Parrot 2.8.0 released | parrot.org Log: irclog.perlgeek.de/parrot/today | Nopaste: nopaste.snit.ch:8001 | close 25 tickets; remove deprecated items (especially CodeString);
Set by moderator on 24 September 2010.
whiteknight two questions: does anybody have a link to that object metamodel paper? and, do we have a TODO list anywhere for a new metamodel? 00:02
kid51 msg nwellnhof In testing the gc_ms2_tuning branch on my small-physical-memory machine, is there any command-line setting I should use, say, for gc_threshold? 00:05
aloha OK. I'll deliver the message.
kid51 which object metamodel paper?
plobsing p&w? 00:10
00:10 kid51 is now known as kid51_at_dinner
plobsing purl, p&w? 00:10
ENOPURL 00:11
dukeleto a blessing or a curse? only time will tell. 00:12
cotto leans toward a mild curse
plobsing purl knew that link
cotto docs.google.com/viewer?a=v&q=ca...6bTrVoSfRA 00:14
cotto suspects that that link may be self-aware
00:14 theory joined
whiteknight cotto++ 00:15
I can never remember the title of that paper
I had a copy of it on my other box, but it's broken into tiny pieces 00:16
00:16 jhelwig joined
whiteknight yes, jnthn's work reminds me a lot of this paper 00:17
cotto which work?
or are you referring to the general idea rather than the specifics 00:18
whiteknight yes 00:21
plobsing whiteknight: (wrt pdd post) does that mean for subsystems were we've more or less declared "mission accomplished" we should not have any PDD? 00:26
dukeleto kid51++ # 'make smoke' knows about TEST_JOBS now 00:29
00:32 petdance left
dalek TT #1806 closed by dukeleto++: "make smoke" does not respect TEST_JOBS 00:37
TT #1806: trac.parrot.org/parrot/ticket/1806
rrot: r49347 | dukeleto++ | trunk (2 files):
[t][doc] Teach 'make smoke' about TEST_JOBS, kid51++ and update NEWS
 00:44
whiteknight plobsing: first, name me a subsystem which is "mission accomplished"
plobsing AFAIK, we're pretty satisfied about PCC
whiteknight plobsing: I can't imagine a situation where any system is so perfect that we're not planning any improvements
The PCC interface is decent, but the internals are the subject of many complaints 00:45
plus, how is it going to work with JIT?
and how will it work with Lorito?
and can we modify PCC to produce a precise GC instead of a stack-walking system?
plobsing my understanding of lorito as it pertains to PCC is that PCC sits on top of it more or less. meaning its almost a direct translation.
whiteknight that's my vision for it too, but it may not be that way 00:46
I mean, it's not written anywhere
I'm not trying to be argumentative, I just don't think any systems are really in a state where they are "complete" 00:47
in that case though, I suspect the PDDs turn into philosophical justification documents
00:47 Kulag joined
bacek_at_work ~~ 00:48
plobsing ok, fair enough. what about systems that aren't problems at the present? why bother spending time designing them?
sure it'd be nice if we didn't run around putting out fires, but that's the way it seems we're doing things 00:49
bacek_at_work seen nwellnhof
aloha nwellnhof was last seen in #parrot 2 hours 18 mins ago joining the channel.
whiteknight plobsing: my thought is that we set specific priorities. Second half of 2010 and maybe first half of 2011 we're obviously working on GC. It would be nice to have a comprehensive design for it to work for 00:51
we don't need to go crazy rewriting design documents for systems we're not planning to do serious work on 00:52
plobsing I guess my question boils down to: do we keep or delete the PDDs we're not willing to invest time in? 00:53
whiteknight bump them all back to the "drafts" folder 00:54
or a fancy new "old" folder
plobsing I suggest /pdd/outright-lies or /pdd/wishful-thinking
whiteknight specificity. I like that 00:55
00:56 theory left 01:07 kid51_at_dinner is now known as kid51
kid51 whiteknight: What does 'TL;DR' mean? 01:18
whiteknight "Too long, didn't read" 01:20
01:27 hercynium joined
dukeleto TLDRWTFBBQ? 01:30
01:31 whiteknight left
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#268) fulltest) at r49347 - Ubuntu 10.10 beta amd64 (g++-4.5 with --optimize) 01:34
ran smoke with TEST_JOBS=5, the report on smolder.parrot.org/app/projects/rep...etails/268 looks ok to me 01:35
smoke #244 (no TEST_JOBS) real 4m11.161s, user 2m24.200s, sys 0m25.190s 01:40
01:41 petdance joined
mikehh smoke #268 (TEST_JOBS=5) real 1m41.422s, user 2m19.080s, 01:42
sys 0m25.300s
that includes sending to smolder
dalek tracwiki: v66 | cotto++ | ParrotQuotes 01:46
tracwiki: We like specificity.
