#pypy IRC log for Friday, 2012-01-06

stakkars_ (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy.00:04
stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Read error: Connection reset by peer00:04
Nick change: stakkars_ -> stakkars00:04
stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Read error: Connection reset by peer00:06
stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy.00:06
stakkars_ (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy.00:08
stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Read error: Connection reset by peer00:08
Nick change: stakkars_ -> stakkars00:08
dmalcolm (david@nat/redhat/x-gqvurvbbnyyzentj) left irc: Quit: Leaving00:08
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds00:08
nedbat (~nedbat@python/psf/nedbat) joined #pypy.00:12
stakkars_ (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy.00:13
stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Read error: Connection reset by peer00:13
Nick change: stakkars_ -> stakkars00:13
squiddy (~squiddy@g224193131.adsl.alicedsl.de) left irc: Quit: Leaving00:15
bbot24Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/778 [12fijal, better-jit-hooks]00:25
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.00:27
espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 276 seconds00:56
jterrace (~Adium@dynamic-oit-vapornet-b-397.Princeton.EDU) joined #pypy.00:56
espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy.01:02
aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy.01:03
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 260 seconds01:04
bbot23Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64-2/builds/2301:32
davisagli (~davisagli@davisagli.com) joined #pypy.01:34
jterrace (~Adium@dynamic-oit-vapornet-b-397.Princeton.EDU) left irc: Quit: Leaving.01:39
asmeurer_ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) joined #pypy.01:41
hruske (~Gasper@188-230-156-183.dynamic.t-2.net) left irc: Remote host closed the connection01:45
bbot24Failure: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/76701:47
nettok (~quassel@190.148.36.35) joined #pypy.01:58
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/191402:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/32902:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/66602:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/31802:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/66702:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/118302:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/85002:00
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/150202:00
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/77902:00
mwhudson (~mwh@linaro/mwhudson) joined #pypy.02:01
derdon (~derdon@pD9E1CB12.dip.t-dialin.net) left irc: Remote host closed the connection02:11
Nisstyre (~yours@c-208-90-102-250.netflash.net) joined #pypy.02:11
NisstyreSo, any idea why PyPy would run slower on this than CPython? http://pastie.org/313509902:13
Nisstyreit spits out everything at once, whereas CPython seems to do it incrementally02:13
Nisstyredouble_squares.txt is just a textfile of large numbers02:14
Nisstyreseparated by newlines02:14
_habnabittry adding sys.stdout.flush() after the print02:16
Nisstyreokay02:16
Nisstyrethat doesn't change anything afaict02:17
Nisstyrethe time is still the same02:17
NisstyreI'm using the "time" command to time it02:17
NisstyreI'm just curious why it would be slower02:17
Alex_GaynorHow many lines are there02:19
Nisstyreabout 2002:19
NisstyreI can paste the file for you02:19
mikefc2and how much slower do you mean by "runs slower"02:19
Nisstyreoh, not much slower, only like a millisecond :)02:19
mikefc2Nisstyre: your code probably doesn't run for long enough for the JIT to kick in02:20
Nisstyreah okay02:20
mikefc2make a file with 5000 inputs and try again :)02:20
NisstyreI'll try it with a larger trial02:20
ericflo (~ericflo@75.103.8.110) left irc: Quit: ericflo02:21
Action: mikefc2 doesn't mean to steal anyone's thunder. "your code doesn't run for long enough" seems to be the common reason for pypy being slower than cpython. 02:21
Action: mikefc2 thought he'd be the only one awake at this hour.02:22
Alex_GaynorAre these numbers super large (longs)?02:23
espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 248 seconds02:24
mikefc2Nisstyre: can you pastie the original file?02:24
espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy.02:30
Nisstyremikefc2: yes, one sec02:38
Nisstyrehttp://pastie.org/313518302:39
jterrace (~Adium@dynamic-oit-vapornet-b-397.Princeton.EDU) joined #pypy.02:42
mikefc2probably a bit of a cheat, but if you just stick it all in a loop doing it 100 times, cpython takes 15 seconds, and pypy takes 2.02:46
NisstyreI copy and pasted the file contents until it was at about 2000 lines, and PyPy was about 7 seconds, and CPython was about 2602:46
Nisstyreso there you go02:46
NisstyreI guess it does run faster if it runs longer02:47
mikefc2Nisstyre: the actually JITting of the code happens after a section of code has been seen running >1000 times (roughly)02:47
Nisstyrehmm, that's a pretty awesome optimization02:47
mikefc2once it's been seen enough, pypy does it's fancy JIT thing, so then it's faster after that.02:47
mikefc2Nisstyre: yup. pypy is amazing :)02:48
bbot23Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/19502:55
Action: mikefc2 hopes he hasn't characterized pypy jitting incorrectly.03:00
Action: mikefc2 looks for a faq03:00
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.03:03
Kaskuka (~John@osbk-4db06dd0.pool.mediaWays.net) left irc: Ping timeout: 240 seconds03:04
dracman (~draco@212.255.38.129) left irc: Ping timeout: 252 seconds03:19
dracman (~draco@212.255.40.176) joined #pypy.03:25
mat^2 (~mathias@212.130.113.35) left irc: 03:33
anish_ (~anish@123.201.39.166) joined #pypy.03:35
lahwranis there a way to tell pypy to jit *everything* on startup?03:38
mikefc2there is a threshold argument you can pass to pypy.03:39
lahwranso set it to 0 and everything will be JITed immediately?03:39
mikefc2probably won't work that way :)03:39
lahwran:<03:39
mikefc2but you could try03:39
mikefc2do a pypy --help and have a look at the options03:39
MostAwesomeDudeYou probably don't want everything immediately JIT'd anyway.03:40
mikefc2i'm sure that if you set it low, it'll spend too much time jitting useless stuff - thus taking longer.03:40
lahwranjust means longer startup, does it not?03:40
Action: mikefc2 talks through his hat03:41
MostAwesomeDudeYou want PyPy to run through each function a couple dozen times, at least, to get an idea of which types are being passed in.03:41
lahwranah03:41
MostAwesomeDudeThen it's got a better chance of inlining types which will actually be used.03:41
mikefc2lahwran: i say to give it a go and tell us how it works for you :)03:41
lahwranthat's a genius idea03:42
lahwranare there any speed examples of long running, processor-intensive things?03:42
lahwranthe kind of stuff that would be called enough to trigger all the jitting03:42
mikefc2lahwran: the benchmarks?03:42
mikefc2speed.pypy.org03:42
lahwranfrom what I understand, a lot of the stuff on there is very short-running :/03:42
mikefc2bump up the loop/array/image sizes03:43
lahwranin which?03:43
mikefc2e.g. in the raytracer you could just make it trace a huge image instead of the default size.03:43
lahwranI'm unfamiliar with the running of these benchmarks myself, where can i find them?03:44
anish_ (~anish@123.201.39.166) left irc: Ping timeout: 240 seconds03:45
anish_ (~anish@123.201.39.166) joined #pypy.03:46
mikefc2*shrug*03:47
mikefc2try this: https://bugs.pypy.org/file481/smallpt.py03:47
mikefc2a path tracer03:48
mikefc2pypy smallpt.py 200 200 403:50
Action: mikefc2 gets back to work03:50
JaredW (~JaredW@118.142.17.238) joined #pypy.03:50
mikefc2you can  see when the JIT kicks in.03:52
JaredWhi. so what's the progress of zeromq on pypy03:55
mikefc2JaredW: isn't there a lot of cython code in zeromq?03:56
ernop (~ernie@114.250.129.15) joined #pypy.03:57
mikefc2JaredW: isn't there a lot of cython code in the python bindings to zeromq?03:57
JaredWthere was an early pypy ctypes version of it I remember03:58
mikefc2JaredW: you're right: https://github.com/svpcom/pyzmq-ctypes   .  Give it a try and tell us how it goes :)04:00
__import__ (import@python/builtins/import) left #pypy ("Leaving").04:03
gutworth (~benjamin@64-131-9-52.usfamily.net) left irc: Ping timeout: 240 seconds04:32
nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 248 seconds04:32
bbot23Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/100404:37
thinker` (~thinker@218-167-101-92.dynamic.hinet.net) joined #pypy.04:39
Nick change: thinker` -> Sinker04:40
JaredWI got ImportError: No module named py04:48
durin42 (~durin@adium/durin42) left irc: Ping timeout: 252 seconds05:00
jimbaker` (~jbaker@c-75-71-80-146.hsd1.co.comcast.net) joined #pypy.05:01
durin42 (~durin@adium/durin42) joined #pypy.05:02
jimbaker (~jbaker@canonical/jimbaker) left irc: Ping timeout: 240 seconds05:02
jimbaker` (~jbaker@c-75-71-80-146.hsd1.co.comcast.net) left irc: Ping timeout: 240 seconds05:07
mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) left irc: Remote host closed the connection05:19
mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) joined #pypy.05:20
mwhudson (~mwh@linaro/mwhudson) left irc: Quit: Leaving05:22
anish_ (~anish@123.201.39.166) left irc: Ping timeout: 252 seconds05:31
bbot23Success: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/191405:39
anish_ (~anish@123.201.39.166) joined #pypy.05:43
anish_ (~anish@123.201.39.166) left irc: Client Quit05:46
anish_ (~anish@123.201.39.166) joined #pypy.05:51
nettok (~quassel@190.148.36.35) left irc: Remote host closed the connection05:53
anish_ (~anish@123.201.39.166) left irc: Ping timeout: 240 seconds06:02
asmeurer_ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) left irc: Quit: asmeurer_06:07
asmeurer_ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) joined #pypy.06:08
anish_ (~anish@123.201.39.187) joined #pypy.06:15
fermianyon (~lane@c-71-229-21-197.hsd1.al.comcast.net) left irc: Ping timeout: 240 seconds06:20
anish_ (~anish@123.201.39.187) left irc: Remote host closed the connection06:22
anish_ (~anish@123.201.39.187) joined #pypy.06:26
anish_ (~anish@123.201.39.187) left irc: Remote host closed the connection06:35
anish_ (~anish@123.201.39.187) joined #pypy.06:38
bbot23Success: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/150206:42
bbot23Success: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/85006:45
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/32906:48
bbot23Success: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/118306:58
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/66707:04
apollo13 (~apollo13@unaffiliated/apollo13) left irc: Quit: SIGILL07:08
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/66607:08
asmeurer_ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) left irc: Quit: asmeurer_07:10
espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 276 seconds07:25
apollo13 (~apollo13@unaffiliated/apollo13) joined #pypy.07:26
anish_ (~anish@123.201.39.187) left irc: Remote host closed the connection07:29
espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy.07:31
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds07:35
anish_ (~anish@123.201.39.187) joined #pypy.07:35
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.07:36
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/31807:41
anish_ (~anish@123.201.39.187) left irc: Ping timeout: 260 seconds07:48
asmeurer_ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) joined #pypy.08:02
anish_ (~anish@123.201.39.187) joined #pypy.08:03
bbot23Success: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/77908:08
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy.08:11
lucian (~lucian@cpc1-newc15-2-0-cust84.gate.cable.virginmedia.com) left irc: Ping timeout: 240 seconds08:11
mattip (~chatzilla@109.65.107.23) joined #pypy.08:15
mattipfijal: ping08:17
mattipI've made progress on numpy-axisops08:17
mattipI turned it inside-out and got much better zen08:18
fijalpong08:18
mattipbut08:18
fijalzen is good08:18
mattip_create_iter() drags this chunklist around that is only used really in VirtualSliceSignature08:19
mattipI need another argument to _create_iter(): dim08:19
mattipfor ReduceSignature08:19
fijaldim is passed at each next right?08:20
mattipwhich is the dim to iterate over (column, row, ...)08:20
fijaloh that one08:20
fijalsounds fair08:20
mattipThat's a possibility I didn't think of: passing it in at each next.08:20
mattipIt might be expensive in my implementation08:21
mattipI would prefer to pass it to _create_iter08:21
mattiplike chunklist08:21
fijalI can't really see why one would be better over other without code :)08:22
fijalso please do whatever is best for you08:22
mattipWhat is the cost of passing all the xtra args to create_frame, _create_iter as kwargs not as a list of args?08:22
mattip(all the previous chatter was backround to that question)08:22
mattipI guess I'll just do whatever is best for me and then you can look at it.08:23
fijalnone08:23
mattip:)08:23
fijalif you always name then it's always fine08:24
fijalif you try playing with *args and keyword args or something, it'll stop compiling08:24
fijalnote that all the calls in the loops are inlined08:25
fijalif you look at test_zjit, this is the list of *all* operations performend in the inside loop08:26
PiotrSikora (~none@nginx/adept/piotrsikora) left irc: Excess Flood08:26
mattipahh. OK. That's the answer I was looking for. Thanks.08:26
fijaldid you run test_Zjit with --viewloops?08:28
PiotrSikora (~none@nginx/adept/piotrsikora) joined #pypy.08:28
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/667 [12fijal, better-jit-hooks]08:30
mattipyes, but I was overwhelmed. Can you help me understand it a bit? Here's what I have so far:08:31
fijalI can explain a bit08:31
mattiptest_sum in zjit runs the jitdriver08:31
fijaljitdriver is just a marker08:32
mattipand check_simple_loop reports everything inside the marker?08:32
fijalyou run code for a while08:32
fijalsec08:33
fijalsok08:35
fijalso you run code until it becomes hot, typically after 1619 iterations08:35
fijalthen you reach the jit_merge_point and you start tracing08:35
fijaltracing = recording operations executed by the interpreter08:35
fijalafter you reach the same jit_merge_point with the same greens (that's one loop iteration in our case) you close the loop08:36
fijalyou optimize the recorded operations (and this is what you see in viewloops)08:36
fijaland in the compiled case (not test_zjit, but full pypy) you will compile that to efficient assembler08:37
fijalwith me so far?08:37
mattipyes08:37
fijalwell, this is pretty much it :)08:37
fijalthe trick is that optimization copies loop twice08:37
fijalmoving loop-invariant code before the loop08:37
mattipand check_simple_loop reports the resulting optimized code?08:37
fijalso you see two iterations of the loop, btu actually only the latter will be executed08:38
fijalyes, but only the operations08:38
fijalso int_add is a processor level int addition08:38
fijaland ther are no calls for example08:38
fijaland no allocations08:38
mattipright. That's a "good" loop: small, no calls, no allocations.08:38
fijalnow I think you can see why having tight single loop (And not two) is a bit crucial08:39
mattipYes. thanks. 08:40
mattipThat scary basement is not so bad after all.08:41
fijal:)08:41
fijalI would strongly suggest you read our jit docs08:41
fijalit would shed some light08:41
fijalhttp://doc.pypy.org/en/latest/jit/index.html08:42
fijalit's not all up-to-date and no really low level enough08:42
fijalbut should give some info08:42
mattipI really tried, but that was before I saw the light. I'll try again.08:42
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy08:44
mattipAt the beginning I needed to learn rpython, pep8, test writing, translation, the whole pypy ecosystem.08:47
mattipIt took quite a while, but now I am almost ready for the next level.08:48
fijalspend your skill points wisely :)08:52
lessshaste (~lesshaste@87-194-206-189.bethere.co.uk) left irc: Quit: Leaving08:54
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/667 [12fijal, better-jit-hooks]08:54
tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) joined #pypy.08:55
tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) left irc: Changing host08:55
tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy.08:55
kenaan12fijal better-jit-hooks 11b7cde06c0fc6 15/pypy/module/pypyjit/: work in progress on improving the JITted ops representation08:57
kenaan12fijal better-jit-hooks 1145d0a0377adb 15/pypy/module/pypyjit/interp_resop.py: bah, fix translation08:57
fijalhm08:58
fijalcan I make a mixed-module level constant thats' a dict?08:58
fijalI guess in setup_after_space_initialization08:58
kenaan12fijal better-jit-hooks 11967f730f60fb 15/pypy/module/pypyjit/: expose a bit more and improve tests09:06
kenaan12fijal better-jit-hooks 1137c20c819ee1 15/pypy/module/pypyjit/interp_resop.py: an attempt to fix translation (likely not working)09:09
fijalbah :/09:10
fijalmattip: note that it took quite a few iterations to get iterators cooperating with optimizer well enough so there are no allocations or extra field reads or anything like that09:12
squiddy (~squiddy@g224197217.adsl.alicedsl.de) joined #pypy.09:14
fijalAlex_Gaynor: ping?09:17
antocuni (~antocuni@host208-85-dynamic.7-79-r.retail.telecomitalia.it) joined #pypy.09:22
lesshaste (~lesshaste@87-194-206-189.bethere.co.uk) left irc: Quit: Leaving09:24
kenaan12fijal better-jit-hooks 110002a1e634e4 15/pypy/module/pypyjit/interp_resop.py: hopefully (ugly) fix translation.09:26
fijalantocuni: 09:27
fijalfeel like helping with annotation?09:27
lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) joined #pypy.09:28
antocunifijal: it's a holyday here :-)09:31
fijalheh09:32
fijalwhy do you expect me to know all the italian holidays :)09:32
fijalanyway, also an awesome day to help ;-)09:32
fijalI find myself writing this never-run-code09:33
fijalthat just makes types happier :/09:33
kenaan12fijal better-jit-hooks 114e8f6daccede 15/pypy/module/pypyjit/interp_resop.py: another attempt09:33
