#pypy IRC log for Monday, 2010-02-15

ebo^ (~ebolad@233.195.116.85.dsl.manitu.net) left irc: Quit: Verlassend00:44
santagada (~santagada@201.86.197.164.dynamic.adsl.gvt.net.br) joined #pypy.01:18
nettok (~netto@200.119.153.5) joined #pypy.01:22
Nick change: chesttone -> elbowtone01:31
nshepperd (~nshepperd@2001:44b8:61::23) joined #pypy.01:50
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 260 seconds03:00
fijal (~fijal@rrcs-24-227-109-37.se.biz.rr.com) joined #pypy.03:47
kenaan03fijal 10r7123514 07M(pypy/extradoc/talk/pycon2010/pypyspeed/talk.txt): A bit of progress, probably last checkin before I start real slides ...03:56
ousado (~ousado@port-92-206-139-119.dynamic.qsc.de) joined #pypy.04:05
kindly (~david@cpc3-lewi8-0-0-cust995.bmly.cable.ntl.com) left irc: Ping timeout: 256 seconds04:08
ousado_ (~ousado@port-92-206-114-37.dynamic.qsc.de) left irc: Ping timeout: 252 seconds04:08
kenaan03fijal 10r7123614 pypy/extradoc/talk/pycon2010/keynote 07M(talk.pdf talk.tex): typos and minor fixes ...04:16
gutworthfijal: you say "for example" too much :)04:20
fijalgutworth: feel like fixing that?04:21
fijaland hi04:22
gutworthhi04:22
kenaan03benjamin 10r7123714 07M(pypy/extradoc/talk/pycon2010/crossinterp/talk.txt): remove a little awkwardness ...04:24
fijalgutworth: that's samuele04:24
gutworthah, sorry then04:27
fijalthat'04:31
fijals my talk though04:31
fijaltoo bad you can't make it to pycon04:32
bja (~bja@c-98-222-1-220.hsd1.il.comcast.net) left irc: Read error: Connection reset by peer04:43
kenaan03fijal 10r7123814 pypy/extradoc/talk/pycon2010 05A(/common/ beamerouterthememy.sty beamerthemeWarsaw.sty merlinux-logo.png pypy-logo.png beamerouterthememy.sty beamerthemeWarsaw.sty) 06D(beamerouterthememy.sty beamerthemeWarsaw.st ...04:46
stakkars (~tismer@i59F7E9F9.versanet.de) left irc: Read error: Connection reset by peer05:16
stakkars (~tismer@i59F7E9F9.versanet.de) joined #pypy.05:17
nsf (~nsf@jiss.convex.ru) joined #pypy.05:19
panni_ (hannes@ip-95-222-52-93.unitymediagroup.de) left irc: Read error: Connection reset by peer05:28
stakkars (~tismer@i59F7E9F9.versanet.de) left irc: Read error: Connection reset by peer05:28
hikoz (~hikoz@27.183.244.43.ap.yournet.ne.jp) joined #pypy.05:29
stakkars (~tismer@i59F7E9F9.versanet.de) joined #pypy.05:29
stakkars (~tismer@i59F7E9F9.versanet.de) left irc: Read error: Connection reset by peer05:38
Alex_Gaynor (~alex@nidoran-37.dynamic.rpi.edu) left irc: Read error: Connection reset by peer05:39
stakkars (~tismer@i59F7E9F9.versanet.de) joined #pypy.05:39
Alex_Gaynor (~chatzilla@mankey-36.dynamic.rpi.edu) joined #pypy.05:42
fijalAlex_Gaynor: hi05:43
Alex_Gaynorhey05:43
stakkars (~tismer@i59F7E9F9.versanet.de) left irc: Read error: Connection reset by peer05:45
stakkars (~tismer@i59F7E9F9.versanet.de) joined #pypy.05:46
kenaan03fijal 10r7123914 pypy/extradoc/talk/pycon2010/pypyspeed 05A(beamerouterthememy.sty beamerthemeWarsaw.sty talk.pdf talk.tex): start working on talk ...05:50
Alex_Gaynor (~chatzilla@mankey-36.dynamic.rpi.edu) left irc: Ping timeout: 240 seconds06:20
Alex_Gaynor (~chatzilla@mankey-36.dynamic.rpi.edu) joined #pypy.06:21
fijal (~fijal@rrcs-24-227-109-37.se.biz.rr.com) left irc: Ping timeout: 256 seconds06:34
Alex_Gaynor (~chatzilla@mankey-36.dynamic.rpi.edu) left irc: Read error: Connection reset by peer06:47
