|
github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm Set by AlexDaniel on 12 June 2018. |
|||
| [Coke] fires up a build. | 00:32 | ||
| passed here, 10.13.6 | 00:43 | ||
| MasterDuke | github.com/perl6/nqp/issues/274 | 00:49 | |
|
02:08
mst joined,
ChanServ sets mode: +o mst
03:15
stmuk joined
03:17
stmuk_ left
04:40
squashable6 joined
06:05
MasterDuke left
06:51
robertle joined
06:52
domidumont joined
06:58
domidumont left
06:59
domidumont joined
|
|||
| masak | jnthn, timotimo: do we still have that AngularJS app for doing metric on GC/JIT etc? | 07:47 | |
| I remember it being slow, and I remember semi-promising to make it faster. :) I've actually done that now with other things. the trick is to remove the AngularJS | 07:48 | ||
| robertle | haha! so the trick to making angular faster is not using angular? I'll remember that... | 07:53 | |
| masak | robertle: you think it's a joke... :P | 07:57 | |
| there's a tweet somewhere that goes further. something like "we hired this JavaScript performance expert to help make our site faster. she advised us to get rid of most of the JavaScript" | 07:58 | ||
|
08:05
zakharyas joined
08:10
lizmat left
08:23
lizmat joined
|
|||
| jnthn | masak: fwiw, I think timotimo is building a new profiler interface (not using Angular :)) | 09:08 | |
| masak | oh! | 09:16 | |
| good for him! :) | |||
| is it published somewhere? | |||
| tadzik | masak: oh, that might be a joke, but reminds me of a story of a certain nosql database :) | 09:17 | |
| where a trick to improve performance by a few orders of magnitute was disabling one of the 8 monster servers powering its cluster | |||
| masak | tadzik: I guess that's why "serverless" is so hyped these days :P | 09:18 | |
| tadzik | just imagine 8 boxes, each with 32 cores (64 threads), 128G of ram and a 1TB of storage, all put together putting up a whooping... 100req/s | ||
| masak: well, if it scales linearly like that... :P | 09:19 | ||
| masak | tadzik: sometimes I think the engineers who built the computers that landed us on the moon would look at our current practices and cry in anguish | 09:22 | |
| tadzik | masak: probably :) | 09:24 | |
| I used to find it funny when my phone was running quake 3 more smoothly than angry birds. Who can one's engineering be *so much* better than the other? These days that phone wouldn't even be able to run Everyone's Favourite Chat App[tm] | 09:25 | ||
| it's a known fact that software is getting slower more rapidly than hardware is getting faster, but we've really accelerated that process in recent years, it feels | 09:26 | ||
| masak | ah, yes. the tower of abstractions is so high you can't see the ground anymore | 09:34 | |
| tadzik | I read that as "can't see the ground above" :P | 09:35 | |
| masak | well, the way we draw trees in CS, it only stands to reason the ground would be found above | 09:40 | |
| tadzik | back to the roots | 09:44 | |
|
09:50
timotimo joined
|
|||
| timotimo | masak: i'd have already published it, were i not so ashamed of the code quality :| | 09:52 | |
| though it ought to be public soon, after all it's grant work for tpf | |||
| masak | timotimo: your call completely -- but consider me interested | 09:54 | |
| "your profiler is intriguing to me and I wish to subscribe to your newsletter" :P | 09:55 | ||
| timotimo | wakelift.de ;) | ||
| tadzik | "here's my seal of consent to you handling my email address" :P | ||
| masak | tadzik: I used to use seals for expressing consent, but the zoo called me and said don't do that | 09:56 | |
| tadzik | queue Ylvis - The language of love | 09:57 | |
| (youtu.be/MCxDJnHiLqs?t=123) | 09:58 | ||
| right, time to get to work... :P | |||
| timotimo | oh lord i haven't seen that one yet | 10:01 | |