Action: fijal is inclined do just do 'push tannit'09:36
kenaan12fijal better-jit-hooks 11d8f5fca2919d 15/pypy/module/pypyjit/interp_resop.py: yet another go09:36
fijalbb is slow :/09:36
antocunifijal: I'll be off all the day, so no hacking today09:36
antocuniwhat are you trying to do btw? Expose optimizations at app level?09:36
fijalso far just exposing resops09:36
fijaldo you want me to describe the problem or not really?09:37
antocunisure09:37
fijalantocuni: I'm also fixing so on_compile is called on all jitdrivers09:37
antocunijust don't expect me to help today :-)09:37
fijal(this is done)09:37
antocuniI'll be off on ~30 minutes09:37
fijalso the problem is annotation time09:37
fijalI annotate pypyjit module way before the JIT09:37
antocuniso the annotator doesn't really know the types of most things?09:38
fijalit never sees the creation of objects in the first place09:38
fijalit has no idea what JitDriver is for example09:38
fijaletc.09:38
antocunipff09:38
antocuniannoying09:38
fijalyes09:38
antocuniyou could use a hack similar to what it's done in backend/llgraph/llimpl09:39
fijalI want a set of functions that have precise annotation and say "we'll fill in details later"09:39
antocuniin which you can statically declare the signature of functions09:39
fijalI guess09:40
fijalwould that even work?09:41
antocuninot sure09:41
fijalbecause now we have rpython/rpython bridge really09:41
antocuniI think there is no support in the annotator for what you want to do09:41
fijallet's hope armin shows up ;-)09:41
antocuniso you have to write it first09:41
fijalwell09:41
fijalI would not need to change annotator09:42
fijalsince functions get flown later anyway09:42
fijalit's sort of delayed function09:42
fijalI'll wait for armin09:43
antocunifijal: btw, if we manage to have support for this09:43
antocunithen it'll be very helpful also for separate compilation09:43
G2P (~G2P@fw-asn1.ornis.com) joined #pypy.09:44
fijalright09:47
fijalexcept not quite09:47
fijalI mean you need quite a bit more for separate comp09:47
fijalbecause I only need to join them at rtyped graph layer09:47
fijalnot compiled C layer09:48
antocunisure09:49
antocuniI just said that it'll be helpful, not the whole solution :-)09:49
lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) left irc: Ping timeout: 268 seconds09:50
lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) joined #pypy.09:51
fijalyeah, I guess I'll need that in the long run anyway09:53
kenaan12fijal better-jit-hooks 116895bfac2d92 15/pypy/module/pypyjit/interp_resop.py: be slightly more abstract09:53
fijalantocuni: I think we can kill that RPython definition changes btw :)09:53
antocunidoesn't it?09:54
fijalit rarely does09:54
antocunie.g. we recently add support for generators09:54
fijalsure, you can add things09:54
fijalbut it almost never changes in a backward incompatible way09:54
antocunitrue09:55
Taggnostr2 (~quassel@host224-64-dynamic.116-80-r.retail.telecomitalia.it) joined #pypy.09:57
Taggnostr (~quassel@host224-64-dynamic.116-80-r.retail.telecomitalia.it) left irc: Ping timeout: 255 seconds09:57
mattiphaha. The joke is on me. I stole fijal's AxisIterator that he made for _fastcopy09:58
fijalmattip: ?09:58
mattipturned it inside out so that it now iters over all array items09:58
fijalcool09:58
fijalmattip: I'll try to write some numpy docs maybe today09:58
mattipwith a flag when it finishes the desired axis09:59
fijalI have a fairly exhautive list of items to do09:59
mattipand now it broke _fastcopy.09:59
mattip(which assumes AxisIterator jumps over the axis)09:59
fijalmattip: can you just provide a different iterator for fastcopy?09:59
fijalso you don't have to worry too much09:59
fijala bit of code duplication is fine09:59
fijalantocuni: is it still true you can't modify a global list?10:00
fijalI think it's not10:00
mattipyes of course, I just thought it was funny.10:00
antocunifijal: I think it has never been true10:00
antocuniyou cannot *rebind* it10:00
fijalyes, of course10:01
Sinker (~thinker@218-167-101-92.dynamic.hinet.net) left irc: Remote host closed the connection10:02
kenaan12fijal better-jit-hooks 111d703b15bd80 15/pypy/module/pypyjit/interp_resop.py: bah, can be none10:07
fijalantocuni: http://paste.pocoo.org/show/530842/10:09
stakkars_ (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy.10:10
stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Read error: Connection reset by peer10:10
Nick change: stakkars_ -> stakkars10:10
antocuni (~antocuni@host208-85-dynamic.7-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds10:12
kyrre (c010b8a4@gateway/web/freenode/ip.192.16.184.164) joined #pypy.10:20
kenaan12fijal better-jit-hooks 11307dbbf5067d 15/pypy/module/pypyjit/interp_resop.py: wtf10:31
espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 240 seconds10:35
tilgovi (~randall@couchdb/developer/tilgovi) left irc: Read error: Operation timed out10:39
espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy.10:42
lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) left irc: Ping timeout: 240 seconds10:42
voidspace (~voidspace@python/psf/voidspace) joined #pypy.10:46
tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) joined #pypy.10:46
tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) left irc: Changing host10:46
tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy.10:46
antocuni (~antocuni@host208-85-dynamic.7-79-r.retail.telecomitalia.it) joined #pypy.10:49
kenaan12fijal better-jit-hooks 117de1c76c02c4 15/pypy/module/pypyjit/interp_resop.py: grrrr;10:50
voidspace_ (~voidspace@python/psf/voidspace) joined #pypy.10:55
voidspace (~voidspace@python/psf/voidspace) left irc: Ping timeout: 240 seconds10:56
Nick change: voidspace_ -> voidspace10:56
tilgovi (~randall@couchdb/developer/tilgovi) left irc: Remote host closed the connection11:00
aleksi (~aleksi@109-184-50-32.dynamic.mts-nn.ru) joined #pypy.11:06
fijalFUCK11:06
antocuni (~antocuni@host208-85-dynamic.7-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds11:07
kenaan12fijal better-jit-hooks 113826773898b1 15/pypy/module/pypyjit/interp_resop.py: make sure we can modify the list11:08
GoGi (~gogi@p5B10406D.dip.t-dialin.net) joined #pypy.11:16
GoGiyou called your replacement of numpy numpypy11:16
GoGidoes this mean that one can also import the original numpy somehow?11:16
fijalGoGi: no11:17
fijalwe called it because numpy people complained :/11:17
aleksi (~aleksi@109-184-50-32.dynamic.mts-nn.ru) left irc: Remote host closed the connection11:18
anish_ (~anish@123.201.39.187) left irc: Quit: Leaving11:19
mfoord (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy.11:22
mfoord (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host11:22
mfoord (~anonymous@python/psf/voidspace) joined #pypy.11:22
kenaan12fijal better-jit-hooks 1118807d528b49 15/pypy/module/pypyjit/interp_resop.py: another hack...11:22
kenaan12mattip numpypy-axisops 1144fd891a3a1c 15/pypy/module/micronumpy/: turn AxisIterator inside out11:23
kenaan12mattip numpypy-axisops 11188fd8e68904 15/pypy/module/micronumpy/interp_numarray.py: merge11:24
kenaan12mattip numpypy-axisops 115e2f7e1b6648 15/pypy/module/micronumpy/: cleanup11:24
mattipfijal: I'm getting a "AttributeError getting at the binding!" error in test_zjit11:24
fijalnice11:24
fijalbut also some other error no?11:24
mattipbut test_ztranslation passes11:24
mattipas do all the normal tests11:25
fijalI'm relatively sick of translation errors :/11:25
fijalbut ok11:27
mattiphmm. so just give me a hint11:27
aleksi (~aleksi@109-184-50-32.dynamic.mts-nn.ru) joined #pypy.11:28
fijalE       AnnotatorError': annotation of <pypy.annotation.listdef.ListItem object at 0x5058850> degenerated to SomeObject()11:28
fijalthis is the actual error11:28
fijaland it means AxisIterator and other iterators are not sharing base class11:29
fijalif you have a pdb you can look that11:29
mattipOK. I can deal with that.11:29
fijalit's bad that test_ztranslation passes :/11:29
aleksi (~aleksi@109-184-50-32.dynamic.mts-nn.ru) left irc: Remote host closed the connection11:32
mattipthanks.11:35
kenaan12mattip numpypy-axisops 1183403a06fc8c 15/pypy/module/micronumpy/interp_iter.py: translation error11:35
mattipnow I'll try to add a jit_merge_point11:36
mattipand maybe even a zjit test of my own11:36
fijalcool11:39
fijallet me have a quick look btw11:39
fijalmattip: you have unusued variables for one11:40
fijalI have them glow red :)11:41
fijalobjlen11:41
fijalok11:41
fijalthis is good enough to *write* a zjit test :)11:41
fijalyou might need to add features to compile.py btw11:41
kenaan12fijal better-jit-hooks 11e1915f5fd176 15/pypy/module/pypyjit/interp_resop.py: I'm getting sick of this slowly11:43
mattipthanks. Yes, it's still a bit rough. Trying to learn by doing.11:50
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Quit: Leaving11:50
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.11:50
kenaan12fijal better-jit-hooks 11a469f09278ce 15/pypy/module/pypyjit/interp_resop.py: screw hacks, provide annotation by hand. I wonder if this is enough (very likely not)12:15
kenaan12fijal better-jit-hooks 115b4f622c57aa 15/pypy/module/pypyjit/test/test_ztranslation.py: better than nothing12:16
