Vision for 2.0: Production Users | Priority for 1.9: Performance and Testing | trac.parrot.org/parrot/wiki/Propos...chProtocol | Note: This channel is only for our Tuesday status meetings; you probably want #parrot instead.
Set by moderator on 3 December 2009.
02:19 TimToady joined 09:27 Tene joined 17:24 mikehh joined
japhb DONE: 17:28
* Hacking docs! Reviews requested: gitorious.org/parrot-plumage/parrot...cs/hacking , start with the contributing.pod doc.
* More metadata updates from fperrad++
* Add accessor methods to Plumage::Project
* Add project-dir command to plumage
* Move responsibility for determining build sequence to Plumage::Project
* Move responsibility for mkpath($build_root) to Plumage::Project's fetch action
* Excise %STAGES and supporting code from plumage.nqp 17:29
* Improve API of perform_actions{,_on_projects}
* Add stub of update stage to Plumage::Project, reusing (and falling back to) fetch
* Make harness exit(1) after failing test runs
* Fix broken test
* Tell plumage command_ subs what command invoked them;
* Collapse all plumage project action commands into one sub
NEXT UP:
* More new project actions and new plumage commands
BLOCKERS:
* NQP-rx repo prep by pmichaud so that I can move Glue and Util into NQP-rx
EOR
Tene Minor updates to rakudo, working on exceptions and failures. 17:31
EOR
17:41 NotFound joined 17:42 NotFound left
mikehh What I did since my last report: 17:43
17:43 NotFound joined
mikehh * building and testing parrot on amd64/i386, with gcc/g++, with and without --optimize 17:43
* fixing codetest and manifest_tests failures
* implement some patches from bubaflub++ and kurahapo++
with some modifications converting tests to pir
* fixing some tests
What I intend to do in the next week:
* testing and fixing
* investigate gc
.eor
17:46 bubaflub joined 17:47 particle joined 17:53 allison joined
NotFound What I did (last two weeks): 17:55
- parrot
* Just some testing.
- Winxed
* Renamed the parser example to stage 1 compiler.
* Modified harness to use it with both stages.
17:55 kj joined
NotFound * Improved the stage 1 compiler. 17:55
* Some fixes in stage 0.
What I will do:
* More work in Winxed stage 1
EOR
kj report for kjs: * nothing to report due to moving apt., deadlines, another move to another apt. and yet more deadlines. kjs is still lurking, though. EOR 17:56
allison - I've been absorbed by course work again this week. Classes end on Friday for a month, so I'll have more time available. 17:57
EOR
17:57 mikehh joined 17:58 plobsing joined 18:03 kj left
dukeleto EIRLBUSYNOREPORT 18:04
bubaflub Done 18:05
* 10 Perl+PIR tests converted to pure PIR
* Broke make coretest :-(
This week
* Fix coretest!
* More tests converted to PIR
EOR
q1q
cotto_w0rk #done: 18:06
* (re)factored the profiling runcore
* output is configurable and extensible, options atm are pprof and none (binary and possibly callgrind are planned)
* env var PARROT_PROFILING_OUTPUT now determines the output type, PARROT_PROFILING_FILENAME determines the filename
* updated pprof2cg.pl for some minor pprof format changes
* applied some patches from JimmyZ, closed associated tickets
#planned:
* more profiling work, figuring out testing
* take care of more patches/tickets from JimmyZ
* work on whatever the priority this week is
#blocking:
* me
.eor
q1q
plobsing What I Did: 18:07
- tried to get libjit_framebuilder working with a more recent parrot
- freeze/thaw Eval issues
- read up on freeze/thaw
- came up with a plan to clean up and fix issues with freeze/thaw Eval (TT #1359)
- started implementing TT #1359
What I Plan:
- finish implementation, submit patches
- get libjit_framebuilder working on a recent parrot
- steal parts from libjit_framebuilder for other projects (libffi for example)
EOR
Util # Done: 18:21
* Read/reading the Git Parable (how a Git-like system could be built from first principles), which is helping with my SVN biases. Highly recommended!
= tom.preston-werner.com/2009/05/19/t...rable.html
# Plan for this week:
* Update TT#1217 , TT#1218; turn back over to bubaflub++ for re-write
* Write up ideas for new PIR todo() API in TT#1302.
# Blockers:
* None
.end
18:25 barney joined 18:27 chromatic joined
chromatic Fixed some bugs. 18:29
Profiled and optimized.
18:29 simcop2387_ joined
chromatic Figured out how to make :immediate work with PBC; hopefully this will help NQP-rx and PCT with performance. 18:29
allison hi 18:32
NotFound hola
chromatic Hello, everyone. 18:33
Util hello
bubaflub hello.
