| bbot2 (~bbot2@wyvern.cs.uni-duesseldorf.de) left irc: Ping timeout: 244 seconds | 00:09 | |
| bbot2 (~bbot2@wyvern.cs.uni-duesseldorf.de) joined #pypy. | 00:10 | |
| K-os (~gabriel@cl-3210.ham-01.de.sixxs.net) left irc: Quit: Ex-Chat | 00:15 | |
| lac_ (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) joined #pypy. | 00:39 | |
| lac (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) left irc: Ping timeout: 248 seconds | 00:40 | |
| CIA-13 | 03r.lobb 07roundup * 10#949/1.7 Sandbox problems: | 00:53 |
|---|---|---|
| CIA-13 | [new] I've having problems getting the sandbox build of pypy-1.7 working. Has anyone | 00:53 |
| CIA-13 | used it successfully? I've done several builds using both p ... * 14https://bugs.pypy.org/issue949 | 00:53 |
| Dulak (~michael@unaffiliated/dulak) left irc: Quit: Leaving | 00:54 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-win-x86-64/builds/78 [12ctismer, win64-stage1] | 00:54 |
| Dulak (~michael@unaffiliated/dulak) joined #pypy. | 01:05 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/156 | 01:06 |
| bbot2 | 13Exception: 15http://buildbot.pypy.org/builders/own-win-x86-64/builds/78 [12ctismer, win64-stage1] | 01:10 |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection | 01:15 | |
| kenaan | 12ctismer win64-stage1 11f26e12e71561 15/pypy/rlib/rarithmetic.py: corrected ovfcheck for the final time! ;-) | 01:18 |
| kenaan | 12ctismer win64-stage1 11c0364c9ad58f 15/pypy/: correction to struct.pack and test_gc | 01:18 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-win-x86-64/builds/79 [12ctismer, win64-stage1] | 01:20 |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) joined #pypy. | 01:24 | |
| voidspace (~voidspace@python/psf/voidspace) left irc: Quit: Uhm... gotta go | 01:32 | |
| Nick change: Gulaway -> Gulopine | 01:35 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1874 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/607 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/629 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1135 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/814 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1456 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/734 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/280 | 01:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/290 | 01:38 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/731 | 01:39 |
| rekamso (~textual@216.190.27.14) left irc: Ping timeout: 244 seconds | 01:43 | |
| davisagli (~davisagli@davisagli.com) joined #pypy. | 01:47 | |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Ping timeout: 240 seconds | 01:47 | |
| Nick change: buug -> bugfree | 01:51 | |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) joined #pypy. | 01:56 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/280 | 01:56 |
| thinker (~thinker@123-194-52-90.dynamic.kbronet.com.tw) left irc: Ping timeout: 244 seconds | 02:04 | |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection | 02:05 | |
| craigkerstiens (~craigkers@204.14.152.118) left irc: Quit: craigkerstiens | 02:06 | |
| cheyuni (~user@125.129.20.246) joined #pypy. | 02:13 | |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) joined #pypy. | 02:14 | |
| panni_ (smuxi@descent.fragstore.net) left irc: Read error: Operation timed out | 02:14 | |
| smuxi (smuxi@descent.fragstore.net) joined #pypy. | 02:17 | |
| Nick change: smuxi -> panni_ | 02:17 | |
| thinker (~thinker@123-194-52-90.dynamic.kbronet.com.tw) joined #pypy. | 02:32 | |
| Shanita (~John@osbk-4db06d9c.pool.mediaWays.net) joined #pypy. | 02:43 | |
| Kaskuka (~John@osbk-4db169db.pool.mediaWays.net) left irc: Ping timeout: 248 seconds | 02:43 | |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/961 | 02:45 |
| Nick change: Gulopine -> Gulaway | 02:46 | |
| ericflo (~ericflo@75.103.8.110) left irc: Quit: ericflo | 02:47 | |
| Taos (~jakob@host-92-20-152-142.as13285.net) joined #pypy. | 02:51 | |
| Taos | is ipython workingon pypy yet? | 02:51 |
| Da_Blitz | on 1.7? | 02:52 |
| Taos | mh? | 02:52 |
| Taos | yes | 02:53 |
| Action: Da_Blitz had to make a change to ipython to get it to work on pypy 1.7 | 02:53 | |
| Da_Blitz | let me find the fix | 02:54 |
| Taos | Thanks | 02:56 |
| rekamso (~textual@c-71-197-238-154.hsd1.wa.comcast.net) joined #pypy. | 02:58 | |
| Taos | Da_Blitz: TraitError: The 'session_number' trait of a HistoryManager instance must be an i | 02:58 |
| Taos | nteger, but a value of 29L <type 'long'> was specified. | 02:58 |
| Taos | Is the current exception i have | 02:58 |
| Da_Blitz | yeah, there is an Int you have to change to CInt | 02:59 |
| Taos | Da_Blitz: hm? | 03:00 |
| Taos | CInt? | 03:00 |
| Da_Blitz | line 66 in core/history.py | 03:01 |
| Da_Blitz | should be in site-packages/IPython/core/history.py | 03:01 |
| Taos | What is CInt? | 03:02 |
| Da_Blitz | session_number = Int() should become session_number = CInt() | 03:02 |
| Taos | and add that to import right? | 03:02 |
| Da_Blitz | ah yes, forgot that one ;) | 03:02 |
| Taos | Da_Blitz: new wonderful error :P | 03:03 |
| Taos | OSError: python27: The specified module could not be found. | 03:03 |
| tilgovi (~randall@couchdb/developer/tilgovi) left irc: Remote host closed the connection | 03:03 | |
| Da_Blitz | thats odd, doing that is enough for it to work on my system | 03:03 |
| Taos | Da_Blitz: should I get it to locad libpypy-c.dll? | 03:04 |
| Taos | (windows box ftw) | 03:04 |
| Da_Blitz | worth a try | 03:04 |
| Action: Da_Blitz hasent touched windows in awhile but that sounds semi correct | 03:04 | |
| Taos | Damn | 03:06 |
| Taos | AttributeError: No symbol PyMem_Malloc found in library C:/pypy/libpypy-c.dll | 03:06 |
| Taos | whats libexpat? | 03:07 |
| Da_Blitz | xml processing lib | 03:07 |
| Taos | damn | 03:09 |
| Taos | no | 03:09 |
| Taos | Well looks like no ipypy | 03:09 |
| Sho_ (~EHS1@kde/hein) joined #pypy. | 03:10 | |
| Taos | Da_Blitz: hmm | 03:15 |
| Taos | Da_Blitz: Ill take another stab later | 03:15 |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection | 03:16 | |
| craigkerstiens (~craigkers@c-76-126-162-38.hsd1.ca.comcast.net) joined #pypy. | 03:19 | |
| cheyuni (~user@125.129.20.246) left irc: Ping timeout: 248 seconds | 03:23 | |
| Ademan (~dan@adsl-71-141-224-79.dsl.snfc21.pacbell.net) joined #pypy. | 03:38 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1874 | 03:45 |
| nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 260 seconds | 04:08 | |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) joined #pypy. | 04:09 | |
| durin42 (~durin@adium/durin42) left irc: Ping timeout: 258 seconds | 04:12 | |
| durin42 (~durin@adium/durin42) joined #pypy. | 04:14 | |
| Nick change: Gulaway -> Gulopine | 04:18 | |
| durin42 (~durin@adium/durin42) left irc: Ping timeout: 252 seconds | 04:43 | |
| durin42 (~durin@adium/durin42) joined #pypy. | 04:45 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1456 | 04:46 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/629 | 04:53 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/814 | 04:54 |
| tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) joined #pypy. | 05:09 | |
| tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) left irc: Changing host | 05:09 | |
| tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy. | 05:09 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1135 | 05:12 |
| aleksi (~aleksi@85.235.191.82) joined #pypy. | 05:17 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/607 | 05:18 |
| fijal (~fijal@197.173.127.165) joined #pypy. | 05:57 | |
| fijal | hi | 05:57 |
| bbot2 | 13Exception: 15http://buildbot.pypy.org/builders/own-win-x86-64/builds/79 [12ctismer, win64-stage1] | 06:06 |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy. | 06:14 | |
| Ademan (~dan@adsl-71-141-224-79.dsl.snfc21.pacbell.net) left irc: Quit: leaving | 06:16 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/734 | 06:17 |
| Nick change: Gulopine -> Gulaway | 06:19 | |
| setmeaway | yeah i built pypy-c in mingw32 using trunk | 06:38 |
| mattip (~chatzilla@bzq-79-177-116-180.red.bezeqint.net) joined #pypy. | 06:38 | |
| mattip | hi | 06:38 |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 244 seconds | 06:39 | |
| CIA-13 | 03hpaulj 07roundup * 10#950/startup_hook in readline: | 06:41 |
| CIA-13 | [new] Ipython running under pypy (1.6) does not autoindent or 'rep'. The problem is | 06:41 |
| CIA-13 | with pypy readline.raw_input. It calls startup_hook before ... * 14https://bugs.pypy.org/issue950 | 06:41 |
| fijal | mattip: hi | 06:42 |
| fijal | mattip: I've found issues with the reshape merge | 06:42 |
| fijal | look at my coming commit | 06:43 |
| kenaan | 12fijal matrixmath-reshape-merge 11501d675cc368 15/pypy/module/micronumpy/: Create a branch to merge matrixmath-reshape. A bit of a mess because of the way the branch was cre... | 06:43 |
| fijal | setmeaway: nice! | 06:43 |
| hruske (~Gasper@188-230-156-183.dynamic.t-2.net) joined #pypy. | 06:47 | |
| asmeurer__ (~asmeurer@dhcp-baca-230.resnet.nmt.edu) left irc: Quit: asmeurer__ | 06:48 | |
| Action: mattip pulling | 06:51 | |
| fijal | there is a failing test | 06:52 |
| mattip | I can reproduce the failure. I will take a look. | 06:54 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/290 | 06:58 |
| apollo13 (~apollo13@unaffiliated/apollo13) left irc: Quit: SIGILL | 07:02 | |
| kenaan | 12fijal numpy-dtype-refactor 118bfca97d2a5c 15/pypy/module/micronumpy/REVIEW: add REVIEW notes | 07:07 |
| tilgovi (~randall@couchdb/developer/tilgovi) left irc: Ping timeout: 255 seconds | 07:09 | |