santagada (~santagada@201.86.197.164.dynamic.adsl.gvt.net.br) left irc: Remote host closed the connection06:50
Alex_Gaynor (~chatzilla@mankey-36.dynamic.rpi.edu) joined #pypy.07:02
ElectronicRU (~Electroni@pppoe2478.mv.ru) left irc: Remote host closed the connection07:11
kursor (~kursor@p4FC0A27E.dip0.t-ipconnect.de) joined #pypy.07:13
mwhudson (~mwh@canonical/launchpad/mwhudson) left irc: Ping timeout: 276 seconds08:09
nettok (~netto@200.119.153.5) left irc: Quit: Ex-Chat08:13
victorpoluceno (~victorpol@187.52.151.142) left irc: Ping timeout: 256 seconds08:23
kursor (~kursor@p4FC0A27E.dip0.t-ipconnect.de) left irc: Read error: Connection reset by peer08:45
kursor (~kursor@p4FC0A27E.dip0.t-ipconnect.de) joined #pypy.08:46
kursor_ (~kursor@p4FC0A27E.dip0.t-ipconnect.de) joined #pypy.08:55
kursor (~kursor@p4FC0A27E.dip0.t-ipconnect.de) left irc: Ping timeout: 256 seconds08:55
Nick change: kursor_ -> kursor08:55
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy.09:02
kursor (~kursor@p4FC0A27E.dip0.t-ipconnect.de) left irc: Ping timeout: 256 seconds09:11
tav (~tav@78.149.149.195) left irc: Quit: tav09:13
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 256 seconds09:18
Alex_Gaynor (~chatzilla@mankey-36.dynamic.rpi.edu) left irc: Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158]09:37
antocuni (~antocuni@host71-123-dynamic.6-79-r.retail.telecomitalia.it) joined #pypy.09:47
lucian (~lucian@cpc6-newc9-0-0-cust64.gate.cable.virginmedia.com) left irc: Remote host closed the connection09:51
antocunihi09:54
pedronis (~user@2001:16d8:ffca:0:20f:b5ff:fe44:40b4) joined #pypy.10:12
arigato (~arigo@adsl-188-155-126-154.adslplus.ch) joined #pypy.10:48
ebo^ (~ebolad@233.195.116.85.dsl.manitu.net) joined #pypy.10:53
hikoz (~hikoz@27.183.244.43.ap.yournet.ne.jp) left irc: Ping timeout: 260 seconds11:33
stakkars_ (~tismer@i59F7E9F9.versanet.de) joined #pypy.11:37
zejn (~Gasper@217.72.91.237) joined #pypy.11:38
stakkars__ (~tismer@i59F7E9F9.versanet.de) joined #pypy.11:39
stakkars_ (~tismer@i59F7E9F9.versanet.de) left irc: Ping timeout: 252 seconds11:42
zejn (~Gasper@217.72.91.237) left irc: Ping timeout: 245 seconds11:52
antocuniso, who makes me a short summary of what happened in pypy last week? :-)11:55
arigatohi antonio.  I would be glad to hear it too :-)12:00
Action: arigato fights a bit with Psyco to support the new bytecodes of Python 2.712:01
zejn (~Gasper@217.72.91.237) joined #pypy.12:01
arigatothey added stuff like POP_JUMP_IF_FALSE which is a bit messy12:02
arigatobecause there are so many semi-implicit constrains in psyco12:02
arigatoI'm busy trying to remember them all :-(12:02
pedronisantocuni: not much, some pycon preparations, cfbolz at some point maybe before merged oprofile support12:03
antocunipedronis: ok, thanks12:03
antocuniI see that html5lib is much faster now12:04
antocuniwhat happened?12:04
antocunihi armin!12:04
pedroniscfbolz rewrote some codecs at interp level12:04
antocuniah, nice12:04
pedroniswe looked a bit into generators as well, some ideas but nothing on trunk yet12:04
pedronisI arigato is still trying to debug some crashes12:05
pedroniss/I arigato/I think arigato/12:05
arigatoyes :-(12:06
pedronisI'm leaving tomorrow for pycon so not much going on here12:06
zejn (~Gasper@217.72.91.237) left irc: Ping timeout: 245 seconds12:13
