| ronny | re | 00:00 |
|---|---|---|
| Unhelpful | ronny: you did distribute --user? because i use that :) | 00:04 |
| ronny | Unhelpful: yes | 00:07 |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/916 | 00:09 |
| Turicas (~alvaro@mail.justen.eng.br) left irc: Quit: Leaving | 00:16 | |
| kennethreitz (~kennethre@c-24-127-96-129.hsd1.va.comcast.net) joined #pypy. | 00:20 | |
| asmeurer (~asmeurer@216.239.45.19) joined #pypy. | 00:44 | |
| asmeurer (~asmeurer@216.239.45.19) left irc: Ping timeout: 240 seconds | 00:50 | |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1802 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/248 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/536 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/589 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1064 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/775 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1407 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/668 | 00:50 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/239 | 00:50 |
| bbot2 | 3Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/113 | 01:47 |
| Moku (~John@osbk-4db06492.pool.mediaWays.net) joined #pypy. | 01:55 | |
| Shinita (~John@osbk-4d08bc06.pool.mediaWays.net) left irc: Ping timeout: 240 seconds | 01:55 | |
| Nick change: Moku -> Guest28331 | 01:55 | |
| aericson (~andre@186.212.140.66) joined #pypy. | 01:55 | |
| aericson | I get errors when running pypy when pwd is home | 01:58 |
| aericson | wait pasting | 01:58 |
| aericson | http://paste.pocoo.org/show/497131/ | 01:59 |
| aericson | is it a known problem? | 01:59 |
| timonator | is "pypy" a symlink? or did you copy the pypy binary over? | 02:07 |
| timonator | does pypy -v give more information? | 02:07 |
| aericson | $ file /usr/bin/pypy /usr/bin/pypy: symbolic link to `/opt/pypy/pypy-c' | 02:09 |
| aericson | pypy -v output the same thing | 02:10 |
| timonator | oh? interesting, i'd've expected /opt/pypy/bin/pypy-c | 02:10 |
| timonator | well, hmm. don't know then | 02:10 |
| timonator | i'm just a newbie myself :| | 02:10 |
| aericson | so it's probably a package bug right? | 02:10 |
| aericson | will check with the distro | 02:11 |
| timonator | is it only when pwd is home? | 02:11 |
| timonator | there might be something in your ~ that pypy thinks it's supposed to import, which doesn't actually work or something? | 02:11 |
| aericson | seems like only ~ | 02:12 |
| aericson | could be | 02:12 |
| timonator | have you tried with a completely fresh user? straced the process maybe? | 02:12 |
| timonator | that is: strace -e open pypy | 02:12 |
| aericson | works fine with other user | 02:13 |
| timonator | then it's probably something that's wrong in your ~, some file or folder that confuses pypy i suppose? | 02:13 |
| aericson | looking for it | 02:14 |
| aericson | strace doesn't mention anything about home but does show some access errors | 02:15 |
| aericson | http://paste.pocoo.org/show/497136/ | 02:15 |
| timonator | anyway, i'm going to get some sleep | 02:15 |
| aericson | sure | 02:15 |
| aericson | thanks for the help | 02:15 |
| aericson | will try to figure it also | 02:15 |
| aericson | figure it out* | 02:15 |
| timonator | i get open("/opt/pypy/lib-python/modified-2.7/site.py", O_RDONLY) = 3 | 02:15 |
| timonator | some permission problem or similar i guess | 02:16 |
| timonator | uh, you get that, too, sorry. already tired | 02:16 |
| aericson | it's ok | 02:16 |
| timonator | you could try cding into modified-2.7 and import site.py from there and see if an exception happens, but that's just my typical shotgun-debugging | 02:17 |
| timonator | good luck! | 02:17 |
| aericson | thanks | 02:17 |
| tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) joined #pypy. | 02:26 | |
| tilgovi (~randall@c-98-210-155-124.hsd1.ca.comcast.net) left irc: Changing host | 02:26 | |
| tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy. | 02:26 | |
| canta (~canta@77-20-123-240-dynip.superkabel.de) left irc: Quit: Leaving. | 02:32 | |
| alexr_web (80dc9f14@gateway/web/freenode/ip.128.220.159.20) joined #pypy. | 02:42 | |
| alexr_web | Is there any reason why pypy would give a different answer than normal cpython when doing many floating point additions? | 02:43 |
| Unhelpful | Just additions? And how are they performed? Are the values in a list, a dict, etc? | 02:45 |
| alexr_web | i just checked again, and it's not the additions | 02:46 |
| alexr_web | i'm writing a probabilistic parser, and python vs pypy come up with two different weights for a particular string/grammar | 02:47 |
| alexr_web | i thought the parses were the same | 02:47 |
| alexr_web | (which would have pointed to the additions being weird) | 02:47 |
| alexr_web | but I just looked again, and the parses are different | 02:47 |
| alexr_web | is this sort of thing considered a bug? (or interesting to developers at all) | 02:48 |
| Unhelpful | It's possible pypy may order dictionary keys differently, which is why I ask what produces the values. | 02:48 |
| alexr_web | k | 02:49 |
| alexr_web | that sounds plausible | 02:49 |
| alexr_web | i am using dictionaries | 02:49 |
| whyking (~quassel@ip68-1-185-182.ri.ri.cox.net) joined #pypy. | 02:50 | |
| Unhelpful | Are you iterating their contents? | 02:50 |
| alexr_web | no; just doing lookups | 02:50 |
| alexr_web | iterating sets very rarely | 02:51 |
| Arfrever (~Arfrever@apache/committer/Arfrever) left irc: Quit: Ex+re | 02:51 | |
| alexr_web | i shouldn't be relying on order of iteration anywhere, though | 02:52 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1802 | 02:52 |
| Unhelpful | Ah, iterating a set is similar. I would try to see at what point in the code the results actually differ. | 02:53 |
| Unhelpful | If the difference isn't in set or dict ordering, or something else the language doesn't specify, then yes, it might be a bug. There's a tracker where you can file it, preferably with simple example code. | 02:56 |
| alexr_web | I think I found it, and it is in set ordering. I'll look into it more; it indicates a bug in my code either way. | 02:58 |
| alexr_web | thanks | 02:58 |
| alexr_web (80dc9f14@gateway/web/freenode/ip.128.220.159.20) left irc: Quit: Page closed | 02:58 | |
| tilgovi (~randall@couchdb/developer/tilgovi) left irc: Read error: Connection reset by peer | 03:03 | |
| whyking (~quassel@ip68-1-185-182.ri.ri.cox.net) left irc: Read error: Connection reset by peer | 03:27 | |
| jonanin (~jonanin@24-183-50-140.dhcp.mdsn.wi.charter.com) left irc: Read error: Connection reset by peer | 03:44 | |
| mat^2 (~mathias@212.130.113.35) left irc: Ping timeout: 276 seconds | 03:50 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1407 | 03:51 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/589 | 03:52 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/775 | 03:53 |
| ericholscher (~eric@173.203.223.63) left irc: Ping timeout: 252 seconds | 04:01 | |
| kalenz (~kalenz@ping.kalenz.fr) left irc: Read error: Operation timed out | 04:01 | |
