| antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 00:18 | |
| aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy. | 00:28 | |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) left irc: Ping timeout: 252 seconds | 00:32 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/758 | 00:33 |
|---|---|---|
| [mat^2] (mathias@87.55.0.198) left irc: Ping timeout: 244 seconds | 00:36 | |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) joined #pypy. | 00:37 | |
| nickik (~nick@2001:67c:20a1:1018:226:9eff:feef:8369) joined #pypy. | 00:40 | |
| tellone (~tellone@81-232-82-223-no32.tbcn.telia.com) left irc: Quit: Lämnar | 00:50 | |
| nickik (~nick@2001:67c:20a1:1018:226:9eff:feef:8369) left irc: Quit: Verlassend | 00:52 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64-2/builds/14 | 00:52 |
| Trundle (~andy@python/site-packages/trundle) left irc: Ping timeout: 240 seconds | 00:55 | |
| lincolnn (~lincolnn0@cpe-98-154-215-147.socal.res.rr.com) left irc: | 01:01 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/186 | 01:01 |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) left irc: Ping timeout: 244 seconds | 01:07 | |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) joined #pypy. | 01:28 | |
| kenaan | 12mattip numpypy-axisops 114e517ebe55c0 15/pypy/module/micronumpy/: fix test_axis_iterator, start to see some results | 01:29 |
| kenaan | 12mattip numpypy-axisops 11ff482782c5fe 15/pypy/module/micronumpy/interp_numarray.py: bugfix, passes tests | 01:30 |
| kenaan | 12mattip numpypy-axisops 111714b0167e37 15/pypy/: merge with default | 01:30 |
| kenaan | 12mattip numpypy-axisops 11e3824eb2fc4c 15/pypy/module/micronumpy/: more finely distinguish between different 'promote_?' s | 01:30 |
| kenaan | 12mattip numpypy-axisops 11f1a8c490b279 15/pypy/module/micronumpy/interp_numarray.py: whoops, start fixes for translate | 01:30 |
| mattip (~chatzilla@bzq-79-177-107-57.red.bezeqint.net) joined #pypy. | 01:30 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1903 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/655 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/319 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/309 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/658 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1174 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/841 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1493 | 01:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/768 | 01:30 |
| mattip | what does it mean if "pytest.py pypy/module/micronumpy/test/test_numarray.py -x" bonks on test_argmax, | 01:31 |
| mattip | but "pytest.py pypy/module/micronumpy/test/test_numarray.py -k argmax" succeeds? | 01:32 |
| mattip | (on numpypy-axisops branch) | 01:32 |
| mattip (chatzilla@bzq-79-177-107-57.red.bezeqint.net) left #pypy. | 01:35 | |
| derdon (~derdon@pD9E1D53F.dip.t-dialin.net) left irc: Remote host closed the connection | 01:37 | |
| mimi_ (~mimi@97-88-28-179.dhcp.roch.mn.charter.com) left irc: Ping timeout: 240 seconds | 01:48 | |
| mimi_ (~mimi@97-88-28-179.dhcp.roch.mn.charter.com) joined #pypy. | 02:03 | |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) left irc: Ping timeout: 244 seconds | 02:08 | |
| aboudreault (~alanb@osgeo/member/aboudreault) left irc: Ping timeout: 240 seconds | 02:09 | |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) joined #pypy. | 02:09 | |
| chickamade (~chickamad@118.69.67.203) joined #pypy. | 02:10 | |
| chickamade (~chickamad@118.69.67.203) left irc: Client Quit | 02:15 | |
| chickamade (~chickamad@118.69.67.203) joined #pypy. | 02:22 | |
| chickamade (~chickamad@118.69.67.203) left irc: Client Quit | 02:23 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/995 | 02:44 |
| nettok (~quassel@200.119.189.163) joined #pypy. | 02:53 | |
| asmeurer (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) left irc: Quit: asmeurer | 03:44 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1903 | 03:47 |
| anish_ (anish@nat/redhat/x-gyzorzzhjupilurg) joined #pypy. | 04:10 | |
| durin42 (~durin@adium/durin42) left irc: Ping timeout: 240 seconds | 04:21 | |
| durin42 (~durin@adium/durin42) joined #pypy. | 04:23 | |
| kennethr_ (~kennethre@c-24-127-96-129.hsd1.va.comcast.net) left irc: Quit: Textual IRC Client: http://www.textualapp.com/ | 04:26 | |
| kennethreitz (~kennethre@c-24-127-96-129.hsd1.va.comcast.net) joined #pypy. | 04:26 | |
| ShadowE989 (~ShadowE98@74-47-218-61.dr02.rmny.wv.frontiernet.net) left irc: Ping timeout: 252 seconds | 04:40 | |
| nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 248 seconds | 04:42 | |
| zain (~textual@c-67-160-201-63.hsd1.ca.comcast.net) left irc: Quit: Textual IRC Client: http://www.textualapp.com/ | 04:50 | |
| ShadowE989 (~ShadowE98@74-36-227-143.br1.rmn.wv.frontiernet.net) joined #pypy. | 04:53 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1493 | 04:53 |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/658 | 05:00 |
| dracman (~draco@31-17-70-40-dynip.superkabel.de) left irc: Ping timeout: 252 seconds | 05:05 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/841 | 05:11 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/655 | 05:14 |
| Vorpal (~AnMaster@unaffiliated/anmaster) left irc: Ping timeout: 276 seconds | 05:16 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1174 | 05:19 |
| lahwran | drat, there is no ppa for ubuntu maverick | 05:30 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/768 | 06:19 |
| lahwran | is pypy's stdlib implementation of marshal the same one that will be reached via https://gist.github.com/37898376e48f3d54582b ? | 06:36 |
| lahwran | because `import marshal` seems to perform significantly faster than that does | 06:40 |
| lahwran | twice as fast, even | 06:40 |
| Arach (~arach@128.75.168.204) left irc: Quit: bye | 06:51 | |
| Arach (~arach@128.75.168.204) joined #pypy. | 06:51 | |
| aleksi (~aleksi@85.235.191.82) joined #pypy. | 06:56 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/319 | 06:57 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/309 | 07:10 |
| lahwran | it seems the second one (which appears to be pure python?) calls file.read, whilst a direct `import marshal` does not | 07:21 |