lucian_ (~80f0e503@gateway/web/freenode/x-iwzfupvhqeadlnwz) joined #pypy.12:20
cfbolz (~cfbolz@dhcp-41-80.vpn.uni-hamburg.de) joined #pypy.12:47
pedroniscfbolz: hi12:49
cfbolzhi12:50
tmu (~tmu@evil.oulu.nixu.fi) joined #pypy.12:50
stakkars__ (~tismer@i59F7E9F9.versanet.de) left irc: Quit: stakkars__12:55
zejn (~Gasper@217.72.91.237) joined #pypy.13:04
tav (~tav@78.149.149.195) joined #pypy.13:08
stakkars_ (~tismer@i59F7E9F9.versanet.de) joined #pypy.13:08
lucian_ (~80f0e503@gateway/web/freenode/x-iwzfupvhqeadlnwz) left irc: Quit: Page closed13:12
ElectronicRU (~Electroni@pppoe2478.mv.ru) joined #pypy.13:21
amaury_ (~5bc34844@gateway/web/freenode/x-rffoegydjlssuuxv) joined #pypy.13:23
lizardo (~lizardo@189.2.128.130) joined #pypy.13:25
ElectronicRU (~Electroni@pppoe2478.mv.ru) left irc: Remote host closed the connection13:27
amaury_ (~5bc34844@gateway/web/freenode/x-rffoegydjlssuuxv) left irc: Ping timeout: 252 seconds13:30
ElectronicRU (~Electroni@pppoe2478.mv.ru) joined #pypy.13:43
ElectronicRU (~Electroni@pppoe2478.mv.ru) left irc: Remote host closed the connection13:47
ElectronicRU (~Electroni@pppoe2478.mv.ru) joined #pypy.13:51
fuzzyman (~voidspace@87-194-212-65.bethere.co.uk) joined #pypy.14:05
ElectronicRU (~Electroni@pppoe2478.mv.ru) left irc: Remote host closed the connection14:12
panni_ (hannes@ip-95-222-52-93.unitymediagroup.de) joined #pypy.14:22
stakkars (~tismer@i59F7E9F9.versanet.de) left irc: Ping timeout: 276 seconds14:36
stakkars_ (~tismer@i59F7E9F9.versanet.de) left irc: Ping timeout: 248 seconds14:36
CIA-10 (cia@208.69.182.149) joined #pypy.14:40
stakkars (~tismer@i59F7CD44.versanet.de) joined #pypy.14:42
stakkars_ (~tismer@i59F7CD44.versanet.de) joined #pypy.14:43
arigo (~arigo@adsl-188-155-250-89.adslplus.ch) joined #pypy.14:45
arigato (~arigo@adsl-188-155-126-154.adslplus.ch) left irc: Disconnected by services14:45
Nick change: arigo -> arigato14:45
zejn (~Gasper@217.72.91.237) left irc: Quit: Leaving15:02
voidspace (~voidspace@87-194-212-65.bethere.co.uk) joined #pypy.15:03
fuzzyman (~voidspace@87-194-212-65.bethere.co.uk) left irc: Ping timeout: 240 seconds15:03
Nick change: voidspace -> fuzzyman15:03
sim2409 (~kvirc@92.117.184.46) joined #pypy.15:26
Action: tav waves anyone around ?15:30
cfbolzsomeone always is15:31
tavgot a stable: how stable would a non-jit'd pypy interpreter running in sandboxed "mode" be? i.e. will it run w/o crashing?15:31
tavcfbolz =)15:31
tavs/stable:/question:/15:31
nshepperd (nshepperd@2001:44b8:61::23) left #pypy.15:32
cfbolznon-jitted is very stable by now15:32
tavcool15:33
tavknow how importing works when sandboxed? how do modules get imported when the interpreter doesn't have access to the filesystem ?15:33
cfbolzyou can fake a file-system15:34
lucian (~lucian@cpc6-newc9-0-0-cust64.gate.cable.virginmedia.com) joined #pypy.15:36
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy.15:42
cfbolz (~cfbolz@dhcp-41-80.vpn.uni-hamburg.de) left irc: Quit: Leaving15:53
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 265 seconds16:00
fijal (~fijal@rrcs-24-227-109-37.se.biz.rr.com) joined #pypy.16:13
fijalhi16:18
pedronishi16:18
fijalpedronis: when do you fly?16:18
pedronistomorrow16:18
arigatohi16:19
fijalpedronis: thanks for cross interp slides16:21
fijalI think this is mostly it16:21