lesshaste (~lesshaste@ikaris.cs.bris.ac.uk) joined #pypy.12:17
fijalpfffffffffffffffff12:28
fijalI fuckin give up :/12:29
fijalwhere is armin when you need one?12:29
derdon (~derdon@pD9E1C4AA.dip.t-dialin.net) joined #pypy.12:48
asksol_ (asksol@2a01:7e00::f03c:91ff:fedf:af30) joined #pypy.13:03
asksol (~asksol@sanedefaults.net) left irc: Quit: ZNC - http://znc.sourceforge.net13:03
Nick change: asksol_ -> asksol13:04
kyrre (c010b8a4@gateway/web/freenode/ip.192.16.184.164) left irc: Quit: Page closed13:08
mitsuhiko (~mitsuhiko@ubuntu/member/mitsuhiko) left irc: Ping timeout: 240 seconds13:11
lesshaste:)13:27
lesshastefijal, hi13:27
fijalhi13:28
derdon (~derdon@pD9E1C4AA.dip.t-dialin.net) left irc: Remote host closed the connection13:29
mitsuhiko (~mitsuhiko@ubuntu/member/mitsuhiko) joined #pypy.13:33
thinker (~thinker@123-194-52-90.dynamic.kbronet.com.tw) left irc: Remote host closed the connection13:53
thinker (~thinker@123-194-52-90.dynamic.kbronet.com.tw) joined #pypy.13:53
bfirsh (u1308@gateway/web/irccloud.com/x-bqfcxygkbimowclp) joined #pypy.13:53
mat^2 (~mathias@212.130.113.35) joined #pypy.13:56
therve (~therve@2a01:e0b:1:116:ca0a:a9ff:fe03:3be7) joined #pypy.14:12
thervehey there14:12
fijaltherve: hi14:12
therveI want to contribute a small feature to the curses module14:12
fijalcool14:12
thervewhat's the best way to do it?14:13
fijalwrite a test :)14:13
fijaland put a patch on bugtracer?14:13
fijalif it's really small, just paste it14:13
therveit's surprisingly big14:13
therveI changed the _minimal_curses module14:13
therveand there are different "instances" of it, I don't really what they do14:14
fijal"instances"?14:15
whitelynx (~whitelynx@63.241.75.144) joined #pypy.14:15
thervehttp://paste.ubuntu.com/794884/ is what it looks like14:16
thervefijal, lib_pypy/_minimal_curses.py and pypy/module/_minimal_curses/14:16
fijalis the lib one even used?14:16
fijalI think no14:16
fijalit's definitely missing tests :)14:17
thervesure :)14:17
nedbat (~nedbat@python/psf/nedbat) joined #pypy.14:17
therveso you say the change in lib_pypy is unnecessary?14:17
fijalyeah, just ignore it14:18
fijalthere are very bad reasons why you can't use ctypes for module like this14:18
therveok :)14:18
therveother question: how can you just compile relevant change when you make a change in pypy/module/?14:19
amaury_ (amaury_@nat/google/x-qqzxicvylnlmctqu) joined #pypy.14:24
Rhy0lite (dje@nat/ibm/x-bzlldijoxwcartuu) joined #pypy.14:25
exarkunignore it, or delete it?14:25
exarkuntherve: You can't.  No partial/incremental translation.14:26
therveah ok. Screw the baby polar bears then.14:26
aboudreault (~alanb@osgeo/member/aboudreault) left irc: Quit: Leaving14:26
thervewhere should the tests should reside? I'm modifying the 2.7 curses module, should I change test_curses as well?14:29
exarkunis it in the modified lib dir?14:30
exarkuneither way, you should probably also try to contribute the improved tests to cpython14:30
exarkununless there's some reason that it would only apply to pypy14:31
lesshasteanyone looked at the machine learning library http://scikit-learn.org/ in the context of pypy?14:31
therveone step at a time :)14:31
therveexarkun, I changed the 2.7 one. I guess I should changed modified-2.7 instead?14:32
exarkuntherve: I /think/ so.  I'm not totally clear on how duplicates across 2.7 and modified-2.7 interact.14:32
exarkunlesshaste: Does scikit depend on scipy?14:32
exarkunlesshaste: coz scipy is a big hairy mess with lots of problems wrt running on pypy14:33
lesshasteah.. scikit-learn is a Python module integrating classic machine learning algorithms in the tightly-knit world of scientific Python packages (numpy, scipy, matplotlib).14:33
lesshastefirst get numpy to work then move on from there :)14:33
lesshasteis there any point in numpy when you have pypy ?14:34
lesshasteexcept for compability14:34
lesshastecompatibility14:34
exarkunsure14:34
exarkunnumpy implements some useful things14:34
gtaylor (~gtaylor@108-196-160-156.lightspeed.tukrga.sbcglobal.net) joined #pypy.14:34
lesshasteexarkun, oh right.. I meant more that numpy is always supposed to give a speedup14:34
lesshasteexarkun, when doing the same things as you would in python14:34
arigato (~arigo@xdsl-188-154-144-172.adslplus.ch) joined #pypy.14:35
lesshastewhich I assumed was because the numerical arrays are implemented more efficeintly14:35
lesshasteis that wrong?14:35
exarkunI certainly hope that someday numpy will be implemented purely in Python and be faster on PyPy than the current version ever was on CPython14:35
exarkunWho knows if that will happen?14:35
fprimex (~fprimex@brent-macbook.sc.fsu.edu) joined #pypy.14:36
lesshastewell.. I have lots of students who need projects.. so if we can find ways to make suitable projects14:36
lesshastethen that would help14:36
lesshasteexcept they need hand holding normally14:36
lesshasteso maybe not14:36
exarkunStudents are excellent at writing high quality software that stands the test of time.14:36
lesshasteexarkun, I take your sarcasm and agree with it :)14:36
exarkun(I am so bitter, I typed that with a straight face)14:36
lesshastethey are even better at finishing projects properly in such a way that they can be handed over to the next person14:37
exarkunlesshaste: Maybe a task of such reduced scope as adding better unit test coverage to numpy could be taken on by some competent students14:38
lesshastethe problem is that they won't get the marks14:38
exarkunHaving full test coverage will help PyPy fully support numpy, someday, in some way14:38
exarkunlesshaste: The marks?14:38
lesshasteexarkun, students want marks14:39
lesshasteif the project is not exciting enough they don't get many14:40
lesshastegrades?14:40
lesshasteI am not sure which dialect you speak :)14:40
lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) joined #pypy.14:45
aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy.14:55
mattip (~chatzilla@109.65.107.23) left irc: Ping timeout: 252 seconds15:03
jimbaker (~jbaker@c-75-71-80-146.hsd1.co.comcast.net) joined #pypy.15:04
jimbaker (~jbaker@c-75-71-80-146.hsd1.co.comcast.net) left irc: Changing host15:04
jimbaker (~jbaker@canonical/jimbaker) joined #pypy.15:04
fijalexarkun: need some more help?15:05
exarkunfijal: me?  I'm not doing anything right now.15:06
fijaltherve: you15:07
fijalexarkun: sorry15:07
exarkunoh, maybe I need help with this15:08
exarkunis evaluation order of keyword argument value expressions in a function call defined?15:09
exarkunis it defined that b is always evaluated before d in f(a=b, c=d)?15:09
arigatoI think so, but relying on it seems a bit relying on a corner case15:09
exarkunit's rather convenient :/15:10
arigatomaybe I'm wrong, and relying on it is fine15:10
exarkunhttp://codepad.org/GD4Sh6pp15:11
arigatoI see15:11
exarkunarigato: You sound right to me, I was hoping you'd strongly feel it was defined behavior though :)15:11
thervefijal, so I should change modified-2.7, not 2.7 ?15:11
arigatoexarkun: I think your use case merits saying it's defined behavior15:12
arigatoit sounds similar to building a big dict with the syntax {'a': b, 'c': d},15:12
fijalarigato: hi15:13
fijalarigato: I have quite some issues :/15:13
arigatowhich also happens to guarantee that b is evaluated before d15:13
fijalmaybe you can help15:13
arigatofijal: hi15:13
arigato:-)15:13
fijalso15:13
fijalI'm working on better jit hooks15:13
fijalproblem is - how do I expose ResOperation interface at annotation time?15:13
fijalit always conflicts with some later annotations from jitting15:14
arigatolooks like a mess15:14
lucianexarkun: there's got to be a better way to write that :)15:14
fijalarigato: yes :/15:15
fijalarigato: any idea how to get around the mess?15:15
arigatocan you delay annotation of the corresponding parts?15:15
exarkunlucian: Probably15:16
arigatoor else, the pypyjit module needs to consider these as really opaque objects15:16
arigatoand inspect them using pypy.rlib.jit.some_interface()15:17
arigatoyes, that sounds saner15:17
fijalarigato: how?15:17
fijalif they're exposed?15:17
arigatothey are gcrefs, and you cannot cast them explicitly to ResOperation15:18
arigatoyou need to use some interface that converts initially to llops15:18
arigatoand then replace the llops when applying the jit15:18
GoGi (~gogi@p5B10406D.dip.t-dialin.net) left irc: Ping timeout: 248 seconds15:20
fijalpypy.rlib.jit.some_interface is insane15:24
fijalit has to be modified ni 17 places to do anything15:24
fijalI don't even want to think what happens if I want to use more advanced data structures :(15:24
arigatoif you don't want that, then you need to figure out a way to delay annotation15:24
lahwranperhaps refactoring time, eh?15:25
arigatoor else, define your own classes in rlib.jit, with some reasonable interface, similar but different from ResOperation15:27
arigatothen when the hook is called, the jit needs to make instances of that class15:27
GoGi (~gogi@pippo.9.6.gogi.tv) joined #pypy.15:28
arigatoit also has the advantage that we can change details to the ResOperation class without breaking pypy.module.pypyjit15:28
fijalarigato: but that's extremely fragile15:29
