|
Parrot 6.6.0 "Parrothead" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC Set by moderator on 23 July 2014. |
|||
|
01:40
FROGGS_ joined
02:11
kid51_ joined
02:15
kid51_ joined
06:51
basiliscos joined
07:30
basiliscos joined
08:50
bighugedog joined
10:14
he joined
11:28
kid51 joined
14:24
jsut_ joined
14:32
rurban1 joined
14:51
Chirag joined
|
|||
| Chirag | rurban: Hey! | 14:56 | |
| rurban | Hey | 15:05 | |
| we still need a nice reproducer for the perl6 CallContext error | |||
| or it could be nqp or perl6 bug only | |||
| Chirag | oh.. I see | 15:11 | |
| its not because of our last task right? | 15:12 | ||
|
15:14
basiliscos joined
|
|||
| rurban | I don't think so | 15:21 | |
| Chirag | ok.. My branch's current failure is really tricky | 15:23 | |
| rurban | yes, some fundamendal error. you need to check the generated callcontext.c against our hand-fixed one I guess | 15:24 | |
| Chirag | i suspect Parrot_CallContext_class_init | 15:25 | |
| rurban | Chirag: yes, most likely | 15:43 | |
| nope. Parrot_CallContext_class_init is called correctly 2x, with pass 0 and pass 1 | 15:48 | ||
|
15:57
Chirag joined
|
|||
| Chirag | hmm.. | 16:08 | |
| so, I replaced my callcontext.c and .h files with the ones from your branch but still got the same error... I think pmc2c has become faulty ... checking now | 16:25 | ||
| rurban | I rather think it's some new replacement code in the call sources | 17:17 | |
| Chirag | yes.. one of these changes (my latest commit vs ur branch's latest commit) | 17:21 | |
| gist.github.com/ZYROz/0139e9287165407ed437 | |||
| args.c is the culprit | 17:31 | ||
| I guess one of the methods is behaving there | 17:32 | ||
|
17:48
khisanth_ joined
|
|||
| rurban | This works now? | 18:32 | |
|
18:32
Psyche^ joined
|
|||
| rurban | nope, there are just a few wrong direct Parrot_CallContext calls in arrays | 18:36 | |
| on arrays | |||
| fixed it now | 18:37 | ||
| Chirag | check my latest commit .. works | 18:41 | |
| still bringing back the exported methods | |||
| rurban | I'm squashing now the various commits | 18:42 | |
| Chirag | oh.. give me a while.. I will finish with the conversions | 18:43 | |
| rurban: not sure about VTABLE_get*, VTABLE_set* and VTABLE_exists_keyed_str ... | 18:54 | ||
| rurban | there were some arrays treated as CallCopntect | 18:55 | |
| Chirag | would you like to see a git diff with my current changes? | 18:56 | |
| just for args.c | |||
| rurban | yep | ||
| Chirag | gist.github.com/ZYROz/0b31f2c775bd152d1263 | 19:05 | |
| rurban | about 5 or so wrong calls | ||
| Chirag | in this? | 19:06 | |
| but builds correctly | |||
| rurban | nope, in the old [pcc] More Direct Methods GH #1083 | ||
| Chirag | oh | ||
| rurban | collect_named was wrong | ||
| and current_array | |||
| the rest was good | 19:07 | ||
| I'm benchmarking now | |||
| I hope we are still at 3% | |||
| Chirag | wait | ||
| you are using my args.c ? | |||
| rurban | a mix of all, I squashed all together into 2 commits, one for pmc2c and one for the the rest | 19:08 | |
| Chirag | because my args.c is incomplete | ||
| havent covered gets and sets | |||
| rurban | oh | ||
| Chirag | also not sure about VTABLE_exists_keyed_str | 19:09 | |
| maybe we can first optimize this and then test | |||
| rurban | oh, fill_params actually allows NULL call_object | 19:13 | |
| Chirag | ok.. | 19:14 | |
| I am not sure which VTABLE_set* to change | |||
| i am getting a SEGV | 19:15 | ||
| rurban | I think I've covered all VTABLE cases with CallContext already. | ||
| dalek | rrot/pcc-gh1083: 476dd48 | ZYROz++ | lib/Parrot/Pmc2c/ (2 files): [pmc2c] Faster CallContext ATTR accessors + non-static for GH #1083 Omit the obj check in the CallContext ATTR accessors. Put CallContext non-static decl into .h as they are not static anymore, so that we can call them now directly internally in src/call/*.c to avoid VTABLE redirection for those. work together with Reini Urban |
||
| rrot/pcc-gh1083: 844ffaa | rurban++ | src/call/ (3 files): [pcc] use the direct CallContext methods GH #1083 and do not use the CallContext vtable methods anymore. also inline one Continuation_set_pointer call. all tests still passing, 2.98% faster. |
|||
| Chirag | great! | 19:16 | |
| now the tests | |||
| rurban | you've also found a couple more VTABLE calls, just a bit too many | 19:20 | |
| not sure about the slurpy cases. named looks good | 19:21 | ||
| Chirag | I was actually debugging by replacing a set of VTABLE calls | ||
| at a time | |||
| rurban | 9.662807002 down to 9.507910017 | ||
| Chirag | 1.6% only | 19:22 | |
| rurban | 1.63% faster | 19:23 | |
| this looks closer to reality | |||
| Chirag | when I was debugging args.c, I got a feeling that if we used a mix of VTABLE and CallContext, we might get better speed ... (just a feeling) | 19:24 | |