| kalenz (~kalenz@ping.kalenz.fr) joined #pypy. | 04:01 | |
| Nick change: kalenz -> Guest1278 | 04:01 | |
| __class__ (~class@99-105-56-217.lightspeed.sntcca.sbcglobal.net) left irc: Read error: Operation timed out | 04:02 | |
| aericson (~andre@186.212.140.66) left irc: Ping timeout: 252 seconds | 04:02 | |
| Rhyolite (~dje@pool-108-6-25-71.nycmny.fios.verizon.net) left irc: Ping timeout: 252 seconds | 04:02 | |
| chickamade (~chickamad@118.69.67.203) joined #pypy. | 04:02 | |
| Alex_Gaynor (u1246@gateway/web/irccloud.com/x-avmqhggwnokubikw) left irc: Excess Flood | 04:02 | |
| ericholscher (~eric@173.203.223.63) joined #pypy. | 04:02 | |
| aericson (~andre@186.212.140.66) joined #pypy. | 04:02 | |
| Rhyolite (~dje@pool-108-6-25-71.nycmny.fios.verizon.net) joined #pypy. | 04:02 | |
| __class__ (~class@99.105.56.217) joined #pypy. | 04:06 | |
| DasIch_ (~dasich@p4FFDFEA7.dip.t-dialin.net) joined #pypy. | 04:07 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/1064 | 04:10 |
| DasIch (~dasich@p4FFDD5A5.dip.t-dialin.net) left irc: Ping timeout: 248 seconds | 04:10 | |
| Alex_Gaynor (u1246@gateway/web/irccloud.com/x-ecpcksyavcturcto) joined #pypy. | 04:13 | |
| #pypy: mode change '+o Alex_Gaynor' by ChanServ!ChanServ@services. | 04:13 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/536 | 04:28 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/248 | 04:52 |
| asmeurer (~asmeurer@c-71-198-74-130.hsd1.ca.comcast.net) joined #pypy. | 05:07 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/668 | 05:34 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/239 | 05:47 |
| mwhudson_ (~mwh@121-73-77-183.cable.telstraclear.net) joined #pypy. | 06:00 | |
| aleksi (~aleksi@85.235.191.82) joined #pypy. | 06:07 | |
| JaRoel|4d (~jaroel|4d@sink.jaroel.nl) left irc: Remote host closed the connection | 06:11 | |
| Nick change: Guest1278 -> kalenz | 06:13 | |
| kushal (kdas@nat/redhat/x-zcvgkvobfmsbknyd) joined #pypy. | 06:13 | |
| kushal (kdas@nat/redhat/x-zcvgkvobfmsbknyd) left irc: Changing host | 06:13 | |
| kushal (kdas@fedora/kushal) joined #pypy. | 06:13 | |
| mvt (~mvantelli@87.213.45.85) joined #pypy. | 06:19 | |
| canta (~canta@77-20-123-240-dynip.superkabel.de) joined #pypy. | 06:30 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 06:41 | |
| asmeurer (~asmeurer@c-71-198-74-130.hsd1.ca.comcast.net) left irc: Quit: asmeurer | 07:14 | |
| canta (~canta@77-20-123-240-dynip.superkabel.de) left irc: Quit: Leaving. | 07:27 | |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) joined #pypy. | 07:33 | |
| tlynn (~tlynn@cpc6-cmbg14-2-0-cust121.5-4.cable.virginmedia.com) joined #pypy. | 07:33 | |
| ojii (~ojii@40-34.60-188.cust.bluewin.ch) joined #pypy. | 07:36 | |
| EnCuKou (~encukou@a91-156-243-246.elisa-laajakaista.fi) left irc: Quit: Mizím. | 07:48 | |
| teknico (~quassel@88-149-209-113.dynamic.ngi.it) joined #pypy. | 07:59 | |
| amaury_ (amaury_@nat/google/x-jdpggdnyowfvjwea) joined #pypy. | 08:09 | |
| lucian (~lucian@78-86-111-129.zone2.bethere.co.uk) joined #pypy. | 08:20 | |
| G2P (~G2P@fw-asn1.ornis.com) joined #pypy. | 08:29 | |
| lucian (~lucian@78-86-111-129.zone2.bethere.co.uk) left irc: Ping timeout: 240 seconds | 08:35 | |
| kenaan | 12arigo concurrent-marksweep 116a06b46f0e3f 15/pypy/rpython/memory/gc/concurrentgen.py: Major collections. | 08:39 |
| kenaan | 12arigo concurrent-marksweep 112c6f2010d4af 15/pypy/rpython/memory/: Fixes. | 08:39 |
| voidspace (~voidspace@python/psf/voidspace) joined #pypy. | 08:43 | |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) left irc: Ping timeout: 260 seconds | 08:47 | |
| tlynn (~tlynn@cpc6-cmbg14-2-0-cust121.5-4.cable.virginmedia.com) left irc: Ping timeout: 240 seconds | 09:03 | |
| bogner (~bogner@2600:3c03::f03c:91ff:fedf:7ef4) joined #pypy. | 09:10 | |
| kenaan | 12cfbolz int-tag-untag-as-operations 1163dfc29454a2 15/: a branch where to make int tagging and untagging explicit operations | 09:12 |
| kenaan | 12cfbolz int-tag-untag-as-operations 11300e42e0ee80 15/pypy/: introduce int_tag (which can overflow) and int_untag | 09:12 |
| kenaan | 12cfbolz int-tag-untag-as-operations 1152e92200bf93 15/pypy/jit/metainterp/optimizeopt/: each the optimizer that int_tag and int_untag are each others opposites | 09:12 |
| kenaan | 12cfbolz int-tag-untag-as-operations 11c499ddaefd52 15/pypy/jit/metainterp/optimizeopt/: add a more precise .mul operation on integer bounds. | 09:12 |
| kenaan | 12cfbolz int-tag-untag-as-operations 11c7b823622f57 15/pypy/jit/metainterp/optimizeopt/: check that operations on untagged integers get their overflow checks removed | 09:12 |
| kenaan | 12cfbolz int-tag-untag-as-operations 11766299f6042c 15/pypy/: rename the operation to int_tag_ovf | 09:12 |
| kenaan | 12cfbolz int-tag-untag-as-operations 111a183750ebf7 15/pypy/jit/metainterp/: introduce an no-overflow-check variant int_tag for when the bounds-optimizer finds out that so... | 09:12 |
| bbot2 (~bbot2@wyvern.cs.uni-duesseldorf.de) left irc: Ping timeout: 260 seconds | 09:19 | |
| mvt (~mvantelli@87.213.45.85) left irc: Quit: Leaving | 09:28 | |
| ericvrp_ (~ericvrp_@ericvrp.xs4all.nl) joined #pypy. | 09:29 | |
| mvt (~mvantelli@87.213.45.85) joined #pypy. | 09:31 | |
| derdon (~derdon@pD9E1C2F0.dip.t-dialin.net) joined #pypy. | 09:46 | |
| ronny | meh | 09:51 |
| arigato (~arigo@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 09:51 | |
| ronny | seems like the author of pexpect cant be reached timely, porting pyrepl will be harder without it on py3 | 09:51 |
| voidspace_ (~voidspace@87-194-212-65.bethere.co.uk) joined #pypy. | 09:57 | |
| voidspace_ (~voidspace@87-194-212-65.bethere.co.uk) left irc: Changing host | 09:57 | |
| voidspace_ (~voidspace@python/psf/voidspace) joined #pypy. | 09:57 | |
| voidspace (~voidspace@python/psf/voidspace) left irc: Ping timeout: 240 seconds | 09:59 | |
| Nick change: voidspace_ -> voidspace | 09:59 | |
| ericvrp_ (~ericvrp_@ericvrp.xs4all.nl) left irc: Quit: ericvrp_ | 10:01 | |
| mvt (mvantelli@87.213.45.85) left #pypy ("Leaving"). | 10:06 | |
| mwhudson_ (~mwh@121-73-77-183.cable.telstraclear.net) left irc: Ping timeout: 260 seconds | 10:09 | |
| mvt (~mvantelli@87.213.45.85) joined #pypy. | 10:10 | |
| voidspace_ (~voidspace@87-194-212-65.bethere.co.uk) joined #pypy. | 10:10 | |
| voidspace_ (~voidspace@87-194-212-65.bethere.co.uk) left irc: Changing host | 10:10 | |
| voidspace_ (~voidspace@python/psf/voidspace) joined #pypy. | 10:10 | |
| voidspace (~voidspace@python/psf/voidspace) left irc: Ping timeout: 240 seconds | 10:11 | |
| Nick change: voidspace_ -> voidspace | 10:11 | |
| witulski (~stupsi@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 10:14 | |
| aericson (~andre@186.212.140.66) left irc: Ping timeout: 240 seconds | 10:14 | |
| witulski (stupsi@fwstups.cs.uni-duesseldorf.de) left #pypy. | 10:14 | |
| aericson (~andre@186.212.140.66) joined #pypy. | 10:16 | |
| ericflo (~ericflo@c-76-126-10-15.hsd1.ca.comcast.net) left irc: Read error: Connection reset by peer | 10:19 | |
| ericflo (~ericflo@c-76-126-10-15.hsd1.ca.comcast.net) joined #pypy. | 10:19 | |