| lahwran | I am running the script in pypy, so it's pure pypy code either way | 07:21 |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) joined #pypy. | 07:35 | |
| kkris (~kris@93-82-42-211.adsl.highway.telekom.at) joined #pypy. | 07:35 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 07:40 | |
| setmeaway (~setmeaway@119.201.52.190) left irc: Ping timeout: 240 seconds | 07:43 | |
| antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) joined #pypy. | 07:45 | |
| mat^2 (mathias@87.55.0.198) joined #pypy. | 07:53 | |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) left irc: Ping timeout: 264 seconds | 07:53 | |
| JaRoel (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 08:06 | |
| mimi_ (~mimi@97-88-28-179.dhcp.roch.mn.charter.com) left irc: Quit: Leaving | 08:09 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Ping timeout: 248 seconds | 08:09 | |
| bivab (~david@ip-62-143-59-5.unitymediagroup.de) joined #pypy. | 08:17 | |
| fijal (~fijal@196-210-199-8.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds | 08:25 | |
| kenaan | 12fijal default 1187d1c652d40b 15/pypy/tool/jitlogparser/parser.py: oops, fix test_pypy_c | 08:43 |
| kenaan | 12fijal default 11f641bf7c9be8 15/pypy/rlib/jit.py: fix some tests | 08:45 |
| fijal (~fijal@196-210-199-8.dynamic.isadsl.co.za) joined #pypy. | 08:46 | |
| stakkars_ (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy. | 09:35 | |
| mikefc | hi fijal | 09:35 |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Ping timeout: 240 seconds | 09:36 | |
| Nick change: stakkars_ -> stakkars | 09:36 | |
| nettok (~quassel@200.119.189.163) left irc: Remote host closed the connection | 09:38 | |
| fijal | mikefc: hi | 10:00 |
| mitsuhiko (~mitsuhiko@ubuntu/member/mitsuhiko) left irc: Quit: ZNC - http://znc.sourceforge.net | 10:08 | |
| sunoano (~sa@unaffiliated/sunoano) left irc: Read error: Connection reset by peer | 10:09 | |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Quit: hasta la vista | 10:15 | |
| mitsuhiko (~mitsuhiko@ubuntu/member/mitsuhiko) joined #pypy. | 10:29 | |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy. | 10:42 | |
| azanella (~azanella@189.6.80.207) joined #pypy. | 10:51 | |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Quit: stakkars | 11:00 | |
| anish_ (anish@nat/redhat/x-gyzorzzhjupilurg) left irc: Quit: Leaving | 11:08 | |
| setmeaway (setmeaway3@119.201.52.190) joined #pypy. | 11:08 | |
| fijal | Alex_Gaynor: I'm fighting with "what do I want as a summary" | 11:18 |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy. | 11:28 | |
| derdon (~derdon@pD9E1C6E8.dip.t-dialin.net) joined #pypy. | 11:35 | |
| reiner (~reiner@46.128.136.223.dynamic.cablesurf.de) joined #pypy. | 11:49 | |
| Nick change: reiner -> squiddy | 11:49 | |
| nedbat (~nedbat@python/psf/nedbat) joined #pypy. | 12:11 | |
| danishman (~kvirc@0x5da0ec2a.cpe.ge-0-1-0-1104.ronnqu1.customer.tele.dk) joined #pypy. | 12:30 | |
| voidspace (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy. | 12:41 | |
| voidspace (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host | 12:41 | |
| voidspace (~anonymous@python/psf/voidspace) joined #pypy. | 12:41 | |
| Nick change: DasIch_ -> DasIch | 13:18 | |
| Arach (~arach@128.75.168.204) left irc: Quit: bye | 13:32 | |
| lahwran | nedbat: so when I use the builtin marshal, it cuts it down from a second startup/teardown time to a half second | 13:36 |
| lahwran | that's running in pypy | 13:37 |
| nedbat | lahwran: yes, because you don't have to import as much stuff. | 13:37 |
| lahwran | no | 13:37 |
| lahwran | I ran it through a profiler | 13:37 |
| nedbat | oh? | 13:37 |
| lahwran | the builtin marshal.py does not call file.read | 13:37 |
| lahwran | instead, it does whatever file.read does, directly | 13:37 |
| lahwran | err, drop '.py' from that sentence | 13:38 |
| lahwran | whereas the .py version seems to be twice as slow apparently just due to calling file.read | 13:38 |
| nedbat | lahwran: learn something new every day. | 13:39 |
| lahwran | so either we'll want to optimize marshal.py, which would amount to cutting down calls to file.read as much as possible, or ensure that it's safe to use the builtin marshal | 13:40 |
| lahwran | it also would be good to cut down on what parts of the stdlib the sandbox slave imports | 13:41 |
| nedbat | I'm curious about this: "Warning: cannot find your CPU L2 cache size in /proc/cpuinfo" | 13:41 |
| lahwran | fijal said that was originally a serious warning, but they decided to disregard it, and it's currently on his todo list of things to silence | 13:42 |
| lahwran | and silencing it is currently on his todo list* | 13:42 |
| tumbleweed | the minimark GC doesn't determine the L2 cache size any more | 13:42 |
| Action: tumbleweed has a patch to remove it entirely in the bug tracker | 13:42 | |
| lahwran | nedbat: a -Ojit recompile will disable the warning; it hasn't been removed from -O2 yet, I think | 13:43 |
| fijal | tumbleweed: hybrid does though | 13:43 |
| fijal | lahwran: it's not about silencing | 13:43 |
| fijal | it' | 13:43 |
| lahwran | oh? | 13:43 |
| fijal | s about pickup more more reasonable default | 13:43 |
| lahwran | ah | 13:43 |
| kenaan | 12ned nedbat-sandbox-2 11c03e4d0528e4 15/pypy/translator/sandbox/sandlib.py: Py is now required again. | 13:53 |
| kenaan | 12ned nedbat-sandbox-2 11cd7b359d0a00 15/pypy/translator/sandbox/pypy_interact.py: Change pypy_interact's default to quiet, you have to ask for verbose. | 13:53 |
| nedbat | fijal: I change pypy_interact's default to quiet. | 13:54 |
| fijal | nedbat: feel free to merge this stuff to default | 13:54 |
| nedbat | *changed | 13:54 |
| nedbat | fijal: thanks. | 13:54 |
| Arach (~arach@128.75.168.204) joined #pypy. | 13:55 | |
| kenaan | 12ned default 1142fbbcbfc2ac 15/pypy/translator/sandbox/: Merge sandbox tweaks from nedbat-sandbox-2 | 13:56 |
| fijal | close the branch | 13:56 |
| nedbat | fijal: you don't mind that I keep opening new ones? | 13:57 |
| fijal | not at all | 13:57 |
| fijal | we have 74 open branches or so, do you think one would make a difference? | 13:58 |