18:33 moritz joined
cotto_w0rk hi 18:33
chromatic Let's review last week's priorities.
mikehh hello
chromatic Did anyone work on vtable overrides? 18:34
cotto_w0rk confusion reigned
18:34 davidfetter joined, whiteknight joined, davidfetter left
chromatic How so? 18:34
cotto_w0rk I think that the tasks on that page needed to be spelled out more explicitly.
istr dukeleto tried and got frustrated 18:35
chromatic A fair point, but I have a concern that we haven't met any of our weekly priorities in quite a while. 18:36
allison that could be a sign that we're not setting the right priorities
that is, it seems there has been good progress in speeding parrot up 18:37
cotto_w0rk q1q (total 2) 18:38
chromatic If we can't identify a priority, describe it in sufficient detail that any of several people could do it, and have one or more of those people do it, we'll block on that priority for the one person who described it.
That hurt us on the PCC refactor until we solved it with some applied heroics. 18:39
allison does the vtable override priority need a better description?
chromatic Perhaps; this is the first I heard about anyone poking at it. 18:40
allison I missed last week, so fill me in: why was it selected as the priority? 18:41
chromatic It was 1) documented 2) achievable in a week 3) a performance improvement 4) a cleanliness improvement
allison good reasons 18:43
chromatic Other thoughts? 18:44
allison volunteer time 18:45
(it's limited)
chromatic Thoughts from other people? 18:46
Util Tasks, when made a priority, need a clear "next action" defined; this is a GTD/project management concept. 18:47
Taking the time to do this up-front lowers the barrier to starting on the task.
chromatic Is that what we're missing? 18:48
Util that is just a thought.
allison breaking tasks down to make them more approachable is good 18:49
chromatic I'm happy to do that, but it seems like there's little buyin from other developers on the priorities.
allison how about we put this on the agenda for Sunday? 18:50
Util next-action is not even a full breakdown; just "what is the first bite we can tear off".
allison review if the weekly priorities have helped and if we want to keep them
chromatic Okay, that's a good agenda item.
mikehh I thgink it would work better if we had clearly defined tasks - at least for me 18:51
bubaflub and i don't mind having people say "hey, you, do this..."
and if it's clearly defined, all the better
Util When a priority is set, we get nods, but not a sense of who expects to devote time to it, nor how much time.
chromatic I will improve the task definitions on that page.
Other priorities for this week? 18:52
allison define the roadmap for next year
(that's the priority)
chromatic Questions. 18:53
cotto_work?
cotto_w0rk What's keeping plobsing from getting a commit bit?
chromatic +1 to giving him a bit 18:54
+1 to Gerd too, if he doesn't already have one.
allison do we have CLAs?
I see one for Gerd 18:55
I don't know plobsing's real name
particle received CLAs from whiteknight and cotto yesterday
peter lobsinger iirc
allison that sounds familiar...
NotFound +1 18:56
chromatic Any concerns about either?
allison yes, we have a CLA for plobsing too
dukeleto is back
allison do we have mentors for them?
dukeleto i am willing to mentor
chromatic Okay, who can flip their commit bits? 18:57
dukeleto chromatic: you?
chromatic: i don't know that i can flip a bit
whiteknight I'll mentor
(with what little time I have) 18:58
co-mentor
particle docs/project/metacommitters_guide.pod
Tene The time I've done the most work for Parrot was when I was listed as "Tene will implement this" on the roadmap, and I was asked about it specifically, by name, at #ps meetings.
That's correlation, not causation, but maybe relevant.
allison that's helpful data 18:59
chromatic Other questions? 19:00
bubaflub yep. did ya'll receive my CLA?
dukeleto what is this weeks priority?
cotto_w0rk I forgot my other question.
dukeleto Also, do people want me to continue sending benchmarks to parrot-dev? Are they useful? 19:01
allison dukeleto: the roadmap for next year, and documentation
whiteknight I find the benchmarks useful
chromatic The priorities are vtable overrides, documentation, and figuring out the roadmap for 2010.
I find the benchmarks useful too, but I still would like to see Callgrind numbers more than times.
Util q1q
dukeleto chromatic: i will get you valgrind output for the array_access.pir slowdown 19:02
allison do we have a specific time for Sunday?
dukeleto what is happening sunday?
allison maybe a 3 hour block? 19:03
chromatic Three hours sounds useful.
allison dukeleto: see Will's message to the list about roadmap review workshops 19:04
spreadsheets.google.com/ccc?key=0A...&hl=en 19:05
19:05 mikehh joined
allison bubaflub: drop me an email and I'll check on it 19:06
japhb is bak
bubaflub allison: thanks.
dukeleto it would be nice if our meeting was later in the day so that bacek could join in 19:07
the sunday meeting, i mean
allison the same time as parrot sketch would be easy to remember, but is probably too early 19:08
how about 2 hours later?
cotto_w0rk That'll be Monday morning for him, won't it?
allison that's about as late as I can go for a 3 hour meeting 19:09
where is bacek?
Tene q1q
cotto_w0rk allison, He's in eastern Australia 19:10
allison so, wouldn't he need earlier in the day?
sounds like we'd better take this to the list 19:11
chromatic Let's do that.
Util, you had a question?