fijalwell, to start with the call sites change15:29
arigato?15:29
arigatoto me, that sounds far less fragile than passing directly the ResOperation instances15:30
fijalsomething, like a method15:30
fijalwill become SomePBC(known_descr={..1..})15:30
fijalinto SomePBC(known_descr={..2..})15:30
fijalno?15:30
arigatono clue what you're really talking about...?15:31
mat^2 (~mathias@212.130.113.35) left irc: Ping timeout: 252 seconds15:31
fijalannotation of methods15:31
fijalsay I create jit.FakeResOp15:31
fijaland pass around FakeResOp15:31
arigatoI proposed to really have code in the JIT that really makes new instances of pypy.rlib.jit.MyResOperation, copying the ResOperations15:31
fijalok15:32
fijalwell, even ignoring the fact that it's inefficient15:32
fijalhow this survives say list resized vs unresized?15:32
hruske (~Gasper@188-230-156-183.dynamic.t-2.net) joined #pypy.15:32
fijalbecause I'll need mock-only annotated code that's never run right?15:32
arigatono15:32
mat^2 (~mathias@212.130.113.35) joined #pypy.15:32
fijalno?15:33
arigatoor yes, but that's a general problem of all the jit hooks, not just this one15:33
fijalwell, yes15:34
arigatothen maybe you really need to find a way to delay all annotations of them15:34
fijalI guess I can provide annotations of the interface15:35
fijalsay15:35
fijalgetopnum() returns int15:35
fijaletc.15:35
arigato? that's what I proposed first, but you said no15:35
fijalhow do I do that?15:36
fijaler15:36
fijalI said "no" to the current way this is implemented15:36
fijalwhich is invent a hint15:36
fijalemit a special operation15:36
fijalrewrite operation to a call15:36
fijalimplement call15:36
fijalunless we have an idea how to automate it15:36
arigatosorry, I'm running out of ideas, I think you have them all and you have to find the one that seems best for you15:39
fijalI like none of them :/15:44
fijalit's all extremely fragile15:44
fijalI just spent half a day trying to write code that's annotated and not run15:44
fijalthat has correct types15:45
fijaland I only wrote 3 functions15:45
Rhy0litefijal: if you sweep the ugliness under the carpet, it has to appear as a lump somewhere :-)15:48
Rhy0litefor PyPy, that means modifying 17 places15:49
dmalcolm (~david@c-24-61-12-82.hsd1.ma.comcast.net) joined #pypy.15:55
oal (u4126@gateway/web/irccloud.com/x-xmvfknsrhzjydqmi) left irc: Remote host closed the connection16:05
fijalarigato: I guess I can also simply express my general unhapiness with the situation :/16:11
jterrace (~Adium@dynamic-oit-vapornet-b-397.Princeton.EDU) left irc: Quit: Leaving.16:12
GoGi (~gogi@pippo.9.6.gogi.tv) left irc: Ping timeout: 260 seconds16:15
fijalarigato: ok, can I have one more question regarding hints?16:18
arigatoyes?16:18
oal (u4126@gateway/web/irccloud.com/x-lulpqtjvtjrokqyk) joined #pypy.16:18
fijalhow do I keep state?16:19
fijalthe problem is that - say I have a hint that gets a number out of a resop16:19
fijalso say pypy.rlib.jit.get_resop_no()16:19
fijalhow do I store the resop somewhere?16:19
arigato?16:19
fijalmaybe I'm not expressing myself correctly16:19
fijalUH16:20
aleksi (~aleksi@109-184-50-32.dynamic.mts-nn.ru) joined #pypy.16:20
fijalso I have a hook that calls python function16:20
fijalspace.call_function(w_python_function, some args)16:20
fijalnow I have a resop here16:20
oal (u4126@gateway/web/irccloud.com/x-lulpqtjvtjrokqyk) left irc: Client Quit16:20
fijaland I want to pass it to some args16:20
fijalI can create say a class WrappedOp(Wrappable): that would have some interface16:21
fijalbut I can't do WrappedOp(some_resop)16:21
fijalbecause the annotator does not know about resop16:21
fijalhm16:21
fijalshould I simply pass a llmemory.GCREF and have a hint that accepts that?16:21
oal (u4126@gateway/web/irccloud.com/x-wjaiclderiyrhclj) joined #pypy.16:22
arigatoyes16:22
kenaan12fijal better-jit-hooks 11d498e3a9f2e2 15/pypy/rlib/jit.py: shuffle stuff a bit16:23
oal (u4126@gateway/web/irccloud.com/x-wjaiclderiyrhclj) left irc: Remote host closed the connection16:23
fijalit's fairly messy :/16:23
Alex_Gaynor (u1246@gateway/web/irccloud.com/x-tirznnnobywzpxgd) left irc: Read error: Connection reset by peer16:24
bfirsh (u1308@gateway/web/irccloud.com/x-bqfcxygkbimowclp) left irc: Remote host closed the connection16:24
alexsuraci (u1962@gateway/web/irccloud.com/x-iaebohqulkqcxjvh) left irc: Remote host closed the connection16:24
GoGi (~gogi@pippo.9.6.gogi.tv) joined #pypy.16:24
fijalarigato: we should invent some sort of stronger typing system for RPython16:25
fijalso you can specify types16:25
fijalit would help us with things like that, separate compilation and debugging16:25
arigatoyes16:26
fijalon a slightly unrelated note, I tried improving docs of RPython, feel like looking - http://paste.pocoo.org/show/530842/16:26
fijal?16:26
arigatowhy did you kill "global lists cannot be modified"?  did I forget that we improved the situation?16:28
lucianfijal: could you do it at function call boundaries with the py3 annotation syntax?16:28
lucianlike def bla(a: int, b: float)16:28
fijalarigato: uh?16:29
fijalI always modified global lists and never had a problem16:29
fijalsame for anto16:29
arigatoI mean lists that appear as a global variable16:29
fijalyes16:30
Action: arigato tries again16:30
sven_hager (~sven@ip-178-202-227-26.unitymediagroup.de) joined #pypy.16:31
arigatono, it doesn't work16:31
Rhy0litearigato: any more success with GCC?16:31
arigatohttp://paste.pocoo.org/show/530947/16:31
arigatoprints 016:31
arigatoRhy0lite: no16:31
fijalarigato: ok16:32
fijalarigato: so I'll revive this16:32
Rhy0litehow can I help you debug the build problem?16:32
Rhy0liteHi, sven_hager16:32
arigatoRhy0lite: I suppose I would be happy with a version of gcc where "../trunk/configure; make" works, in some standard distribution like Ubuntu or Debian, 32 or 6416:33
arigatoor even Gentoo16:33
arigato(didn't try Gentoo so far)16:33
Rhy0liteyes, but I think I need to help you build it on your system16:34
arigatowell, I'd be fine if it worked on any of the systems listed above16:34
fijalarigato: would you object if I invent a way to make more rigid typing in RPython?16:34
Rhy0litearigato: I think we need to look at the error messages and fix the problem16:35
Rhy0liteI'm not sure I can make it work out of the box16:35
arigatofijal: no, but it might be a sizeable project, definitely a new branch16:36
fijalarigato: I have a policy of not doing anything with annotation on trunk :)16:36
fijalit's just way too easy to break stuff16:36
arigato:-)16:36
arigatoRhy0lite: ok, I can try again.  which of the above systems should I try it on?  is ubuntu 64 ok?  or would a 32-bit system offer less causes for crashes?16:37
Rhy0litearigato: Per Arneng claims to be able to build GCC 4.6 on Ubuntu 10.01: http://blog.scalebit.com/2011/04/compiling-gcc-46-in-ubuntu-1010-for-c11.html16:38
sven_hagerhello david16:38
Rhy0litearigato: might be better to start on a 32 bit system16:39
asmeurer_ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) left irc: Quit: asmeurer_16:39
Rhy0litearigato: remember to configure in a separate build directory, not in the gcc source code16:39
Rhy0liteand you might try adding --disable-multilib to the configure command16:40
arigatofijal: re changes to the RPython docs: very good16:40
fijalarigato: ok, I'll commit it then16:41
arigatoRhy0lite: ok, trying16:41
Rhy0litearigato: I'm not sure why the person needed some of the other configure commands16:41
Rhy0litearigato: if you skip multilib, you may be able to avoid problems with missing libraries16:43
Rhy0liteyour 64 bit system may be missing multilib versions of development libraries16:44
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) joined #pypy.16:45
Alex_Gaynor (u1246@gateway/web/irccloud.com/x-tdrapaosntmwnjdb) joined #pypy.16:45
#pypy: mode change '+o Alex_Gaynor' by ChanServ!ChanServ@services.16:45
antocuni (~antocuni@host208-85-dynamic.7-79-r.retail.telecomitalia.it) joined #pypy.16:54
nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 276 seconds16:55
oal (u4126@gateway/web/irccloud.com/x-ilvlueygtnkahbhx) joined #pypy.17:01
lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) left irc: Ping timeout: 240 seconds17:02
TuomasT (~tuomas@hoasnet-fe23dd00-20.dhcp.inet.fi) joined #pypy.17:02
G2P (~G2P@fw-asn1.ornis.com) left irc: Quit: Leaving.17:03
TuomasTI have this threaded Python app, which is some 15-35% faster with PyPy than it is with Psyco and Python2.6. However, running with pypy it sometimes segfaults randomly. GDB does not produce any meaningfull information. How does one go about debugging something like this?17:04
TuomasTI guess debugging enabled pypy executable would be a start17:05
arigatoTuomasT: yes...17:06
fijalchecking C extension usage also17:07
arigatoTuomasT: if you can, build yourself a pypy from sources, and compile it with "make lldebug"17:08
arigatoi.e. "translate.py -Ojit --source", move the source code outside /tmp/usession-yourname/, and type "make lldebug" in the "testing_1" subdir17:08
arigatoRhy0lite: ...still compiling...17:09
Rhy0litegood17:09
espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 240 seconds17:10