| fijal | mattip: dot seems like a mess, but there are more things that operate per-axis | 07:09 |
| mvt (~mvantelli@87.213.45.85) joined #pypy. | 07:14 | |
| stakkars_ (~tismer@p5DDB7E68.dip.t-dialin.net) left irc: Quit: schnarch | 07:29 | |
| mattip | fijal: ping | 07:50 |
| fijal | mattip: pong | 07:51 |
| mattip | Reshape is also a mess. | 07:51 |
| mattip | It seems that when things get too complicated, numpy makes a copy of the original array | 07:52 |
| mattip | when doing a reshape | 07:52 |
| fijal | ok | 07:53 |
| kenaan | 12fijal default 119f0ae6c94a99 15/pypy/doc/index.rst: Bump the release number | 07:53 |
| fijal | you can test it | 07:53 |
| mattip | Yes. But if we wish for numpy comatablity, sometimes reshape on a slice of an array will return a view, and sometimes a new array. | 07:54 |
| fijal | isn't that fine? | 07:55 |
| fijal | you can even check if it's incompatible or compatible | 07:55 |
| mattip | OK, just will take me more time to implement. | 07:55 |
| mattip | And lots more tests. | 07:56 |
| mattip | Back to dot implementation. | 07:57 |
| mattip | It seems we need some kind of framework for per-axis ufuncs. | 07:57 |
| fijal | yes | 07:57 |
| mattip | So, add a whole new framework of generated functions? or modify existing ufunc generator to accept an argument? | 07:58 |
| mattip | The second seems more numpy-like | 07:59 |
| mattip | but would require modification of start_iter | 07:59 |
| mattip | ? | 08:00 |
| Nick change: Guest48309 -> masquerade | 08:03 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 08:03 | |
| mattip | Ahh, progress on (re)shape. | 08:04 |
| mattip | Obviously, when you stare at it enough. a.shape = () will never modify the data buffer. | 08:05 |
| mattip | but b=a.reshape() will return a copy of a contiguous buffer if the strides do not match. | 08:06 |
| mattip | So if b = a.reshape() may or may not add to a.invalidates | 08:08 |
| mattip | s/if // | 08:08 |
| fijal | this all has to be tested btw | 08:12 |
| fijal | isn't reshape forcing the array anyway? | 08:12 |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy | 08:12 | |
| fijal | if you force you clean up invalidates no? | 08:12 |
| kenaan | 12fijal default 114b2e7974fd92 15/pypy/jit/backend/: Be slightly more vigilant when it comes to allocating registers for variables at the beginning of the loop. Not *mu... | 08:13 |
| fijal | good | 08:13 |
| Action: mattip still learning about forcing and invalidates | 08:14 | |
| arigato (~arigo@82.113.121.54) joined #pypy. | 08:15 | |
| fijal | arigato: hi | 08:15 |
| arigato | hi | 08:15 |
| Action: fijal is fighting the register allocator | 08:15 | |
| rekamso (~textual@c-71-197-238-154.hsd1.wa.comcast.net) left irc: Quit: Computer has gone to sleep. | 08:16 | |
| fijal | pom pom | 08:28 |
| fijal | do we want a gzip2 benchmark? | 08:28 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Read error: Connection reset by peer | 08:37 | |
| bivab (~david@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 08:37 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 08:40 | |
| lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) joined #pypy. | 08:51 | |
| antocuni (~antocuni@host200-125-dynamic.25-79-r.retail.telecomitalia.it) joined #pypy. | 08:57 | |
| apollo13 (~apollo13@unaffiliated/apollo13) joined #pypy. | 08:57 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 08:59 | |
| mattip | It would be nice to have access to something like unladen swallow's perf.py http://code.google.com/p/unladen-swallow/wiki/Benchmarks but adapted for arbitrary benchmark code. | 09:03 |
| fijal | we have pypy-benchmarks | 09:03 |
| fijal | good, I just shaved 10% by improving regalloc | 09:04 |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection | 09:04 | |
| arigato | uh, good | 09:04 |
| fijal | for this particular loop | 09:04 |
| fijal | (tight loops) | 09:04 |
| arigato | fwiw, I noticed that a typical tight loop is around 450 bytes of assembler, the last 150 of them being the JUMP reordering | 09:05 |
| fijal | nice :/ | 09:05 |
| fijal | in the bay in front of me there is party ship with soundsystem and DJ and everything that looks like a pirate boat | 09:06 |
| arigato | :-) | 09:06 |
| fijal | ok | 09:07 |
| fijal | I'll make a surf break and try to attack sharing of iterators | 09:07 |
| fijal | that would reduce the guard number | 09:07 |
| fijal | and int_adds | 09:08 |
| mattip | fijal: never noticed https://bitbucket.org/pypy/benchmarks, thanks. | 09:08 |
| fijal | arigato: FYI, the list-strategies caused regressions on things like | 09:08 |
| fijal | a, b, c = l | 09:08 |
| fijal | where l is a list | 09:08 |
| fijal | unfortunately not caught by test_pypy_c :/ | 09:09 |
| PiotrSikora (~none@nginx/adept/piotrsikora) left irc: Excess Flood | 09:09 | |
| PiotrSikora (~none@nginx/adept/piotrsikora) joined #pypy. | 09:11 | |
| G2P (~G2P@fw-asn1.ornis.com) joined #pypy. | 09:24 | |
| Fluxid | how do i run tests after translation to test if it built correclty and works? | 09:28 |
| fijal | py.test -A | 09:28 |
| fijal | generally if you checked out a version that passed tests it works :) | 09:28 |
| Fluxid | i want to make sure | 09:28 |
| fijal | go to pypy/objspace/std | 09:28 |
| fijal | and do ../../../pytest.py -A -x | 09:28 |
| kenaan | 12arigo pypy.org[extradoc] 11ab85b0339954 15/: kill reference to --stackless (thanks Fluxid) | 09:29 |
| Fluxid | 1 failed, 12 passed, 136 skipped in 3.40 seconds | 09:29 |
| Fluxid | that's all? | 09:29 |
| ojii (~ojii@40-34.60-188.cust.bluewin.ch) joined #pypy. | 09:30 | |
| fijal | that means that one failed :) | 09:32 |
| fijal | not very good | 09:32 |
| fijal | can you paste a traceback? | 09:32 |
| fijal | also if you run without -x it won't stop at the first failure | 09:33 |
| Fluxid | http://paste.pocoo.org/show/515697/ | 09:33 |
| fijal | ffffff | 09:37 |
| fijal | ok, that's silly | 09:37 |
| fijal | but well, a bug | 09:37 |
| fijal | if you run with -x how many failures do you get? | 09:37 |
| fijal | feel free to file a bug report with that | 09:37 |
| Fluxid | a bit more, but paste.pocoo is slooow | 09:39 |
| Fluxid | http://pastebin.com/jwSWkw2b | 09:41 |
| Fluxid | at the end is hg summary | 09:42 |
| Fluxid | maybe i should use branch head and not release tag? | 09:42 |
| fijal | uh | 09:43 |
| fijal | Fluxid: are you running it using the compiled pypy by chance? | 09:43 |
| fijal | it should have been | 09:44 |
| fijal | pypy ../../../pytest.py -A -x | 09:44 |
| arigato (~arigo@82.113.121.54) left irc: Quit: See you | 09:44 | |
| Fluxid | aha! | 09:45 |
| Fluxid | sorry, i assumed it would "just use" the compiled pypy | 09:45 |
| Fluxid | no failures. thank you! | 09:48 |
| apanda (~apanda@c-24-7-101-62.hsd1.ca.comcast.net) joined #pypy. | 09:48 | |
| fijal | good :) | 09:49 |
| fijal | those are still bugs though | 09:50 |
| fijal | some of those | 09:50 |
| fijal | bytearray ones for sure | 09:50 |
| voidspace (~voidspace@python/psf/voidspace) joined #pypy. | 09:50 | |
| fijal | new_array(Const(0)) | 09:58 |
| verte (~verte@python/site-packages/verte) joined #pypy. | 10:02 | |
| JaRoel|4d (~jaroel|4d@guest.bit.nl) joined #pypy. | 10:02 | |
| lizardo (~lizardo@189.2.128.130) joined #pypy. | 10:05 | |
| amaury_ (amaury_@nat/google/x-gadbabndulxyzvtp) joined #pypy. | 10:07 | |
| witulski (~stupsi@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 10:11 | |
| witulski (stupsi@fwstups.cs.uni-duesseldorf.de) left #pypy. | 10:11 | |
| JaRoel|4d (~jaroel|4d@guest.bit.nl) left irc: Remote host closed the connection | 10:36 | |
| JaRoel|4d (~jaroel|4d@2001:7b8:610:0:25fc:9078:8b44:1e1) joined #pypy. | 10:37 | |
| JaRoel|4d (~jaroel|4d@2001:7b8:610:0:25fc:9078:8b44:1e1) left irc: Remote host closed the connection | 10:38 | |
| kenaan | 12hager ppc-jit-backend 11d3e4ca39849d 15/pypy/jit/backend/ppc/ppcgen/test/test_stackframe.py: added force index to description of stackframe | 10:44 |
| cfbolz (~cfbolz@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 10:48 | |
| kenaan | 12cfbolz default 1109d322fa3784 15/pypy/: also add an explicit way to record a class | 10:49 |
| kenaan | 12cfbolz default 118b3befe1bd20 15/pypy/translator/c/src/support.h: jit_record_known_class does nothing in the C backend | 10:49 |
| kenaan | 12cfbolz default 11ba3ca8aaaef4 15/pypy/rlib/jit.py: (arigo, cfbolz): some sanity checks | 10:49 |
| kenaan | 12cfbolz default 110c65719691d8 15/pypy/: merge | 10:49 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1136 [12Carl Friedrich Bolz] | 10:56 |
| Squarepy (~Squarepy@wcw-staff-214-150.wireless.uva.nl) joined #pypy. | 11:05 | |
| Squarepy (~Squarepy@wcw-staff-214-150.wireless.uva.nl) left irc: Changing host | 11:05 | |
| Squarepy (~Squarepy@unaffiliated/squarepy) joined #pypy. | 11:05 | |
| craigkerstiens (~craigkers@c-76-126-162-38.hsd1.ca.comcast.net) left irc: Read error: Connection reset by peer | 11:15 | |
| craigkerstiens (~craigkers@c-76-126-162-38.hsd1.ca.comcast.net) joined #pypy. | 11:15 | |
| setmeaway | http://pastebin.com/iPbruYzG got some warning. and how can i package to redistribute pypy? any docs? | 11:33 |
| setmeaway | how can i exit from pypy shell? exit() not works | 11:36 |
| ronny | setmeaway: ctlr+d? | 11:37 |
| setmeaway | SyntaxError: Unknown character | 11:37 |
| setmeaway | ctrl+z worked | 11:37 |
| JaRoel (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 11:37 | |
| ronny | kill %1 then | 11:37 |
| setmeaway | start . | 11:37 |
| ronny | setmeaway: did you take the binary directory tree apart and put its stuff around to other paths? | 11:38 |