| fijal | if you have something small like flip a flag, commit it to default | 13:58 |
| fijal | if something bigger, preferred way is to create a branch which is named say sandbox-featurename | 13:58 |
| kenaan | 12ned nedbat-sandbox-2 1193bb4d305fdb 15/: Close nedbat-sandbox-2 | 13:58 |
| nedbat | fijal: i've named mine this way because I don't know exactly what I'll be changing in the sandbox, maybe sandbox-improvements ? | 13:59 |
| fijal | ok | 13:59 |
| nedbat | fijal: is there a way in hg to close the branch without having to make one last commit like I just did? Would --close-branch on the merge commit do the right thing? | 14:00 |
| fijal | no | 14:01 |
| fijal | --close-branch is just a commit | 14:01 |
| fijal | it's enough, you don't have to change anything | 14:01 |
| fijal | and you do | 14:01 |
| fijal | hg ci --close-branch | 14:01 |
| nedbat | right, that's what I just did, it just feels like there should be a way to say, "Merge and close this branch". OK. | 14:01 |
| fijal | I don't think there is | 14:01 |
| fijal | I'm nto aware at least | 14:02 |
| sunoano (~sa@unaffiliated/sunoano) joined #pypy. | 14:06 | |
| jyp (~interrupt@175.197.245.105) joined #pypy. | 14:09 | |
| jyp (~interrupt@175.197.245.105) left irc: Remote host closed the connection | 14:14 | |
| Alex_Gaynor | fijal: pong | 14:15 |
| fijal | Alex_Gaynor: so, I was thinking how to present the loops | 14:15 |
| fijal | we have 3 slices per-trace | 14:15 |
| fijal | we have before-the-first-label | 14:15 |
| fijal | between labels | 14:16 |
| fijal | and bridges | 14:16 |
| fijal | now bridges can have labels and loops as well | 14:16 |
| fijal | note that they don't all share the counter | 14:16 |
| Alex_Gaynor | I'm not sure, sorry, for jitviewer it's probably easiest to stay closest to what we have now though | 14:16 |
| fijal | well, the truth is we can't | 14:18 |
| fijal | ok, well, would be great if you had some inspiration :) | 14:20 |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) left irc: Read error: Connection reset by peer | 14:31 | |
| stakkars (~tismer@p5DDB7A14.dip.t-dialin.net) joined #pypy. | 14:31 | |
| Rhy0lite (dje@nat/ibm/x-ajkxaovjxvlqloer) joined #pypy. | 14:33 | |
| fijal | Alex_Gaynor: are you making mockups? | 14:40 |
| Alex_Gaynor | fijal: I'm going to convert my first one to be proper django templates, and then I'm going to start drawing up the trace viewer | 14:41 |
| fijal | Alex_Gaynor: so, did you find the "inlined function" any useful or was that a useless feature? | 14:41 |
| fijal | also brdiges | 14:41 |
| fijal | how did you find the bridge button working | 14:42 |
| Alex_Gaynor | you mean being able to jump into an inlined function? | 14:42 |
| Alex_Gaynor | that was useful | 14:42 |
| Alex_Gaynor | the bridges, a medium amount, but bridges were always so under-optimized it wasn't that useful | 14:42 |
| fijal | instead of having it inlined | 14:42 |
| Alex_Gaynor | I don't see how you display the function inlined with the sourc code | 14:42 |
| fijal | you put a hline or something | 14:44 |
| fijal | and insert a piece of source code | 14:44 |
| fijal | *very confusing* | 14:44 |
| fijal | you can also expand the inlined function in a box | 14:44 |
| fijal | so you get the source inside a little box | 14:44 |
| fijal | but I guess this is to be found | 14:44 |
| fijal | not that those options are that different | 14:44 |
| fijal | so what I was thinking about - split stuff into loops and linear traces of code | 14:45 |
| fijal | and display them in one chunk | 14:45 |
| fijal | then loops from bridges would be separate | 14:45 |
| fijal | but with bridge working as it is that links later to a loop with a link | 14:45 |
| fijal | do I make any sense? | 14:45 |
| Alex_Gaynor | i think so | 14:46 |
| fijal | ok | 14:46 |
| fijal | I'll try to wrap it up in a nice API | 14:46 |
| fijal | and then back to jitviewer | 14:46 |
| fijal | hakanardo: ping? | 14:46 |
| hakanardo | hi | 14:47 |
| fijal | hakanardo: how do I get a bridge that contains a label? | 15:00 |
| fijal | with a jump back to this label? | 15:01 |
| hakanardo | fijal: try test_multiple_specialied_zigzag in test_ajit for example | 15:04 |
| bivab_ (~david@ip-62-143-59-5.unitymediagroup.de) joined #pypy. | 15:05 | |
| bivab (~david@ip-62-143-59-5.unitymediagroup.de) left irc: Read error: Connection reset by peer | 15:05 | |
| Nick change: bivab_ -> bivab | 15:05 | |
| fijal | any way from python code? | 15:06 |
| fijal | or I should not really rely on how specific python code compiles to traces? | 15:06 |
| hakanardo | the same princeple should work at python level | 15:09 |
| hakanardo | getting retracing to work at applevel is however a bit tricky right now | 15:09 |
| hakanardo | (and have always been) | 15:10 |
| fijal | ok | 15:10 |
| hakanardo | you have to make sure that the virtuals states match | 15:10 |
| hakanardo | otherwise you'll end up with a jump to the preamble | 15:10 |
| hakanardo | instead of with a jump to the intermediate label in the bridge | 15:10 |
| fijal | nice | 15:11 |
| fijal | ok | 15:11 |
| fijal | I give up | 15:11 |
| hakanardo | you'll still get a bridge containing a label though if that was enough :) | 15:11 |
| fijal | n o it's not :) | 15:12 |
| Rhy0lite | bivab: hi david | 15:13 |
| fijal | the inner-outer-loop thingie also greatly increase the complexity of putting that into python code context | 15:13 |
| fijal | since you might have unrolled loops | 15:13 |
| bivab | Rhy0lite: hi | 15:13 |
| fijal | anyway, /me off surf | 15:13 |
| fijal | bbl | 15:13 |
| fijal | Rhy0lite: hi | 15:13 |
| Rhy0lite | fijal: hi | 15:13 |
| Rhy0lite | enjoy your surfing | 15:14 |
| mcdonc_ (~mcdonc@ip72-209-213-54.dc.dc.cox.net) left irc: Ping timeout: 244 seconds | 15:18 | |
| danishman (~kvirc@0x5da0ec2a.cpe.ge-0-1-0-1104.ronnqu1.customer.tele.dk) left irc: Quit: KVIrc 4.1.3 Equilibrium http://www.kvirc.net/ | 15:26 | |
| kkris1 (~kris@80-123-37-148.adsl.highway.telekom.at) joined #pypy. | 15:35 | |
| Kaskuka (~John@osbk-4db1432e.pool.mediaWays.net) left irc: Read error: Connection reset by peer | 15:36 | |
| Nick change: brownan -> abrown | 15:36 | |