fijalpedronis: is "inter peter" something real?16:23
fijalor just a typo?16:23
pedronisoops typo16:24
kenaan03fijal 10r7124214 07M(pypy/extradoc/talk/pycon2010/vmsummit/talk.txt): my affiliation & typo ...16:26
bja (~bja@c-98-222-1-220.hsd1.il.comcast.net) joined #pypy.16:26
pedronisfijal: the crossinterp ones needs to be transformed into a real slide format16:27
arigato (~arigo@adsl-188-155-250-89.adslplus.ch) left irc: Read error: Connection reset by peer16:27
fijalpedronis: yeah16:27
fijalpedronis: I'm working on pypyspeed one now16:27
fijalbut will do that16:27
pedronisok16:27
pedronisI will soonish leave to go home finish packing16:27
fijalsure16:28
Alex_Gaynor (~alex@nidoran-37.dynamic.rpi.edu) joined #pypy.16:30
pedronisfijal: seems we should try to be at the sync meeting thurs16:32
pedronis(release planning again)16:32
fijalfine16:33
pedronisit's going to be during the lang summit, no?16:33
fijalyes16:33
fijalholger won't be there16:33
fijalsince he has a tutorial16:33
sim2409 (~kvirc@92.117.184.46) left irc: Quit: KVIrc 3.2.0 'Realia'16:33
pedroniswell he is going to be traveling around during the release window anyway, no16:34
pedronisisn't he back properly only in april?16:34
fijalI don't know16:35
arigato (~arigo@adsl-188-155-250-89.adslplus.ch) joined #pypy.16:48
magcius (~jstpierre@c-24-61-243-212.hsd1.ma.comcast.net) left irc: Read error: Connection reset by peer16:49
magcius (~jstpierre@c-24-61-243-212.hsd1.ma.comcast.net) joined #pypy.16:50
fijalpedronis: did you see my mail?16:51
kursor (~kursor@p4FC082A6.dip0.t-ipconnect.de) joined #pypy.16:52
gutworth (~benjamin@64-131-1-181.usfamily.net) left irc: Ping timeout: 240 seconds16:54
pedronisfijal: yes16:55
fijalany thoughts?16:55
pedronisI don't know, the problem with trunk is that is of course of moving target16:55
fijalI sort of don't believe myself16:55
fijalI might have local changes or something16:56
fijaleven though I think I checked16:56
fijalso would be good if someone can reproduce it16:56
fijalwell, but I guess it's not for today16:56
pedronisthe wait to have assembler stuff?16:56
pedroniscfbolz started working a bit on that branch16:56
voidspace (~voidspace@87-194-212-65.bethere.co.uk) joined #pypy.16:56
fijaler16:57
fijalparse error16:57
fijal"the wait to have assembler stuff?"16:57
fijalwhat do you mean?16:57
fijalyou mean abort_no_asm branch?16:57
pedronisyes16:57
fijalhow is it twisted-related?16:58
pedronisI don't know16:58
pedronisjust wondering if you had that change still in your working copy16:58
arigato(sorry, are you talking about a mail to samuele only or is there an issue with my mailbox?)16:59
pedronisit's a mail only to me16:59
fuzzyman (~voidspace@87-194-212-65.bethere.co.uk) left irc: Ping timeout: 272 seconds16:59
Nick change: voidspace -> fuzzyman16:59
pedronisthough is really a general question16:59
pedroniswhether to swith the twisted benchmarks to twisted trunk16:59
pedronisbecause fijal says they are faster with the newer code16:59
fijalarigato: the issue being that on twisted benchmark (twisted_web) I observe speedups on trunk17:00
pedroniswhat kind of speedups?17:00
pedronisthat's relevant too17:00
fijalabout 2x17:02
fijalalthough I think we should investigate this first17:02
fijalor at least someone should be able to reproduce that17:02