| mvt (~mvantelli@87.213.45.85) left irc: Read error: Connection reset by peer | 10:19 | |
| ericflo (~ericflo@c-76-126-10-15.hsd1.ca.comcast.net) left irc: Client Quit | 10:19 | |
| mvt (~mvantelli@87.213.45.85) joined #pypy. | 10:20 | |
| whitelynx (~whitelynx@75.111.197.204) left irc: Quit: Leaving | 10:22 | |
| aericson (~andre@186.212.140.66) left irc: Ping timeout: 240 seconds | 10:24 | |
| voidspace (~voidspace@python/psf/voidspace) left irc: Quit: Uhm... gotta go | 10:26 | |
| kenaan | 12cfbolz int-tag-untag-as-operations 1130319db8738f 15/pypy/jit/: (arigo, cfbolz): implement int_tag, int_tag_ovf and int_untag in the x86 backend | 10:32 |
| ramusara (~ramusara@220.156.210.236.user.e-catv.ne.jp) joined #pypy. | 10:33 | |
| Vorpal (~AnMaster@unaffiliated/anmaster) joined #pypy. | 10:36 | |
| verte (~verte@python/site-packages/verte) joined #pypy. | 10:37 | |
| voidspace (~anonymous@python/psf/voidspace) joined #pypy. | 10:38 | |
| chickamade (~chickamad@118.69.67.203) left irc: Quit: chickamade | 10:48 | |
| arigato (~arigo@fwstups.cs.uni-duesseldorf.de) left irc: Ping timeout: 244 seconds | 10:51 | |
| amaury_ (amaury_@nat/google/x-jdpggdnyowfvjwea) left irc: Remote host closed the connection | 10:52 | |
| witulski (~stupsi@134.99.16.23) joined #pypy. | 10:57 | |
| witulski (stupsi@134.99.16.23) left #pypy. | 11:02 | |
| arigato (~arigo@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 11:02 | |
| witulski (~stupsi@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 11:23 | |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) joined #pypy. | 11:26 | |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy | 11:28 | |
| fijal (c156f311@gateway/web/freenode/ip.193.86.243.17) joined #pypy. | 11:29 | |
| fijal | hi | 11:30 |
| fijal (c156f311@gateway/web/freenode/ip.193.86.243.17) left irc: Ping timeout: 265 seconds | 11:37 | |
| arigato | hi | 11:39 |
| arigato | er, too late :-) | 11:39 |
| witulski (stupsi@fwstups.cs.uni-duesseldorf.de) left #pypy. | 11:45 | |
| JaRoel (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 11:47 | |
| verte (~verte@python/site-packages/verte) left irc: Quit: ~~~ Crash in JIT! | 11:48 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Ping timeout: 258 seconds | 11:50 | |
| exarkun | What do the "More information" icons on https://bitbucket.org/pypy/compatibility/wiki/Home mean? | 11:56 |
| antocuni | exarkun: that there is a wiki page with more infos about that specific package | 11:58 |
| JaRoel (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 12:00 | |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy. | 12:02 | |
| arigato | bah, how do you do "cast_float_to_uint" in x86 assembler... | 12:03 |
| amaury_ (amaury_@nat/google/x-mtsptvyaitsaqitq) joined #pypy. | 12:03 | |
| Action: arigato noticed that our code is wrong | 12:03 | |
| Action: arigato looks at what gcc produces and is scared | 12:05 | |
| exarkun | antocuni: How do you find that wiki page? | 12:10 |
| antocuni | click on the link? | 12:10 |
| exarkun | I would have thought. But it's not a link, just an icon. | 12:11 |
| antocuni | exarkun: no, you can click on the package name | 12:11 |
| antocuni | e.g., "Trac" has the icon, and if you click on "Trac" you find more infos | 12:11 |
| exarkun | Ah. Thanks. | 12:12 |
| antocuni | if you click on other packages, you get en empty page | 12:12 |
| Rhy0lite (dje@nat/ibm/x-losizlekoquzszlt) joined #pypy. | 12:14 | |
| stakkars_ (~tismer@82.113.121.133) joined #pypy. | 12:25 | |
| Unhelpful | erm, *cast* or *convert*? | 12:27 |
| arigato | "cast" in the C sense, i.e. really convert | 12:28 |
| arigato | in this case | 12:28 |
| Unhelpful | ah. hrm. | 12:28 |
| Action: arigato doesn't even want to look at how on 32-bit gcc does a casst from "double" to "unsigned long long" | 12:29 | |
| Unhelpful | erm, looks to be one instruction, basically... with sse anyway :) | 12:32 |
| arigato | to unsigned long long? | 12:32 |
| Unhelpful | oh. well, i was doing int! :) | 12:33 |
| arigato | yes, cast_float_to_int works | 12:33 |
| arigato | what doesn't are all other cases | 12:33 |
| arigato | i.e. cast_float_to_X, for an integer type X that is not contained within 'int' | 12:33 |
| kenaan | 12arigo concurrent-marksweep 11d5d9c2cef342 15/pypy/rpython/memory/gc/concurrentgen.py: Reorder the tests for performance. | 12:34 |
| kenaan | 12arigo default 114241a19c0141 15/pypy/: Fix the declaration of the lloperation 'cast_float_to_int' to no longer pretend it raises OverflowError, which is w... | 12:34 |
| Unhelpful | to unsigned int apparently is a one-op too. double->unsigned long long looks messier. | 12:35 |
| Unhelpful | and those one-ops all assume use of sse and storage of floats in xmm registers | 12:35 |
| arigato | to unsigned int? on 64-bit maybe, because there 'unsigned int' is contained into the C long | 12:35 |
| arigato | when I was talking about 'cast_float_to_int' it's with the Python names, so in C names it's really 'cast_double_to_long' | 12:36 |
| Unhelpful | "long" meaning how many bits, here? | 12:37 |
| Unhelpful | always 64? always 32? or depends on architecture? | 12:37 |
| arigato | a C long on any platform supported by pypy right now | 12:38 |
| bgola (~bgola@c95185bc.virtua.com.br) joined #pypy. | 12:38 | |
| Unhelpful | so whatever C *calls* long, not necessarily the largest single-register machine word? | 12:39 |
| arigato | a C long on any platform supported by pypy right now *is* the largest single-register machine word | 12:44 |
| kenaan | 12arigo concurrent-marksweep 11503589dfa9f4 15/: hg merge default | 12:45 |
| Unhelpful | ugh, yeah, and apparently there's no way to do but x87 fpu. | 12:48 |
| ojii_ (~ojii@40-34.60-188.cust.bluewin.ch) joined #pypy. | 12:48 | |
| santagada (~leonardo@201.86.242.147.dynamic.adsl.gvt.net.br) joined #pypy. | 12:49 | |
| ojii (~ojii@40-34.60-188.cust.bluewin.ch) left irc: Ping timeout: 240 seconds | 12:49 | |
| Nick change: ojii_ -> ojii | 12:49 | |
| G2P (~G2P@fw-asn1.ornis.com) left irc: Read error: Connection reset by peer | 12:52 | |
| Action: arigato broke micronumpy translation | 12:52 | |
| arigato | sorry sorry | 12:52 |
| jnoller (~jnoller@vpn.nasuni.com) joined #pypy. | 12:55 | |
| Unhelpful | arigato: after a crash course in x87... it looks like a lot of this mess is about storing the fpu control word, setting it, and restoring it. if i'm following correctly, the actual conversion happens with fistpll (%esp), storing the result to stack. | 12:59 |
| arigato | we don't use the x87 in pypy. | 12:59 |
| arigato | only xmm registers | 13:00 |
| Unhelpful | well, it looks like x86-32 really wants to use x87, or at least that's what gcc does. | 13:00 |
| Unhelpful | specifically for conversion to unsigned long | 13:00 |
| arigato | no, we compile with -msse2 -mfpmath=see | 13:01 |
| arigato | no, we compile with -msse2 -mfpmath=sse | 13:01 |
| Unhelpful | yes, i just added mfpmath=sse... that gets me a *different* mess :) | 13:02 |
| canta (~canta@77-20-123-240-dynip.superkabel.de) joined #pypy. | 13:07 | |
| G2P (~G2P@fw-asn1.ornis.com) joined #pypy. | 13:09 | |
