Camelia, the Perl 6 bug

IRC log for #parrot, 2012-03-29

Parrot | source cross referenced

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:00 kid51 joined #parrot
00:09 whiteknight dukeleto?
00:10 whiteknight which one, from jaswanth?
00:10 whiteknight That email from him was just an introduction, a real proposal is coming soon, I'm told
00:10 whiteknight and I can send an update about PLA to the list
00:14 schmooster joined #parrot
01:04 jashwanth joined #parrot
01:14 jashwanth dukeleto:hello
01:22 whiteknight bacek: ping
01:24 whiteknight msg bacek_at_work dalek is missing. Is that something you can fix?
01:24 aloha OK. I'll deliver the message.
01:33 bacek_at_work whiteknight, I can't. But afaik Infinoid or sorear can.
01:34 sorear the machine dalek is hosted on is being decommissioned.
01:34 sorear moritz++ is moving all the services to feather3
01:35 whiteknight ah, okay. That explains it
01:35 whiteknight I can never remember who runs that bot
01:35 sorear with help from diakopter++ and a tiny bit from me
01:51 benabik joined #parrot
01:57 benabik ~~
02:44 kid51 joined #parrot
03:03 bacek_at_work joined #parrot
03:46 jsut_ joined #parrot
05:57 fperrad joined #parrot
06:05 mj41 joined #parrot
06:43 dalek joined #parrot
06:45 dalek joined #parrot
06:47 dalek joined #parrot
07:13 nine moritz: yes the mandelbrot you posted on parrot-dev when we merged green_threads. It's totally weird. Results vary by several seconds. Just now I did an init S to have a completely quiet system and now the benchmark takes about 53 seconds compared to 44 yesterday evening.
07:15 moritz nine: without any code changes?
07:19 sorear msg whiteknight dalek lives
07:19 aloha OK. I'll deliver the message.
07:41 moritz an admin of parrot.org/github/ needs to change the host name of the post-commit hook
07:41 moritz it's now feather3.perl6.nl
07:42 wagle joined #parrot
07:43 sorear ITYM github.org/parrot
07:45 mj41 joined #parrot
07:46 moritz erm, yes, other way round :-)
08:01 lucian joined #parrot
08:32 nine moritz: yep, without any changes whatsoever using the same perl6 binary
08:34 nine moritz: and now with a running desktop I'm back to 44 seconds. I'm baffled
08:35 moritz maybe that runmode disables some memory management techniques, or caching, or something?
08:36 nine maybe turbo mode?
08:36 moritz my PC doesn't have that button anymore :-)
08:39 nine If it was a button, I could at least make sure it was pressed all the time :)
08:40 moritz anyway, the mandelbrot benchmark worked pretty well for me in the past, though it wasn't really designed as a benchmark
08:40 tadzik and then wings of fury will run too fast and the plane will not take off :(
08:41 moritz if you find something better, be sure to let me know
08:41 nine Another question: do I actually have to rebuild rakudo to test parrot's performance?
08:43 moritz hm, maybe not; libparrot.so is dynamically linked
08:43 moritz though of course rebuilding rakudo is also a parrot benchmark with RL relevancy
08:44 nine not having to rebuild rakudo would save lots of time testing different implementations
08:53 alvis`` joined #parrot
09:10 Maddingu1 joined #parrot
09:15 nine Seems like small changes do work, but when switching branches I have to rebuild rakudo
09:18 moritz then it might make sense to have a separate copy of parrot+rakudo for each branch that you work on regularly
09:20 nine indeed
09:24 mdupont joined #parrot
09:30 bacek ~~
09:47 bacek msg whiteknight I started new branch pcc_reorder_signatures to just split Singatures from CallContext. Actual reorder of CallContext creation will be done later after this one.
09:47 aloha OK. I'll deliver the message.
09:57 moritz didn't we unify the two earlier to get less GC pressure?
10:02 bacek moritz, yes. I did it. But new Parrot_Signature isn't PMC.
10:02 bacek and it's 1-to-1 owned by CallContext.
10:03 bacek So, it will increase memory pressure _slightly_
10:03 bacek But it will allow to allocate CallContext on callee side when needed.