| kkris (~kris@93-82-42-211.adsl.highway.telekom.at) left irc: Ping timeout: 252 seconds | 15:37 | |
| Kaskuka (~John@osbk-4db176e0.pool.mediaWays.net) joined #pypy. | 15:40 | |
| asmeurer (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) joined #pypy. | 15:45 | |
| Nick change: abrown -> brownan | 16:01 | |
| quasarj (~QAJarosz@170.94.194.242) joined #pypy. | 16:03 | |
| mat^2 (mathias@87.55.0.198) left irc: Ping timeout: 276 seconds | 16:04 | |
| aleksi (~aleksi@85.235.191.82) left irc: Remote host closed the connection | 16:06 | |
| dracman (~draco@31-17-70-40-dynip.superkabel.de) joined #pypy. | 16:06 | |
| Ademan (~dan@adsl-71-141-224-79.dsl.snfc21.pacbell.net) left irc: Quit: leaving | 16:07 | |
| setmeaway (setmeaway3@119.201.52.190) left irc: Read error: Connection reset by peer | 16:18 | |
| setmeaway (setmeaway@119.201.52.190) joined #pypy. | 16:22 | |
| JaRoel (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 16:22 | |
| mat^2 (mathias@87.55.0.198) joined #pypy. | 16:26 | |
| Vorpal (~AnMaster@unaffiliated/anmaster) joined #pypy. | 16:31 | |
| mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) joined #pypy. | 16:41 | |
| voidspace (~anonymous@python/psf/voidspace) left irc: Quit: voidspace | 16:49 | |
| lahwran | what in the world | 16:50 |
| lahwran | https://gist.github.com/e82a9ca40d786af707a0 | 16:51 |
| lahwran | nedbat: prod, if you have a moment; you have any idea why getopt might not be understanding those arguments I fed it? | 16:52 |
| lahwran | I added a `print sys.argv` in pypy_interact after it didn't work the first time; it output ['.../pypy/translator/sandbox/pypy_interact.py', '--timeout', '8', '--heapsize', '20M', '--tmp', '/tmp/sandboxtEWHrC', '/.../pypy/translator/goal/pypy-c', '/tmp/sandboxtEWHrC/slave.py'] | 16:52 |
| lahwran | I replaced the full path to my working dir with ... because it's quite long; I did this by hand in a text editor | 16:53 |
| nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 245 seconds | 16:53 | |
| lahwran | nope, no moments available from him | 16:53 |
| jterrace (~jterrace@aegis.CS.Princeton.EDU) joined #pypy. | 16:55 | |
| Rhy0lite | bivab: ping | 16:57 |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) joined #pypy. | 17:03 | |
| nettok (~quassel@200.119.171.54) joined #pypy. | 17:03 | |
| nettok_ (~quassel@200.119.164.208) joined #pypy. | 17:14 | |
| nettok (~quassel@200.119.171.54) left irc: Ping timeout: 276 seconds | 17:16 | |
| jterrace | how come cpython and pypy's multiprocessing.connection Listener/Client are not compatible? | 17:18 |
| jterrace | I get IOError: bad message length - is it something with pickling not being compatible? | 17:19 |
| nedbat (~nedbat@python/psf/nedbat) joined #pypy. | 17:19 | |
| lcubo (~leonardo@190.94.99.125) joined #pypy. | 17:23 | |
| voidspace (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy. | 17:24 | |
| voidspace (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host | 17:24 | |
| voidspace (~anonymous@python/psf/voidspace) joined #pypy. | 17:24 | |
| voidspace (~anonymous@python/psf/voidspace) left irc: Client Quit | 17:25 | |
| Alex_Gaynor | jterrace: this was fixed a few days ago | 17:27 |
| Alex_Gaynor | jterrace: it was an endianess issue, we forgot to call htonl | 17:27 |
| jterrace | ooh let me try a nightly | 17:27 |
| jterrace | i just found the pypy-dev list | 17:27 |
| jterrace | it works | 17:29 |
| jterrace | nice | 17:29 |
| lesshaste (~lesshaste@87-194-206-189.bethere.co.uk) joined #pypy. | 17:29 | |
| lesshaste | hi all | 17:29 |
| lesshaste | I was reading http://pyinsci.blogspot.com/2010/12/efficcient-mcmc-in-python.html | 17:29 |
| lesshaste | as far as I can tell, pypy is now as fast as the super optimised cython version | 17:29 |
| Alex_Gaynor | yay :) | 17:30 |
| lesshaste | although I can't compile the gsl version.. the non -gsl version is almost 3 times slower than pypy :) | 17:31 |
| espes (espes@2001:470:1f05:ca1::) left irc: Remote host closed the connection | 17:31 | |
| lesshaste | I was wondering if anyone here could get the gsl version to work? | 17:32 |
| Action: fijal back | 17:35 | |
| fijal | lesshaste: what is gsl? | 17:36 |
| lesshaste | http://www.gnu.org/software/gsl/ I beleive | 17:36 |
| lesshaste | believe | 17:36 |
| lesshaste | I get undefined reference to `gsl_rng_mt19937' .. I think it's just an annoying ubuntu installation problem | 17:37 |
| fijal | lesshaste: the background is scary | 17:37 |
| lesshaste | fijal: the background of the webpage? | 17:38 |
| jterrace | GSL is widely used and very well optimized afaik | 17:38 |
| fijal | lesshaste: yes | 17:39 |
| fijal | gsl_rng_mt19937 | 17:39 |
| lesshaste | ah yes | 17:39 |
| lesshaste | got it to work | 17:39 |
| fijal | what an awesome variable name | 17:39 |
| lesshaste | just needed gcc -O3 -I/usr/include/python2.6 -o fasttest fasttest.c -lpython2.6 -lpthread -lm -lutil -ldl -fno-strict-aliasing -lgsl -lgslcblas | 17:39 |
| lesshaste | :) | 17:40 |
| lesshaste | wow that's fast | 17:40 |
| fijal | lesshaste: results? | 17:40 |
| fijal | lesshaste: are you by chance comparing various RNGs? | 17:40 |
| lesshaste | the version before the gsl one on http://pyinsci.blogspot.com/2010/12/efficcient-mcmc-in-python.html | 17:41 |
| ousado | well, that's the correct name for that version of mersenne twister | 17:41 |
| lesshaste | about 3 mins in cpython | 17:41 |
| lesshaste | 1 min in pypy | 17:41 |
| lesshaste | 6 seconds using gsl | 17:41 |
| fijal | lesshaste: yeah, but you're having different random sources | 17:41 |
| fijal | that's kinda boring | 17:41 |
| jterrace | did you try PyGSL? | 17:43 |
| jterrace | or is your cython wrapper basically the same as PyGSL | 17:45 |
| lesshaste | jterrace: I didn't | 17:48 |
| fijal | lesshaste: it's really apples and oranges though | 17:50 |
| Alex_Gaynor | fijal: well, I'm guessing it's namespacing it with gsl, rng is from the random number generator package there, and mt19937 is a messerne twister is a a 19937 period or something | 17:50 |
| fijal | Alex_Gaynor: still, comparing python stdlib against GSL is kind of silly | 17:50 |