TuomasTIs it a problem that I use "import cPickle" instead of "import Pickle" ?17:10
arigatono17:10
arigatobut do you use any C extension module from CPython?17:10
arigato(i.e. a module that you compiled with "pypy setup.py install")17:11
arigatoa default installation of pypy doesn't contain any: all the modules it has are purely for pypy17:11
TuomasTarigato: no17:11
arigatook, then it's strange indeed17:11
TuomasTI only have added MySQLdb equivalent for pypy and I use feedparser, which does not use any 3rd party libraries17:12
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Ping timeout: 240 seconds17:12
arigatoany usage of "import ctypes"?17:12
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy.17:12
arigatoRhy0lite: success :-)17:13
Rhy0liteyea!17:13
arigatothank you17:13
Rhy0litenp17:13
Rhy0litenow you can try 6417:13
TuomasTarigato: no, only the cPickle17:13
Rhy0liteusing --disable-multilib17:13
Rhy0liteI think the problem is missing development libraries for the 32 bit multilib on 64 bit system17:14
arigatoI already tried with --disable-multilib17:15
kenaan12edelsohn ppc-jit-backend 115227411080a8 15/pypy/jit/backend/ppc/ppcgen/regalloc.py: Cast value in convert_to_imm, which fixes test_new_with_vtable.17:15
arigatoI didn't have --disable-bootstrap --enable-languages=c,c++ though17:15
fermianyon (~lane@c-71-229-21-197.hsd1.al.comcast.net) joined #pypy.17:15
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Read error: Connection reset by peer17:15
Rhy0litewell, --disable-bootstrap is not really correct17:15
Rhy0lite--disable-bootstrap means treat the system like a cross-compiler17:16
arigatobah17:16
Rhy0litebuild the new gcc with the system gcc, but don't try to build gcc with itself17:16
arigatoI also didn't have bison installed17:16
espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy.17:16
Rhy0litebison should not be necessary because the generated files are included in the tar file17:17
nick125 (~nick@pdpc/supporter/student/nick125) joined #pypy.17:17
Rhy0liteat least progress is good17:17
Rhy0liteso you see parts of it can work!17:17
arigatoah, I'll try to keep --disable-bootstrap then17:17
Rhy0litetry to keep?17:18
arigatosome of the errors I used to get where in stage 3 compilation17:18
arigatowere17:18
Rhy0liteah, hmm17:18
Rhy0litecomparison failures?17:18
arigatono17:18
arigatoI don't remember which error were where :-)17:18
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.17:19
lesshaste (~lesshaste@ikaris.cs.bris.ac.uk) left irc: Quit: Leaving17:25
alexsuraci (u1962@gateway/web/irccloud.com/x-mxejbgidhfmbotfo) joined #pypy.17:26
Husio (husiatyn@oceanic.wsisiz.edu.pl) left irc: Ping timeout: 240 seconds17:28
sven_hager (~sven@ip-178-202-227-26.unitymediagroup.de) left irc: Quit: Leaving17:28
arigatoRhy0lite: http://paste.pocoo.org/show/530978/17:31
pgollakota (c646a2a2@gateway/web/freenode/ip.198.70.162.162) joined #pypy.17:32
pgollakotaHello!17:33
arigatohi17:34
pgollakotaI have been using Python for a while and recently got interested in PyPy17:34
pgollakotaI want to browse the source for Python implemented in PyPy (not the RPython toolchain) 17:34
pgollakotaespecially as a way to understand how the builtin data types of Python have been implemented in PyPy17:35
arigato:-)17:35
pgollakotaI can't figure out where the source code for Python implementations of List, Tuple etc. are located. 17:35
pgollakotaCan you please help me?17:36
arigatothat's pypy.objspace.std.listobject17:36
arigatoand tupleobject etc.17:36
whyking_ (~quassel@p5B3DD544.dip.t-dialin.net) joined #pypy.17:37
pgollakotaDomo arigato arigato! :)17:37
arigato:-)17:37
pgollakotaBTW, Is there any resource that explains the directory structure of PyPy. I tried searching for "PyPy directory structure", etc. with no real luck.17:39
arigatonot really17:40
TuomasT (~tuomas@hoasnet-fe23dd00-20.dhcp.inet.fi) left irc: Quit: leaving17:41
arigatoah yes17:41
Husio (husiatyn@oceanic.wsisiz.edu.pl) joined #pypy.17:41
arigatohttp://doc.pypy.org/ , bottom of the page17:41
pgollakota:o Embarrassing17:42
pgollakotaThanks!17:42
amaury_http://doc.pypy.org/en/latest/architecture.html#pypy-architecture ?17:42
pgollakotaThanks amaury_17:43
pgollakota (c646a2a2@gateway/web/freenode/ip.198.70.162.162) left #pypy.17:45
Action: fijal is a bit afraid of people who randomly pop up and ask about directory structure17:52
amaury_ (amaury_@nat/google/x-qqzxicvylnlmctqu) left irc: Read error: Operation timed out17:53
arigato:-)17:56
fijal_ (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.17:56
GoGi (~gogi@pippo.9.6.gogi.tv) left irc: Ping timeout: 260 seconds17:59
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds17:59
Nick change: fijal_ -> fijal17:59
kenaan12arigo concurrent-marksweep 11719f30faa809 15/pypy/rpython/memory/gc/: Start implementing the new section.17:59
kenaan12arigo concurrent-marksweep 114f6d92354eb2 15/pypy/rpython/memory/gc/: test_many_objects passes again :-)17:59
kenaan12fijal separate-applevel-numpy 1116ea77edcb5e 15/: close the branch will be redone18:01
kenaan12fijal import-numpy 11be18c0104736 15/pypy/module/micronumpy/__init__.py: rename the interplevel module18:01
kenaan12fijal default 11b67e65d709e1 15/pypy/doc/coding-guide.rst: make coding guide a bit more up to date when it comes to RPython definition18:01
GoGi (~gogi@pippo.9.6.gogi.tv) joined #pypy.18:01
panni_ (smuxi@descent.fragstore.net) left irc: Remote host closed the connection18:02
antocuni (~antocuni@host208-85-dynamic.7-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds18:02
panni_ (smuxi@descent.fragstore.net) joined #pypy.18:03
panni_ (smuxi@descent.fragstore.net) left irc: Remote host closed the connection18:03
panni_ (smuxi@descent.fragstore.net) joined #pypy.18:04
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Ping timeout: 240 seconds18:09
xcombelle (~xcombelle@AToulouse-551-1-132-44.w86-201.abo.wanadoo.fr) joined #pypy.18:11
tos9 (~tos9@unaffiliated/tos9) left irc: Ping timeout: 240 seconds18:12
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.18:14
tos9 (~tos9@unaffiliated/tos9) joined #pypy.18:16
oal (u4126@gateway/web/irccloud.com/x-ilvlueygtnkahbhx) left irc: Remote host closed the connection18:25
Alex_Gaynor (u1246@gateway/web/irccloud.com/x-tdrapaosntmwnjdb) left irc: Remote host closed the connection18:25
alexsuraci (u1962@gateway/web/irccloud.com/x-mxejbgidhfmbotfo) left irc: Read error: Connection reset by peer18:25
alexsuraci (u1962@gateway/web/irccloud.com/x-yqxqgvwcoyescdbr) joined #pypy.18:28
Alex_Gaynor (u1246@gateway/web/irccloud.com/x-fqpckjddpnsprhuy) joined #pypy.18:30
#pypy: mode change '+o Alex_Gaynor' by ChanServ!ChanServ@services.18:30
Alex_Gaynorfijal: pong18:31
fijalAlex_Gaynor: mess :/18:31
Alex_Gaynorwith annotation?18:31
fijalyes18:32
andrewfr (~andrew@modemcable211.101-70-69.static.videotron.ca) joined #pypy.18:34
andrewfr@arigato added the gc=none and it worked!18:34
fijalAlex_Gaynor: I'm trying to document numpy instead18:35
Alex_Gaynorfijal: numpypy you mean?18:35
fijalyes18:35
Alex_Gaynordocs for the code you mean?18:36
Alex_Gaynorthat is, internals18:36
fijal_ (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.18:36
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Ping timeout: 240 seconds18:37
Alex_Gaynorfijal_: don't we have a way to specify types in RPython? objectmodel.enforce_annotation or something18:37
fijal_yeah?18:38
fijal_so specify me unmodifiable list of ResOps18:38
Alex_Gaynorheh, indeed18:38
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds18:39
fijal_if we have stuff like sepcomp would not be such a mess18:40
Nick change: fijal_ -> fijal18:40
Alex_Gaynorfijal: you still have class numbering to deal with18:41
fijalyeah18:41
fijalbut that's easy compared to everything else18:41
Alex_Gaynorsure18:42
fijalso to have good better-jit-hooks we need to address that issue I fear18:42
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.18:42
fijalarigato: is there any point any more to keep faking around?18:42
fijal--no-nofaking is my favorite py.py option :)18:43
fijalheh18:44
fijalimporting numpy on cpython does not work either18:44
andrewfr (andrew@modemcable211.101-70-69.static.videotron.ca) left #pypy.18:45
oal (u4126@gateway/web/irccloud.com/x-opcpdwvmovryeapc) joined #pypy.18:45
fijal"Thank you for your willingness to help make NumPy the best array system18:45
fijalavailable."18:45
arigato--no-nofaking starts in 14 seconds, whereas --nofaking starts in 0.1 seconds less18:45
fijalarigato: I would like to file a bug report, faking does not work on pypy :)18:46
fijalarigato: for what is worth, I completely fail to understand how do you develop numpy18:47
fijal"you should not try to import numpy from18:48
fijal        its source directory; please exit the numpy source tree, and relaunch18:48
fijal        your python intepreter from there.18:48
fijal"18:48
Alex_Gaynorpara ingles?18:48
mfoord (~anonymous@python/psf/voidspace) left irc: Quit: mfoord18:48
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy.18:49
fijalAlex_Gaynor: so I have to install it to do anything?18:49
fijalwhy?18:49