10:03 bacek Huge win for tailcalls for example
10:03 bacek Or C methods
10:03 bacek When we don't need separated CallContext.
10:07 PacoAir joined #parrot
10:08 moritz ok
11:49 benabik joined #parrot
12:00 benabik ~
12:22 jlaire huh?
12:22 benabik jlaire: huh what?
12:23 jlaire benabik: why was I op'd
12:23 moritz huh. Things happen. Trust is placed.
12:25 jlaire nah, I'm just a lurker who's never contributed anything
12:25 jlaire but thanks :)
12:26 moritz jlaire: well, if you lurk long enough, we start to trust you :-)
12:28 whiteknight joined #parrot
12:28 whiteknight good morning, #parrot
12:29 benabik 'lo whiteknight
12:29 whiteknight hello benabik
12:29 whiteknight how are you doing today?
12:30 benabik whiteknight: Doing alright.  Lacking sleep.  Have to do homework and proposals.  :-D
13:35 benabik I hate nearly finishing an assignment before noticing a small comment near the end that says "be sure to use the $Id$ and $Log$ tags in your code".  Of course I used git, so I have a log, just not anything automatically inserted.
13:35 benabik I'm tempted to just take the `git log` output and munge it until it looks like CVS.
13:37 whiteknight that's a weird requirement. Did the teacher expect you to use CVS?
13:38 moritz SVN probably
13:38 benabik SCCS, RCS, or CVS.  The class is "Software Development Tools" and we just had a lecture on them, so I should have expected it.
13:38 benabik But that requirement was after the "how to submit" so I glossed over it.  :-(
13:39 moritz benabik: well, you can add $Id$ tags in git
13:39 moritz benabik: they just won't get expanded
13:39 benabik moritz: Well setting the ident attribute will expand the Id tag, but it won't help with the log.
13:40 moritz well, but even if they don't get expanded, you met the requirement to "use them in your code" :-)
13:40 benabik Hah!
13:50 szbalint People still teaching CVS are unfit to lecture :\
13:51 benabik It's important to learn what's come before.  And if you want to understand the theory behind SCM, then CVS is much simpler to explain.
13:51 whiteknight I had several teachers who insisted all the world would ever need is FORTRAN77
13:52 benabik My prof. doesn't insist that CVS is good, just that it was used and what it does.  :-)
13:52 whiteknight In my senior year, when we were compiling DOS programs in C using Borland 4.52, FORTRAN77 didn't look so bad
13:53 NotFound Real programmers can write Fortran in any language.
13:54 szbalint it's important to learn about version control but RCS/CVS is a bad, obsolete example
13:54 lateau joined #parrot
13:56 szbalint unlearning CVS takes some effort too, if you ever want to use proper branches and think about changesets, distributed repositories etc...
13:56 atrodo NotFound++
13:56 atrodo szbalint> I had an advantage when git came on the scene.  I never "learned" cvs, it never made sense to me.
13:58 whiteknight At my last job they used an old version of visual source safe, which was awesome until you had multiple people working on things at once
13:58 benabik VSS considered harmful
13:58 szbalint kind of defeats the purpose :)
13:59 whiteknight we switched over to svn when we got enough people in the office who understood the advantages
14:11 jsut joined #parrot
14:12 cosimo joined #parrot
14:57 nbrown joined #parrot
15:28 jashwanth joined #parrot
16:17 rich joined #parrot
16:27 Coke do we still need to update dalek's url for #parrot?
16:28 benabik Coke: probably
16:29 Coke szbalint: a case can be made that cvs is better at many things than subversion is.
16:29 Coke I see a post-receive URL of http://host04.appflux.net:​5000/dalek?t=magnet,parrot
16:32 benabik It apparently needs to be feather3.perl6.nl
16:32 benabik walk &
16:33 Coke same port?
16:33 * Coke checks perl6 github admin stuff.
16:34 Coke .. except I'm not a perl6 admin.
16:34 * Coke shrugs.
16:42 dukeleto ~~
16:42 Coke ~~
16:44 whiteknight ~~
16:51 nine ~~
16:57 benabik ~~
16:58 benabik Coke: All I know is that the host needs to change. moritz is the one who mentioned it here.