| kushal (kdas@fedora/kushal) left irc: Ping timeout: 245 seconds | 13:11 | |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) left irc: Ping timeout: 260 seconds | 13:17 | |
| Nick change: DasIch_ -> DasIch | 13:19 | |
| aboudreault (~alanb@osgeo/member/aboudreault) joined #pypy. | 13:19 | |
| kenaan | 12arigo default 119977c322a23b 15/pypy/jit/: cast_uint_to_float and cast_float_to_uint are needed for micronumpy. Implement them (hopefully correctly now) as a... | 13:23 |
| lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) joined #pypy. | 13:27 | |
| gtaylor (~gtaylor@108-196-160-120.lightspeed.tukrga.sbcglobal.net) joined #pypy. | 13:38 | |
| lucian_ (~lucian@93-97-174-114.zone5.bethere.co.uk) joined #pypy. | 13:40 | |
| whitelynx (~whitelynx@li117-47.members.linode.com) joined #pypy. | 13:41 | |
| lucian (~lucian@93-97-174-114.zone5.bethere.co.uk) left irc: Ping timeout: 240 seconds | 13:41 | |
| Remilia_Scarlet (~proton@124.131.64.28) joined #pypy. | 13:50 | |
| Remilia_Scarlet | test | 13:51 |
| kushal (~kdas@fedora/kushal) joined #pypy. | 13:53 | |
| santagada | Remilia_Scarlet: its working | 13:53 |
| Remilia_Scarlet | oh thank you~ | 13:53 |
| Remilia_Scarlet | new to irc -_- | 13:54 |
| Remilia_Scarlet | i've been reading the src of flowgraph part of pypy | 13:56 |
| Remilia_Scarlet | there are 2 classes called SpamBlock and EggBlock | 13:56 |
| Remilia_Scarlet | could anyone tell me why they are named this? this confuses me a lot | 13:56 |
| Remilia_Scarlet | in pypy/objspace/flow/flowcontext.py | 13:58 |
| exarkun | "spam and eggs" is a monty python skit | 13:59 |
| timonator | i've only ever seen spam blocks, what do egg blocks do? | 14:00 |
| exarkun | https://www.youtube.com/watch?v=anwy2MPT5RE | 14:00 |
| timonator | and are all flow graphs full of spam? i thought a spamblock would cause translation to stop? | 14:00 |
| arigato | ...cast_float_to_longlong... | 14:00 |
| cwillu_at_work (~cwillu@cwillu-1-pt.tunnel.tserv13.ash1.ipv6.he.net) left irc: Ping timeout: 244 seconds | 14:03 | |
| Remilia_Scarlet | well, sounds interesting... but i can't watch youtube videos. thank you @exarkun | 14:03 |
| Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 248 seconds | 14:05 | |
| Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy. | 14:06 | |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) joined #pypy. | 14:07 | |
| antocuni | arigato: what's the point of the int-tag-untag-* branch? | 14:15 |
| jnoller (~jnoller@vpn.nasuni.com) left irc: Quit: Leaving... | 14:17 | |
| Remilia_Scarlet | and are all flow graphs full of spam? ah yes | 14:19 |
| Remilia_Scarlet | AFAIK [SpamBlock]s are the 'normal' blocks | 14:20 |
| Remilia_Scarlet | and [EggBlock]s are generated when there is a 'is_true' operation, | 14:21 |
| Remilia_Scarlet | to fool the opcode interpreter it's True or False returned | 14:22 |
| Remilia_Scarlet | this is the trick online doc refers to | 14:23 |
| lucian_ (~lucian@93-97-174-114.zone5.bethere.co.uk) left irc: Ping timeout: 260 seconds | 14:33 | |
| arigato | antocuni: it's easier to optimize operations on them, and the backend writes more compact code than with int_lshift/int_add | 14:35 |
| antocuni | yes, but why do we need tagged ints in the first place? | 14:35 |
| antocuni | to save memory? | 14:35 |
| arigato | that's for carl right now, in the Prolog interpreter | 14:36 |
| arigato | but it could be applied to pypy too | 14:36 |
| arigato | to store ints in instances | 14:36 |
| arigato | it is faster because we never need to allocate the int | 14:37 |
| arigato | and it doesn't come with the drawback of requiring everywhere a type check | 14:37 |
| antocuni | it's faster, in theory :-) | 14:37 |
| arigato | in fact it requires no type check at all, because it is information stored in the map | 14:37 |
| antocuni | haven't we already tried with tagged ints with no luck in the past? | 14:38 |
| arigato | so yes, it's faster in theory | 14:38 |
| arigato | it's very different for all reasons I just explained | 14:38 |
| antocuni | true | 14:38 |
| antocuni | well, we will see I suppose :-) | 14:38 |
| arigato | yes | 14:38 |
| EnCuKou (~encukou@193.167.65.94) joined #pypy. | 14:40 | |
| durin42 (~durin@173-15-14-174-Illinois.hfc.comcastbusiness.net) joined #pypy. | 14:44 | |
| durin42 (~durin@173-15-14-174-Illinois.hfc.comcastbusiness.net) left irc: Changing host | 14:44 | |
| durin42 (~durin@adium/durin42) joined #pypy. | 14:44 | |
| Remilia_Scarlet (proton@124.131.64.28) left #pypy. | 14:45 | |
| mvt (~mvantelli@87.213.45.85) left irc: Quit: Leaving | 14:58 | |
| lac (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) joined #pypy. | 15:01 | |
| lac_ (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) left irc: Ping timeout: 240 seconds | 15:02 | |
| kenaan | 12edelsohn ppc-jit-backend 11b0d393de7dd6 15/pypy/jit/backend/ppc/ppcgen/test/test_ppc.py: Update test_call_function for PPC64 to use load_from_addr | 15:18 |
| kenaan | 12edelsohn ppc-jit-backend 11128dccb72575 15/pypy/jit/backend/ppc/ppcgen/codebuilder.py: Update load_from_addr and store_reg for PPC64 | 15:18 |
| kenaan | 12edelsohn ppc-jit-backend 1190737207852a 15/pypy/jit/backend/ppc/ppcgen/opassembler.py: Add PPC64 support to emit_setfield_gc and emit_getfield_gc. Correct typo in emit_setfield_gc for size == 4. | 15:18 |
| aleksi (~aleksi@85.235.191.82) left irc: Remote host closed the connection | 15:27 | |
| witulski (~stupsi@134.99.16.23) joined #pypy. | 15:44 | |
| voidspace (~anonymous@python/psf/voidspace) left irc: Quit: voidspace | 15:46 | |
| ltratt (~ltratt@87.113.118.244) joined #pypy. | 15:46 | |
| voidspace (~voidspace@python/psf/voidspace) joined #pypy. | 15:48 | |
| kenaan | 12hager ppc-jit-backend 11155a57d1d251 15/pypy/jit/backend/ppc/ppcgen/: Implemented basic array operations. | 15:48 |
| kenaan | 12hager ppc-jit-backend 1138f0e1f8b91e 15/pypy/jit/backend/ppc/ppcgen/: merge | 15:48 |
| witulski (stupsi@134.99.16.23) left #pypy. | 15:48 | |
| kenaan | 12arigo default 11054d434f5e82 15/pypy/jit/: Support casts between floats and (u)longlongs written as a force_cast. | 15:49 |
| ojii (~ojii@40-34.60-188.cust.bluewin.ch) left irc: Quit: Leaving | 15:49 | |
| cfbolz (~cfbolz@fwstups.cs.uni-duesseldorf.de) joined #pypy. | 15:51 | |
| ltratt | Hello all. I have a stupid question. | 15:54 |
| dmalcolm (david@nat/redhat/x-aryquvcngwunikfv) joined #pypy. | 15:55 | |
| mcdonc (~mcdonc@cabana.palladion.com) joined #pypy. | 15:55 | |
| ltratt | Is it possible to access os.path.realpath, or an equivalent filename canonicalising function, from RPython? | 15:55 |
| Alex_Gaynor | ltratt: at least some portions of os.path are RPython, let me see if I can find out which | 15:57 |
| gutworth | thought this is an accident... | 15:58 |
| Alex_Gaynor | gutworth: nope, look at ll_os_path.py | 15:58 |
| rekamso (~textual@67.51.82.66) joined #pypy. | 15:59 | |
| ltratt | Alex_Gaynor: I fear realpath isn't RPython | 15:59 |
| Alex_Gaynor | it doesn't appear to be | 15:59 |
