|
github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm Set by AlexDaniel on 12 June 2018. |
|||
|
00:47
squashable6 left
01:44
leont left
02:20
hoelzro_ joined,
Altreus_ joined
02:27
moon-child left,
hoelzro left,
Altreus left
02:33
moon-child joined
03:08
Kaiepi joined
03:15
vrurg left
|
|||
| nwc10 | good *, #moarvm | 06:32 | |
|
06:59
domidumont joined
07:00
samcv left
07:05
sena_kun joined
|
|||
| nine | Good coffee time! | 07:11 | |
| nwc10 | yes, apart from the point when I spilled a bit | 07:16 | |
|
07:46
zakharyas joined
08:22
Altai-man joined
08:25
sena_kun left
09:02
Altreus_ is now known as Altreus
09:06
zakharyas left
09:08
zakharyas joined
|
|||
| jnthn | I pressed the hot water button instead of the coffee button. This is not useful. | 09:12 | |
| nwc10 | just add tea | 09:13 | |
| jnthn | I realized before it produced much water ("huh, this coffee looks weak") :) | ||
| nwc10 | or alternative "decaf lite"! | ||
| jnthn | And there is a cancel button :) | ||
| timotimo | praise the engineers for offering the ability to retry | 09:15 | |
| nwc10 | it was an Abort Retry Fail | 09:16 | |
| moritz | Abort To Orbit :D | 09:21 | |
|
10:09
domidumont left
|
|||
| timotimo | if it feels like you're sent into orbit, then what you just made wasn't coffee | 10:25 | |
| MasterDuke | canceling coffee is sort of like a deopt, right? | 10:35 | |
|
10:35
MasterDuke left
10:37
MasterDuke joined
10:53
leont joined
11:18
MasterDuke left
11:19
zakharyas left
11:30
MasterDuke joined
|
|||
| jnthn wonders if lunch will have made him any smarter... | 12:09 | ||
| Getting all the bits lined up for resume is tricky | |||
| nwc10 | you're resuming new-disp with resume? | 12:10 | |
| jnthn | Indeed | ||
| It's the big missing piece | 12:11 | ||
| And it's not just about "make it work" but also "make it work decently fast" | 12:12 | ||
| And make it specialize well, both with and without inlining | |||
| nwc10 | and you have to solve all these at the same time? | 12:13 | |
| jnthn | And dealing with the fact that callsame is the actual point that we do the `resume`, but that's megamorphic and the place we'd really like to attach any kind of PIC to is a level up | ||
| Well, yes if I don't want to throw lots away later. | 12:14 | ||
| I don't have to actually implement all the spesh/inline bits right off, but I need to know what, approximately, it should do | |||
| nwc10 | ah right. yes, that sort of design skill - "this looks like it's the right direction" | 12:15 | |
| "no obvious traps" | |||
|
12:18
domidumont joined
|
|||
| jnthn | That plus wanting relatively few mechanisms that solve many problems, rather than lots of special-case stuff | 12:19 | |
| Which'd be a pity to end up with at this point, given how nicely general the mechanism is up to this point in time | |||
| Also because it's easier to optimize something without too many semantic details | 12:20 | ||
|
12:23
sena_kun joined
12:24
zakharyas joined
12:25
Altai-man left
|
|||
| Geth | MoarVM: MasterDuke17++ created pull request #1352: Simplify and optimize isprime_I |
13:18 | |
|
13:32
domidumont left
|
|||
| jnthn | MasterDuke: We ended up with libtommath partly out of license concerns (although LGPL is probably OK; I always forget exactly what it enforces), but in a bigger part because of libtommath being far easier to build across platforms. | 15:02 | |
| MasterDuke: Of note, stock gmp doesn't build on the MSVC compiler. | 15:03 | ||
| There is these days a fork that does: mpir.org/ | |||
| (Noting that at the time MSVC was my primary compiler, and there was no way there was time to go and shave that yak) | 15:05 | ||
| Phew, I think I've finally got most of a design that might actually work out | 15:13 | ||
| And doesn't have the implementation problems my first idea had | |||
| Well, first N ideas :) | 15:16 | ||
| Think I'll sleep on this and have a got at implementation starting tomorrow. | |||