19:11 mikehh joined
Util Some trig tests on N registers in t/op/inf_nan.t are todo'ed, stating: "cot/coth/acot not implemented for real numbers". 19:11
This statement asserts that these trig ops are planned for, just not implemented yet. Is this assertion correct? 19:12
(perhaps "implies" is a better word than "asserts" in this audience)
allison it was correct, but we're now trying to move advanced math ops to dynops 19:13
so, if we add them, they should be added as dynops
Util Does IMCC work with parsing dynops, or must we wait for PIRC? 19:14
chromatic If you load the dynops, it works fine.
Util KTHX. EOQ
dukeleto Util: i planned to implement those as dynops and haven't gotten around to it
bubaflub dukeleto: after a bit of hand holding, i can probably help ya out 19:15
dukeleto bubaflub: sounds good!
chromatic Tene, question?
dukeleto q1q
19:16 treed joined
Tene chromatic: treed was voicing some significant concerns about his issues as a HLL dev being ignored. 19:17
dukeleto Tene: where were these concerns raised? 19:18
Tene Where should he voice those concerns to get them noticed as pain for an HLL dev?
treed On an unrelated IRC channel.
dukeleto Tene: parrot-dev and parrot-users ?
Tene dukeleto: That's what this question is about.
treed I can try to make a list and do something about it.
Post it to both of those?
dukeleto treed: if you are having issues using Parrot because there are bugs, then it is best to cc parrot-dev and parrot-users 19:19
allison parrot users is for user questions
treed k
dukeleto treed: if you find out that the discussion only belongs on one of the lists, then you can specialize
treed I thought parrot users was for end users though?
Not HLL devs?
allison developer needs should go to parrot-dev
dukeleto allison: yes, do HLL devs count as users?
allison well, HLL devs are users
dukeleto allison: i consider HLL devs "users" 19:20
Tene treed: the "end users" of parrot are HLL devs.
treed Users of what?
Tene: I mean end users of parrot-based products.
allison the way to think about it is parrot-users is for "helpdesk" questions
parrot-dev is for development tasks
dukeleto allison: i agree. so i guess just send it to parrot-users first
treed: if it ends up being a parrot bug, then cc parrot-dev 19:21
allison: does that sounds reasonable?
allison good start
treed: if it's a list of bugs you need us to bump up in priority, then parrot-dev 19:22
chromatic dukeleto, question?
treed Okay, I'll see if I can put together a list. I can only thing of two issues right now. (Maybe a third depending on if it's a bug (which I think is likely) or me being a retard (not unpossible))
dukeleto chromatic: How slow is too slow? 19:23
chromatic In what context?
dukeleto chromatic: i.e. does Parrot have any benchmark guidelines? Are we going to ship 2.0 if we haven't fixed arrays being twice as slow in trunk as 1.8.0 ?
chromatic: my quesiton is more general though 19:24
chromatic We don't have any benchmark guidelines yet.
dukeleto chromatic: should/do benchmarks effect our releases or priorites in any formal way? or is it just "oh crap, that is slow, we should look into that"
chromatic: i am thinking of writing some benchmark tests 19:25
allison dukeleto: we don't have a hard target of "no release unless we're X fast"
dukeleto allison: not so much as "we want to be X fast" as "we used to be X fast and now we are X/2" 19:26
allison: just wondering if we need some guidelines
allison in terms of general goals, they're more relative than absolute. Like, "Pynie is faster than CPython"
being faster on, say, a fibonacci benchmark isn't actually terribly helpful 19:27
dukeleto allison: i guess i am trying to put examples/benchmarks to some quantitive use
allison you know what would be useful? getting us back into the alioth debian benchmark shootout
I've been meaning to look into what that would take 19:28
dukeleto allison: i am talking about array access being twice as slow. that slows down parsing and *everything*
mikehh bear in mind what pmichaud said about that
dukeleto allison: i am not talking about optimizing useless scripts
chromatic dukeleto, what if we started with some metrics (weekly) and evaluated their utility after a few weeks?
allison dukeleto: right, but we don't have any way of knowing what to test for speed until we find one of those
dukeleto chromatic: that would be great. i just want us to be thinking about how we should act on benchmark info 19:29
allison: well, i just test everything and only tell parrot-dev the bad news ;)
allison it's really benchmark regression testing, so we start with problems we see now and add more as we find them
dukeleto allison: yes, i am talking about regression benchmarks, i guess 19:30
chromatic Other questions? cotto, did you have another? 19:31
dukeleto allison: i can't seem to find out how to add a language to the debian shootout 19:32
allison dukeleto: we used to be on there, then dropped off
chromatic Is there anything else? Shall we move the benchmark discussion to #parrot? 19:35
Sounds like it (or I've netsplit). Thanks, everyone. 19:39
19:39 moritz left
cotto_w0rk cotto_w0rk, nothing more from me 19:40
dukeleto ENODISHES
19:41 bubaflub left 19:43 NotFound left 20:02 plobsing left 20:40 PacoLinux left 21:19 chromatic left 22:51 Whiteknight joined 23:13 mikehh joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!