| Alex_Gaynor | FWIW the pure python version looks to be almost RPython, if you futzed with it you could proibably make it work | 16:00 |
| ltratt | So does that mean editing the version in lib-python/2.7/posixpath.py? Or copying it somewhere else and editing that version? | 16:01 |
| Arfrever (~Arfrever@apache/committer/Arfrever) joined #pypy. | 16:03 | |
| gutworth | Alex_Gaynor: I was referring to "import os.path" | 16:03 |
| cfbolz | Alex_Gaynor: except for the windows version | 16:03 |
| G2P (~G2P@fw-asn1.ornis.com) left irc: Quit: Leaving. | 16:04 | |
| Alex_Gaynor | ltratt: copying it somewhere else and using that version | 16:05 |
| EnCuKou (~encukou@193.167.65.94) left irc: Remote host closed the connection | 16:05 | |
| Alex_Gaynor | cfbolz: Do you have an opinion on the two test failures I pushed the other day? | 16:05 |
| cfbolz | Alex_Gaynor: link? | 16:06 |
| Alex_Gaynor | http://buildbot.pypy.org/summary/longrepr?testname=TestLLtype.%28%29.test_ptr_eq_str_constant&builder=own-linux-x86-32&build=1802&mod=jit.metainterp.optimizeopt.test.test_optimizebasic | 16:06 |
| fschwidom (~fschwidom@2.213.71.29) joined #pypy. | 16:06 | |
| Alex_Gaynor | http://buildbot.pypy.org/summary/longrepr?testname=TestLLtype.%28%29.test_ptr_eq_str_constants&builder=own-linux-x86-32&build=1802&mod=jit.metainterp.test.test_ajit | 16:06 |
| ltratt | Alex_Gaynor: thanks. I will have a look and if I can work out where to ;) | 16:07 |
| mcdonc (~mcdonc@cabana.palladion.com) left irc: Ping timeout: 258 seconds | 16:07 | |
| cfbolz | Alex_Gaynor: apart from "unroll.py is a frickin' mess and my eyes bleed whenever I look at it"? | 16:08 |
| Alex_Gaynor | cfbolz: ;) the bug is unrelated to unroll.py, the issue is the PTR_{EQ,NE} optimizer tries to get the RPython class of any constant, not just rclass instances | 16:08 |
| cfbolz | yes yes | 16:08 |
| cfbolz | just saw it in the stack trace | 16:09 |
| cfbolz | well, it's clearly a bug | 16:09 |
| cfbolz | you cannot expect everything pointer to have a class | 16:09 |
| Alex_Gaynor | I started working on http://paste.pocoo.org/show/497402/ | 16:10 |
| Alex_Gaynor | but I"m not really convinced this is correct | 16:10 |
| cfbolz | Alex_Gaynor: no, this is the wrong level. really, what needs to be fixed is the optimizer | 16:10 |
| cfbolz | in a fairly fundamental way it seems | 16:10 |
| Alex_Gaynor | how would you fix it though? I guess we still want this optimization if we know that they're both instances | 16:11 |
| Alex_Gaynor | I think armin suggested maybe having INSTANCE_PTR_{EQ,NE} and PTR_{EQ,NE} | 16:11 |
| cfbolz | would work | 16:12 |
| cfbolz | but the whole concept of LEVEL_* is broken | 16:12 |
| cfbolz | imo | 16:12 |
| cfbolz | for ptrs that are not instances | 16:13 |
| Alex_Gaynor | probably, there's no way to refer to "rpy_string" type right? | 16:14 |
| cfbolz | no | 16:15 |
| Alex_Gaynor | hmm | 16:16 |
| cfbolz | Alex_Gaynor: the problem is really that getting the class is reading the first word of the object, which is not a vtable in the case of strings | 16:16 |
| cfbolz | anyway, I'm going home now, sorry | 16:17 |
| Alex_Gaynor | ok, I guess I'll try doing an INSTANCE_PTR_EQ solution | 16:17 |
| cfbolz | Alex_Gaynor: as I said, that's just dealing with the symptoms | 16:17 |
| Alex_Gaynor | do you have a solution that targets the real problem? | 16:17 |
| cfbolz | no | 16:18 |
| Alex_Gaynor | me neither :) | 16:18 |
| cfbolz | needs a thorough review, or something | 16:18 |
| cfbolz (~cfbolz@fwstups.cs.uni-duesseldorf.de) left irc: Quit: Leaving | 16:18 | |
| Alex_Gaynor | arigato: do you have an opinoin on which approach for the PTR_EQ issue? | 16:25 |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) left irc: Quit: Leaving | 16:25 | |
| stakkars_ (~tismer@82.113.121.133) left irc: Quit: schnarch | 16:28 | |
| arigato | INSTANCE_EQ seems the right solution | 16:31 |
| dgl (~dgl@109.86.165.231) joined #pypy. | 16:31 | |
| Alex_Gaynor | ok | 16:31 |
| JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection | 16:32 | |
| mcdonc (~mcdonc@cabana.palladion.com) joined #pypy. | 16:32 | |
| Alex_Gaynor | arigato: at which level, do they need a seperate llop as well, or just pyjitpl op? | 16:33 |
| arigato | just pyjitpl | 16:33 |
| Alex_Gaynor | ok | 16:33 |
| arigato | in jtransform | 16:33 |
| Alex_Gaynor | sounds good | 16:33 |
| fschwidom (~fschwidom@2.213.71.29) left irc: Ping timeout: 245 seconds | 16:35 | |
| kenaan | 12hager ppc-jit-backend 11064df4d72ec4 15/pypy/: Fixed hack in treating bools. | 16:39 |
| kenaan | 12hager ppc-jit-backend 1192cd323471d9 15/pypy/jit/backend/ppc/ppcgen/regalloc.py: Remove comment | 16:40 |
| davisagli (~davisagli@davisagli.com) left irc: Excess Flood | 16:52 | |
| davisagli (~davisagli@davisagli.com) joined #pypy. | 16:53 | |
| rekamso (~textual@67.51.82.66) left irc: Ping timeout: 258 seconds | 16:56 | |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 16:56 | |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer | 16:59 | |
| arigato | Alex_Gaynor: while you're at it, you should review also {GET,SET}FIELD_GC_PURE | 17:01 |
| arigato | and non-PURE | 17:01 |
| Alex_Gaynor | arigato: in what sense? | 17:01 |
| Action: Alex_Gaynor has been procrastinating a bit | 17:01 | |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 17:01 | |
| arigato | to know if the optimizer assumes that it has a class just because it's used in a GETFIELD_GC | 17:01 |
| arigato | which is wrong in general | 17:02 |
| Alex_Gaynor | AFAIK it only assumes it is nonnull | 17:03 |
| arigato | yes, but better check than assuming :-) | 17:03 |
| stakkars_ (~tismer@p5DDB5E98.dip.t-dialin.net) joined #pypy. | 17:06 | |
| stakkars (~tismer@p5DDB5E98.dip.t-dialin.net) left irc: Read error: Connection reset by peer | 17:06 | |
| Nick change: stakkars_ -> stakkars | 17:06 | |
| Alex_Gaynor | arigato: AFAIK we don't have a state of "it is an instance", just "it is nonnull" and "it is a specific class" | 17:07 |
| EnCuKou (~encukou@a91-156-243-246.elisa-laajakaista.fi) joined #pypy. | 17:07 | |
| etrepum (~bob@75-101-96-144.dsl.static.sonic.net) left irc: Quit: etrepum | 17:08 | |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Ping timeout: 260 seconds | 17:09 | |
| arigato | "being an instance" should not really be a state | 17:11 |
| amaury_ (amaury_@nat/google/x-mtsptvyaitsaqitq) left irc: Ping timeout: 244 seconds | 17:11 | |
| Alex_Gaynor | arigato: well, I don't htink it is, so I guess we're safe :) | 17:11 |
| arigato | what I'm saying above is: you should check that it's not doing stuff like | 17:12 |
| arigato | "oh, we're reading or writing field "X.name" in object P, so I know that the class of P is at least X" | 17:12 |
| arigato | I'm not too sure how helpful this optimization would be, nor how exactly you would do it, so I guess it's not done | 17:13 |
| arigato | but I didn't check recently :-) | 17:13 |
| kushal (kdas@fedora/kushal) left #pypy ("Leaving"). | 17:14 | |
| arigato | argh | 17:16 |
| Alex_Gaynor | arigato: what's the best way to write "is this type an rclass.OBJECT ptr" | 17:18 |