arigatobtw, would you two be around in a Leysin sprint after Easter?17:02
pedroniswhen is easter again?17:03
Action: fijal did not know the sprint date was decided17:03
fijalarigato: no17:03
arigato"I can't answer yet" is fine :-)17:03
arigatopedronis: that would be e.g. 10-18 april17:03
pedronisvery unlikely at this point17:03
fijalarigato: I could be beginning of may, probably17:04
arigatook, then I suppose I need to ask the next question: does it make any sense at all to continue organizing sprints or are we all kind of dropping the style?17:04
pedronisI don't know, I fear a will have more things to say after I'm back from my trips17:05
pedronisI suppose it's a really a question for a much larger group17:07
pedronisthough17:07
arigatook, sorry about that17:08
fijalarigato: you can also view that as a "series of personal questions"17:08
fijaland I don't know answer to any of those17:09
arigatoI'm planning to come back to Gtbg for one month until Easter17:10
arigato(just so you know)17:10
fijalarigato: what do you do next?17:11
fijal(if you know)17:11
arigatowell, back in Switzerland for Easter anyway,17:12
arigatothen probably a trip to Düsseldorf, but after that it's likely to be Gtbg again17:13
arigatothat also means that I'm open to other suggestions17:15
arigato:-)17:15
fijalheh, ok :-)17:15
fijalarigato: from my side, I'm likely to stay in the US till end April17:16
fijaland since I did not decide what to do with my life afterwards, I won't make any promises17:16
arigatowell, same here, really17:16
victorpoluceno (~victorpol@187.5.181.26) joined #pypy.17:20
pedronisfijal: going home to finish packing, see you in Atlanta17:30
fijalpedronis: see you17:30
pedronissee you17:30
pedronis (~user@2001:16d8:ffca:0:20f:b5ff:fe44:40b4) left irc: Quit: -> home17:30
cfbolz (~cfbolz@62.206.45.186) joined #pypy.17:31
arigatohi Carl Friedrich17:31
cfbolzhi17:31
cfbolzhi armin17:32
cfbolzhow was your vacation?17:32
gutworth (~benjamin@64-131-1-181.usfamily.net) joined #pypy.17:34
arigatoit was rather nice17:38
arigatoI'm having trouble digging inside pypy again now :-/17:38
fijalcfbolz: hi17:39
arigatofor 3 days I've mostly been running translations and failing to understand the crash of pypy-c-jit17:39
cfbolzyes, that's not much fun :-(17:39
cfbolzamaury had a crash a while ago as well17:39
cfbolzit looked like this: http://paste.pocoo.org/show/176173/17:39
cfbolz(but I don't know the actual program)17:39
cfbolzlooks like a stack overflow, somehow17:40
arigatoah, this one is probably easy to understand17:40
arigatoit's indeed just a stack overflow17:40
cfbolzok, I think I can fix that17:41
arigatoprobably the real program consumes a lot of the stack17:41
cfbolzah17:41
arigatoand then the JIT doesn't have much left17:41
arigatonot completely sure what to do about it, to be honest17:41
cfbolzarigato: I think it's another issue also. ResumeDataLoopMemo_number is recursive in the saved stack17:42
arigatoyes, I see17:42
cfbolzso if you inline a lot, you get this17:42
cfbolzwhich sounds bad, somehow17:42
arigatoit's a minor issue, but still an issue17:43
cfbolznot that hard to fix, I think17:43
arigatoif we could rewrite it to avoid the recursion it would probably be good17:43
cfbolzshould be easy17:43
arigatoyes, but that leaves open the question of what to do with real stack overflows in the jit17:43
