds]
<fijal>
pjenvey: should we just reuse the cert store in wrap_socket?
<fijal>
that seems easy enough
<pjenvey>
well libs need to do it
<pjenvey>
some libs definitely need to be improved
<fijal>
why libs?
<pjenvey>
fijal: pypy not keeping up with finalizing these things though is a problem
<fijal>
as in, why can't we just cache it?
<fijal>
well, it's also a bit insane
<fijal>
like you get 500 sockets, puff you have 500M of shit
<pjenvey>
i dunno. maybe ssl.wrap_socket could cache
<pjenvey>
libs should probably be smarter about them anyway though, like allowing you to pass in a custom one, otherwise creating its own 1 context, reusing it