| mtigas (~Adium@users.spokesman.com) joined #pypy. | 17:19 | |
| ltratt (~ltratt@87.113.118.244) left irc: Ping timeout: 245 seconds | 17:20 | |
| arigato | "lltype._castdepth(rclass.OBJECT, T) >= 0"... | 17:21 |
| Alex_Gaynor | ok | 17:22 |
| arigato | er, no | 17:22 |
| arigato | ..."lltype._castdepth(T, rclass.OBJECT) >= 0" | 17:22 |
| Alex_Gaynor | arigato: is T = v.concretetype or v.concretetype.TO? | 17:25 |
| Alex_Gaynor | .TO apparently | 17:26 |
| arigato | yes | 17:27 |
| stakkars_ (~tismer@89.204.153.80) joined #pypy. | 17:27 | |
| Action: arigato -> off | 17:28 | |
| kenaan | 12arigo concurrent-marksweep 1192beb8ad501b 15/pypy/rpython/memory/gc/concurrentgen.py: Tweaks. | 17:29 |
| kenaan | 12arigo concurrent-marksweep 11229af044431a 15/pypy/rpython/memory/gctransform/framework.py: It seems that not using "borrowed vars" is much more efficient. | 17:29 |
| kenaan | 12arigo concurrent-marksweep 11b661552e1f80 15/pypy/rpython/memory/gc/concurrentgen.py: Move this outside the collector thread, to ensure that it is correctly nested in a debug_print("gc-*") | 17:29 |
| arigato (~arigo@fwstups.cs.uni-duesseldorf.de) left irc: Quit: See you | 17:29 | |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 17:35 | |
| etrepum (~bob@accessnat4.mochimedia.net) joined #pypy. | 17:37 | |
| ramusara (~ramusara@220.156.210.236.user.e-catv.ne.jp) left irc: Quit: Leaving... | 17:41 | |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) joined #pypy. | 17:47 | |
| dgl (~dgl@109.86.165.231) left irc: Quit: Leaving... | 17:51 | |
| JaRoel|4d (~jaroel|4d@2001:980:7133:1:741b:d3c2:3098:7dfa) joined #pypy. | 17:59 | |
| fijal (~fijal@78.9.163.32) joined #pypy. | 18:04 | |
| fijal | hi | 18:05 |
| fijal | Alex_Gaynor: I think I have a nice idea how to vectorize stuff | 18:08 |
| kenaan | 12fijal lightweight-finalizers 116949b37f7cb8 15/pypy/rpython/memory/gc/minimark.py: rewrite obvious nonsense, needs tests | 18:09 |
| kenaan | 12fijal default 119ea261642156 15/pypy/pytest.ini: add some options | 18:10 |
| kenaan | 12fijal default 1176d0bea6c4e7 15/pypy/: merge heads | 18:10 |
| teknico (~quassel@88-149-209-113.dynamic.ngi.it) left irc: Remote host closed the connection | 18:10 | |
| Alex_Gaynor | fijal: oh yeah? | 18:12 |
| fschwidom (~fschwidom@2.213.45.31) joined #pypy. | 18:16 | |
| fijal | so the idea is to have an optimizer pass that recognizes a couple of consecutive operations | 18:17 |
| fijal | and fold them into one vector operation | 18:17 |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer | 18:17 | |
| fijal | and then unroll the loop N times, where N is the width of vectorizer operations | 18:17 |
| Alex_Gaynor | fijal: sounds very sane, I guess you should read papers on how GCC or llvm (if it does) do it | 18:18 |
| fijal | we don't want to do what they do | 18:18 |
| fijal | for a couple of reasons | 18:18 |
| fijal | a) they recognize the entire loop | 18:18 |
| Alex_Gaynor | I guess you need to make sure you have no guards between operations | 18:18 |
| fijal | yes | 18:18 |
| Rhy0lite | fijal: GCC also recognizes straight line parallelism (SLP) | 18:18 |
| fijal | b) they bail out if soemthing is suspicious (like mod) | 18:19 |
| Alex_Gaynor | fijal: FWIW I think there are now 0 guards for signals, so that helps if you want to do this on pure python stuff | 18:19 |
| fijal | while in my model you would simply end up with an unrolled loop for suspect operations | 18:19 |
| fijal | Rhy0lite: oh nice | 18:19 |
| fijal | Alex_Gaynor: I don't care about pure python | 18:19 |
| fijal | thanks for merging inline-dict-ops | 18:19 |
| Alex_Gaynor | fijal: np, it had bugs with virtuals, I don't know if you saw | 18:20 |
| fijal | Alex_Gaynor: reading pypy-svn now | 18:20 |
| Alex_Gaynor | fijal: well if you write this as a JIT optimizer, instead of special rpython classes, it eventually becomes possible | 18:20 |
| etrepum (~bob@accessnat4.mochimedia.net) left irc: Read error: No route to host | 18:20 | |
| Alex_Gaynor | fijal: I disabled virtuals of arrays-of-structs for now, I'll fix on a branch | 18:20 |
| fijal | yes, but probably also very unlikely | 18:20 |
| etrepum (~bob@accessnat4.mochimedia.net) joined #pypy. | 18:20 | |
| fijal | they didn't work anyway right? | 18:20 |
| fijal | as in any get/set interiorfield destroyed them | 18:20 |
| Alex_Gaynor | fijal: yes, but if you had an empty virtual array of structs and you hit blackhole it exploded | 18:21 |
| fijal | Rhy0lite: I think what I have in mind is in-line with paper you presented, but more meta | 18:21 |
| Rhy0lite | probably | 18:21 |
| Rhy0lite | the paper also might not be completely up to date | 18:21 |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 18:21 | |
| Rhy0lite | fijal: http://gcc.gnu.org/projects/tree-ssa/vectorization.html#slp | 18:22 |
| fijal | yeah, I read that | 18:22 |
| Alex_Gaynor | fijal: ATM I'm shaving yaks about PTR_EQ optimizations :/ | 18:22 |
| fijal | Alex_Gaynor: ok :) | 18:22 |
| fijal | Alex_Gaynor: I found the bug in lightweight finalizers | 18:22 |
| Alex_Gaynor | fijal: did you fix it? | 18:22 |
| fijal | yes | 18:22 |
| Rhy0lite | fijal: I think traditional SLP is the second phase of your algorithm | 18:23 |
| Rhy0lite | recognizing sequences after unrolling | 18:23 |
| fijal | Rhy0lite: ok | 18:23 |
| Rhy0lite | you additionally want to collect instructions and unroll them to the vectorization factor | 18:23 |
| fijal | I'm wondering what to do with tails/heads | 18:24 |
| Rhy0lite | fijal: with runtime analysis, you should be able to decide when it is profitable | 18:24 |
| fijal | we don't yet do this kind of stuff | 18:24 |
| Rhy0lite | enough iterations for unrolling + head/tail peeling to be profitable | 18:24 |
| fijal | like "compile and see how it performs" | 18:24 |
| Rhy0lite | why compile it? | 18:25 |
| Rhy0lite | you should be able to count iterations and instructions | 18:25 |
| fijal | yes I guess | 18:25 |
| fijal | but not instruction cache misses | 18:25 |
| fijal | or so | 18:25 |
| Rhy0lite | data cache misses or instruction cache misses? | 18:25 |
| Alex_Gaynor | if we unroll arbitrary loops icache misses is the conern | 18:25 |
| fijal | instruction, data is real time anyway | 18:25 |
| Rhy0lite | Alex: if it's hot, that should be overcome, no? | 18:26 |
| Rhy0lite | one would not want to do this unless the trace already was determined to be hot | 18:26 |
| Alex_Gaynor | we don't have relative hotness of traces :) | 18:26 |
| Rhy0lite | Alex: it's a simple matter of engineering ;-P | 18:27 |
| Alex_Gaynor | http://www.cs.purdue.edu/homes/dec/essay.criticize.html | 18:27 |
| Rhy0lite | fijal: you also can add guards to bail if the loop does not have enough loops or is not sufficiently aligned | 18:28 |
| fijal | it's always aligned | 18:29 |
| fijal | the cool thing is we might simply want to run the interpreter at the beginning/end | 18:29 |
| fijal | so we don't compile extra assembler for it | 18:30 |
| Rhy0lite | yes | 18:30 |
| fijal | and we can even do more cool funky stuff ;-) | 18:30 |
| fijal | for small arrays | 18:30 |