| Alex_Gaynor | fijal: yes of course, I just meant about the variable name | 17:51 |
| fijal | I would not have figured this out | 17:51 |
| fijal | someone is doing us fairly black PR | 17:51 |
| Alex_Gaynor | ? | 17:52 |
| fijal | http://groups.google.com/group/golang-nuts/browse_thread/thread/30dc63db6927646c | 17:52 |
| fijal | grep for "in my experience" | 17:52 |
| Alex_Gaynor | "the first result was so unexpected that I abandoned further tests" that's usually a way to do good science /sigh | 17:52 |
| espes (espes@2001:470:1f05:ca1::) joined #pypy. | 17:53 | |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 17:54 | |
| fijal | science is haaard | 17:54 |
| mtigas (~mtigas@users.spokesman.com) joined #pypy. | 17:54 | |
| jterrace | that's a terrible benchmark in the first place | 17:54 |
| fijal | jterrace: yes, but also noone even run it on pypy and then is posting stuff | 17:55 |
| kenaan | 12hager ppc-jit-backend 1144255e0c9ffe 15/pypy/jit/backend/ppc/ppcgen/: (edelsohn, bivab) fix calls to leave_jitted_hook on PPC64 as well as restoring the LR to return from the co... | 17:55 |
| jterrace | hmm? | 17:55 |
| jterrace | he said he ran it | 17:55 |
| fijal | grep for "in my experience" | 17:55 |
| jterrace | i did | 17:55 |
| jterrace | read the next sentence | 17:56 |
| jterrace | "PyPy runs Reinhard's table sort benchmark at the same speed as CPython2.7 (32-bit x86 CPU). PyPy uses 80% more memory than CPython2.7. " | 17:56 |
| fijal | well ok :) | 17:56 |
| jterrace | the benchmark only runs it once though | 17:57 |
| jterrace | it seems | 17:57 |
| bivab (~david@ip-62-143-59-5.unitymediagroup.de) left irc: Quit: bivab | 17:57 | |
| fijal | doh | 17:59 |
| fijal | it's super silly anyway | 17:59 |
| fijal | http://wikimediafoundation.org/wiki/RFP/XML_Dumps_Help | 18:00 |
| fijal | wtf | 18:00 |
| fijal | wikimedia is doing xml dumps | 18:01 |
| Rhy0lite | fijal: edits done | 18:01 |
| fijal | and then complains it's slow | 18:01 |
| fijal | .... | 18:01 |
| fijal | Rhy0lite: thx | 18:01 |
| fijal | anything interesting that you changed? | 18:01 |
| Rhy0lite | fijal: I credited David for PPC backend help | 18:02 |
| fijal | ugh, sorry | 18:02 |
| fijal | I thought I mentioned everyone | 18:02 |
| squiddy (~reiner@46.128.136.223.dynamic.cablesurf.de) left irc: Quit: Verlassend | 18:02 | |
| Rhy0lite | just some minor English phrases | 18:02 |
| Rhy0lite | drastic -> dramatic | 18:03 |
| Rhy0lite | (drastic progress -> dramatic progress) | 18:03 |
| Rhy0lite | angles -> facets / aspects | 18:03 |
| fijal | "In case of codes unrelated to the tests PyPy team is dealing with, the outcome of using PyPy instead of CPython is questionable." | 18:04 |
| fijal | lesshaste: I do consider this black PR | 18:04 |
| jterrace | lesshaste: i think you could also write that gibbs function using numpy | 18:04 |
| jterrace | and would be way faster | 18:04 |
| fijal | Alex_Gaynor: note than numeric_4 and numeric_5 from laplace benchmark are slow because of invalidation :/ | 18:04 |
| lesshaste | fijal: sorry my point was how great pypy was as it is 3 times faster than the basic cython implementation :) | 18:05 |
| lesshaste | fijal: the gsl thing was just a distraction for those who like to talk about optimatisation | 18:05 |
| fijal | ok | 18:05 |
| derdon (~derdon@pD9E1C6E8.dip.t-dialin.net) left irc: Remote host closed the connection | 18:05 | |
| Alex_Gaynor | fijal: ugh right, we still don't have a solution there do we? | 18:05 |
| fijal | lesshaste: I was referring to the golang-nuts though | 18:05 |
| fijal | Alex_Gaynor: no :/ | 18:05 |
| lesshaste | fijal: oh ok..I haven't been following closely, sorru | 18:05 |
| fijal | np | 18:06 |
| Alex_Gaynor | fijal: can you show numbers with and without the hack? | 18:06 |
| Alex_Gaynor | numpypy.pypy.remove_invalidates | 18:06 |
| fijal | yes | 18:08 |
| fijal | 5.3 vs 1.9 or so | 18:08 |
| lahwran | does pypy keep any hidden caches of any kind? | 18:14 |
| lahwran | such as something that could possibly cause my code to not update when I edit it and run it again | 18:14 |
| Rhy0lite | lahwran: do you mean caches in the file system? | 18:14 |
| lahwran | yes | 18:14 |
| Rhy0lite | no | 18:14 |
| lahwran | are you completely sure | 18:15 |
| Rhy0lite | it will completely re-trace and re-jit your code | 18:15 |
| lahwran | dammit then what is happening | 18:15 |
| lahwran | gaaah this shouldn;t be possible | 18:15 |
| Rhy0lite | there are discussions about how to make PyPy that smart | 18:15 |
| Rhy0lite | but it currently isn't | 18:15 |
| Rhy0lite | are you sure nothing has changed in your data? | 18:15 |
| quasarj (~QAJarosz@170.94.194.242) left irc: Quit: Leaving. | 18:15 | |
| Rhy0lite | nothing got overwritten? | 18:15 |
| Rhy0lite | if your program using any external input | 18:16 |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) joined #pypy. | 18:29 | |
| espes (espes@2001:470:1f05:ca1::) left irc: Ping timeout: 268 seconds | 18:33 | |
| kenaan | 12edelsohn ppc-jit-backend 114b4775dd052f 15/pypy/jit/backend/ppc/ppcgen/arch.py: BACKCHAIN_SIZE is 6 | 18:34 |
| espes (~espes@123-243-221-248.tpgi.com.au) joined #pypy. | 18:34 | |
| fijal | lahwran: .pyc files though are kept | 18:38 |
| lesshaste | jterrace: that would be interesting | 18:42 |
| lesshaste | jterrace: I would personally find any version that could be pypy'ed to run in something like 6 seconds very interesting too | 18:42 |
| lesshaste | fijal: going back to the gsl version.. is your main "apples versus oranges" point just that the random number generators are different or more that gsl is a specialised library and so bound to be super fast | 18:43 |
| fijal | lesshaste: it's not "bound" to be super fast | 18:44 |
| fijal | but I would expect gsl authors to pay more attention than stdlib's random number authors | 18:44 |
| fijal | especially that by the time rng for python was written, there was no pypy even | 18:44 |
| fijal | lesshaste: note that it's not a given that for any algo gsl will be faster than pypy | 18:45 |
| lesshaste | fijal: right... what I meant was could you advise a related but fair comparison :) | 18:45 |
| lesshaste | fijal: I am in the mood for benchmarking :) | 18:45 |
