|
Post closed tickets in your report. | Note: This channel is for our weekly status meetings (Tuesdays at 19:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/ Set by moderator on 20 March 2012. |
|||
|
01:12
alvis```` joined
02:06
PerlJam joined
02:07
Util joined
02:12
Coke joined
06:15
alvis joined
07:25
lucian_ joined
08:40
lucian joined
14:04
JimmyZ joined
14:46
JimmyZ left
15:42
lucian joined
17:20
lucian_ joined
18:01
alvis joined
|
|||
| nine | 19:30 UTC | 19:30 | |
| dukeleto | hola | ||
| dukeleto is very close to closing on a house and has EVERYFEWTUITS | |||
| Util | Hello | 19:31 | |
| alvis | hello | ||
| Util has no report; also micro-tuits for next 2 weeks. | 19:34 | ||
|
19:38
whiteknight joined
|
|||
| whiteknight | hello | 19:38 | |
| (belatedly) | |||
| nine | I made some progress on threads: fixed quite a few bugs, got task.'send' and receive working and implemented some optimizations. I also implemented my first threading tests in winxed and a new benchmark multiplying matrices with vectors. | 19:41 | |
| First results look ok with about 25 % performance increase using 4 threads compared to one. It's better than this sounds because much of the time is spent single threaded creating the 10000x10000 matrix. The actual multiplication does not take that long. But there is quite some overhead merging the results back... | 19:42 | ||
| whiteknight | okay, that's not bad | 19:43 | |
| 25% is an improvement, and can likely get better | |||
|
19:44
benabik joined
|
|||
| cotto | howdy | 19:44 | |
| nine | Results would look much more favourable if calculations would take longer. The couple of multiplications and additions are just too fast :) | ||
| whiteknight | Are there benchmarks we cold put together that are more calculation-intensive and are less skewed by setup times? | 19:45 | |
| cotto | nine, can you benchmark just the multiplication part? | ||
| nine | I also updated my thesis with the latest code changes. | ||
| cotto | yeah, that. ;) | ||
| nine | cotto: if someone tells me how to get the current time in winxed, sure ;) | ||
| cotto | I'm pretty sure there's an op for that. | 19:46 | |
| time | 19:48 | ||
| whiteknight | float i; ${ time i }; say(i); | 19:49 | |
| There might be a builtin too. Rosella provides a few pretty wrappers for things like that | |||
| benabik | floattime() is a builtin, I think. | 19:50 | |
| As is time(), I think. But that returns an int. | |||
| whiteknight | ah, right. floattime() | ||
| I forget about the builtins | |||
|
19:59
marcel_r joined
|
|||
| nine | Without setup, I'm at about 50 % speedup | 20:00 | |
| If now I could only get rid of these strange GC problems :/ | 20:04 | ||
| whiteknight | I'll help get those under control | 20:05 | |
| nine | :) | 20:06 | |
| Time for me to got o bed | 20:13 | ||
| alvis | qlq | 20:21 | |
| Util | Go ahead, alvis | ||
| alvis | I note the release_manager_guide.pod has cotto as the Release Manager for the April release, but there's no-one listed 'til dukeleto in October. Someone probably needs to put out a call for Release Managers. I'm happy todo another or two, but .... | ||
| Util | Will you put out the call, then? | 20:22 | |
| alvis | Sure | 20:23 | |
| Util | Great! It is better to ask for volenteers, knowing that you might fill a unanswered need yourself, rather than just doing the work; it leads to less burnout. | 20:24 | |
| Util is reminding himself of that, in fact. | 20:25 | ||
| alvis | You're right there. :) Gotta go. Baby's waking! | 20:26 | |
| Util | thanks again | ||
| dukeleto | alvis++ # put out the call for relmanagers | 20:29 | |
| alvis: also, congrats on the Baby.new()! | |||
|
20:58
lucian_ joined
23:02
benabik left
23:57
whiteknight joined
|
|||