tracwiki: trac.parrot.org/parrot/wiki/ParrotQ...ction=diff
01:53 Kovensky joined
Kovensky hai Kulag 01:53
sorear mikehh: 25s sys? what. 01:56
02:06 kid51 left
mikehh soarer: ? 02:18
bacek_at_work 25s of sys is way too much 02:19
mikehh it has to connectr to smolder, send the archive file, etc
I get pretty much the same results running make test TEST_JOBS=5 02:22
02:31 Drossel joined
mikehh time make coretest TEST_JOBS=4 => real 0m33.712s, user 0m52.450s, sys 0m14.340s 02:31
02:33 Kulag left 02:35 janus left 02:36 Kulag joined 02:37 Drossel left
mikehh time make coretest => real 1m37.251s, user 0m46.600s, sys 0m14.200s 02:37
02:41 janus joined
sorear mikehh: sending to sockets isn't sys 02:58
03:01 theory joined
mikehh soarer: look at the results from make coretest above - sys 14s in both cases 03:03
so I don't know what is going on 03:04
sorry sorear 03:06
dukeleto hola 03:08
mikehh howdy dukeleto 03:09
mikehh needs some sleep if I am going to function later - it's after 4am here 03:15
03:53 cognominal left 03:54 ash_ left 03:58 ash_ joined 04:04 cognominal joined 04:05 ash_ left 04:07 cognominal left 04:08 cognominal joined 04:12 s1n joined 04:22 cognominal left 04:23 s1n left 05:01 theory left 05:30 plobsing left 05:41 hercynium left 06:08 petdance left 06:23 bacek left 06:41 cognominal joined 06:49 esskar left 06:55 fperrad joined 06:57 KatrinaTheLamia left 07:19 lucian joined 07:21 contingencyplan left 07:40 lucian left 07:43 tadzik joined 07:44 lucian joined 08:05 mikehh left 08:10 bacek joined 08:31 jsut_ joined 08:36 jsut left 08:40 mikehh joined
dalek kudo: f523366 | moritz++ | docs/ChangeLog:
[docs] update ChangeLog
 08:47
rrot: r49348 | bacek++ | branches/generational_gc/include/parrot/list.h:
Remove wrong semicolon from LIST_APPEND macro
 08:54
rrot: r49349 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Fix bug with swapping root_objects and objects[0]
rrot: r49350 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Reduce gc_threshold to 1M to increase load on GC for testing purposes
09:02 dip joined 09:38 contingencyplan joined
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#275) fulltest) at r49350 - Ubuntu 10.04 i386 (g++) 10:22
10:23 lucian left
dalek rrot: r49351 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Factor out mark_string_header. We do need to move strings between
 10:35
rrot: r49352 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Skip older PMC in mark
rrot: r49353 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Remove too expensive PARROT_ASSERT
10:39 ruoso left 11:00 esskar joined
mikehh bacek: ping 11:06
bacek mikehh, pong
mikehh bacek: got some MANIFEST and codetest fixups for generational_gc branch 11:07
bacek mikehh, go for it 11:08
jnthn bacek: ooh, hacking on generational GC? :-) 11:12
bacek jnthn, indeed :)
sorear The #1 storage feature I'd like to see is the PMC/Object unification 11:13
our current Object -> Object_attrs -> RPA -> RPA_attrs -> PMC** tango is probably responsible for a lot of the pressure on the GC
jnthn sorear: 6model does better. 11:15
sorear jnthn: 2 blocks or 3? 11:16
jnthn 3 'cus I'm lazy 11:17
will be 2 later
dalek rrot: r49354 | mikehh++ | branches/generational_gc/MANIFEST:
[generational_gc] re-generate MANIFEST
 11:26
rrot: r49355 | mikehh++ | branches/generational_gc (2 files):
[generational_gc] add svn properties and bring t/op/gc-non-recursive.t into line with trunk
rrot: r49356 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
First real generational GC attempt.