| lesshaste | (please) | 18:45 |
| fijal | you would need to write mersenne twister using numpy I believe | 18:46 |
| fijal | and pay attention how you do that | 18:46 |
| fijal | right now even our trace viewer is broken :/ | 18:46 |
| lesshaste | http://docs.scipy.org/doc/numpy/reference/generated/numpy.random.mtrand.RandomState.html ? | 18:46 |
| fijal | this one is not implemented though :) | 18:47 |
| fijal | btw | 18:47 |
| fijal | what you would end is to call gsl probably | 18:47 |
| fijal | it's pretty boring | 18:47 |
| fijal | I meant some composition of what gsl offers | 18:48 |
| fijal | I bet that gsl will be faster on any algo that' | 18:48 |
| fijal | 's a single function call | 18:48 |
| lesshaste | ok.. oh well :) | 18:48 |
| lesshaste | I just wanted to have some scientific computing comparison | 18:48 |
| lesshaste | benchmarks are always hard to design well | 18:49 |
| fijal | I'm working on laplace now | 18:49 |
| lesshaste | cool! | 18:49 |
| lesshaste | I have a particular interest in mcmc however | 18:49 |
| fijal | ok | 18:49 |
| lesshaste | fijal: while I have the guru here.. :) Is it of any interest that shedskin is able to produce faster code than pypy sometimes for the tiny subset of python that it covers | 18:50 |
| lesshaste | the tiny subset that shedskin covers that is | 18:50 |
| fijal | no, it's not | 18:50 |
| fijal | it's usually very boring | 18:50 |
| fijal | and usually rpython is faster than shedskin anyway | 18:50 |
| fijal | besides, can you *still* find an example where shedskin is faster? | 18:51 |
| lesshaste | why is that? I mean from one perspective pypy has much higher goals. But from another shedskin is much less sophisticated and yet does a "better job" on toy examples | 18:51 |
| fijal | not specifically optimized for shedskin preferably | 18:51 |
| fijal | does it? | 18:51 |
| lesshaste | fijal: yes.. the random little toy scripts I write :) | 18:51 |
| fijal | example? | 18:51 |
| fijal | well, it's not supporting full python that's why | 18:51 |
| fijal | RPython will be faster | 18:51 |
| lesshaste | I mean a little bit better.. nothing great | 18:51 |
| lesshaste | fijal: I understand it does much less than pypy but is rpython a fair comparison? I mean I really write these toy scripts in python | 18:52 |
| lesshaste | and I assume lots of other people do to | 18:52 |
| fijal | yes, shedskin and rpython are very comparable | 18:52 |
| fijal | I think RPython covers slightly more | 18:53 |
| fijal | but they are a very compatible subset | 18:53 |
| fijal | and if you say a little better, you can easily blame it on our assembler generation not being as great | 18:53 |
| fijal | but also a bit of random checks that have to be there | 18:54 |
| lesshaste | http://pastebin.com/22sRCkkP is slightly faster in shedskin, for example | 18:54 |
| fijal | lesshaste: can you please use pastebin without captcha? | 18:54 |
| lesshaste | http://www.pastebin.ca/2097054 ? | 18:55 |
| fijal | well ok, I've seen that | 18:55 |
| lesshaste | right | 18:55 |
| fijal | this is a matter of looking into random module | 18:56 |
| fijal | maybe we can do something about it | 18:56 |
| Alex_Gaynor | fijal: I looked at inlining random module, didn't help almost at all (< 5%), even though it would have saved an allocation+box/unbox per iteration | 18:57 |
| Alex_Gaynor | fijal: It would probably have benefited from less insane call_assembler though :) | 18:58 |
| lesshaste | fijal: let me dig out another one.. just running some tests. .pypy may win every time of course :) | 18:58 |
| fijal | Alex_Gaynor: is random a module? | 18:58 |
| Alex_Gaynor | fijal: random is a .py file, it's _random that's a module | 18:58 |
| Alex_Gaynor | a MixedModule that is | 18:58 |
| [mat^2] (mathias@87.55.0.198) joined #pypy. | 18:59 | |
| mat^2 (mathias@87.55.0.198) left irc: Read error: Connection reset by peer | 18:59 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:00 | |
| nettok_ (~quassel@200.119.164.208) left irc: Ping timeout: 240 seconds | 19:00 | |
| nettok (~quassel@200.119.179.17) left irc: Read error: Connection reset by peer | 19:01 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:02 | |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer | 19:03 | |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) joined #pypy. | 19:04 | |
| nettok (~quassel@200.119.179.17) left irc: Read error: Connection reset by peer | 19:04 | |
| espes (~espes@123-243-221-248.tpgi.com.au) left irc: Ping timeout: 244 seconds | 19:04 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:05 | |
| nettok (~quassel@200.119.179.17) left irc: Read error: Connection reset by peer | 19:06 | |
| fijal | Alex_Gaynor: ok | 19:07 |
| fijal | lesshaste: then it's probably just general RPython inefficiency or so | 19:07 |
| fijal | nothing *that* interesting | 19:07 |
| Alex_Gaynor | fijal: no, _random is ok,99% of the code is in rlib.rrandom | 19:08 |
| lesshaste | fijal: ok.. I can't even compile shedskin from git source cuurently :) | 19:08 |
| [mat^2] (mathias@87.55.0.198) left irc: Read error: Connection reset by peer | 19:08 | |
| mat^2 (mathias@87.55.0.198) joined #pypy. | 19:08 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:08 | |
| nettok (~quassel@200.119.179.17) left irc: Read error: Connection reset by peer | 19:09 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:10 | |
| espes (espes@2001:470:1f05:ca1::) joined #pypy. | 19:10 | |
| nettok (~quassel@200.119.179.17) left irc: Read error: Connection reset by peer | 19:11 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:12 | |
| nettok (~quassel@200.119.179.17) left irc: Read error: Connection reset by peer | 19:13 | |
| nettok (~quassel@200.119.179.17) joined #pypy. | 19:14 | |
| `fox` (~fox@host139-108-dynamic.245-95-r.retail.telecomitalia.it) left irc: Ping timeout: 240 seconds | 19:21 | |
| jterrace | lesshaste: a numpy version: http://paste.pocoo.org/show/527034/ | 19:25 |
| jterrace | not pretty and im not 100% sure it's correct, but i think it is | 19:26 |
| lesshaste | ooh :) | 19:26 |
| Alex_Gaynor | jterrace: "numpy.lib.stride_tricks" I can't say that's exactly what I'd name the library, but what the hell! :) | 19:26 |