| ronny | setmeaway: what exactly do you intend to do? | 11:38 |
| setmeaway | ronny are you saying about redistribute pypy? | 11:38 |
| ronny | setmeaway: kind of | 11:39 |
| ronny | setmeaway: what exactly did you do to the directory tree that was in the binary archive? | 11:40 |
| setmeaway | i just ran pypy-c. and then tried quiting pypy. is this what you asked for? | 11:41 |
| nedbat (~nedbat@python/psf/nedbat) joined #pypy. | 11:41 | |
| ronny | setmeaway: so you just unpacked and tried to run? | 11:42 |
| setmeaway | ronny. no, i didn't. i built pypy on mingw32 and pypy-c is in temporary directory %TMP%\usession-unknown-1\testing_1 | 11:43 |
| ronny | setmeaway: oh, then that the misstake, it should get copyed into the pypy tree | 11:44 |
| ronny | usually its located in the same dir as translate.py | 11:44 |
| ronny | else it cant find the stdlib and stuff like that' | 11:44 |
| setmeaway | ronny, yes. i'll try that. translate.py raised compilation error after it did generated c source. | 11:45 |
| ronny | why didnt you tell that to begin with -_- | 11:45 |
| setmeaway | ronny sorry for that. | 11:46 |
| ronny | at least my pizza is delicious | 11:48 |
| ronny | setmeaway: wrt making a package, there is something about that in pypy/tool/release/package.py | 11:49 |
| Squarepy (~Squarepy@unaffiliated/squarepy) left irc: Read error: Connection reset by peer | 11:51 | |
| Squarepy (~Squarepy@wcw-staff-214-150.wireless.uva.nl) joined #pypy. | 11:51 | |
| Squarepy (~Squarepy@wcw-staff-214-150.wireless.uva.nl) left irc: Changing host | 11:52 | |
| Squarepy (~Squarepy@unaffiliated/squarepy) joined #pypy. | 11:52 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1136 [12Carl Friedrich Bolz] | 11:55 |
| kenaan | 12cfbolz default 11c8c6a9ce7e04 15/pypy/jit/: disable the use of cast_pointer in the JIT for now, to find out whether it caused the test failures. | 12:04 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1137 [12Carl Friedrich Bolz] | 12:04 |
| setmeaway | ronny thanks, i got working pypy | 12:06 |
| Nick change: shawn_ -> shawn | 12:08 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 12:19 | |
| JaRoel (~jaroel|4d@office.fourdigits.nl) left irc: Ping timeout: 260 seconds | 12:22 | |
| squiddy (~squiddy@g224197246.adsl.alicedsl.de) joined #pypy. | 12:25 | |
| mattip (chatzilla@bzq-79-177-116-180.red.bezeqint.net) left #pypy. | 12:37 | |
| _habnabit (~habnabit@python/site-packages/habnabit) left irc: Quit: ZNC - http://znc.sourceforge.net | 12:38 | |
| _habnabit (~habnabit@2001:470:1f0e:bdd::2) joined #pypy. | 12:40 | |
| kenaan | 12hager ppc-jit-backend 11d1741320b829 15/pypy/jit/backend/ppc/ppcgen/: started refactoring of stackframes | 12:41 |
| fijal | hello | 12:50 |
| ramusara (~ramusara@220.156.210.236.user.e-catv.ne.jp) joined #pypy. | 12:56 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 13:00 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 13:01 | |
| cfbolz | fijal: hi | 13:02 |
| fijal | hi | 13:02 |
| fijal | cfbolz: have you seen my concerns with list-strategies? | 13:03 |
| fijal | well "concerns" | 13:03 |
| cfbolz | no | 13:03 |
| fijal | more "things that were overlooked because lacked tests" | 13:03 |
| fijal | so a, b, c = l | 13:03 |
| fijal | is no longer unrolled when l is a list | 13:03 |
| fijal | according to alex it used to be | 13:03 |
| cfbolz | fijal: there seems to be a general problem with unrolling right now | 13:03 |
| lucian | i can't figure out if http://www.nuitka.net/blog/2011/12/nuitka-release-0-3-15/ supports more of python than cython | 13:03 |
| cfbolz | I don't know why yet | 13:03 |
| fijal | lucian: less I would say | 13:04 |
| fijal | lucian: or a different subset | 13:04 |
| fijal | cfbolz: well yes, in general if you change code around hints are not nicely preserved | 13:04 |
| fijal | nto sure if we have any other option than "write more test_pypy_c tests" | 13:04 |
| cfbolz | fijal: no, it broke last night | 13:04 |
| fijal | did it? | 13:04 |
| lucian | fijal: it if did so without extending (like Cython does), it might be useful as an interim solution for making CPython programs faster | 13:04 |
| cfbolz | not when we introduced list-strategies | 13:04 |
| fijal | last night = when float strategy was introduced? | 13:04 |
| fijal | lucian: I don't think it's usable for anything | 13:05 |
| cfbolz | fijal: http://buildbot.pypy.org/summary/longrepr?testname=TestDicts.%28%29.test_list&builder=pypy-c-jit-linux-x86-32&build=1136&mod=pypy.module.pypyjit.test_pypy_c.test_containers | 13:05 |
| cfbolz | and this: http://buildbot.pypy.org/summary/longrepr?testname=TestCall.%28%29.test_stararg_virtual&builder=pypy-c-jit-linux-x86-32&build=1136&mod=pypy.module.pypyjit.test_pypy_c.test_call | 13:05 |
| cfbolz | (which does the equivalent of the unpacking you wrote above) | 13:05 |
| fijal | oh, cool, so we have tests :) | 13:05 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1137 [12Carl Friedrich Bolz] | 13:05 |
| fijal | good | 13:05 |
| cfbolz | yes, but it's completely unclear why it broke | 13:05 |
| cfbolz | now it's clear :-) | 13:06 |
| verte | lucian: if gcc and guile are the poster children for GPL runtimes, nuitka is an attempt at insanity | 13:06 |
| fijal | ? | 13:06 |
| cfbolz | it has to do with the cast_pointer thing | 13:06 |
| cfbolz | because the run I just did worked | 13:06 |
| fijal | pfff | 13:06 |
| lucian | verte: heh. oh yeah, it's GPL. stupid | 13:08 |
| cfbolz | fijal: anyway, if you find things that are slow, please write tests, I will fix them | 13:08 |
| verte | lucian: that in itself wouldn't be so bad; but the author is considering the output of the compiler a derivitive work | 13:11 |
| lucian | verte: huh? how's that even possible? | 13:11 |
| verte | the result includes code written by the compiler author. it's a long shot, but that's the problem, not the GPL itself | 13:12 |
| verte | imao | 13:12 |
| Rhy0lite (dje@nat/ibm/x-zzexcoiegummbmof) joined #pypy. | 13:13 | |
| fijal | cfbolz: cool | 13:13 |
| fijal | cfbolz: but this seems we have tests so far :) | 13:13 |
| Rhy0lite | hi | 13:14 |
| cfbolz | fijal: when did alex discover this | 13:14 |
| cfbolz | Rhy0lite: hey | 13:14 |
| fijal | cfbolz: I discovered this | 13:14 |
| fijal | yesterday I believe? | 13:14 |
| cfbolz | fijal: it only broke in tonights run | 13:14 |
| fijal | Rhy0lite: hi | 13:14 |
| fijal | so maybe today | 13:14 |
| Circlefusion (~cf@74-138-199-109.dhcp.insightbb.com) left irc: Ping timeout: 244 seconds | 13:15 | |
| fijal | cfbolz: I have an OO quest | 13:16 |
| fijal | I have a DAG represented by objects | 13:16 |
| fijal | it starts at root, but is not a tree | 13:16 |
| verte | oh great, a questions about DAGs | 13:16 |
| fijal | to be precise it's like a tree, but some of leafs can be merged | 13:16 |
| fijal | to simplify things, let's say there is a class Node that has children | 13:17 |
| fijal | and class Leaf | 13:17 |
| fijal | how do I write a method that makes a copy of it, preserving structure | 13:17 |
| cfbolz | hahaha. that's hard :-) | 13:17 |
| cfbolz | you need a dict | 13:17 |
| cfbolz | exactly the prolog-problem | 13:17 |
| fijal | if I have a dict I can't make it virtual :) | 13:18 |
| verte | dicts (of sets) are great for this sort of thing, because you can have several graphs with the same nodes in them | 13:18 |
| verte | ah :) | 13:18 |
| fijal | the thing is the structure is known in advance | 13:19 |
| fijal | so for each iteration it'll be the same | 13:19 |
| fijal | I can (somehow) encode the structure in the greenkey | 13:19 |
| cfbolz | fijal: then you can statically number the elements | 13:19 |
| cfbolz | and use a list of bools | 13:19 |
| cfbolz | that stays virtual | 13:19 |
| fijal | why a list of bools? | 13:20 |
| fijal | ah I see | 13:20 |
| stakkars_ (~tismer@i59F76401.versanet.de) joined #pypy. | 13:20 | |
| cfbolz | that's kind of equivalent what pyrolog does | 13:20 |
| fijal | that requires a bit of refactoring, but probably a good thing to do | 13:20 |
| fijal | this are expression trees in numpypy | 13:21 |
| cfbolz | it's a bit annoying, I guess | 13:21 |
| cfbolz | right | 13:21 |
| cfbolz | I guess not list of bools then, but list of values? | 13:21 |
| fijal | yes, a bit | 13:21 |
| cfbolz | or something | 13:21 |
| fijal | yes yes | 13:21 |
| fijal | I tried cheating and storing "next version" on nodes | 13:22 |
| fijal | but the optimizer was not impressed | 13:22 |
| G2P (~G2P@fw-asn1.ornis.com) left irc: Remote host closed the connection | 13:23 | |
| G2P (~G2P@fw-asn1.ornis.com) joined #pypy. | 13:24 | |
| Circlefusion (~cf@74-138-199-109.dhcp.insightbb.com) joined #pypy. | 13:27 | |
| aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy. | 13:28 | |
| arigato (~arigo@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 13:32 | |
| verte (~verte@python/site-packages/verte) left irc: Quit: ~~~ Crash in JIT! | 13:35 | |
| arigato | Rhy0lite: I'm told you mentioned that in order to build gcc trunk, I need to use a known working revision? | 13:35 |
| arigato | if so, how do I find out? :-) | 13:35 |
| Rhy0lite | I think the recent problems have been fixed | 13:36 |
| Rhy0lite | I'm not sure about the context of that comment | 13:36 |
| arigato | I just got a svn checkout from 2 hours ago | 13:36 |
| arigato | and it fails to build, for me | 13:36 |
| Rhy0lite | fails in what way? | 13:36 |
| arigato | fatal error: gnu/stubs-32.h: No such file or directory | 13:36 |
| Rhy0lite | for x86? | 13:37 |
| arigato | yes, linux64 | 13:37 |
| arigato | regular ubuntu | 13:37 |
| Rhy0lite | that's not a gcc error | 13:38 |