16:58 Coke benabik: aye.
16:59 moritz Coke: it needs to change to http://feather3.perl6.nl:5​000/dalek?t=magnet,parrot
17:01 Coke ok. I entered, hit "test", nothing happened. I clicked save.
17:06 elmex joined #parrot
17:08 rich joined #parrot
17:30 dmalcolm joined #parrot
17:32 dalek parrot/ayardley/parrot_github_release: 66a7335 | alvis++ | tools/release/parrot_github_release.pl (2 files):
17:32 dalek parrot/ayardley/parrot_github_release: Start work on parrot_github_release.pl
17:32 dalek parrot/ayardley/parrot_github_release: review: https://github.com/parrot/parrot/commit/66a7335650
17:32 dalek parrot/ayardley/parrot_github_release: 76b9b10 | alvis++ | tools/release/parrot_github_release.pl~:
17:32 dalek parrot/ayardley/parrot_github_release: Remove emacs' temporary working file.
17:32 dalek parrot/ayardley/parrot_github_release: review: https://github.com/parrot/parrot/commit/76b9b1015e
17:32 dalek parrot/ayardley/parrot_github_release: 177fc78 | alvis++ | docs/project/release_parrot_github_guide.pod:
17:32 dalek parrot/ayardley/parrot_github_release: Add 'docs/projecet/release_parrot_github_guide.pod' into the branch.
17:32 dalek parrot/ayardley/parrot_github_release: review: https://github.com/parrot/parrot/commit/177fc788b6
17:32 dalek parrot/ayardley/parrot_github_release: f8d97af | alvis++ | / (222 files):
17:32 dalek parrot/ayardley/parrot_github_release: Merge branch 'master' into ayardley/parrot_github_release
17:32 dalek parrot/ayardley/parrot_github_release: review: https://github.com/parrot/parrot/commit/f8d97af81d
17:32 dalek parrot/ayardley/parrot_github_release: ee1662f | alvis++ | tools/release/parrot_github_release.pl:
17:32 dalek parrot/ayardley/parrot_github_release: Automation of the first part of the process -- to archive 'docs/' to the parrot-docsx repo -- complete.
17:33 dalek parrot/ayardley/parrot_github_release: review: https://github.com/parrot/parrot/commit/ee1662f591
17:33 dalek parrot/ayardley/parrot_github_release: 8f686bc | alvis++ | tools/release/parrot_github_release.pl:
17:33 dalek parrot/ayardley/parrot_github_release: autogen script to package parrot.github.com and supporting repositories - 99 percent complete and tested.
17:33 dalek parrot/ayardley/parrot_github_release: review: https://github.com/parrot/parrot/commit/8f686bc617
17:33 benabik ohai
17:34 dukeleto alvis++
17:35 benabik Oh, hey.  That branch even looks really useful.  alvis++
17:37 whiteknight alvis++
17:40 nine Damn...my thesis is not even finished yet but it's already outdated in some places
17:40 benabik hah
17:41 nine I guess that's the way of all written documentation
18:17 jsut_ joined #parrot
18:23 rich joined #parrot
18:33 mj41 joined #parrot
18:46 dukeleto nine: that is why I like math. Correct proofs are never outdated
18:46 dukeleto nine: but sometimes people find shorter ones :)
18:47 nine aspell++ # fixing my terrible spelling
18:51 elmex joined #parrot
18:55 nine vim++ # set spell # awesome!
19:12 jsut joined #parrot
19:27 benabik I've aliased \s to set togglespell
19:35 nine So...enough for today. Good night, #parrot!
19:36 benabik G'night, nine!
19:50 contingencyplan joined #parrot
20:05 whiteknight well, we've received our first proposal for GSOC
20:05 benabik That's on my weekend todo list.
20:06 benabik Is it an interesting one?
20:06 jsut_ joined #parrot
20:06 whiteknight LAPACK bindings for PLA
20:06 whiteknight I think it's pretty interesting
20:07 whiteknight I've gotten at least half a dozen emails from other interested students, but very little follow-up so far
20:07 benabik I see how it could be useful.
20:08 whiteknight It's a necessary part of the runtime for a MATLAB-alike compiler, which I still plan to work on eventually
20:09 NotFound lapack + threading has potential for number crunching applications.