| lesshaste | jterrace: have you tried it with numpypy? | 19:26 |
| Alex_Gaynor | lesshaste: this won't run on it yet | 19:26 |
| lesshaste | ah ok | 19:26 |
| lesshaste | something to aim for :) | 19:27 |
| DasIch_ (~dasich@p4FFDCF99.dip.t-dialin.net) joined #pypy. | 19:27 | |
| Alex_Gaynor | jterrace: samples should obviously be a numpy record array :) | 19:27 |
| jterrace | results: pypy 7.8s, cpython + numpy 16.6s, cpython regular 1m3s | 19:27 |
| jterrace | so not even very good for speed | 19:27 |
| jterrace | and yeah, it wont work on numpypy | 19:27 |
| Alex_Gaynor | yet | 19:27 |
| lesshaste | jterrace: what was the pypy timed version? | 19:28 |
| jterrace | pypy 7.8s | 19:28 |
| jterrace | on 1.7 | 19:28 |
| lesshaste | fijal: I just tested a random script in shedskin.. pypy 30 seconds.. shedskin still hasn't completed :) | 19:28 |
| lesshaste | fijal: I think I hit a weak spot :) | 19:28 |
| lesshaste | jterrace: I mean given that it can't handle numpy | 19:28 |
| jterrace | oh that was the regular gibbs function | 19:29 |
| lesshaste | got you | 19:29 |
| lesshaste | so in short.. pypy rocks :) | 19:29 |
| lesshaste | but in this particular case is slightly slower than shedskin | 19:29 |
| Alex_Gaynor | fijal: I have the HTML mock all django'd up now. Now I have to think about the traceviewer part again. | 19:29 |
| jterrace | i dont think numpy is that great at those rolling window stride tricks either | 19:29 |
| lesshaste | even though shedskin is appallingly bad on this other bit of code I am testing | 19:29 |
| fijal | Alex_Gaynor: want some help/discussions? | 19:30 |
| jterrace | and actually, this might be evaluating the inside of the dstack before assigning | 19:30 |
| lesshaste | it's so bad I think I will report it to them as a bug :) | 19:30 |
| jterrace | so it might not even be correct | 19:30 |
| fijal | I didn't do anything today :/ | 19:30 |
| lesshaste | jterrace: ah | 19:30 |
| Alex_Gaynor | fijal: I want to sit down with a pen first | 19:30 |
| fijal | killed some monsters in flash | 19:30 |
| fijal | Alex_Gaynor: ok | 19:30 |
| DasIch (~dasich@p3E99135B.dip.t-dialin.net) left irc: Ping timeout: 252 seconds | 19:30 | |
| kenaan | 12edelsohn ppc-jit-backend 111d80c59520d2 15/pypy/jit/backend/ppc/ppcgen/opassembler.py: assert offset is_reg in cmp_guard_class. | 19:30 |
| kenaan | 12edelsohn ppc-jit-backend 116b75e619e2d4 15/pypy/jit/backend/ppc/ppcgen/codebuilder.py: Fix typo in loadx. | 19:31 |
| jterrace | question | 19:37 |
| fijal | ? | 19:37 |
| jterrace | could part of numpypy's rpython be compiled to CUDA? | 19:37 |
| jterrace | to make it... insanely fast | 19:37 |
| MostAwesomeDude | jterrace: There is a common misconception that GPUs are automatically fast. | 19:38 |
| Alex_Gaynor | jterrace: well, in theory you could write a CUDA JIT backend and then NumPy exprs could be compiled for it | 19:38 |
| jterrace | i mean there would be overhead, so you wouldnt want to do it for small ops | 19:38 |
| fijal | jterrace: yes | 19:38 |
| fijal | for some operations | 19:38 |
| MostAwesomeDude | The thing is that you need to include the time taken to send things out to the GPU and retrieve them. | 19:38 |
| Alex_Gaynor | jterrace: there's some engineering in there (you want to have two different JIT backends), but yes, in theory it's possible | 19:38 |
| fijal | and yes, what MAD says | 19:38 |
| MostAwesomeDude | So I don't think there's very many numpy things which are big enough to warrant it. Then again, I haven't done GPUs in a while. | 19:38 |
| jterrace | interesting | 19:38 |
| fijal | jterrace: we can invent manual hints though | 19:39 |
| fijal | like numpy.array(storage='GPU') | 19:39 |
| fijal | or so | 19:39 |
| fijal | then you can do better | 19:39 |
| fijal | so not quite automatically | 19:39 |
| fijal | would be a first step | 19:39 |
| fijal | SSE first though :) | 19:39 |
| peter_zfs (~a@ppp-94-65-237-41.home.otenet.gr) joined #pypy. | 19:39 | |
| Alex_Gaynor | +1 | 19:39 |
| jterrace | just thinking about it :) | 19:39 |
| fijal | jterrace: thinking is nice, doing is better :) | 19:40 |
| antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 19:41 | |
| Alex_Gaynor | fijal: is there a reason jit-running doesn't log which trace it's in? | 19:45 |
| Alex_Gaynor | I guess that doesn't totally work with bridges/call assembler though. | 19:45 |
| fijal | yes | 19:45 |
| fijal | not only | 19:45 |
| fijal | you have jumps and whatnot | 19:45 |
| Alex_Gaynor | eh? | 19:45 |
| fijal | but generally you would enter at some entry bridge | 19:46 |
| fijal | and go somewhere else | 19:46 |
| fijal | like to the loop | 19:46 |
| fijal | but indeed, it' | 19:46 |
| Action: Alex_Gaynor wants more profiling data | 19:46 | |
| fijal | s not too interesting | 19:46 |
| fijal | I was thinking about measuring time | 19:46 |
| fijal | can we do a neat trick somehow? | 19:46 |
| Alex_Gaynor | in theory you need to emit ASM to log at every entry-point, so the first label, and anything *some other trace* can jump to | 19:47 |
| Alex_Gaynor | or maybe the other trace should have the ASM | 19:47 |
| Alex_Gaynor | maybe more than that | 19:47 |
| Alex_Gaynor | I want to know how much time I spent in any label I guess | 19:47 |
| fijal | we have counters | 19:47 |
| Alex_Gaynor | so you really need it withthe count | 19:47 |
| fijal | but time is trickier | 19:47 |
| Alex_Gaynor | yes, because you need stop time | 19:47 |
| Alex_Gaynor | well, something to think about | 19:48 |
| fijal | and you can't just execute rtld at each iteration | 19:48 |
| fijal | that would kill your perf | 19:48 |
| Alex_Gaynor | well, not by default | 19:48 |
| Alex_Gaynor | err, yes not on every iteration | 19:48 |
| fijal | we have counters on every iteration | 19:48 |
| Alex_Gaynor | you want to do it 1) at entry bridge, 2) at start of loop, 3) when you exit loop | 19:48 |
| fijal | and that's "not too bad" | 19:48 |
| Alex_Gaynor | or so | 19:48 |
| fijal | but rtld is more expensive I think | 19:48 |
| fijal | well | 19:49 |
| fijal | the simplest-possible-thing | 19:49 |