| arigato | it seems to try to build x86_64-unknown-linux-gnu/32/libgcc, which I don't care about, because the 64bit version is enough for me | 13:38 |
| Rhy0lite | it means you don't have enough development libraries installed on ubuntu | 13:38 |
| arigato | ok | 13:38 |
| Rhy0lite | you don't have part of glibc installed | 13:39 |
| arigato | I don't have the 32-bit version of everything | 13:39 |
| Rhy0lite | there you go | 13:39 |
| rguillebert | arigato, you can probably use the Linux From Scratch book | 13:39 |
| arigato | I really don't want to have them, too | 13:39 |
| Rhy0lite | it might work if you configure with --disable-multilib | 13:39 |
| arigato | for 32-bit I'm using a different chroot with gentoo | 13:39 |
| arigato | thanks | 13:39 |
| fijal | arigato: :-) | 13:39 |
| rguillebert | at the begin there's a guide on how to build a compiler toolchain from scratch | 13:40 |
| arigato | rguillebert: I'm not from scratch at all | 13:40 |
| rguillebert | no but it's a preamble to linux from scratch | 13:41 |
| arigato | I'm just trying to figure out if it's enough to do "configure; make" in gcc's svn checkout, and it seems to be, assuming enough packages are installed :-) | 13:41 |
| rguillebert | oh ok | 13:41 |
| jnoller (~jnoller@vpn.nasuni.com) joined #pypy. | 13:43 | |
| fijal | jnoller: hi jesse | 13:46 |
| jnoller | hola | 13:46 |
| cfbolz | hey jesse | 13:47 |
| fijal | how's speed.python.org doing? | 13:47 |
| aleksi (~aleksi@85.235.191.82) left irc: Remote host closed the connection | 13:48 | |
| Rhy0lite | arigato: http://gcc.gnu.org/install/prerequisites.html | 13:50 |
| Rhy0lite | also, we suggest that you do not build in the source directory | 13:51 |
| jnoller | fijal: got side tracked by pycon stuff | 13:53 |
| fijal | cool that someone works on it btw | 13:54 |
| fijal | sorry that I had no time | 13:54 |
| kenaan | 12ctismer win64-stage1 11cf81cefc5208 15/pypy/jit/backend/llsupport/test/test_descr.py: struct.pack issues with the JIT | 13:57 |
| jnoller | fijal: I'm just trying to keep things moving. | 13:57 |
| fijal | good :) | 13:57 |
| arigato | fatal error: asm/errno.h: No such file or directory | 13:59 |
| Nick change: Gulaway -> Gulopine | 14:00 | |
| arigato | obscure, that one seems to be a bug in my ubuntu installation | 14:00 |
| fijal | arigato: gcc in ubuntu is not quite the vanilla gcc | 14:02 |
| fijal | cfbolz: do you think it's ok to say unroll_safe on hash__Tuple? | 14:03 |
| Action: arigato doesn't think so | 14:05 | |
| cfbolz | fijal: nope | 14:05 |
| fijal | because tuples can be large? | 14:05 |
| cfbolz | yes | 14:05 |
| cfbolz | I think it's much saner to finish specialized tuples | 14:06 |
| fijal | yes | 14:06 |
| fijal | probably | 14:06 |
| fijal | Segmentation Fault | 14:08 |
| fijal | nice | 14:08 |
| fijal | arigato: are you doing STM experiments? | 14:14 |
| arigato | right now? no | 14:14 |
| fijal | I mean, is this why you try to build gcc | 14:14 |
| arigato | ah, yes | 14:14 |
| fijal | arigato: do we have gdb scripts to reconstruct the jitted stack? | 14:17 |
| fijal | antocuni: ? | 14:17 |
| antocuni | I don't think so | 14:17 |
| antocuni | fijal: but you are welcome to improve pypy/tool/gdb_pypy.py :-) | 14:17 |
| antocuni | it's kind of easy to script gdb in python | 14:18 |
| fijal | ok, so I have a segfaulting program | 14:18 |
| fijal | only after jitting | 14:18 |
| fijal | arigato: do you remember how do we get the stack? | 14:19 |
| arigato | unsure what the question is | 14:21 |
| fijal | jitted frames have a return pointer in some register | 14:23 |
| fijal | or somewhere on stack | 14:23 |
| fijal | but gdb fails to find them | 14:23 |
| fijal | right? | 14:23 |
| whitelynx (~whitelynx@63.241.75.144) joined #pypy. | 14:28 | |
| arigato | ah, it depends on the precise platform | 14:29 |
| Action: fijal 's computer is a very precise platform | 14:30 | |
| kenaan | 12arigo SpecialisedTuples 11acd6e35b5711 15/pypy/objspace/std/: - rename ANY to Any to avoid confusion - use a more pragmatic approach of having 10 specialized versio... | 14:31 |
| fijal | eh | 14:32 |
| fijal | so I have a failing guard_class | 14:32 |
| fijal | "failing" | 14:32 |
| fijal | segfaulting guard class, because stuff is None | 14:32 |
| fijal | or NULL | 14:32 |
| sven_hager (~sven@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 14:33 | |
| kenaan | 12hager ppc-jit-backend 116ad92645f0e4 15/pypy/jit/backend/ppc/ppcgen/test/test_stackframe.py: slightly changed stackframe layout so that encoding area can be accessed with fixed offset from SPP | 14:34 |
| derdon (~derdon@pD9E1D2AF.dip.t-dialin.net) joined #pypy. | 14:34 | |
| fijal | ? | 14:35 |
| legogris (~legogris@c80-217-207-157.bredband.comhem.se) left irc: Ping timeout: 252 seconds | 14:35 | |
| fijal | arigato: can you explain a bit? | 14:37 |
| fijal | it's a 64bit linux | 14:37 |
| legogris (~legogris@c80-217-207-157.bredband.comhem.se) joined #pypy. | 14:37 | |
| apollo13 (~apollo13@unaffiliated/apollo13) left irc: Ping timeout: 240 seconds | 14:38 | |
| cfbolz | fijal: he's pairing with bivab | 14:39 |
| fijal | ok | 14:39 |
| fijal | uh, our assembler is a mess sometimes :/ | 14:41 |
| fijal | cfbolz: for example if we create a new frame | 14:41 |
| fijal | we execute write barrier per item | 14:42 |
| Nick change: Gulopine -> Gulaway | 14:42 | |
| fijal | if it's the bad case scenario, each of them pushes and pops all the registers | 14:42 |
| legogris (~legogris@c80-217-207-157.bredband.comhem.se) left irc: Ping timeout: 258 seconds | 14:42 | |
| fijal | our call_assembler is also a mess | 14:43 |
| apollo13 (~apollo13@unaffiliated/apollo13) joined #pypy. | 14:43 | |
| legogris (~legogris@c80-217-207-157.bredband.comhem.se) joined #pypy. | 14:43 | |
| mat^2 (~mathias@212.130.113.35) joined #pypy. | 14:46 | |
| kenaan | 12hager ppc-jit-backend 11de0569ccdb88 15/pypy/jit/backend/ppc/ppcgen/test/test_stackframe.py: remove parameter save area from stacklayout | 14:47 |
| fijal | cfbolz: http://paste.pocoo.org/show/515774/ | 14:47 |
| fijal | you can scroll visually to see how much of a relatively tight loop is occupied by various dances around jump | 14:48 |
| antocuni | gooood | 14:51 |
| antocuni | I finished to develop my boring django app | 14:51 |
| antocuni | the good news is that from now I can do pypy again | 14:51 |
| cfbolz | fijal: well, our register allocator sucks | 14:51 |
| antocuni | the bad one is that I'm not getting paid, but well :-) | 14:51 |
| fijal | cfbolz: yes, but it's a bit hard to do better with passing a billion arguments | 14:52 |
| fijal | also it made sense the day it was written :) | 14:52 |
| Rhy0lite | fijal: any more ideas about Laplace? | 14:58 |
| fijal | Rhy0lite: yeah, but I'm debugging a segfault | 14:59 |
| craigkerstiens (~craigkers@c-76-126-162-38.hsd1.ca.comcast.net) left irc: Quit: craigkerstiens | 15:00 | |
| Action: fijal wonders how much new_array(0) makes sense | 15:00 | |
| mat^2 (~mathias@212.130.113.35) left irc: | 15:02 | |
| jterrace (~jterrace@aegis.CS.Princeton.EDU) joined #pypy. | 15:12 | |
| fijal | uh uh uh | 15:12 |
| fijal | cfbolz: tons of ideas how to improve :) | 15:12 |
| fijal | good I suppose ;-) | 15:12 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 15:13 | |
| fijal | I think one of those is attacked by armin in op_malloc branch | 15:14 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 15:16 | |
| Nick change: ojii -> definitelynotoji | 15:22 | |
| Nick change: definitelynotoji -> ojii | 15:22 | |
| cfbolz (~cfbolz@fwstups.cs.uni-duesseldorf.de) left irc: Ping timeout: 248 seconds | 15:24 | |
| setmeaway | how to increase maxrecursiondepth? Warning: uncaught exception in callback: bootstrap <StackOverflow object at 0x63577fa0> | 15:33 |
| setmeaway | i modified pypy.exe by giving --stack=102400000, but doesn't work | 15:34 |
| asmeurer_ (~asmeurer@dhcp-baca-230.resnet.nmt.edu) joined #pypy. | 15:40 | |
| jterrace (jterrace@aegis.CS.Princeton.EDU) left #pypy. | 15:45 | |
| Squarepy_ (~Squarepy@wcw-staff-214-150.wireless.uva.nl) joined #pypy. | 15:49 | |
| Squarepy (~Squarepy@unaffiliated/squarepy) left irc: Ping timeout: 252 seconds | 15:50 | |
| fijal | pom pom pom | 15:50 |
| fijal | Alex_Gaynor: ping? | 15:50 |
| mvt (~mvantelli@87.213.45.85) left irc: Quit: Leaving | 15:54 | |
| antocuni (~antocuni@host200-125-dynamic.25-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 15:55 | |
| fijal | arigato: ping? | 15:57 |
| apanda (~apanda@c-24-7-101-62.hsd1.ca.comcast.net) left irc: Remote host closed the connection | 15:59 | |
| arigato | Rhy0lite: what is the way in C on PowerPC to invalidate the instruction cache, after we generated code? | 16:05 |
| Rhy0lite | there already is code in the branch for that | 16:06 |
| arigato | (that was missing at least on ARM, it was the cause of the random rare bugs) | 16:06 |
| arigato | ah, good | 16:06 |
| arigato | fijal: pong | 16:06 |
| Rhy0lite | see backend/ppc/_flush_icache.c | 16:06 |
| fijal | arigato: I have a very strange segfault, but the only thing that I have is a combination of long loops and micronumpy :/ | 16:07 |
| fijal | arigato: I suppose I'll try to reduce the problem before I bother you | 16:07 |
| arigato | Rhy0lite: thank you | 16:07 |
| Rhy0lite | arigato: I'm not sure if it is used correctly. I believe that David said it only was used at the beginning | 16:07 |
| arigato | fijal: :-( thank you | 16:07 |
| arigato | Rhy0lite: ah$ | 16:07 |
| arigato | /usr/include/unistd.h:275:21: error: two or more data types in declaration specifiers | 16:08 |
| Rhy0lite | I don't think the ppc jit is patching code yet. | 16:09 |