| Alex_Gaynor | fijal: for the last 7 iterations or so? | 18:30 |
| fijal | yes | 18:31 |
| Alex_Gaynor | cool | 18:31 |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Ping timeout: 240 seconds | 18:31 | |
| fijal | so it'll become something like | 18:31 |
| Alex_Gaynor | while num_iters > 8: do_vector_magic(); finish_in_interp | 18:32 |
| asmeurer_ (~asmeurer@64.168.229.50) joined #pypy. | 18:33 | |
| fijal | http://paste.pocoo.org/show/497499/ | 18:33 |
| Alex_Gaynor | why not do the unroll in the jit? | 18:34 |
| Rhy0lite (dje@nat/ibm/x-losizlekoquzszlt) left irc: Quit: Leaving | 18:35 | |
| fijal | "in the jit" being? | 18:35 |
| fijal | the optimizer? | 18:35 |
| Alex_Gaynor | yes | 18:36 |
| Alex_Gaynor | especially since different vector ops have different widths I think? | 18:36 |
| fijal | yes they do | 18:37 |
| fijal | but you can have a constant/variable from the jit somewhere | 18:37 |
| fijal | meh, it's a mess in the optimizer | 18:37 |
| fijal | comparable to short preamble mess | 18:37 |
| Alex_Gaynor | carl was complaining about said mess earlier today | 18:37 |
| santagada (~leonardo@201.86.242.147.dynamic.adsl.gvt.net.br) left irc: Ping timeout: 240 seconds | 18:38 | |
| CIA-25 | 03stefanor 07roundup * 10#920/Lost data with async I/O: | 18:39 |
| CIA-25 | [new] The attached file, async_sub.py, is a slightly modified for debugging copy, of a | 18:39 |
| CIA-25 | file of the same name in pygame, which is a descendant of ... * 14https://bugs.pypy.org/issue920 | 18:39 |
| CIA-25 | 03stefanor 07roundup * 10#920/Lost data with async I/O: | 18:41 |
| CIA-25 | [chatting] I have a fix in progress in https://bitbucket.org/stefanor/pypy (see | 18:41 |
| CIA-25 | https://bitbucket.org/pypy/pypy/pull-request/12/dont-lose-data-w ... * 14https://bugs.pypy.org/issue920 | 18:41 |
| asmeurer_ (~asmeurer@64.168.229.50) left irc: Quit: asmeurer_ | 18:41 | |
| fijal | Alex_Gaynor: so I would refrain from doing that for now | 18:43 |
| Alex_Gaynor | okey dokey | 18:43 |
| fijal | also it would be relatively hard to reuse interpreter for tail | 18:44 |
| Alex_Gaynor | you could use blackhole intepreter I guess | 18:46 |
| Alex_Gaynor | that's slow though | 18:46 |
| fijal | when it comes to writing a few extra lines in numpy vs refactoring the entire jit to do what I want.... | 18:48 |
| fijal | guess how would i decide | 18:48 |
| Alex_Gaynor | refactor the entire jit? | 18:48 |
| Alex_Gaynor | :) | 18:48 |
| timonator | sounds like a good plan | 18:52 |
| etrepum (~bob@accessnat4.mochimedia.net) left irc: Remote host closed the connection | 18:53 | |
| etrepum (~bob@accessnat4.mochimedia.net) joined #pypy. | 18:53 | |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) joined #pypy. | 18:54 | |
| stakkars_ (~tismer@89.204.153.80) left irc: Ping timeout: 240 seconds | 18:55 | |
| stakkars_ (~tismer@89.204.153.109) joined #pypy. | 18:56 | |
| kkris (~kris@80-123-32-36.adsl.highway.telekom.at) joined #pypy. | 18:58 | |
| fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy | 19:02 | |
| Alex_Gaynor | mitsuhiko: is paste.pocoo down? | 19:09 |
| antocuni (~antocuni@host152-120-dynamic.11-79-r.retail.telecomitalia.it) left irc: Ping timeout: 260 seconds | 19:09 | |
| Alex_Gaynor | fijal: feel like reviewing a yak? http://paste.pocoo.org/show/497520/ | 19:09 |
| mitsuhiko | works here Alex_Gaynor | 19:10 |
| fijal | Alex_Gaynor: and what;'s the problem? | 19:11 |
| Alex_Gaynor | mitsuhiko: i guess my internet was randomly slow | 19:11 |
| Alex_Gaynor | fijal: the PTR_EQ optimizer tries to take the class of any thing it sees, including things that aren't rclass instances | 19:11 |
| fijal | pff | 19:11 |
| Alex_Gaynor | armin suggested introducing PTR_EQ vs. INSTANCE_PTR_EQ | 19:12 |
| fijal | return lltype._castdepth(v.concretetype.TO, rclass.OBJECT) >= 0 | 19:12 |
| fijal | this is not RPython I bet | 19:12 |
| fijal | oh jtransform | 19:12 |
| Alex_Gaynor | yup | 19:15 |
| fijal | Alex_Gaynor: LGTM? | 19:15 |
| Alex_Gaynor | fijal: cool, thanks | 19:16 |
| kenaan | 12alex_gaynor default 112d1620b150fa 15/pypy/jit/: (alex, armin/carl for idea and discussion, fijal review) create instance_ptr_{eq,ne} resops, which are for co... | 19:18 |
| kenaan | 12alex_gaynor default 114c9bd9aa3b11 15/pypy/pytest.ini: merged default | 19:18 |
| aat (~aat@cpe-72-225-174-173.nyc.res.rr.com) left irc: Quit: Computer has gone to sleep. | 19:21 | |
| fzzzy (~donovan@nat/mozilla/x-plliwhrgiombkrjl) joined #pypy. | 19:22 | |
| kenaan | 12alex_gaynor virtual-dicts 1171c19063db1e 15/pypy/: merged in default | 19:22 |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy. | 19:26 | |
| Alex_Gaynor | fijal: time for class, then more virtual-array-of-structs, it already works if you don't have guards :) | 19:28 |
| fijal | cool | 19:28 |
| mtigas (~Adium@users.spokesman.com) left irc: Quit: Leaving. | 19:28 | |
| Action: Alex_Gaynor supposes guards are kind of a common case | 19:28 | |
| fijal | I'll probably sleep | 19:28 |
| Alex_Gaynor | sounds fun | 19:28 |
| mtigas (~Adium@users.spokesman.com) joined #pypy. | 19:28 | |
| fijal | SSE sounds fun too though :) | 19:29 |
| stakkars_ (~tismer@89.204.153.109) left irc: Quit: schnarch | 19:33 | |
| stakkars_ (~tismer@89.204.153.109) joined #pypy. | 19:39 | |
| stakkars_ (~tismer@89.204.153.109) left irc: Ping timeout: 240 seconds | 19:43 | |
| Alex_Gaynor | fijal: so work on that, or features for numpy, or wahtever | 19:45 |
| stakkars_ (~tismer@89.204.137.79) joined #pypy. | 19:45 | |
| canta (~canta@77-20-123-240-dynip.superkabel.de) left irc: Quit: Leaving. | 19:48 | |
| Alex_Gaynor | hakanardo: ping | 19:52 |
| kenaan | 12alex_gaynor virtual-dicts 117e285a4099fa 15/pypy/jit/metainterp/optimizeopt/virtualize.py: progress | 19:52 |
| fschwidom (~fschwidom@2.213.45.31) left irc: Ping timeout: 256 seconds | 19:56 | |
| Trundle (~andy@89.204.137.162) joined #pypy. | 19:58 | |
| Trundle (~andy@89.204.137.162) left irc: Changing host | 19:58 | |
| Trundle (~andy@python/site-packages/trundle) joined #pypy. | 19:58 | |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Read error: Operation timed out | 19:58 | |
| mcdonc (~mcdonc@cabana.palladion.com) left irc: Ping timeout: 258 seconds | 19:59 | |
| fzzzy (~donovan@nat/mozilla/x-plliwhrgiombkrjl) left irc: Ping timeout: 260 seconds | 20:01 | |
| tilgovi (~randall@199.36.158.16) joined #pypy. | 20:06 | |
| tilgovi (~randall@199.36.158.16) left irc: Changing host | 20:06 | |
| tilgovi (~randall@couchdb/developer/tilgovi) joined #pypy. | 20:06 | |
| fijal | meh | 20:06 |
| Action: fijal is unsure about lightweight-finalizers | 20:06 | |
| Alex_Gaynor | fijal: do you understand virtualstate code? | 20:06 |
| fijal | I don't think so | 20:07 |
| fijal | I can try | 20:07 |
| kkris (~kris@80-123-32-36.adsl.highway.telekom.at) left irc: Remote host closed the connection | 20:07 | |
| Alex_Gaynor | if you feel like it, I need to write a new one for virtual array of structs, I have no idea what it should look like | 20:07 |
| fzzzy (~donovan@nat/mozilla/x-cldjexracoxlngze) joined #pypy. | 20:07 | |