| Alex_Gaynor | and all of this should be only with the appropriate logging flag | 19:49 |
| fijal | would be to modify things we have now | 19:49 |
| fijal | Alex_Gaynor: btw, this is what happens with counters | 19:49 |
| fijal | so we have *some* infrastructure | 19:49 |
| fijal | but "stop" timing would be tricky | 19:50 |
| fijal | we can also try to improve on cProfile | 19:50 |
| fijal | but that does not correspond to traces too well | 19:50 |
| Alex_Gaynor | right, ATM I"m thinking about trace timings | 19:51 |
| nekto0n (~nekto@v10-167-138.yandex.net) joined #pypy. | 19:54 | |
| nekto0n | hi there! tried to donate via google checkout but failed - no shipping to russia :) | 19:55 |
| kenaan | 12hakanardo jit-improve-nested-loops 11a7a84c778e28 15/pypy/: hg merge default | 20:00 |
| kenaan | 12hakanardo jit-improve-nested-loops 113f020e9bd542 15/pypy/translator/sandbox/: hg merge default | 20:00 |
| kenaan | 12hakanardo jit-improve-nested-loops 11c30513c22175 15/: closing to be merged branch | 20:00 |
| kenaan | 12hakanardo default 1169095778cbfd 15/pypy/jit/metainterp/: Merge jit-improve-nested-loops. It allows bridges to end with a jump to the the top of a loop and not only to t... | 20:00 |
| kenaan | 12edelsohn ppc-jit-backend 11cb6d39cadc56 15/pypy/jit/backend/ppc/ppcgen/opassembler.py: Implement emit_cond_call_gc_wb for PPC64. Fix typo in function address variable name. | 20:01 |
| fijal | nekto0n: hahaha :) | 20:08 |
| fijal | nekto0n: isn't paypal a bit better? | 20:09 |
| fijal | yeah, let's hope bitcoin becomes more popular | 20:10 |
| nekto0n | fijal: i guess so, but my I don't have paypal account. I bought some android apps and my card is automagically tied to google account :) | 20:10 |
| Alex_Gaynor | fijal: yes, that way we can have a currency which makes even less sense than our current ones | 20:11 |
| nekto0n | :) | 20:11 |
| fijal | nekto0n: nice | 20:17 |
| fijal | Alex_Gaynor: try living outside of the US | 20:17 |
| fijal | and buying anything online | 20:17 |
| Rhy0lite | fijal: I can't buy things in other countries either | 20:19 |
| Rhy0lite | trying sending a gift to a friend in another country from a story in that country | 20:19 |
| `fox` (~fox@host142-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 20:21 | |
| Nick change: DasIch_ -> DasIch | 20:22 | |
| Rhy0lite (dje@nat/ibm/x-ajkxaovjxvlqloer) left irc: Quit: Leaving | 20:24 | |
| aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy. | 20:24 | |
| nekto0n | things are getting better (at least here in russia) | 20:25 |
| nettok (~quassel@200.119.179.17) left irc: Remote host closed the connection | 20:26 | |
| ShadowE989 (~ShadowE98@74-36-227-143.br1.rmn.wv.frontiernet.net) left irc: Read error: Connection reset by peer | 20:31 | |
| steakknife (~steakknif@pfsense.hackerdojo.com) joined #pypy. | 20:57 | |
| nekto0n (nekto@v10-167-138.yandex.net) left #pypy. | 21:20 | |
| zain (~textual@cpe-68-173-70-124.nyc.res.rr.com) joined #pypy. | 21:22 | |
| lcubo (~leonardo@190.94.99.125) left irc: Ping timeout: 252 seconds | 21:23 | |
| Dulak (~michael@unaffiliated/dulak) left irc: Read error: Connection reset by peer | 21:26 | |
| Dulak (~michael@unaffiliated/dulak) joined #pypy. | 21:26 | |
| fox_ (~fox@host142-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 21:29 | |
| `fox` (~fox@host142-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer | 21:29 | |
| kkris1 (~kris@80-123-37-148.adsl.highway.telekom.at) left irc: Quit: Leaving. | 21:37 | |
| azanella (~azanella@189.6.80.207) left irc: Ping timeout: 240 seconds | 21:42 | |
| lesshaste | Rhyolite: I find ebay is the easiest way | 21:50 |
| lesshaste | they have them all over the world | 21:50 |
| JaRoel|4d (~jaroel|4d@sink.jaroel.nl) joined #pypy. | 21:53 | |
| K-os (~gabriel@cl-3210.ham-01.de.sixxs.net) joined #pypy. | 21:53 | |
| espes (espes@2001:470:1f05:ca1::) left irc: Read error: Connection timed out | 21:54 | |
| espes (espes@2001:470:1f05:ca1::) joined #pypy. | 21:54 | |
| K-os (~gabriel@cl-3210.ham-01.de.sixxs.net) left irc: Client Quit | 21:57 | |
| fox_ (~fox@host142-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 21:57 | |
| fijal (~fijal@196-210-199-8.dynamic.isadsl.co.za) left irc: Ping timeout: 240 seconds | 21:58 | |
| nettok (~quassel@200.119.177.100) joined #pypy. | 22:00 | |
| mtigas (~mtigas@users.spokesman.com) left irc: Quit: mtigas | 22:08 | |
| mat^2 (mathias@87.55.0.198) left irc: | 22:08 | |
| voidspace (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy. | 22:16 | |
| voidspace (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host | 22:16 | |
| voidspace (~anonymous@python/psf/voidspace) joined #pypy. | 22:16 | |
| unbit (~unbit@93-58-162-185.ip159.fastwebnet.it) left irc: Read error: Connection reset by peer | 22:16 | |
| unbit (~unbit@93-58-162-185.ip159.fastwebnet.it) joined #pypy. | 22:16 | |
| claude_ (~claude@178235011238.warszawa.vectranet.pl) joined #pypy. | 22:22 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/759 | 22:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/187 | 22:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/996 | 22:30 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64-2/builds/15 | 22:30 |
| jterrace (~jterrace@aegis.CS.Princeton.EDU) left irc: Quit: Leaving. | 22:30 | |
| lesshaste (~lesshaste@87-194-206-189.bethere.co.uk) left irc: Quit: Leaving | 22:30 | |
| Nick change: masquera1e -> masquerade | 22:39 | |
| forrestv (~forrestv@unaffiliated/forrestv) joined #pypy. | 22:40 | |
| JaRoel|4d (~jaroel|4d@sink.jaroel.nl) left irc: Remote host closed the connection | 22:50 | |
| claude_ (~claude@178235011238.warszawa.vectranet.pl) left irc: Quit: Leaving | 23:12 | |
| rguillebert (~rguillebe@2a01:e34:eea7:c690:21f:c6ff:fe12:4dee) left irc: Ping timeout: 268 seconds | 23:16 | |
| rguillebert (~rguillebe@fec76-1-78-234-124-105.fbx.proxad.net) joined #pypy. | 23:31 | |
| tos9 (~tos9@unaffiliated/tos9) left irc: Quit: Changing server | 23:38 | |
| tos9 (~tos9@unaffiliated/tos9) joined #pypy. | 23:39 | |
| peter_zfs (~a@ppp-94-65-237-41.home.otenet.gr) left irc: | 23:44 | |
| --- Thu Dec 29 2011 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!