| *go | 15:17 | ||
| MasterDuke | jnthn: thanks, useful info | 15:21 | |
| [Coke] wonders why hacktoberfest repo is not using artistic. | 15:35 | ||
| ww | |||
| MasterDuke | fwiw, github.com/ShiftMediaProject/gmp seems more up-to-date | 15:37 | |
| jnthn | ah, nice :) | 16:12 | |
| Given that exists, we maybe can think about a move to gmp, though of course it needs a vict^Wvolunteer... | 16:13 | ||
| And it's indeed a non-small task | |||
| MasterDuke | my guess is that getting it integrated into the build is going to be more work than the code changes | 16:14 | |
| since we aren't doing very complicated math ourselves, mostly just offering a passthrough to their functions | 16:15 | ||
| jnthn | Yeah, mine too | ||
| Build stuff is always horrible :( | |||
| MasterDuke | yeah. i would be interested in attempting the code changes, but not sure if i want to try and get it building... | 16:17 | |
| i wonder how much work it would be to just get it building locally, make any necessary code changes, and then leave it up to someone else to do the full build integration | 16:19 | ||
|
16:22
Altai-man joined
16:25
sena_kun left
16:37
lucasb joined
|
|||
| tobs | Out of curiosity, what would be required build-integration-wise to use GMP? Looking at Configure.pl, you look for a system libtommath and otherwise build the bundled one, right? Are there any unusual snares here or is it "this must work on all operating systems" kind of worries? | 18:39 | |
| tobs just made Perl "alien" modules for many things at work, including a custom version of GMP | 18:40 | ||
| nwc10 | I don't know for sure, but I'm assuming it's just the two you say - use a system one if found, and then "must work on all operating systems" | 18:44 | |
| tobs | oh, it seems that all 3rdparty libraries currently bundled are pretty simple to build, judging from github.com/MoarVM/MoarVM/blob/mast...L585-L616. GMP would definitely need more encouragement from there but I'm inclined to try. | 18:53 | |
| MasterDuke | oooo, that would be very cool | 18:55 | |
| tobs: how long do you think it would take you? | 19:08 | ||
| tobs | MasterDuke: I already started, but I have to understand Configure.pl too. This would be my first contribution to moar. In total I'd say 24 hours from now there is either a first candidate or I have discovered how hard it really is :P | 19:11 | |
| MasterDuke | oh nice, no complaints here! | 19:12 | |
|
19:12
zakharyas left
|
|||
| MasterDuke | then i'm not even going to bother trying to get a local only build right now | 19:12 | |
|
19:26
lucasb left
|
|||
| [Coke] | tobs++ | 19:32 | |
| jnthn | tobs++ MasterDuke++ | 19:44 | |
|
20:20
lucasb joined
20:23
sena_kun joined
20:25
Altai-man left
20:49
tbrowder left
20:50
kawaii left,
chansen_ left
20:51
tbrowder joined
20:52
kawaii joined
20:54
chansen_ joined
20:59
sena_kun left
|
|||
| tobs | I just had the weirdest thing happen to me. While compiling moar my laptop wanted to go into hibernation but froze instead. For 30 minutes I tried every mechanism on the web to get out of this cleanly because this is my work laptop with 12 workspaces and 200 days uptime worth of shell history, if nothing else, and an encrypted disk that would not be cleanly unmounted. | 21:19 | |
| After 30 minutes and no avail I held down the power button for 5s and rebooted. Amazingly, when it started back up, it continued right where it apparently failed to hibernate, all workspaces and files intact (as far as I can see). | 21:20 | ||
| jnthn | Phew | ||
| AlexDaniel | tobs: how about using infinite bash history? | 21:21 | |
| timotimo | it could have been "suspend to both" where it suspends to ram but also to disk | ||
| AlexDaniel | something like | 21:22 | |
| export HISTSIZE='' | |||
| export HISTFILESIZE='' | |||
| export HISTCONTROL=ignoredups:erasedups # avoid duplicates | |||
| shopt -s histappend # append history entries | |||