arigatowe could just capture it and ignore it for now, I suppose17:44
cfbolzright17:44
arigatoand add 1 to some debugging counter :-)17:44
arigatojust so we know, in case we get an example that triggers this a lot17:45
cfbolzI think iResumeDataLoopMemo.number is one of the few places in the JIT that is recursive17:45
gutworth_ensure_framestack_....17:45
arigatogood17:46
arigatoit's possible to write a test for this situation, too17:46
arigatosomething like a Python program that adds one recursion level and causes more JIT compilation, and repeats forever17:46
cfbolzarigato: ?17:47
arigatodef f():17:47
arigato   ...do stuff that forces some new JIT compilation...17:47
arigato   f()17:47
bja (~bja@c-98-222-1-220.hsd1.il.comcast.net) left irc: Ping timeout: 240 seconds17:48
cfbolzah17:49
Action: cfbolz tries17:49
arigatoah, it's probably also good to just let the interp-level RuntimeError propagate17:49
fijalexcept when it's NotImplementedError...17:49
arigatoas it really means "we are out of stack space, I give up"17:50
arigatofijal: right17:50
bja (~bja@c-98-222-1-220.hsd1.il.comcast.net) joined #pypy.18:00
arigatofor now I'm busy waiting for a crash in some extremely slow runs of pypy-c-jit18:06
arigatothey are slow because I made it generate a check for every getfield and setfield everywhere18:06
arigatohoping to find out where this garbage object is set or read out first18:06
arigatoI suppose it's great that we can easily add this kind of checks to the whole of pypy18:07
arigatobut it takes ages anyway18:07
arigatojust to retranslate and run the pypy-c-jit18:07
Action: arigato stops distracting other people by passing the time in describing what he did18:08
kenaan03fijal 10r7124314 pypy/extradoc/talk/pycon2010/pypyspeed 05A(/examples/) 07M(talk.pdf talk.tex): Finish slides ...18:10
fijalarigato: actually I appreciate that18:10
arigato:-)18:11
fijalyou often don't describe what you do, so oververbosity is not a bad thing18:11
fijalfor you, that is18:11
arigatook18:11
kenaan03fijal 10r7124414 05A(pypy/extradoc/talk/pycon2010/pypyspeed/examples/tracing.py): missing file ...18:11
arigatofijal: just curious:18:15
arigatox = y + z18:15
arigatowhere are the 5 frame accesses?18:15
cfbolzread y, read z, push stuff on stack, read from stack, write to x18:16
fijal (~fijal@rrcs-24-227-109-37.se.biz.rr.com) left irc: Ping timeout: 246 seconds18:16
arigatocfbolz: that way of counting gives a result of 618:16
cfbolzwhy 6?18:17
arigatowell no, even much more18:17
arigatoread y, push it, read z, push it, pop(2x), push total, pop it, write to x18:17
arigato918:17
cfbolzah, true18:17
ElectronicRU (~Electroni@pppoe2478.mv.ru) joined #pypy.18:19
cfbolzarigato: who said 5, anyway?18:20
cfbolzthne talk?>18:20
cfbolzyes, seems so18:20
arigatoyes18:20
ElectronicRU (~Electroni@pppoe2478.mv.ru) left irc: Remote host closed the connection18:20
ElectronicRU (~Electroni@pppoe2478.mv.ru) joined #pypy.18:20
kenaan03benjamin 10r7124514 07M(pypy/trunk/pypy/jit/metainterp/pyjitpl.py): remove incorrect comment ...18:26
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy.18:36
froosch_ (~froschi@147.172.224.61) joined #pypy.18:38
froosch (~froschi@147.172.224.61) left irc: Read error: Operation timed out18:38
kursor (~kursor@p4FC082A6.dip0.t-ipconnect.de) left irc: Quit: kursor18:59