Alex_Gaynorfijal: I dunno, they alsao like C, cython, and fortran though, so maybe they aren't totally sane by our standards18:49
Alex_Gaynorwe like python quite a bit :)18:49
fijalnumpy is 12M18:51
fijalof python18:51
Alex_Gaynoro_O18:51
fijalah no18:51
fijalnot just python18:51
ericflo (~ericflo@75.103.8.110) joined #pypy.18:52
Alex_Gaynoraccording to sloccount it is: 112k c, 82k python, 800 cpp18:52
Alex_Gaynorwho knows if that's accurate though18:52
Action: nick125 mumbles something about fortran...18:52
Alex_Gaynornick125: 300 fortran, 200 f90m if this is to be trusted18:52
nedbat (~nedbat@python/psf/nedbat) joined #pypy.18:53
fijalAlex_Gaynor: I'm importing numpy18:53
fijalor pieces of it18:53
nick125I was rewriting some code from fortran 66 to Python yesterday...yeah.18:53
Alex_Gaynorfijal: it would be really cool if we could ship _numpy and say "pip install numpy" to actually have it be usable18:53
Alex_Gaynornick125: David Beazley was contemplating rewriting some Fortran/C (I forget which) he wrote 20 years ago and won some scientific computing performance award for in Python, just to benchmark it under pypy :)18:54
fijalAlex_Gaynor: doubt that18:54
fijalpip install numpypy though?18:54
Alex_Gaynorfijal: seems reasonable, it's just dumb to ship a bunch of NumPy sources in pypy tree18:55
fijalyeah18:55
fijalwe can skip that18:55
fijalI suggest putting them under pypy/numpypy on bitbucket18:55
fijaland uploading to pypi?18:55
fijalarigato: do you have opinions?18:55
Alex_Gaynormakes sense to me18:55
etrepum (~bob@75-101-96-144.dsl.static.sonic.net) left irc: Quit: etrepum18:55
Action: fijal expects armin to carefully not have opinion18:55
amaury_we already have lib-python and lib_pypy18:57
Alex_Gaynorthe idea is for this to be outside the pypy sourcetree18:58
amaury_so is lib_pypy18:58
Alex_Gaynoroutside the repo then :)18:58
amaury_ah ok18:58
amaury_makes sense then18:58
amaury_our fork of (part of) numpy18:59
fijalso we'll only ship _numpypy right?18:59
fijalit might be an API mess18:59
fijalbut at least lets make sure releases are fine18:59
Alex_Gaynork18:59
ojii (~ojii@50-57-189-190.static.cloud-ips.com) left irc: Ping timeout: 260 seconds19:13
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Ping timeout: 240 seconds19:14
ojii (~ojii@50-57-189-190.static.cloud-ips.com) joined #pypy.19:15
etrepum (~bob@accessnat4.mochimedia.net) joined #pypy.19:23
tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy.19:27
aboudreault (~alanb@osgeo/member/aboudreault) left irc: Quit: Leaving19:31
fijalunlike my pypy checkout which is 16G19:33
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection19:33
Alex_Gaynorfijal: o_O19:33
Alex_Gaynormine is 507MB19:34
fijalcompiled C19:34
fijalmostly19:34
fijalhm19:36
fijalAlex_Gaynor: we should probably bundle it19:37
fijalnightly runs of tests?19:37
Alex_Gaynorwhy?19:37
Alex_Gaynorwe can just have our tests import _numpy19:37
Alex_Gaynorand assume the pure python we didn't write is fine19:37
Alex_Gaynorunless you mean nightly runs of numpy's tests, in which case that should be  a seperate run19:37
Alex_GaynorI think19:38
Alex_Gaynor(because they'll mostly fail ATM)19:38
fijalhm19:39
fijalI think it's still easier to import it19:39
fijalbut fine19:39
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.19:41
fijalcan someone make bbhook report that?19:43
Alex_Gaynorreport what?19:43
fijalreport commits to pypy/numpypy on bitbucket19:44
fijalAlex_Gaynor: can you up to import-numpy branch and see if tests pass for you?19:50
fijaltest_module specifically19:50
kenaan12fijal import-numpy 114fa4cf3c8f35 15/pypy/module/micronumpy/test/: fix tests19:50
kenaan12fijal import-numpy 11d06444d3b317 15/pypy/module/micronumpy/test/test_ztranslation.py: fix test_ztrasnlation19:50
Alex_Gaynorsure19:50
Alex_Gaynoris push finished?19:51
fijalyes19:51
fijaloh wow19:52
fijalAlex_Gaynor: you know that docstrings are actually added at applevel?19:52
Alex_Gaynorfijal: fine with me :)  it's not like __doc__ has _immutable_fields_?19:53
Alex_Gaynorfijal: http://paste.pocoo.org/show/531037/19:54
tilgovi (~randall@couchdb/developer/tilgovi) left irc: Remote host closed the connection19:54
fijalAlex_Gaynor: do you have any vision as to *why* this is failing?19:55
Alex_Gaynorfijal: in pdb I see that there's an IMPORT_NAME opcode with "numpypy" as teh str19:55
Alex_Gaynorfijal: ah, app_numpy.py19:56
fijalah19:56
fijalthat is a very confusing message :)19:56
Alex_Gaynorit's only confusing because appnumpy has no frame19:56
Alex_Gaynorin a traceback19:56
fijalyes19:56
kenaan12fijal import-numpy 119b39d93ee949 15/pypy/module/micronumpy/app_numpy.py: fix more tests, thanks alex19:57
kenaan12fijal import-numpy 1131e655394304 15/pypy/module/micronumpy/test/test_ufuncs.py: fix more tests19:59
fijal_ (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.20:01
fijal_wtf?20:01
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Read error: Connection reset by peer20:01
fijal_my computer just turned itself off20:01
Nick change: fijal_ -> fijal20:02
CIA-5303jmr 07roundup * 10#983/Publish hashes for source archives: 20:04
CIA-53[new] The download page currently lists hashes only for the binaries. Please provide20:04
CIA-53hashes for the source archives as well. (GPG signatures for ... * 14https://bugs.pypy.org/issue98320:04
nickik (~nick@217.150.234.222) joined #pypy.20:05
arigato(issue983: essential request, but that's my sarcastic opinion only)20:05
fijalarigato: it's not like we check whoever commits code20:07
fijalI can easily imagine someone slipping in some backdoor at least in short term20:07
arigatoyes20:08
Alex_Gaynorwell, I at least read every commit that comes in :)20:08
arigatoon the other hand, anyone concerned about this should get the mercurial source and update to the known revision of release-1.720:08
fijalAlex_Gaynor: yes, but not immediately20:08
fijalarigato: you can probably to man-in-the-middle?20:08
fijalno, not really20:08
fijalI have no idea what those people are after20:09
arigatonot really, you can't easily forge mercurial numbers20:09
fijalit's also https20:09
Action: arigato computes the hashes20:09
arigato...by downloading the file and computing the hashes without even trying to check that the file is absolutely correct20:10
arigatoof course20:10
CIA-5303arigo 07roundup * 10#983/Publish hashes for source archives: 20:11
CIA-53[chatting] fd0ad58b92ca0933c087bb93a82fda9e release-1.7.tar.bz220:11
CIA-53b4be3a8dc69cd838a49382867db3c41864b9e8d9 release-1.7.tar.bz2 * 14https://bugs.pypy.org/issue98320:11
fijalno, I expect you to read the entire pypy source code :)20:11
arigatoI could at least check that the file is exactly the same as the one originally produced from hg20:11
arigatobut it's so likely to be a waste of time that I won't investigate that20:12
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Read error: Connection reset by peer20:14
kenaan12arigo pypy.org[extradoc] 113b4a745b2e65 15/: Add hashes.20:14
kenaan12arigo pypy.org[extradoc] 11c809887b1d6e 15/: merge heads20:14
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.20:15
CIA-5303arigo 07roundup * 10#983/Publish hashes for source archives: 20:23
CIA-53[chatting] Closing either as "resolved" or as "wontfix", depending on your position. I20:23
CIA-53suspect that the position of many pypy developers is "res ... * 14https://bugs.pypy.org/issue98320:23
andrewfr (~andrew@modemcable211.101-70-69.static.videotron.ca) joined #pypy.20:25
oal (u4126@gateway/web/irccloud.com/x-opcpdwvmovryeapc) left irc: Read error: Connection reset by peer20:25
oal (u4126@gateway/web/irccloud.com/x-dnnrsleqejrakwcd) joined #pypy.20:31
JaRoel|4d (~jaroel|4d@sink.jaroel.nl) joined #pypy.20:33
andrewfr (andrew@modemcable211.101-70-69.static.videotron.ca) left #pypy.20:38
Rhy0lite (dje@nat/ibm/x-bzlldijoxwcartuu) left irc: Quit: Leaving20:44
craigkerstiens (~craigkers@204.14.152.118) joined #pypy.20:50
lukegb (~lukegb@2a01:4f8:101:2403::2:1) joined #pypy.20:50
aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy.20:52
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds21:00
tilgovi (~randall@75-149-58-169-SFBA.hfc.comcastbusiness.net) joined #pypy.21:03
tilgovi (~randall@75-149-58-169-SFBA.hfc.comcastbusiness.net) left irc: Changing host21:03
tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy.21:03
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.21:06
fijalAlex_Gaynor: ping21:07
Alex_Gaynorfijal: pong21:07
fijalso...21:07
fijalwe can't import numpy21:07
fijalgive up21:07
Alex_Gaynorwhy not?21:07
fijalbut we have to tell people not to copy pieces21:07
Alex_Gaynorerr, what license is numpy anyways?21:07
fijalbecause it's shitload of work to just mock stuff21:07
fijalthis code is insane21:08
fijalor maybe if they want we can import temporary hacks21:08
Alex_Gaynorwhat if we modify their code slightly?21:08
fijalmikefc: ping21:08
fijal"slightly"21:08
fijalAlex_Gaynor: you do that :)21:08
fijalno, you can't modify it slightly21:08
fijalbecause everything happens at import time21:08
fijalyou can try to copy pieces though21:08
fijalif you care21:08
fijalthat's what mikefc was doing21:08
fijaland I complained :)21:08
Alex_Gaynorwhat license is nump?21:08