| tobs | AlexDaniel: I do but it still won't survive a system crash where it is never written to disk. | ||
| jnthn | I wish my work machine would manage 200 days without freezing up (I get a week or two :/). Pretty sure it's something graphics related, given I get some fascinating mis-renders of things shortly before a lot of them. | ||
| AlexDaniel | tobs: ohhh that's probably true | ||
| timotimo | mis-renderings can often be a heat related thing | ||
| jnthn | Hm, interesting. | 21:23 | |
| timotimo | this is a desktop machine? nvidia or amd gpu or intel built-in graphics? | 21:24 | |
| tobs | Guess this also calls for a refresher backup to appease the computer gods :) | ||
| timotimo | there's usualy a simple-ish way to get at the gpu temperature | ||
| jnthn | timotimo: Desktop. Completely forgot and not at the machine right now to check, though pretty sure not built-in. | 21:25 | |
| AlexDaniel | tobs: maybe also something like this? askubuntu.com/a/67306 | ||
| tobs: I remember losing the last few command and being really bummed when having to type them again :D | |||
| timotimo | i have a munin set up just for fun, i get graphs of, among other things, temperatures for years and such | ||
| AlexDaniel | doing this in PROMPT_COMMAND feels a bit sketchy to me, but hey it should work | 21:26 | |
| jnthn | But yeah, I'd just assumed dodgy graphics driver or something else in the graphics stack (it seems to render stuff in the wrong places, like rendering bits of a different browser tab in another tab, etc.) Heat is an interesting idea. | 21:27 | |
| AlexDaniel | jnthn: if it freezes up because of graphics, shouldn't it be possible to SSH to it anyway? | 21:28 | |
| jnthn | I was going to add "always seems to be Chrome", then realized that everything else I have is either a terminal or an IntelliJ (or such) instance. | ||
| AlexDaniel | not necessarily if the kernel freaks out because of it, but still | ||
| jnthn | AlexDaniel: Didn't try, though I'm not sure what'd be possible even if I can. | 21:30 | |
| Maybe I should try not running an Ubuntu that's barely still in support also... :) | |||
| MasterDuke | i had some excitement this morning after applying a bios update. it rebooted and lost my modified boot order so it booted off my old drive. i had cloned the old onto the new, so i didn't immediately notice. but something was a little off (wrong bash history) and i rebooted and selected the write drive order and everything was fine | 21:31 | |
| AlexDaniel | MasterDuke: xD | ||
| MasterDuke | hm. i should wipe the old drive now that everything has been fine for 8 months so this doesn't happen again | 21:32 | |
|
21:55
SmokeMachine left
21:56
SmokeMachine joined
|
|||
| jnthn | Why oh why does the checks page have the text "Check failure on line 180 in Build log" and then you click it and end up on the files changed tab? :/ | 21:56 | |
| Talking about github.com/MoarVM/MoarVM/pull/1352...1155104345 | |||
| Then hidden at the bottom is the link to the azure page to actually see what's up | 21:57 | ||
| Anyway, there's an NQP test failure | 21:58 | ||
| MasterDuke | jnthn: yeah, the test is calling nqp::isprime_I with two args. locally i just removed the second (now un-needed) arg and it passes | 21:59 | |
| jnthn | ah, I see | ||
| Rakudo will need a patch also? | 22:00 | ||
| MasterDuke | yep, just removing the ',100' here github.com/rakudo/rakudo/blob/mast...t.pm6#L160 | 22:01 | |
| jnthn | 'k | ||
| I'll look over the PR properly tomorrow. | 22:02 | ||
| MasterDuke | cool, i should be around in the evening | ||
| jnthn | Will have a crack at implementing the latest dispatch resumption design tomorrow, to find out why this variant won't work... | 22:05 | |
|
22:30
lucasb left
22:50
SmokeMachine left
22:51
tbrowder left
22:52
SmokeMachine joined
22:54
tbrowder joined
23:09
raku-bridge left,
raku-bridge joined,
raku-bridge left,
raku-bridge joined
23:49
leont left
|
|||