| arigato | Rhy0lite: ah, but yes, it can | 16:09 |
| arigato | at least in theory, an old piece of assembler can be freed, and new assembler then written at the same place | 16:09 |
| arigato | so getting the instruction used now, while we think about it, looks like a better plan than waiting months until we get a few random-looking errors from users :-) | 16:10 |
| fijal | arigato: I suppose something-less-than-800k-log-file | 16:10 |
| fijal | would be a good start | 16:10 |
| fijal | I get an object that in theory should be impossible | 16:10 |
| Squarepy_ (~Squarepy@wcw-staff-214-150.wireless.uva.nl) left irc: Quit: Leaving | 16:11 | |
| fijal | arigato: is there a way to disable tracing of a particular function? | 16:12 |
| arigato | no | 16:12 |
| fijal | :( | 16:13 |
| fijal | hm | 16:13 |
| fijal | arigato: if you get ABORT_TRACE_TOO_LONG (or any other abort) is there anything that makes sure lazy setfields get written? | 16:14 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 16:16 | |
| craigkerstiens (~craigkers@204.14.152.118) joined #pypy. | 16:17 | |
| fijal | arigato: can you check? I think not | 16:17 |
| arigato | sorry, back | 16:18 |
| arigato | setfields are not done lazily at all, during tracing | 16:18 |
| arigato | as far as I can tell | 16:18 |
| fijal | it's only getfields? | 16:19 |
| arigato | no | 16:19 |
| arigato | nothing | 16:19 |
| arigato | during tracing | 16:19 |
| Action: arigato -> bivab again | 16:19 | |
| fijal | ok | 16:20 |
| fijal | sorry | 16:20 |
| Action: fijal is fighting windmills | 16:20 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 16:30 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 16:31 | |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) joined #pypy. | 16:35 | |
| G2P (~G2P@fw-asn1.ornis.com) left irc: Quit: Leaving. | 16:38 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 16:42 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 16:44 | |
| _habnabit (~habnabit@2001:470:1f0e:bdd::2) left irc: Changing host | 16:44 | |
| _habnabit (~habnabit@python/site-packages/habnabit) joined #pypy. | 16:44 | |
| Alex_Gaynor | fijal: pong | 16:45 |
| kkris (~kris@80-123-37-202.adsl.highway.telekom.at) joined #pypy. | 16:52 | |
| lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) left irc: Ping timeout: 248 seconds | 16:53 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 16:54 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 16:57 | |
| amaury_ (amaury_@nat/google/x-gadbabndulxyzvtp) left irc: Ping timeout: 240 seconds | 17:00 | |
| fijal | Alex_Gaynor: I found a segfault | 17:02 |
| Alex_Gaynor | fijal: that's no good | 17:02 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 17:03 | |
| fijal | it's also not an easy one | 17:03 |
| fijal | I get a Scalar that has empty value | 17:03 |
| fijal | (NULL) | 17:03 |
| fijal | only with jit on | 17:03 |
| Alex_Gaynor | find out where it comes from | 17:04 |
| tilgovi (~randall@adsl-67-116-241-49.dsl.pltn13.pacbell.net) joined #pypy. | 17:04 | |
| tilgovi (~randall@adsl-67-116-241-49.dsl.pltn13.pacbell.net) left irc: Changing host | 17:04 | |
| tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy. | 17:04 | |
| fijal | from assembler | 17:04 |
| fijal | I can paste you some if you want | 17:04 |
| Alex_Gaynor | I suppose, sure | 17:07 |
| fijal | pom pom pom | 17:09 |
| fijal | can you just clone a bzr from somewhere? | 17:09 |
| Alex_Gaynor | ? | 17:10 |
| Alex_Gaynor | I think I have bzr installed, maybe, I have no idea how to use it | 17:10 |
| fijal | http://paste.pocoo.org/show/515822/ | 17:11 |
| Alex_Gaynor | no way we could reduce this a little? | 17:14 |
| fijal | not that i managed | 17:15 |
| fijal | I think it has something to do with trace limit exceeded | 17:16 |
| fijal | maybe | 17:16 |
| fijal | let me try some more | 17:17 |
| Alex_Gaynor | fijal: can I explain to you why setitem_w isn't an issue? | 17:19 |
| kenaan | 12alex_gaynor numpy-dtype-refactor 11adbd652dc0ca 15/pypy/module/micronumpy/: __module__ is now correctly, numpypy | 17:19 |
| kenaan | 12alex_gaynor numpy-dtype-refactor 11f1d05c0a4b16 15/pypy/module/micronumpy/: Fix tests, expose int{16,32} at app level, add a test for __int__ | 17:19 |
| fijal | there were more types than int I think | 17:19 |
| fijal | look into dir(numpy) | 17:19 |
| fijal | yes please | 17:19 |
| kenaan | 12alex_gaynor numpy-dtype-refactor 116de4a284490c 15/pypy/: merged default | 17:19 |
| Alex_Gaynor | fijal: because setitem_w now calls self.setitem, isntead of dtype.setitem_w, so self.setitem calls invalidates itself | 17:19 |
| fijal | ok | 17:20 |
| kenaan | 12alex_gaynor numpy-dtype-refactor 118dd8ceded68e 15/pypy/module/micronumpy/REVIEW: remove the review file, I addressed fijal's concern | 17:21 |
| Alex_Gaynor | fijal: I've been exposing types as someone tells me about them/I need them, I don't want to expose them without tests. ok to merge? | 17:22 |
| logan` (~loganchie@canlover.m6.ntu.edu.tw) joined #pypy. | 17:22 | |
| fijal | http://paste.pocoo.org/show/515827/ | 17:23 |
| fijal | ok | 17:23 |
| Alex_Gaynor | fijal: that's n ok to merge? | 17:24 |
| fijal | yes :) | 17:24 |
| Alex_Gaynor | fijal: cool, merging! | 17:25 |
| fijal | note that it does not segfault with --jit off | 17:25 |
| fijal | well, we have a segfault... | 17:25 |
| fijal | but it seems to be unrelated | 17:25 |
| kenaan | 12alex_gaynor numpy-dtype-refactor 11dfe5b3a2b6aa 15/: Close branch for merge. | 17:26 |
| kenaan | 12alex_gaynor default 118c67eefdff41 15/pypy/: Merge the numpy-dtype-refactor branch, it refactors the internals of numpy, and exposes the numpy box classes... | 17:26 |
| bivab (~david@fwstups.cs.uni-duesseldorf.de) left irc: Quit: bivab | 17:29 | |
| arigato | fijal: note: such a segfault could really come from a delayed setfield, but not one done by the tracer | 17:29 |
| fijal | arigato: yes | 17:30 |
| fijal | arigato: although I failed to find in traces how it can happen | 17:30 |
| arigato | I've been fighting the exact same issue previously | 17:30 |
| fijal | I removed most of cruft, but it's still a bit biggish | 17:30 |
| arigato | I think we really need to refactor our approach: right now we say "nothing forces any field, except (some list of exceptions)" | 17:31 |
| timotimo | i don't have a current pypy, but does isinstance(np.zeros(10), np.ndarray) work yet? isinstance(np.zeros(10), np.array) gives an error in original numpy, because array is a function and ndarray is the associated class, but works in pypy, which (at least in 1.6) doesn't have the ndarray name | 17:31 |
| logan` (~loganchie@canlover.m6.ntu.edu.tw) left irc: Quit: «â | 17:31 | |
| arigato | fijal: I'm wondering if we are really safe against some recursive kinds of calls to the tracer | 17:32 |
| Alex_Gaynor | timotimo: no it doesn't work yet, I'll go fix that I suppose, fijal didn't when I asked him to ;) | 17:32 |
| fijal | arigato: well, we did not optimize yet, no? | 17:32 |
| timotimo | ah, cool. this just bit me with my program | 17:33 |
| Action: fijal didn't when alex asked him to | 17:33 | |
| arigato | fijal: no, I mean recursive in the sense that this tracing you are looking at is itself called from a piece of generated assembler | 17:33 |
| arigato | if that piece of assembler delays the setfield over the call that ends up tracing | 17:33 |
| arigato | then it's wrong | 17:34 |
| antocuni (~antocuni@host200-125-dynamic.25-79-r.retail.telecomitalia.it) joined #pypy. | 17:34 | |
| fijal | arigato: I don't think it can happen but who knows | 17:34 |
| fijal | indeed I have a Scalar whos value is NULL | 17:36 |
| fijal | which it should never be | 17:36 |
| fijal | (it's not with --jit off) | 17:36 |
| fijal | arigato: the only reason why I suspect so is because there are long traces with setfield quite a bit delayed | 17:36 |
| arigato | can you go up in the C stack and see if the current tracing has a caller that is generated assembler? | 17:37 |
| fijal | no, because I don't know how to go up the stack | 17:37 |
| fijal | the bt shows crap | 17:37 |
| arigato | ? | 17:37 |
| fijal | I'm not in tracing, I'm in generated assembler | 17:37 |
| fijal | http://paste.pocoo.org/show/515833/ | 17:38 |
| fijal | it looks like this | 17:38 |
| sven_hager (~sven@fwstups.cs.uni-duesseldorf.de) left irc: Quit: Leaving | 17:38 | |
| arigato | ok, I guess I'll stop then :-) | 17:38 |
| arigato | I'm trying to guess what could be wrong based on what you tell me, but without knowing which are fact and which are guesses it's hard :-) | 17:39 |
| fijal | well | 17:39 |
| fijal | you can help me to get to the stack :) | 17:39 |
| fijal | you only told me it's super platform dependent so far | 17:39 |
| arigato | I suppose I'd prefer gdb'ing myself over debugging-over-irc | 17:40 |
| arigato | honestly | 17:40 |
| fijal | ok | 17:40 |
| fijal | there is a gdb in screen on tannit | 17:40 |
| fijal | if you feel like it | 17:40 |
| fijal | <tadaam> | 17:40 |
| fijal | I agree debugging over IRC sucks | 17:41 |
| fijal | better than debugging over phone | 17:41 |
| arigato | yes | 17:41 |
| tilgovi (~randall@couchdb/developer/tilgovi) left irc: Ping timeout: 240 seconds | 17:45 | |
| fijal (~fijal@197.173.127.165) left irc: Read error: Connection reset by peer | 17:45 | |
| fijal (~fijal@41.52.155.130) joined #pypy. | 17:50 | |
| berdario (~Tiibiidii@DLM253-160.unibg.it) joined #pypy. | 17:57 | |
| craigkerstiens (~craigkers@204.14.152.118) left irc: Ping timeout: 244 seconds | 17:59 | |
| lucian (~lucian@cpc1-newc15-2-0-cust84.gate.cable.virginmedia.com) joined #pypy. | 18:03 | |