bacek hollei schitt! make coretest passed on r49356! 11:27
moritz you must be doing something wrong :-)
bacek I'm definitely doing it wrong 11:28
much better now. parrot crashes badly :)
mikehh bacek: got some codetest fixes for gc_ms2.c but waitin' for you 11:46
bacek mikehh, go for it. i'm probably done for today. 11:48
mikehh bacek: 'k 11:49
11:54 whiteknight joined
whiteknight good morning, #parrot 12:11
Coke . 12:17
msg kid51 to open a ticket, support@osuosl.org - I'm not sure the webterface is enabled (just like for rt.perl.org); there is #osuosl on freenode, though.\\ 12:18
aloha OK. I'll deliver the message.
Coke msg kid51 - also, particle's a good person to ping about such things.
aloha OK. I'll deliver the message.
12:24 bkuhn joined 12:32 petdance joined 12:34 Coke left
dalek rrot: r49357 | mikehh++ | branches/generational_gc/src/gc/gc_ms2.c:
[generational_gc] fix codetest failures, macro args, c_parens and add documentation
 12:34
12:35 PhatEddy joined 12:41 petdance left 12:51 plobsing joined 12:54 PhatEddy left
whiteknight plobsing: ping 13:00
plobsing png 13:08
pong even
dalek kudo: 607ec45 | moritz++ | README:
[README] fix smolder link, PhatEddy++
 13:17
13:19 Patterner left
whiteknight plobsing: in bytecode, PMCs are all thawed only once? 13:22
like, when the PBC is loaded?
plobsing yes. thawing is part of loading. 13:23
specifically it happens when a ConstTable segment (usually only one per PBC) is unpacked 13:24
whiteknight okay, that's what I was hoping.
13:25 olgagirl joined
plobsing ooc, what are you hoping to use this for? 13:25
13:27 olgagirl left 13:29 Psyche^ joined, Psyche^ is now known as Patterner 13:31 gottreu joined
whiteknight plobsing: I'm just reading over PDD03 with a fine-toothed comb and I want to make sure the sanity levels are high enough 13:33
I'm reading the code too, but I can't pick some of it apart right now 13:34
plobsing OK. I was somewhat woried that you were going to be depending on the ordering for something. 13:37
We might make thawing constants lazy in the future. And side effects in deserialization are just evil.