| fijal | no idea | 20:09 |
| fijal | test/test_zjit.py Segmentation fault | 20:11 |
| fijal | a bit random | 20:11 |
| fijal | given it's a micronumpy test | 20:11 |
| voidspace (voidspace@python/psf/voidspace) left #pypy. | 20:12 | |
| fijal | hahaha | 20:12 |
| fijal | overflowed C stack | 20:12 |
| Alex_Gaynor | what did you do? | 20:14 |
| fijal | messed up tracing somehow | 20:14 |
| fijal | it did not close | 20:14 |
| fijal | the loop | 20:14 |
| kenaan | 12fijal sse-vectorization 11d1e954a0e84e 15/pypy/module/micronumpy/: break test zjit. This gives traces roughly how I would like them. | 20:18 |
| etrepum (~bob@accessnat4.mochimedia.net) left irc: Quit: etrepum | 20:20 | |
| etrepum (~bob@accessnat4.mochimedia.net) joined #pypy. | 20:21 | |
| felipejcruz (~felipecru@177.19.158.201) joined #pypy. | 20:23 | |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) joined #pypy. | 20:32 | |
| mat^2 (~mathias@212.130.113.35) joined #pypy. | 20:33 | |
| mcdonc (~mcdonc@cabana.palladion.com) joined #pypy. | 20:38 | |
| kenaan | 12alex_gaynor virtual-dicts 11c1ae43540abe 15/pypy/jit/metainterp/optimizeopt/virtualstate.py: small amount of progress | 20:41 |
| kenaan | 12alex_gaynor virtual-dicts 113f982c63e966 15/pypy/jit/metainterp/optimizeopt/virtualstate.py: whitespace | 20:41 |
| aboudreault (~alanb@osgeo/member/aboudreault) left irc: Ping timeout: 240 seconds | 20:44 | |
| gtaylor (~gtaylor@108-196-160-120.lightspeed.tukrga.sbcglobal.net) left irc: Remote host closed the connection | 21:00 | |
| jonanin (~jonanin@24-183-50-140.dhcp.mdsn.wi.charter.com) joined #pypy. | 21:03 | |
| kenaan | 12alex_gaynor virtual-dicts 11ae6a1aa152ae 15/pypy/jit/metainterp/optimizeopt/virtualize.py: better error | 21:08 |
| fijal (~fijal@78.9.163.32) left irc: Ping timeout: 256 seconds | 21:13 | |
| lac | Alex_Gaynor: so what are virtual-dicts? | 21:13 |
| Alex_Gaynor | lac: it just means dicts where all the keys it sees are constants are kept virtual in teh dicts, it's basically an extension of virtual arrays and instances | 21:14 |
| lac | aha. neat. | 21:14 |
| kenaan | 12alex_gaynor virtual-dicts 11325af29a17b8 15/pypy/jit/metainterp/: progress I think | 21:21 |
| gutworth | that's optimistic | 21:21 |
| Alex_Gaynor | gutworth: how so? | 21:22 |
| Rhyolite | "progress" | 21:22 |
| Alex_Gaynor | ah, yes :) | 21:22 |
| Alex_Gaynor | It's definitely progress, it might even be forward progress | 21:22 |
| gutworth | dvcs make it very easy to read over someone's shoulder :) | 21:22 |
| Alex_Gaynor | :) | 21:23 |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy. | 21:23 | |
| Alex_Gaynor | gutworth: feel like implementing my stuff for me- over my shouldder | 21:26 |
| gutworth | wouldn't you rather I whisper the secrets in your ear? :) | 21:27 |
| Alex_Gaynor | gutworth: do you understand the secrets of resume data? | 21:27 |
| gutworth | asuming it hasn't changed since last month | 21:28 |
| Alex_Gaynor | I wouldn't assume that | 21:28 |
| gutworth | are you trying to get some test to work? | 21:29 |
| Alex_Gaynor | gutworth: well I'm trying ot make virtuals work with arrays of structs | 21:30 |
| `fox` (~fox@host229-107-dynamic.45-79-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds | 21:41 | |
| fzzzy (~donovan@nat/mozilla/x-cldjexracoxlngze) left irc: Quit: fzzzy | 21:41 | |
| witulski (~stupsi@176.198.244.135) joined #pypy. | 21:43 | |
| fzzzy (~donovan@nat/mozilla/x-ovzaevegwxixfoxt) joined #pypy. | 21:44 | |
| felipejcruz (~felipecru@177.19.158.201) left irc: Quit: felipejcruz | 21:46 | |
| fzzzy (~donovan@nat/mozilla/x-ovzaevegwxixfoxt) left irc: Quit: fzzzy | 21:51 | |
| Action: Rhyolite is waiting for Alex_Gaynor to say virtual-dicts virtually works | 21:54 | |
| fzzzy (~donovan@nat/mozilla/x-jrdavqovezmxecfh) joined #pypy. | 21:54 | |
| EnCuKou (~encukou@a91-156-243-246.elisa-laajakaista.fi) left irc: Quit: Mizím. | 21:56 | |
| Alex_Gaynor | Rhyolite: they work fine, as long as you have no guards :) | 21:57 |
| timonator | does that mean that guards cause misbehavior or that guards cause pypy to use "real" dicts instead? | 21:59 |
| witulski (stupsi@176.198.244.135) left #pypy. | 22:00 | |
| Alex_Gaynor | no it means that guards crash :) | 22:01 |
| timonator | ok, that's not so good | 22:02 |
| Alex_Gaynor | well it only crashes on this branch | 22:06 |
| bbot2 (~bbot2@wyvern.cs.uni-duesseldorf.de) joined #pypy. | 22:07 | |
| mat^2 (~mathias@212.130.113.35) left irc: Read error: Connection reset by peer | 22:12 | |
| mat^2 (~mathias@212.130.113.35) joined #pypy. | 22:12 | |
| tlynn (~tlynn@cpc6-cmbg14-2-0-cust121.5-4.cable.virginmedia.com) joined #pypy. | 22:17 | |
| JaRoel|4d (~jaroel|4d@2001:980:7133:1:741b:d3c2:3098:7dfa) left irc: Remote host closed the connection | 22:23 | |
| kenaan | 12alex_gaynor virtual-dicts 11365af1bb7686 15/pypy/jit/metainterp/optimizeopt/virtualize.py: semantic fix | 22:27 |
| kenaan | 12alex_gaynor virtual-dicts 11a6e23621d2bf 15/pypy/jit/metainterp/optimizeopt/virtualstate.py: progress (these commit messages suck) | 22:27 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/917 | 22:37 |
| Trundle (~andy@python/site-packages/trundle) left irc: Quit: & | 22:37 | |
| stakkars_ (~tismer@89.204.137.79) left irc: Quit: schnarch | 22:38 | |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/114 | 22:38 |
| derdon (~derdon@pD9E1C2F0.dip.t-dialin.net) left irc: Remote host closed the connection | 22:48 | |
| kenaan | 12alex_gaynor default 11adab424acda7 15/pypy/jit/: fix | 22:49 |
| Alex_Gaynor | ugh, I'm having a terrible day for commit messages | 22:49 |
| Sentynel | Alex_Gaynor: there's always http://whatthecommit.com/ | 22:50 |
| kenaan | 12alex_gaynor virtual-dicts 117a95c14349c0 15/pypy/jit/: merge default in | 22:51 |
| Alex_Gaynor | Sentynel: second one I got was "FIX", HOW DID IT KNOW | 22:51 |
| mcdonc (~mcdonc@cabana.palladion.com) left irc: Remote host closed the connection | 22:51 | |
| Sentynel | haha awesome | 22:51 |
| whitelynx (~whitelynx@li117-47.members.linode.com) left irc: Quit: Ex-Chat | 22:59 | |
| amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 258 seconds | 23:01 | |
| mcdonc (~mcdonc@cabana.palladion.com) joined #pypy. | 23:02 | |
| Alex_Gaynor | gutworth: do you understand generalization_of? | 23:12 |
| kenaan | 12alex_gaynor virtual-dicts 11068b20a399e7 15/pypy/jit/metainterp/: we now have a representation of a list of fielddescrs, I don't understand what generalization_of is sup... | 23:12 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/918 [12alex] | 23:21 |
| bbot2 | Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/115 [12alex] | 23:21 |
| bbot2 | 4Failure: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/691 | 23:37 |
| tlynn (~tlynn@cpc6-cmbg14-2-0-cust121.5-4.cable.virginmedia.com) left irc: Ping timeout: 240 seconds | 23:51 | |
| mtigas (~Adium@users.spokesman.com) left irc: Quit: Leaving. | 23:58 | |
| fzzzy (~donovan@nat/mozilla/x-jrdavqovezmxecfh) left irc: Quit: fzzzy | 23:59 | |
| --- Tue Oct 25 2011 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!