| tadzik | I never thought I'd voluntarily get back to it :P | 10:02 | |
| masak | tadzik: that made my day, thanks | 10:03 | |
| timotimo | baeutiful cgi seal .. sealGI? | ||
| tadzik | I CAN'T UNDERSTAND WHAT YOU SAID | ||
| masak | timotimo: I have now read all your blog posts. please write more of them. | 10:04 | |
| timotimo | masak: yes, it goes by way too fast, doesn't it | ||
| that "asian technique" is something else, though | |||
| masak | I can confirm the asian technique locally. | ||
| masak .oO( works on my box! ) | 10:05 | ||
| tadzik | ...asian technique? | ||
| masak | tadzik: it's in the video you posted | ||
| there are... tongs | 10:06 | ||
| timotimo | tongs on the tongue | ||
| tadzik | oh | ||
| I have to admit, I only watched the sealed part :) | |||
| timotimo | oh, OK | ||
| tadzik | time to catch up | 10:07 | |
|
10:07
domidumont left
|
|||
| tadzik | ......oh | 10:08 | |
| masak | :P | 10:10 | |
|
11:03
zakharyas left
11:30
domidumont joined
11:38
robertle left
11:40
robertle joined
11:59
robertle left
12:10
robertle joined
12:24
diakopter left
12:39
zakharyas joined
14:14
domidumont left
14:15
domidumont joined
14:22
domidumont left
14:23
domidumont joined
|
|||
| timotimo | masak: your blog has a mascot!!! | 14:25 | |
| masak | :D | 14:27 | |
| timotimo | a masakscot! | ||
|
14:31
AlexDaniel joined,
AlexDaniel` joined
14:33
ZofBot joined
14:35
AlexDaniel left
14:36
AlexDaniel joined
14:38
AlexDaniel left,
AlexDaniel joined
15:24
robertle left
15:42
domidumont left
15:52
zakharyas left
15:53
robertle joined
|
|||
| Geth | MoarVM/pea: e7c1855c5b | (Jonathan Worthington)++ | 13 files Minimal escape analysis and scalar replacement This only handles the case where: * No aliasing * No deoptimizing instructions * No control flow (conditionals, loops) ... (7 more lines) |
16:07 | |
| jnthn | gist.github.com/jnthn/580ca984c634...183fb18ae8 runs in 60% of the time that it took before that optimization :) | 16:12 | |
| timotimo | nice improvement | 16:15 | |
| jnthn | Yeah. It's too limited for any really interesting cases yet, but it's nice to have a starting point to bulid upon | 16:19 | |
| timotimo | of course | ||
| masak | I suspect the tricky parts of EA come in the aliasing and the control flow | 16:22 | |
| but it seems to me that having it be partial will actually help much of the time | |||
| timotimo | mhm | ||
| Geth | MoarVM/pea: 137ab0f6e6 | (Carl Masak)++ | src/spesh/usages.h Reset MVM_SPESH_CHECK_DU to 0 |
16:32 | |
| MoarVM/pea: 025bdfd7d5 | (Carl Masak)++ | src/spesh/manipulate.c Add missing comment to a new function |
16:36 | ||
|
17:50
Ven`` joined
17:55
Ven`` left
17:57
Ven`` joined
18:14
Ven`` left
18:18
Ven`` joined
18:22
Ven`` left
|
|||
| lizmat | .ask jnthn in lib/experimental.pm6 for "is cached", is the nqp::bindattr_i still needed ? | 19:07 | |
| yoleaux | lizmat: I'll pass your message to jnthn. | ||
|
19:17
domidumont joined
19:24
domidumont left
19:25
Ven`` joined
19:35
Ven`` left
20:11
MasterDuke joined
20:31
robertle left
20:39
avar joined,
avar left,
avar joined
|
|||
| lizmat | .tell jnthn related to that, I've done github.com/rakudo/rakudo/commit/434ede491d | 20:56 | |
| yoleaux | lizmat: I'll pass your message to jnthn. | ||
| jnthn | .tell lizmat 434ede49 looks reasonable to me | 21:17 | |
| yoleaux | jnthn: I'll pass your message to lizmat. | ||
| 19:07Z <lizmat> jnthn: in lib/experimental.pm6 for "is cached", is the nqp::bindattr_i still needed ? | |||
| 20:56Z <lizmat> jnthn: related to that, I've done github.com/rakudo/rakudo/commit/434ede491d | |||
| lizmat | ok, then I'll remove the bindattr_i from lib/experimental, as that is now handled by Routine.wrap | 21:18 | |
| yoleaux | 21:17Z <jnthn> lizmat: 434ede49 looks reasonable to me | ||