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 276 seconds19:08
arigatopom pom pom19:13
cfbolzarigato: not having any fun?19:16
arigatowell, I should maybe try to find something else to do19:17
arigatowhile waiting19:17
mwhudson (~mwh@canonical/launchpad/mwhudson) joined #pypy.19:17
victorpoluceno (~victorpol@187.5.181.26) left irc: Read error: Connection reset by peer19:19
exarkunis even Deferred too much code/too megamorphic/whatever for the pypy jit right now?19:21
cfbolzarigato: eh19:21
exarkunhttp://twistedmatrix.com/trac/browser/trunk/doc/core/benchmarks/deferreds.py is slightly slower on pypy w/ jit from a couple weeks ago than on cpython19:22
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy.19:22
cfbolzexarkun: :-(19:23
cfbolzwhere's the code of Deffered?19:24
exarkuntwisted/internet/defer.py19:24
Alex_Gaynorhttp://twistedmatrix.com/trac/browser/trunk/twisted/internet/defer.py19:24
victorpoluceno (~victorpol@187.5.181.26) joined #pypy.19:24
Alex_Gaynorexarkun: is it just me or are some styles missing from the code browser19:24
arigatoI suppose it's already not awfully bad if it's only slightly slower than cpython19:25
cfbolzthanks19:25
Alex_Gaynorbeing old styles classes probably doesn't help19:25
witulski (~stupsi@ip-78-94-54-234.unitymediagroup.de) joined #pypy.19:26
witulski (stupsi@ip-78-94-54-234.unitymediagroup.de) left #pypy.19:26
cfbolzexarkun: I fear the traceback stuff in __init__ is not really that nice for the JIT19:27
exarkuncfbolz: self.debug is always False for the benchmark... does that help? 19:27
cfbolzah19:27
cfbolzyes, a lot19:28
cfbolzsorry19:28
exarkunThere's some other traceback stuff that might be happening later on19:28
Action: exarkun checks19:28
exarkunAlex_Gaynor: It looks fine to me, but I might just have the relevant css cached.19:28
exarkunEh.  One of the benchmark functions does invoke Failure's traceback walking code, but the other 9 or 10 don't.19:29
cfbolzI guess that one is kind of expected to be slow19:30
exarkunI also tried making Deferred subclass object.  It helped slightly, but it helped CPython more, I think.19:31
cfbolzhehe19:32
cfbolzdo I see it correctly that those two files are kind of self-dependent?19:33
cfbolzie I don't need all of twisted to try this?19:33
exarkunnot all of twisted, but you probably need a couple more files at least19:36
exarkunthere's a couple twisted imports at the top of defer.py.  you can probably delete most of them and still have a runnable benchmark, but twisted.python.failure may still be necessary19:36
exarkun(or you could skip the failure benchmark, I guess)19:37
cfbolzI guess it makes sense to try to make a nightly benchmark out of this?19:37
exarkunMaybe so.  The reason I started looking at it was to try to figure out how to get some use out of it. :)19:39
exarkunDo you have any thoughts on benchmarks that produce multiple outputs like this one?  (have you seen the output yet?)19:39
exarkunit produces something like this: http://pastebin.com/m140c35d619:40
cfbolzno, right now the infrastructure has no nice support for this :-(19:40
cfbolzexarkun: could you just paste your cpython and pypy results, maybe?19:40
exarkunsure19:41
arigatomake 10 benchmarks :-)19:41
exarkunthose were cpython-classic results19:41