13:38 gottreu left
whiteknight plobsing: I'm thinking about a system where we have dedicated "load_constant" opcodes. and every other opcode and all PCC calls only operate on registers 13:38
we cut our number of opcodes in half or more immediately. Plus, we cut out a lot of unnecessary branches in the PCC code 13:39
plobsing we have special branches in PCC for dealing with constants? that just seems wrong. 13:40
whiteknight it is wrong 13:41
look at src/call/args.c:extract_named_arg_from_op as one example 13:42
plobsing why do we have 2 levels of branching there? not sure whether most compilers would flatten that or not. 13:44
13:44 tadzik left 13:46 lucian joined
plobsing whiteknight: we could get what you're thinking of fairly easily. it is the extreme case of what branches/constant_unfolding aims to do. 13:48
13:49 gottreu joined
whiteknight the problem with a load_constant opcode is that we would have to know the index of the constant in order to look it up, but we don't necessarily have the indices of the constants in the table until the table is completed 13:49
so we either need a smarter assembler, or some kind of magic, or a more interactive assembly process 13:50
plobsing I'm confused. I think we may be talking about different things. I was mentally mapping load_constant to set_i_ic, set_n_nc, etc... 13:52
which we already do. so I don't see how we'd need a smarter assembler
whiteknight oh, I was envisioning something more along the lines of a load_constant_p_i, which would load PMC constant i from the table 13:53
13:53 tadzik joined
whiteknight of course, we we did $P0 = load_constant "foo", IMCC could easily translate that internally to the form that's required 13:53
plobsing we could extend virtual registers to include "virtual constant references" 13:54
eg: $PC0 for PMC constant 0, $NC1 for number constant 1, etc 13:55
but then we'd need a way to associate constants with tehm 13:56
whiteknight right
there are some serious questions to be answered, but I think there are big wins here if we do
plobsing I could try and bring constant_unfolding up to date to see where those wins are. 13:58
14:06 lucian left 14:07 lucian joined 14:13 lucian left, lucian joined 14:22 jsut joined 14:26 jsut_ left
dukeleto waves hello 14:46
moritz solitons hello 14:47
14:48 theory joined
dukeleto decomposes self into normal modes and tunnels through moritz 14:49
14:54 lucian_ joined 14:55 plobsing left 14:57 ash_ joined 14:58 lucian left
gottreu parrot on minix3: does the community care at this time that it doesn't build? 15:48
moritz it'll probably accept patches 15:49
gottreu I do not have those. My assumption is that minix doesn't matter until much later in the life of parrot. Probably a little after it works on VMS. 15:51
tadzik or much later in the life of Minix :) 15:52
gottreu has given up. At least until he finds a debugger for Minix. 15:53
moritz my impression is that parrot has a some fixed "core" platforms (less unusual Unix variants and MS windows), and all other platforms are only supported if there happens to be a parrot developer that maintains it
16:00 ruoso joined
whiteknight that's exactly it. We don't currently have a developer that works on Minix 16:08
I may be able to install minix on a vm for testing 16:11
16:14 tadzik left
dalek rrot: r49358 | mikehh++ | branches/generational_gc/compilers/imcc/pbc.c:
[generational_gc] add cast to get g++ to build
 16:33
16:35 davidfetter joined 16:41 M_o_C joined 17:10 plobsing joined
dukeleto does minix have a package system? 17:28
whiteknight i have no idea 17:30
17:33 ash_ left
PerlJam pkg probably 18:02
whiteknight I can't get minix to install on VirtualBox 18:11
so I'm abandoning this particular project
18:12 fperrad left 18:15 jsut_ joined 18:20 jsut left 18:34 Coke joined
dukeleto mumbles about too many projects... 18:34
18:41 M_o_C left 18:57 davidfetter left 18:59 M_o_C joined 19:06 Andy left 19:07 Andy joined 19:25 lucian joined 19:27 lucian_ left
GeJ Bonjour everyone. 19:32
whiteknight hello GeJ
GeJ Hello whiteknight. 19:33
19:43 whiteknight left 19:57 davidfetter joined 19:58 ruoso left
GeJ clock? 20:03
20:08 Andy left 20:10 Andy joined, ruoso joined 20:43 snarkyboojum left 20:49 M_o_C left 21:10 lucian left 21:29 dngor left, dngor joined 21:31 ruoso left 21:33 tadzik joined 21:38 lucian joined 21:40 KatrinaTheLamia joined 21:45 tadzik left, snarkyboojum joined 21:49 tadzik joined 21:52 wagle left 21:53 wagle joined, wagle left 21:59 snarkyboojum left 22:00 wagle joined 22:04 tadzik left 22:09 bkuhn left 22:25 snarkyboojum joined 22:31 snarkyboojum left 22:40 snarkyboojum joined 22:41 bluescreen joined 23:04 snarkyboojum left 23:17 snarkyboojum joined 23:19 bluescreen left 23:26 snarkyboojum left 23:28 whiteknight joined 23:32 whiteknight left, whiteknight joined 23:41 hercynium joined 23:43 davidfetter left 23:48 ash_ 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!