20:09 whiteknight yes, definitely
20:11 whiteknight and graphics applications
20:22 dukeleto I am all for LAPACK stuff.
20:22 dukeleto But do take note that people that are *insane* about performance laugh at everything that is not Fortran.
20:22 dukeleto if we can get away with very little overhead and basically ffi directly out to Fortran, we have a chance of wooing them
20:23 whiteknight dukeleto: that's what the bindings are
20:23 whiteknight dukeleto: The BLAS wrappers are basically just PMCs to manage the memory buffers and make the gemm calls
20:23 NotFound "insane people aren't useful anyway", said PerlJam yesterday.
20:23 whiteknight the LAPACK bindings will be NCI ties to LAPACK, with some fallback implementations if LAPACK isn't available
20:24 whiteknight so the calculations are still possible, but not as fast
20:24 Coke we have enough trouble being fast as perl, let alone fast as fortran. ;)
20:25 whiteknight fortran takes the approach that if brain is not aerodynamic enough, the brain can go
20:25 dukeleto Coke: huh? We are faster than Perl 5 in many cases. The ones that are fair, anyway. Where are we slower than Perl 5?
20:26 dukeleto Coke: also, comparing Parrot vs. Perl 5 isn't fair on either side. A better comparison would be Winxed vs. Perl 5
20:26 NotFound Many people are using Python for scientific caclulations, not everyone wants or needs the maximum runtime speed.
20:27 Coke dukeleto: no, a better comparison is rakudo vs. perl5
20:27 dukeleto NotFound: yeah, but scipy is very fast and is basically a binding to fortran
20:28 dukeleto Coke: I am more interested in Winxed benchmarks, but yes, Rakudo vs. Perl 5 is fair. I just don't have a horse in that race
20:29 Coke as someone on the board of the parrot foundation, you probably should.
20:29 Coke which reminds me, how long until we have last august's elections?
20:29 NotFound What's the difference between Parrit vs, Perl5 and Winxed vs. Perl 5?
20:29 NotFound You mean PIR?
20:29 Coke (not that I think you should be ousted nor do I intend to run.)
20:31 whiteknight Coke: Whenever somebody is motivated enough to run them
20:32 NotFound If you have some case of hand coded PIR significantly faster than winxed, let me know.
20:34 NotFound (Not that I think it's not possible, just to know where code generation must be improved)
20:39 PacoAir joined #parrot
21:37 kid51 joined #parrot
22:03 bacek joined #parrot
22:23 Coke whiteknight: my main concern is how, if it all, it impacts parrot's legal status.
22:36 whiteknight joined #parrot
22:43 whiteknight good afternoon, #parrot
22:44 dukeleto whiteknight: howdy
22:49 whiteknight hello dukeleto. how are you doing todaY?
23:24 dukeleto whiteknight: decent
23:24 dukeleto whiteknight: i am getting closer to my closing date on the house I am trying to buy
23:24 dukeleto whiteknight: i know you know how stressful that is :)
23:29 whiteknight yes, but very worth it
23:29 whiteknight I wish you all the best of luck
23:30 dukeleto whiteknight: thanks. Closing is april 12th
23:30 dukeleto whiteknight: down to the nitty gritty
23:30 whiteknight dukeleto: the only real advice I have is this: It's not over till it's over. Don't burn any bridges until the papers are signed
23:31 whiteknight because we paid big $$ to get out of our lease early, and settlement fell through, and we were stuck
23:31 dukeleto whiteknight: i don't believe anything until the keys are in my hand.
23:31 whiteknight good man
23:31 dukeleto whiteknight: yes, I read all about your trials and tribulations from your blog :)
23:31 whiteknight I only wrote about the parts that I can type out without falling back to a stream of cursewords
23:31 dukeleto whiteknight: yeah. I feel ya.
23:32 whiteknight otherwise most of my posts would have been "that mother**** realtor needs to **** a **** and shove her **** up her own ****"
23:33 * dukeleto laughs more than usual
23:36 kid51 joined #parrot
23:38 dukeleto kid51: top of the localtime()
23:53 kid51 Hello

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

Parrot | source cross referenced