| berdario (~Tiibiidii@DLM253-160.unibg.it) left irc: Ping timeout: 252 seconds | 18:05 | |
| ramusara (~ramusara@220.156.210.236.user.e-catv.ne.jp) left irc: Quit: Leaving... | 18:09 | |
| kenaan | 12alex_gaynor numpy-ndarray 1155386ade1ad3 15/pypy/module/micronumpy/: a branch to make array dn ndarray work the same way as they do in numpy, I can't figure how this change... | 18:10 |
| Alex_Gaynor | fijal: if you get a chance, feel like taking a look at how that commit I just pushed breaks stuff | 18:10 |
| tlynn (~tlynn@cpc6-cmbg14-2-0-cust121.5-4.cable.virginmedia.com) joined #pypy. | 18:11 | |
| fijal | no | 18:11 |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) joined #pypy. | 18:14 | |
| Alex_Gaynor | fijal: just as well, I got it | 18:16 |
| kdas_ (~kdas@114.143.160.170) joined #pypy. | 18:17 | |
| kenaan | 12alex_gaynor numpy-ndarray 112d3c7db968ec 15/pypy/module/micronumpy/: fix the bug, fix one other bug | 18:21 |
| kenaan | 12alex_gaynor numpy-ndarray 119cfed3afa1cc 15/: CLose branhc for merge | 18:21 |
| kenaan | 12alex_gaynor default 11ccf1f7b9b78b 15/pypy/module/micronumpy/: make numpy.ndarray refer to the class, and numpy.array be a funciton which constructs it | 18:21 |
| Alex_Gaynor | timotimo: numpy.ndarray and numpy.array are now the right htings | 18:21 |
| timotimo | great, thanks! | 18:22 |
| Alex_Gaynor | timotimo: I'll go punch buildbots so real humans can use it | 18:24 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1138 [12alex] | 18:25 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/608 [12alex] | 18:25 |
| timotimo | i don't know what you mean by that :\ | 18:26 |
| Alex_Gaynor | timotimo: I mean so other people don't have to compile it themselves. | 18:29 |
| fijal | also known as mere mortals | 18:33 |
| stakkars_ (~tismer@i59F76401.versanet.de) left irc: Quit: schnarch | 18:33 | |
| fijal | or mere web developers | 18:33 |
| Alex_Gaynor | fijal: hey, I'm a web developer | 18:34 |
| fijal | that's how voidspace called himself | 18:34 |
| kdas_ (~kdas@114.143.160.170) left irc: Quit: This computer has gone to sleep | 18:34 | |
| voidspace | fijal: don't blame me! | 18:34 |
| MostAwesomeDude | Ooh, ooh, blame me! | 18:35 |
| Sho_ | Alex_Gaynor: whee, someone seems to have picked up django py3k | 18:36 |
| Alex_Gaynor | Sho_: yes indeed, now I need to make time to review it | 18:36 |
| Action: Sho_ clones Alex_Gaynor | 18:37 | |
| Sho_ | passing testsuite is a promising first start | 18:37 |
| fijal | I wonder who will maintain all the py3k clones | 18:38 |
| arigato (~arigo@fwstups.cs.uni-duesseldorf.de) left irc: Quit: See you | 18:39 | |
| Alex_Gaynor | fijal: what shoul I do next? complex numbers I guess | 18:39 |
| craigkerstiens (~craigkers@204.14.152.118) joined #pypy. | 18:39 | |
| timotimo | how big is the step from complex numbers to user-defined records? | 18:39 |
| fijal | Alex_Gaynor: frompyfunc? | 18:40 |
| fijal | sse? | 18:40 |
| Alex_Gaynor | timotimo: not big, in the branch I merged a big design goal is that doing both would be easy | 18:40 |
| jnoller (~jnoller@vpn.nasuni.com) left irc: Quit: Leaving... | 18:40 | |
| fijal | I attacked (and failed to reach the goal) sharing iterators | 18:40 |
| fijal | it's annoying | 18:41 |
| fijal | we probably have to refactor the signature first | 18:42 |
| timotimo | that's great, alex | 18:45 |
| Alex_Gaynor | fijal: ok | 18:47 |
| stakkars_ (~tismer@g225044039.adsl.alicedsl.de) joined #pypy. | 18:52 | |
| fijal | stakkars_: hi | 18:54 |
| stakkars_ | fijal: hi! | 18:54 |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection | 18:55 | |
| stakkars_ | struggling with libffi and ll2ctypes&. fun | 18:55 |
| fijal | ll2ctypes is "fun" | 19:04 |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) joined #pypy. | 19:06 | |
| Sho_ (~EHS1@kde/hein) left irc: Quit: Stop leaking memory like it's going out of fashion. | 19:06 | |
| tilgovi (~randall@199.36.158.16) joined #pypy. | 19:06 | |
| tilgovi (~randall@199.36.158.16) left irc: Changing host | 19:06 | |
| tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy. | 19:06 | |
| berdario (~Tiibiidii@host118-77-dynamic.10-79-r.retail.telecomitalia.it) joined #pypy. | 19:11 | |
| berdario (~Tiibiidii@host118-77-dynamic.10-79-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer | 19:12 | |
| ojii (~ojii@40-34.60-188.cust.bluewin.ch) left irc: Remote host closed the connection | 19:14 | |
| arigato (~arigo@89.204.130.72) joined #pypy. | 19:15 | |
| Alex_Gaynor | fijal: if people are going to be using sys.exc_info() for py3k stuff we should find a way to make it fast | 19:17 |
| fijal | well | 19:18 |
| fijal | maybe | 19:18 |
| fijal | but also, geez | 19:18 |
| fijal | I'm happy to say "it is your own problem" | 19:18 |
| Alex_Gaynor | yes of course, but still | 19:18 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Read error: Connection reset by peer | 19:18 | |
| Alex_Gaynor | I'm not, because random people's code will be slow, and if your framework uses it, you're scrweed | 19:18 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 19:19 | |
| Alex_Gaynor | Can't we just unroll the sys_exc_info loop? | 19:19 |
| Alex_Gaynor | and gettomframe_nohidden | 19:19 |
| fijal | it's complexity | 19:20 |
| fijal | not a trivial one | 19:20 |
| Alex_Gaynor | pff, traceback escapes the frame | 19:21 |
| fijal | yes, precisely | 19:21 |
| fijal | I mean seriously | 19:21 |
| Alex_Gaynor | I can think of a hack to avoid this | 19:21 |
| Alex_Gaynor | but it's a hack | 19:21 |
| fijal | you use API that's not relevant to what you're doing | 19:21 |
| Alex_Gaynor | Have sys.exc_info() return some special W_TupleObject subclass | 19:21 |
| fijal | because some people changed syntax | 19:21 |
| Alex_Gaynor | that lazilly creates tracebacks | 19:21 |
| Alex_Gaynor | it's a huge hack though | 19:22 |
| fijal | and it makes your code look crappy | 19:22 |
| fijal | that's what py.test does | 19:22 |
| fijal | so say a loop with py.path.local.check | 19:22 |
| fijal | would blow up the JIT | 19:23 |
| Alex_Gaynor | I agree it makes code look crappy, but what can we do, if people decide to write their code (e.g. django might) this way, we should fix it | 19:23 |
| fijal | we educate the people :) | 19:23 |
| Alex_Gaynor | we're not just a faster pypy for people with obscure scientific stuff, we're a faster pypy for everyone | 19:23 |
| fijal | no seriously | 19:23 |
| fijal | this sort of hacks will introduce subtle segfaults | 19:24 |
| fijal | and we have only one armin | 19:24 |
| fijal | virtualrefs and friends caused a lot of headache | 19:25 |
| Alex_Gaynor | mitsuhiko: is paste.pocoo down or slow? | 19:26 |
| fijal | yop | 19:28 |
| fijal | Alex_Gaynor: this is what vinay is doing there | 19:29 |
| Alex_Gaynor | fijal: yup | 19:29 |
| fijal | Alex_Gaynor: if you guys pull that to django, django would go crawling | 19:29 |
| Alex_Gaynor | it's what virtualenv and pip and py.test do as well | 19:29 |
| fijal | and it'll be your own fault | 19:29 |
| fijal | I'm sorry | 19:29 |
| Alex_Gaynor | fijal: what we might do is remove exception hacks, and wait until we drop 2.5 to merge this | 19:29 |
| fijal | http://sumulael.files.wordpress.com/2011/02/sp_1411_clip07.jpg | 19:29 |
| kenaan | 12hager ppc-jit-backend 11ecbfbef353c7 15/pypy/jit/backend/ppc/ppcgen/: made first tests run again on ppc32 | 19:30 |
| fschwidom (~fschwidom@2.211.214.102) joined #pypy. | 19:30 | |
| fijal | Alex_Gaynor: seriously, there is a really heavy hack required to make this sort of stuff fast | 19:31 |
| Alex_Gaynor | Yes I know. | 19:31 |
| fijal | note that you have all kinds of combinations | 19:31 |
| fijal | like sys.exc_info called from an interpreted frame | 19:31 |
| fijal | you're not tracing and you don't know it's [1] | 19:32 |
| fijal | so you're forcing the entire stack and aborting all the jit you have on the stack | 19:32 |
| fijal | and I hate everything you represent :) | 19:32 |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy. | 19:33 | |
| fijal | Alex_Gaynor: is 'as' syntax compatible with 2.6? | 19:34 |
| Alex_Gaynor | fijal: yes | 19:34 |
| fijal | ya | 19:34 |
| stakkars_ | fijal: did we abandon 2.5? | 19:37 |
| fijal | stakkars_: nope | 19:37 |
| fijal | stakkars_: we abandoned 2.4 though | 19:37 |
| stakkars_ | yes, that's good | 19:37 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1138 [12alex] | 19:40 |
| arigato | Alex_Gaynor: ah, note, rather than sys.exc_info() return some strange W_TupleObject subclass, | 19:41 |
| amaury_ | For py3k I fear we'll have to abandon 2.5 as well | 19:41 |
| arigato | you can also have it return some custom instance | 19:41 |
| arigato | i.e. an instance of some class that we write at app-level | 19:41 |
| arigato | that behaves "enough" like a tuple that nobody would notice the difference | 19:41 |
| Alex_Gaynor | arigato: it'll break someones program somewhere :/ | 19:41 |
| Alex_Gaynor | if we do these hacks we should do them properly, with a true tuple at applevl | 19:42 |
| arigato | I see the point, but it's rather unlikely, IMHO | 19:42 |
| Action: Alex_Gaynor thinks it's important to be committed to your hacks | 19:42 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/608 [12alex] | 19:42 |
| arigato | well, I don't think it's worth a true tuple at app-level, and honestly I doubt very much that someone cares | 19:42 |
| arigato | e.g. CPython switches happily from functions that return tuples to function that return e.g. stat_result or other tupleish | 19:43 |
| fijal | arigato: and what would that tuple give us? | 19:43 |