exarkunpypy-classic: http://pastebin.com/m7c08155f19:41
exarkunanother issue with this benchmark is that it might be totally irrelevant to the real world :\19:42
exarkunno one ever has callback chains that are 10k entries long19:42
cfbolzexarkun: let me guess, instantiateShootErrback is the one that does traceback stuff?19:42
exarkunor even 1k entries long19:42
exarkun10 is perhaps roughly realistic19:42
exarkuncfbolz: yea19:42
cfbolzkind of easy to spot, if you look at the times :-)19:43
cfbolzexarkun: I guess it might be irrelevant to the real world, but if the code doesn't do anything crazy, there is no reason why pypy should be slow19:44
cfbolz(although I guess we will probably find out that .pop(0) on a list is really slow in pypy for some reason)19:46
arigatoI can imagine obscure hacks to get a constant-time pop(0).....19:47
exarkunoh that reminds me, can pypy detect lists that are only ever appended to and pop(0)'d from and use a deque to implement them yet? :)19:47
cfbolzexarkun: no, but would be possible to fix19:48
Alex_Gaynorarigato: do you read python-dev ;)19:48
arigatoAlex_Gaynor: usually not, but I happened to catch a glance at that thread19:48
ebo^there was a patch for that recently and afaik it was shot down19:48
cfbolzebo^: link?19:49
zejn (~Gasper@89-212-104-197.dynamic.dsl.t-2.net) joined #pypy.19:50
ebo^http://article.gmane.org/gmane.comp.python.devel/11011719:50
arigatocfbolz: it's true that it might be worth a try to have an alternate implementation of lists, at least to try it out19:54
arigatosomething whose items are split into bunches19:55
cfbolzarigato: yes, but if we want to do something like that, we should think about multilists again19:55
arigatothat would help pop(0) (or pop(n) in general) and also help the GC issue19:55
arigatoyes19:55
arigatowell either multilists, or just a single implementation if we manage to make it have little impact on small lists19:57
cfbolzI would prefer a clean multilist19:59
cfbolzexarkun: is there a reason why twisted doesn't use a deque in the first place?20:04
exarkundeque's a lot newer than Deferred20:06
cfbolzeh, ok :-)20:06
cfbolzexarkun: I guess if most things have ~10 callbacks, it doesn't matter anyway20:07
Action: exarkun nods20:08
exarkunOf course that's just an educated guess on my part.  As with many things, this would benefit from instrumentation and measurement. :)20:08
Action: cfbolz -> movi20:08
cfbolze20:08
cfbolzexarkun: indeed :-)20:09
cfbolzsee you around20:09
cfbolz (~cfbolz@62.206.45.186) left irc: Quit: Leaving20:09
Nick change: froosch_ -> froosch20:44
kenaan03arigo 10r7124714 07M(pypy/trunk/pypy/translator/c/funcgen.py): Fix a compiler warning (for 64-bit compatibility). ...20:55
arigato (~arigo@adsl-188-155-250-89.adslplus.ch) left irc: Quit: g'night21:07
Gromit (~Gromit_@g226029211.adsl.alicedsl.de) joined #pypy.21:18
hruske (~Gasper@89-212-104-197.dynamic.dsl.t-2.net) joined #pypy.21:28
zejn (~Gasper@89-212-104-197.dynamic.dsl.t-2.net) left irc: Ping timeout: 252 seconds21:32
victorpoluceno (~victorpol@187.5.181.26) left irc: Ping timeout: 272 seconds22:08
lizardo (~lizardo@189.2.128.130) left irc: Quit: Leaving22:54
gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Read error: Operation timed out23:23
--- Tue Feb 16 201000:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!