| rurban | interesting | 19:27 | |
| I'm checking now perl6. we got a CallContext reusage problem there somehere | 19:28 | ||
| somewhere | |||
| Chirag | oh | ||
| FROGGS_ | rurban++ # though note that I worked around it, so perl6-p is usable atm | 19:30 | |
| dalek | rrot/pcc-gh1083: d815693 | rurban++ | src/call/ (3 files): [pcc] use the direct CallContext methods GH #1083 and do not use the CallContext vtable methods anymore. also inline one Continuation_set_pointer call. in fill_params() it is safe to use CallContext directly, as it assumes either NULL or CallContext, but never PMCNULL. all tests still passing, 1.63% faster. |
19:31 | |
| Coke | Chirag++ #gsoc | 19:51 | |
| Chirag | Thanks :) | ||
| rurban | Chirag++ :) | ||
| Chirag | :) | 19:52 | |
| rurban | btw. did you try to come to YAPC::EU Sofia now? | 19:53 | |
| because I cancelled my trip there | |||
| Chirag | I really wanted to.. the only reason I didnt try was because of campus placements.. but now the placement manager wont let me sit for it either | 19:54 | |
| rurban | I see | 19:55 | |
| Isn't it too late to ask TPF for travel grants also? | |||
| Chirag | yes and no visa | 19:56 | |
| rurban | yes, I remember. | ||
| Chirag | maybe next year | ||
| rurban | I have to leave the USA also because I got no visa extension yet | ||
| So I'll be 7 hrs away from Util then :) | 19:57 | ||
| but closer to Chirag, only 3.5 hrs then | |||
| Chirag | oh! your company cant do anything? | ||
| rurban | they tried, but it's the government | ||
| and my wife wants to work, so we go back to Germany | 19:58 | ||
| Chirag | maybe I can bring that time diff to zero.. I was offered a thesis at ETHZ for this semester.. I am planning to approach the prof again | ||
| rurban | ETH is nice, but Zurich is very very expensive | ||
| one of the most expensive cities at all | 19:59 | ||
| Chirag | I know.. thats why I postponed it to next semester.. Now I have GSoC stipend and my internship's stipend so dont need parents' support | ||
| I was trying for Saarbrucken but they wont take undergrads | 20:00 | ||
| rurban | for Zurich you really need a good stipend | ||
| perl6 tests with our pcc-gh1083 branch do look good so far | 20:01 | ||
| Chirag | a long shot but I was trying for a parallel internship at Disney Research to help the cause | 20:02 | |
| rurban | which cause? | ||
| Chirag | financial :) | ||
| rurban | the stipend cause :) | ||
| Because I'm having a serious problem with Disney right now... (Marvel movies) | 20:03 | ||
| Chirag | :) | ||
| rurban | They need to bring Joss Whedon back in as soon as possible | ||
| Chirag | I was at dreamworks.. their art is way better than disney.. but storylines are pathetic | 20:04 | |
| rurban | we just got someone from a big gaming company here who use the same technology as the movie animators. just better | ||
| (real time) | |||
| Chirag | I really want to work on all that :D | 20:05 | |
| Germany has lots of great universities for vision and graphics... Inria, France as well | |||
| rurban | they do 3dMAX/Maja + python or lua scripting | ||
| currently they are outsourcing everything to India | 20:06 | ||
| Chirag | but artists in India are expendables... thats what i gathered... no job security | 20:07 | |
| in US too | |||
| Dreamworks US had a very big layoff season | |||
| TU Graz has this amazing VR technology | 20:09 | ||
| FROGGS_ | rurban: what perl6 revision do you use? | 20:10 | |
| rurban | FROGGS_: nom HEAD | 20:11 | |
| and nqp HEAD and parrot pcc-gh1083 | 20:12 | ||
| FROGGS_ | rurban: as I said I worked around the issue in rakudo, see: github.com/rakudo/rakudo/commits/nom | ||
| so you'd need b17a24b5f245e8c49605421f081dceb3e33a2efa to get the fails | |||
| rurban | I'll check. looked like a Heisenbug in the callcontext reuse to me | ||
| FROGGS_ | that sounds very likely | 20:13 | |
| rurban | I see. I'll check it later with --debugging | 20:15 | |
| I suspect 6model | |||
|
20:16
rurban1 joined
|
|||
| rurban | I only got 4 perl6-p fails now in spectests, good | 20:17 | |
| FROGGS_ | yes, that sounds about right | 20:18 | |
|
20:36
perlfan joined
21:42
rurban1 joined
22:31
Psyche^ joined
22:45
kid51 joined
23:41
perlfan joined,
khisanth_ joined,
jsut_ joined,
FROGGS_ joined,
atrodo joined,
rblackwe joined,
GeJ joined,
Hunger joined,
slavorg joined,
janus joined,
Timbus joined,
Tene joined,
Maddingue joined,
smash joined,
allison joined,
simcop2387 joined,
sorear joined,
kshannon joined,
mtj- joined,
TonyC joined,
Util joined,
dalek joined,
p6eval joined,
cottoo joined,
ingy joined,
Coke joined
23:48
cooper joined,
awwaiid joined,
ggherdov__ joined,
rurban joined,
ivan joined
|
|||