| fijal | I mean - how does the tuple make you not force the frame | 19:45 |
| arigato | ...indeed | 19:49 |
| Taggnostr (~quassel@dyn57-215.yok.fi) left irc: Read error: Connection reset by peer | 19:54 | |
| fschwidom (~fschwidom@2.211.214.102) left irc: Ping timeout: 248 seconds | 19:57 | |
| arigato | *cough* here is a different idea for sys.get_info() | 19:58 |
| arigato | it returns a tuple indeed, but the point is that | 19:58 |
| arigato | if we don't abort tracing just because we see sys.get_info() | 19:58 |
| arigato | then we see the whole trace which virtual-ize the tuple | 19:58 |
| arigato | so we can see by looking at the whole trace that the traceback is never used | 19:58 |
| Taggnostr (~quassel@dyn57-215.yok.fi) joined #pypy. | 19:58 | |
| stakkars_ (~tismer@g225044039.adsl.alicedsl.de) left irc: Quit: schnarch | 19:59 | |
| Alex_Gaynor | arigato: if the traceback becomes virtualized, will the mark_frame_as_escaped()still cause an allocation? | 20:00 |
| arigato | well we need to refactor a few things first | 20:01 |
| arigato | in this case, mark_frame_as_escaped() should not be called at all, if the traceback ends up removed | 20:02 |
| arigato | (ideally) | 20:02 |
| stakkars_ (~tismer@g225044039.adsl.alicedsl.de) joined #pypy. | 20:07 | |
| antocuni (~antocuni@host200-125-dynamic.25-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 20:09 | |
| Alex_Gaynor | fijal: FYI laplace benchmark seems to have gotten slower today, I haven't looked why yet | 20:12 |
| fijal | Alex_Gaynor: it got faster for me | 20:12 |
| fijal | arigato: the problem is also what if this is called *not* from tracing | 20:13 |
| fijal | but from the frame somewhere deep, but that happens to be interpreted | 20:13 |
| fijal | I suppose there are also boring solutions | 20:15 |
| fijal | like detect in bytecode sys.exc_info and compile it into check_if_globals_not_changed() | 20:15 |
| Guest___ (~textual@67.51.82.66) joined #pypy. | 20:18 | |
| Nick change: Guest___ -> rekamso | 20:18 | |
| kenaan | 12ctismer win64-stage1 11df6d8c92cc30 15/pypy/rlib/: test_libffi works almost. Only the two byval tests are left for further investigation | 20:21 |
| arigato | stakkars_: in my opinion it would be cool if win64 got a pypy before fixing all stranger things in the world, | 20:26 |
| arigato | e.g. with --no-allworkingmodules | 20:26 |
| stakkars_ | I was cleaning up some bugs which reduced 400 broken tests to work, and now I looked if it is easy to understand. | 20:28 |
| fijal | also ll2ctypes is not needed for translation | 20:28 |
| Action: fijal sidenotes | 20:28 | |
| stakkars_ | yesterday these tests all raised "setup error". I was grabbing my fruit, so to say. | 20:28 |
| arigato | sure, but still, I would (personally, again) ignore all these failing tests and try to have "translate.py targetpypystandalone --no-allworkingmodules" work | 20:29 |
| stakkars_ | and no, the point is that certain things break even on default win32, in test_rffi, which is not ok! | 20:30 |
| stakkars_ | sorry, test_libffi | 20:30 |
| arigato | yes, it *is* ok, you are allowed to work on one goal ignoring 41 other ones | 20:31 |
| arigato | anyway, I'm not complaining | 20:33 |
| arigato | thanks for fixing the tests :-) | 20:33 |
| stakkars_ | if I am doing things right, then most things would work already. I sometimes work on stuff to see if I'm right. this was libffi, where I failed 2 out of 29. Now I will stop it and postpone. | 20:34 |
| Alex_Gaynor | fijal: have you compiled since I merged my branch? | 20:34 |
| fijal | Alex_Gaynor: no | 20:34 |
| Alex_Gaynor | fijal: I guess maybe my branch did :/ | 20:34 |
| fijal | arigato: indeed the patch fixes it | 20:35 |
| arigato | fijal: good | 20:35 |
| fijal | "good" | 20:35 |
| fijal | better than before | 20:35 |
| fijal | yes, admitedly | 20:35 |
| stakkars_ | arigato: btw., now after I have a better overview what to expect, I am pretty sure that christmas is an unrealistic goal ;-/ | 20:36 |
| Alex_Gaynor | fijal: pff, I don't know who caused this, but calc_strides calls are in this trace | 20:36 |
| arigato | Alex_Gaynor: I did a compilation | 20:36 |
| Alex_Gaynor | arigato: what do you mean? | 20:37 |
| arigato | tannit:~arigo/pypysrc/32compiled/ | 20:37 |
| arigato | "have you compiled since I merged my branch?" yes, I did | 20:37 |
| Alex_Gaynor | arigato: I'm at ccf1f7b9b78b | 20:38 |
| Alex_Gaynor | so yes | 20:38 |
| arigato | yes, same here | 20:38 |
| Alex_Gaynor | arigato: I was just asking fijal if he was benchmarking with all my stuff merged, or not | 20:39 |
| Alex_Gaynor | since he said it got faster, when it's like 30x slower for me | 20:39 |
| arigato | ah, sorry | 20:39 |
| Alex_Gaynor | fijal: ah ok, "./pypy-c -S -mtimeit -s "import numpypy" "numpypy.int64(3) + numpypy.int64(2)"" reproduces the issue for me | 20:42 |
| fijal | Alex_Gaynor: ah | 20:45 |
| fijal | Alex_Gaynor: no, I benchmarked just after I included new register allocator stuff | 20:45 |
| fijal | which made numeric_3 10% faster | 20:45 |
| fijal | but I did not benchmark anything else | 20:45 |
| dracman (~draco@212.255.18.75) joined #pypy. | 20:45 | |
| Alex_Gaynor | fijal: this will only effect "slow" laplace | 20:46 |
| Alex_Gaynor | fijal: http://paste.pocoo.org/show/515947/ look ok? | 20:46 |
| fijal | no | 20:49 |
| fijal | it should be jit.unroll_safe | 20:49 |
| fijal | shape is always a constant | 20:50 |
| fijal | wait a second though | 20:50 |
| fijal | why there is calc_strides call *in the loop*? | 20:50 |
| fijal | this is for creating arrays | 20:50 |
| Alex_Gaynor | fijal: Scalar | 20:50 |
| fijal | should you be creating an array there? | 20:50 |
| Alex_Gaynor | numpy.add(2, 3) creates a Scalar object | 20:50 |
| Alex_Gaynor | this is in BaseArray.__init__ | 20:50 |
| Alex_Gaynor | should there be BaseReallyAnArray? | 20:51 |
| fijal | then change if self.shape is None to if self.shape | 20:51 |
| fijal | self.shape is never None anyway | 20:51 |
| Alex_Gaynor | that says `if self.strides is None` | 20:52 |
| fijal | change it to if self.strides | 20:52 |
| lizardo (~lizardo@189.2.128.130) left irc: Ping timeout: 248 seconds | 20:52 | |
| fijal | where is numpy.add in the loop? | 20:52 |
| fijal | I think you're chasing something weird | 20:53 |
| Alex_Gaynor | fijal: + on int64 objs just goes through the ufunc internaly | 20:53 |
| fijal | maybe it should not? | 20:53 |
| Alex_Gaynor | fijal: change it to `if self.strides` breaks tests | 20:53 |
| fijal | unroll_safe is not good enough | 20:53 |
| Alex_Gaynor | fijal: no, it's very good to have ndarray.__add__ and integer.__add__ have the same code path | 20:54 |
| Alex_Gaynor | why not? | 20:54 |
| fijal | because reverse will be still a call | 20:54 |
| Alex_Gaynor | fijal: no it won't, reverse is unrolled if the list is virtual | 20:54 |
| Alex_Gaynor | which it will be for Scalar | 20:54 |
| fijal | well | 20:54 |
| fijal | this is not going in a good direction | 20:54 |
| fijal | because we're already fighting with trace limits | 20:54 |
| fijal | and billion of seconds of warmups | 20:54 |
| fijal | I don't think it's good that it goes through the same path | 20:55 |
| melinath (~melinath@li341-41.members.linode.com) left irc: Ping timeout: 248 seconds | 20:55 | |
| fijal | it's not ok to write a horribly inefficient code just because the JIT can figure it out | 20:55 |
| Alex_Gaynor | well, can you fix it? I don't understand the strides stuff very well | 20:55 |
| fijal | noo | 20:55 |
| fijal | I mean it's broken that you have to instantiate a scalar in the first place | 20:55 |
| Alex_Gaynor | this was an issue before my branch FWIW, it just didn't show up as easil | 20:55 |
| fijal | well ok | 20:55 |
| fijal | but now if we have a loop that iterates over an array and reads items and does something | 20:56 |
| fijal | we go through billion of calls | 20:56 |
| fijal | it was not the case before | 20:56 |
| mikefc (~mikec@60-242-240-196.static.tpgi.com.au) joined #pypy. | 20:56 | |
| fijal | JIT is not good enough for such stuff | 20:56 |
| fijal | you'll have issues with bridges for one | 20:56 |
| panni_ (smuxi@descent.fragstore.net) left irc: Ping timeout: 248 seconds | 20:56 | |
| fijal | integer.__add__ should return NotImplemented and go via int.__add__ IMO | 20:56 |
| Alex_Gaynor | no, that's really a mess, because you have int8which subclasses integer | 20:57 |
| smuxi (smuxi@descent.fragstore.net) joined #pypy. | 20:57 | |
| Nick change: smuxi -> panni_ | 20:57 | |
| exarkun | mmm, ham and cheese int8which | 20:57 |
| fijal | IMO numpy.int32 should be an int subclass that's relatively efficient | 20:57 |
| fijal | and not a scalar | 20:57 |
| Alex_Gaynor | platform int class is a subclass of int, but it doesn't go through int code, due to RPython multiple inheritance mess | 20:58 |
| fijal | or refactor scalar | 20:58 |
| Alex_Gaynor | refactor scalar seems like the best route | 20:58 |
| fijal | then scalar should not really be an array | 20:58 |
| Alex_Gaynor | it should subclass ReallyBaseClass or something | 20:58 |
| fijal | yes, like W_Wrappable | 20:58 |
| fijal | er, Wrappable | 20:59 |
| fijal | or W_Object or so | 20:59 |
| Alex_Gaynor | no, because you want to have things in CAll2 typed as Base<NumpyClass> | 20:59 |
| fijal | probably | 20:59 |
| fijal | but then you kind of want those things to be slightly leaner | 20:59 |
| fijal | right now creating a numpy array in a loop is a major mess | 20:59 |
| Alex_Gaynor | that's pretty unrelated though | 21:00 |
| fijal | yes, well maybe | 21:01 |
| melinath (~melinath@li341-41.members.linode.com) joined #pypy. | 21:01 | |