Alex_Gaynorlooks to be bsd21:09
fijalno, not license21:09
fijaltechnically you can't modify it slightly21:09
Arfrever (~Arfrever@apache/committer/Arfrever) joined #pypy.21:11
Guest___ (~textual@67.51.82.66) joined #pypy.21:12
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Remote host closed the connection21:15
Action: arigato good night21:18
arigato (~arigo@xdsl-188-154-144-172.adslplus.ch) left irc: Quit: See you21:18
xcombelle (~xcombelle@AToulouse-551-1-132-44.w86-201.abo.wanadoo.fr) left irc: Read error: Operation timed out21:21
GoGican I use ipython with pypy?21:23
Nick change: Guest___ -> rekamso21:25
Alex_GaynorGoGi: yes21:28
GoGican I somehow install it into the pypy site-packages or how does it work?21:28
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.21:29
Alex_GaynorGoGi: I'd create a virtualenv for pypy and just pip install ipython21:37
GoGiI have tried virtualenv21:38
GoGiit actually copies over the pypy binary21:38
GoGiinto the virtual environment21:38
GoGiis this correct?21:38
tumbleweedthat's how virtualenvs work21:38
GoGibut it does not copy the lib_pypy directory21:38
GoGiso how will the pypy binary find lib_pypy?21:39
tumbleweedhrm, it should probably copy it21:40
Action: tumbleweed hasn't played much with pypy and ve yet, so I'll shut up :)21:41
GoGiwell it seems to work21:45
[Arfrever] (~Arfrever@apache/committer/Arfrever) left irc: Read error: Connection reset by peer21:46
[Arfrever] (~Arfrever@apache/committer/Arfrever) joined #pypy.21:47
GoGiit actually created a lib_pypy directory21:47
GoGibut it contains only readline.py21:48
GoGiwhat is going on?21:48
fijalAlex_Gaynor: I'll incorporate people's patches tomorrow then21:53
fijalthey were right I was wrong :)21:53
Alex_Gaynorfijal: ok, I'll continue on tracebin then :)21:53
Alex_GaynorI have the models, and some demo data, now I'm updating templates for real data21:53
fijal_ (~fijal@196-215-5-11.dynamic.isadsl.co.za) joined #pypy.21:54
fijal (~fijal@196-215-5-11.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds21:57
GoGiwell it seems to work21:58
GoGican I somehow install scipy and tell it to use the numpy replacement from pypy?21:58
exarkunno22:01
GoGiis this only the case because numpypy is not complete?22:04
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Read error: Connection reset by peer22:04
aleksi (~aleksi@109-184-50-32.dynamic.mts-nn.ru) left irc: Remote host closed the connection22:05
fprimex (~fprimex@brent-macbook.sc.fsu.edu) left irc: Quit: rushing home22:06
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy22:06
exarkunGoGi: At least in part, yes.22:07
exarkunGoGi: Additionally, scipy contains lots of extra C code that probably won't work.22:07
exarkunGoGi: fijal figured out something that allowed matplotlib to work.  It's possible that solution could be applicable to scipy, but I don't know many of the details.  There's a blog post about it on the PyPy blog, though.22:08
fijal_GoGi: it's possible, but not *right now*22:09
fijal_GoGi: if you're interested in working on making scipy work, you can definitely help22:10
fijal_but you can't use it right now22:10
Nick change: fijal_ -> fijal22:10
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.22:13
Husio (husiatyn@oceanic.wsisiz.edu.pl) left irc: Ping timeout: 240 seconds22:14
Husio (husiatyn@oceanic.wsisiz.edu.pl) joined #pypy.22:15
kenaan12fijal import-numpy 11eb12a969ddf7 15/: abandon this approach22:18
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Read error: Connection reset by peer22:19
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.22:23
stakkars_ (~tismer@i59F6B0A9.versanet.de) joined #pypy.22:28
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Read error: Connection reset by peer22:34
mikefc"abandon this approach"22:37
mikefcoh, how I often I have heard/used those words :)22:37
mikefcfijal: pong22:37
mikefcfijal: no luck doing the fancy numpy import?22:38
fijalmikefc: well, you were right, I gonna just put your patches22:38
fijalyeah22:38
mikefcfijal: so that means lots of manual copy-ing/fixing of the numpy stuff into numpypy?22:38
fijalwell22:38
fijalit depends22:38
fijalwhat's the goal22:38
fijalis the goal to have some functionality now, or when we're done?22:38
mikefcmy goal is a little bit more functionality now, but i understand that that may not be possible.22:39
mikefci'm willing to hold off on doing anything until you have the internals figured.22:39
fijalwell22:40
fijalif you build stuff now you'll get functionality now22:40
fijalinternals are pretty good22:40
fijalbut I won't do that for example22:40
fijalso I'm going to incorporate your patches and you'll have this functionality tomorrow22:40
mikefcok.  I'm open to suggestions on how you think I should proceed.  I want to get a few more applevel things in place - even if it is just docstrings and NotImplemented exceptions. is that OK?22:42
fijalyes22:42
fijaljust submit patches22:42
mikefcI guess my 'from_numeric' patch could be put in a separate file, so that the layout at least mimics numpy's layout.22:42
mikefcit would make app_numpy.py less crowded22:43
mikefcso if you hold off on that fromnumeric patch, i will separate it out sometime this weekend. when/if I get the time.22:44
mikefcso i would create an app_fromnumeric and do the appropriate things in __init__.py22:44
mikefcand an app_numeric.py22:47
nedbat (~nedbat@python/psf/nedbat) left irc: Disconnected by services22:47
mikefci also would like to maybe do a little bit of linalg hacking. maybe just get a couple of functions in there (at applevel)22:47
nedbat_ (~nedbat@python/psf/nedbat) joined #pypy.22:48
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.22:48
mikefcfijal: you said a few hours back to Alex_Gaynor  "but we have to tell people not to copy pieces" in regards to numpy/numpypy.   What do you mean by this?22:49
fijalnothing interesting actually22:49
fijalmikefc: yes, it should not go to app_numpy.py22:50
fijalinstead it should go to lib_pypy/numpypy22:50
mikefcso applevel stuff "copied" over from numpy will go in lib_pypy/numpypy, but there'll still be a mixed module in pypy/module?22:55
mikefc"copied" == copied and manually tweaked22:55
fijalyes22:57
fijalinterp level will stay in mixed module22:57
fijalit does not have to have applevel parts to be mixed :)22:57
mikefcok. I'll have a go when I get the chance.  22:58
Action: mikefc has to go out. afk.22:58
Husio (husiatyn@oceanic.wsisiz.edu.pl) left irc: Ping timeout: 252 seconds22:59
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Read error: Connection reset by peer22:59
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/19623:00
bbot2Started: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/76823:00
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/100523:00
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64-2/builds/2423:00
Husio (husiatyn@oceanic.wsisiz.edu.pl) joined #pypy.23:00
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.23:00
stakkars_howdy!23:01
stakkars_I want to improve IRC logging, have better clients, etc.23:01
stakkars_I'm looking into "Quassel", which seems to be pretty close to what I want.23:02
stakkars_Does anyone have experience with Quassel?23:02
ronnystakkars_: try it for a bit, the client-server idea for up <> bncish thing is a nice idea, its up to you to figure if it works out for you23:05
stakkars_ronny: yes, will try it. I thought to set thngs up using my new server, as a better approach to23:08
stakkars_the IRC logs with pybot. I would like to have integration with the IRC client, so that people can browse IRC with a remote connection that never stops.23:09
stakkars_so I would enable this for all interested #pypy users as a service23:10
whyking_ (~quassel@p5B3DD544.dip.t-dialin.net) left irc: Ping timeout: 255 seconds23:14
tilgoviI'm having trouble finding information on creating a shared library pypy23:18
tilgoviwhere should I be looking in the docs?23:18
espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 252 seconds23:19
tilgovinevermind. found it in configuration.23:19
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Ping timeout: 240 seconds23:22
espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy.23:25
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.23:28
Husio (husiatyn@oceanic.wsisiz.edu.pl) left irc: Ping timeout: 240 seconds23:33
whitelynx (~whitelynx@63.241.75.144) left irc: Quit: Ex-Chat23:34
Husio (husiatyn@oceanic.wsisiz.edu.pl) joined #pypy.23:35
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) left irc: Ping timeout: 240 seconds23:37
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 252 seconds23:44
Husio (husiatyn@oceanic.wsisiz.edu.pl) left irc: Ping timeout: 240 seconds23:50
tilgovihmm... the compatibility matrix in the 1.7 docs seems wrong. doesn't jit+stackless work now? I thought that was the default in the binary download.23:51
Husio (husiatyn@oceanic.wsisiz.edu.pl) joined #pypy.23:51
nedbat_ (~nedbat@python/psf/nedbat) left irc: Ping timeout: 248 seconds23:57
Shanita (~John@osbk-4db16fe1.pool.mediaWays.net) joined #pypy.23:58
--- Sat Jan 7 201200:00

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