| Alex_Gaynor | fijal: I guess I'm kind of included to go with look_inside_iff for now, and refactor scalar later | 21:03 |
| Alex_Gaynor | oh actually, I see a better fix | 21:03 |
| fijal | well | 21:03 |
| fijal | I would like it to be done soonish | 21:03 |
| fijal | you'll run into trace-too-long issues | 21:03 |
| rguillebert (~rguillebe@fec76-1-78-234-124-105.fbx.proxad.net) left irc: Remote host closed the connection | 21:04 | |
| Alex_Gaynor | fijal: http://paste.pocoo.org/show/515956/ | 21:04 |
| fijal | shape is always constant btw | 21:05 |
| fijal | so it's generall unroll_safe | 21:05 |
| fijal | scalar shouldn't have strides in the first place | 21:05 |
| fijal | I was thinking scalar is used to represent array(3) | 21:05 |
| fijal | but apparently not | 21:06 |
| Alex_Gaynor | it is, AFAIK | 21:06 |
| fijal | maybe we should make it simply work | 21:06 |
| fijal | well *olnly* this | 21:06 |
| fijal | wrapping and unwrapping is not a good thing IMO | 21:06 |
| Alex_Gaynor | well, I'm going to commit this, ok? | 21:06 |
| rguillebert (~rguillebe@2a01:e34:eea7:c690:21f:c6ff:fe12:4dee) joined #pypy. | 21:07 | |
| fijal | I hope you'll fix it instead of doing quick hacks :) | 21:07 |
| fijal | if it's not fixed today it's fine | 21:07 |
| Alex_Gaynor | I can look at refactoring Scalar long term, but this is a good fix that doesn't create huge traces for now | 21:08 |
| fijal | but if you *really* want to have it fixed now commit jit.unroll_safe() on calc_strides | 21:08 |
| fijal | fine I suppose | 21:08 |
| fijal | I gonna sleep, good night | 21:08 |
| Alex_Gaynor | you preferunroll_safe to just setting shape and stride correctly? | 21:08 |
| kenaan | 12alex_gaynor default 1187235ee9b8ab 15/pypy/module/micronumpy/interp_numarray.py: Preset correct attributes on scalars, so that a call to calc_strides isn't emitted | 21:10 |
| getxsick (xsx@unaffiliated/getxsick) left irc: Ping timeout: 248 seconds | 21:11 | |
| stakkars_ (~tismer@g225044039.adsl.alicedsl.de) left irc: Quit: schnarch | 21:12 | |
| getxsick (xsx@unaffiliated/getxsick) joined #pypy. | 21:12 | |
| ixokai (~ixokai@rrcs-24-199-8-246.west.biz.rr.com) joined #pypy. | 21:14 | |
| ixokai (~ixokai@rrcs-24-199-8-246.west.biz.rr.com) left irc: Changing host | 21:14 | |
| ixokai (~ixokai@pdpc/supporter/student/ixokai) joined #pypy. | 21:14 | |
| kkris (~kris@80-123-37-202.adsl.highway.telekom.at) left irc: Quit: Leaving. | 21:17 | |
| mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) left irc: Remote host closed the connection | 21:18 | |
| aboudreault (~alanb@osgeo/member/aboudreault) left irc: Quit: Leaving | 21:19 | |
| ericflo (~ericflo@75.103.8.110) joined #pypy. | 21:27 | |
| fijal (~fijal@41.52.155.130) left irc: Ping timeout: 248 seconds | 21:35 | |
| stakkars_ (~stakkars@89.204.139.232) joined #pypy. | 21:35 | |
| stakkars_ | arigato: I am hesitatin ro try targetstandalone before the bugs are below, say, 500 | 21:37 |
| arigato | "targetpypystandalone --no-allworkingmodules" is what I mentioned, not all of targetpypystandalone | 21:38 |
| arigato | moreover you can start with -O1 if Boehm works correctly | 21:39 |
| stakkars_ | From old experience, building pypy takes hours. Do you think it is easy to find errors this way? | 21:39 |
| stakkars_ | Ok, boehm seems to work. But what is the reason? It will work if all is correct | 21:41 |
| stakkars_ | Or am I searching bugs which are not covered by test? | 21:41 |
| stakkars_ | the toolchain is working, I think that is pretty complete and covered by rpystonedalone. Why should I try a translation which is very likely to break? | 21:45 |
| stakkars_ | Ok, I can try, but I don't see an advantage, yet | 21:48 |
| Ademan (~dan@15.sub-97-155-244.myvzw.com) joined #pypy. | 22:01 | |
| ixokai (~ixokai@pdpc/supporter/student/ixokai) left irc: Quit: Bye! | 22:07 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/7 [12stakkars, win64-stage1] | 22:11 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/7 [12stakkars, win64-stage1] | 22:12 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/8 [12stakkars, win64-stage1] | 22:12 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/8 [12stakkars, win64-stage1] | 22:12 |
| arigato (~arigo@89.204.130.72) left irc: Quit: See you | 22:12 | |
| Ademan (~dan@15.sub-97-155-244.myvzw.com) left irc: Quit: leaving | 22:13 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-win-x86-64/builds/80 [12stakkars, win64-stage1] | 22:18 |
| thp (thp@Maemo/community/contributor/thp) left #pypy. | 22:23 | |
| fprimex (~fprimex@fl-67-233-144-239.dhcp.embarqhsd.net) joined #pypy. | 22:23 | |
| hruske (~Gasper@188-230-156-183.dynamic.t-2.net) left irc: Remote host closed the connection | 22:25 | |
| robertj (~Rob@97-81-71-34.dhcp.athn.ga.charter.com) left irc: Ping timeout: 260 seconds | 22:26 | |
| stakkars_ (~stakkars@89.204.139.232) left irc: Quit: Colloquy for iPhone - http://colloquy.mobi | 22:29 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/157 | 22:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/732 | 22:38 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/962 | 22:38 |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 244 seconds | 22:45 | |
| nedbat (~nedbat@python/psf/nedbat) left irc: Disconnected by services | 22:51 | |
| nedbat_ (~nedbat@python/psf/nedbat) joined #pypy. | 22:52 | |
| kenaan | 12alex_gaynor default 11e5ef85031c32 15/pypy/: Mark several fields in numpy as being immutable, and the name of a type object as quassiimmut | 22:56 |
| stakkars_ (~stakkars@89.204.139.232) joined #pypy. | 22:57 | |
| stakkars_ | Ari | 22:57 |
| nedbat_ (~nedbat@python/psf/nedbat) left irc: Ping timeout: 258 seconds | 22:57 | |
| whitelynx (~whitelynx@63.241.75.144) left irc: Quit: Ex-Chat | 22:58 | |
| stakkars_ | I am pretty much bored. Going to close my branch and maybe work non-publically for a while | 23:00 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1139 [12alex] | 23:00 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/609 [12alex] | 23:00 |
| kenaan | 12hager ppc-jit-backend 117805074b94a4 15/pypy/jit/backend/ppc/ppcgen/: made first tests pass again on ppc64 | 23:03 |
| mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) joined #pypy. | 23:06 | |
| mfoord (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy. | 23:06 | |
| mfoord (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host | 23:06 | |
| mfoord (~anonymous@python/psf/voidspace) joined #pypy. | 23:06 | |
| stakkars_ (~stakkars@89.204.139.232) left irc: Quit: Colloquy for iPhone - http://colloquy.mobi | 23:10 | |
| Rhy0lite (dje@nat/ibm/x-zzexcoiegummbmof) left irc: Quit: Leaving | 23:10 | |
| jterrace (~jterrace@aegis.CS.Princeton.EDU) joined #pypy. | 23:11 | |
| jterrace | Alex_Gaynor: is it easy to add new numpy ufuncs? | 23:11 |
| Alex_Gaynor | jterrace: extremely | 23:11 |
| jterrace | ok | 23:11 |
| Alex_Gaynor | which ones are we missing? | 23:11 |
| jterrace | let me see | 23:11 |
| jterrace | okay they're not all ufuncs | 23:18 |
| jterrace | but the full list of ones that i use that don't seem to be in yet: identity, fromstring, nan, asmatrix, asarray, concatenate, cumsum, arange, dstack, hstack, vstack, flat, flatten, swapaxes, transpose, cross, newaxis, allclose, sqrt, equal | 23:19 |
| lac (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) joined #pypy. | 23:19 | |
| jterrace | some of those i think would be really easy (like sqrt or equal) | 23:19 |
| jterrace | other probably hard (*stack, allclose) | 23:19 |
| lac_ (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) left irc: Ping timeout: 245 seconds | 23:19 | |
| mikefc | hi all. new to pypy. looking to have a fiddle with the numpy stuff. | 23:22 |
| mikefc | first question: is a ufunc a function that is mapped to each element in a numpy array? | 23:22 |
| mikefc | I'm looking to add pickling support for numpy so that I can use them with multiprocessing. | 23:22 |
| jterrace | can you pickle regular numpy arrays? | 23:23 |
| mikefc | yes. | 23:23 |
| mikefc | but not numpypy ones. | 23:23 |
| mfoord (~anonymous@python/psf/voidspace) left irc: Quit: mfoord | 23:25 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 23:26 | |
| jterrace | mikefc: i think you just have to implement __getstate__ and __setstate__, which can probably just be put on top of something like numpy.save / numpy.load, but i dont think they exist yet either | 23:33 |
| fprimex (~fprimex@fl-67-233-144-239.dhcp.embarqhsd.net) left irc: Quit: Leaving | 23:34 | |
| jterrace | npy format is documented here: https://github.com/numpy/numpy/blob/master/doc/neps/npy-format.txt would be really cool to get numpypy compatibility for npy to allow interchange | 23:36 |
| jterrace (~jterrace@aegis.CS.Princeton.EDU) left irc: Quit: Leaving. | 23:39 | |
| rekamso (~textual@67.51.82.66) left irc: Ping timeout: 244 seconds | 23:39 | |
| stakkars_ (~stakkars@89.204.139.232) joined #pypy. | 23:45 | |
| apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection | 23:49 | |
| JaRoel|4d (~jaroel|4d@2001:980:7133:1:d97d:16c1:8242:a4d2) joined #pypy. | 23:53 | |
| stakkars_ (~stakkars@89.204.139.232) left irc: Quit: Colloquy for iPhone - http://colloquy.mobi | 23:56 | |
| mfoord (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy. | 23:56 | |
| mfoord (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host | 23:56 | |
| mfoord (~anonymous@python/psf/voidspace) joined #pypy. | 23:56 | |
| Arfrever (~Arfrever@apache/committer/Arfrever) joined #pypy. | 23:59 | |
| --- Sat Dec 3 2011 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!
