00:33 lizmat joined 01:56 ilbot3 joined 02:03 MasterDuke joined 03:13 evalable6 joined 03:33 evalable6 joined 05:33 lizmat joined 06:32 domidumont joined 06:36 robertle joined 06:39 domidumont joined 08:27 domidumont joined 11:37 Geth__ joined 11:55 MasterDuke joined 12:06 Geth joined
Zoffix Would someone with keys check all of MoarVM org's repos that have webhooks for bot commits announcements have them set up to with URL hack.p6c.org:8888/?chan=#moarvm ? I already done it for MoarVM/MoarVM, but don't have access to other repos 12:11
12:38 domidumont joined 15:17 zakharyas joined 15:23 zakharyas joined 16:04 dogbert17 joined 17:21 releasable6 joined 18:00 rba_ joined 18:09 AlexDaniel- joined 19:16 lizmat joined
Zoffix So we got some fork of libtommath or somethin? They made second fix for expmod, how do we pull in this fix? github.com/libtom/libtommath/issue...-336654018 20:06
timotimo yeah we have a libtommath repo under moarvm/ 20:12
20:13 robertle joined
timotimo i'm not entirely sure why we made our own fork 20:13
Zoffix timotimo: do I just cherry pick that commit? into that repo and moarvm will automagically use it? 20:18
timotimo Zoffix: looks like we literally just have the fork for the inclomplete fix for the expmod thing
Zoffix hah
Maybe I shouldn't touch this.
timotimo you'd also want to check out the newer commit in the submodule 20:19
hm. another DJ just started and they got like 2x the volume compared to the previous one
Zoffix Yeah, I remember all the issues we had with libtommath repo last time during builds. So I'm not gonna touch it, 'cause I'm a n00b. But they added a new fix to expmod and asking if it's working, so if someone could try it out and respond that'd be great: github.com/libtom/libtommath/issue...-336654018 20:24
Zoffix opens an Issue 20:25
timotimo oh
was that libtommath, yeah?
that was when it wasn't a submodule and i turned it into one
and nobody could "git pull" because there was this mysterious "libtommath" directory that would have been overwritten with a "libtommath" directory 20:26
and clearly that's a no-go
Zoffix There was alrteady issue for it. I commented there: github.com/MoarVM/MoarVM/issues/54...-336664939 20:27
timotimo i got oropax (earplugs) and a power cord now 20:32
why am i doing this on my laptop %) 20:37
Geth MoarVM/master: 4 commits pushed by (Samantha McVey)++ 20:40
samcv getting that merged
timotimo i'm running spectests with very latest libtommath now 20:42
haha, they really have something called SLURM? 20:46
that's amazing. poznan++ 20:48
Zoffix rofl
timotimo q:to/qapla'/ actually works <Ā³ 20:49
er
<3
samcv slurm?
what is SLURM in libtommath 20:50
20:50 TimToady joined 20:51 robertle joined
timotimo no, it's for scheduling and such 20:51
for clusters
Simple Linux Utility for Resource Management (SLURM) 20:52
i actually wrong-windowed slurm into #moarvm even though i meant it to be in #perl5
...
#perl6
huh. i grabbed the newest libtommath stuff and the expmod test file now hangs? 20:53
oooooh 20:54
it's in a branch
let's try that again 20:55
i assume i'd have to add an extra test or two to the file which used to hang it 20:57
that was a bad formulation 20:58
mst I was assuming it was another joke reference to Futurama 21:01
timotimo it probably i 21:05
is
there's just one fudge ni the expmod file and that passes with the fixed branch of libtom 21:07
Zoffix timotimo: in roast? Yeah, there's one fudge I wrote for some ticket where I spotted that it hangs with more than just `-1` for an arg (or whatever it was) 21:08
timotimo right 21:09
should i systematically try lots of different arg combinations? 21:10
Zoffix prolly
timotimo i can also push The Stuff if you'd like to try yourself?
Zoffix push 21:11
oh wait
I don't wanna try
Just push so it's awailable for everyone :) 21:12
timotimo :)
junctions ftw
expmod(any(^100), any(^100), any(^100))
Zoffix :D
timotimo it prints a result
so that means it doesn't hang
it doesn't parallelize though 21:13
Zoffix you didn't exercise the bug with ^100 tho
it needed negatives or something
timotimo oh
which one?
ok now with ^1000 three times it either hangs or takes a long time, and i'm not sure which :D 21:16
Zoffix m: expmod(42, -1, 7) 21:17
camelia (timeout)
timotimo expmod(any(^100), -1, any(^100)) doesn't hang 21:18
Zoffix \o/ 21:19
timotimo hm. it's rather slow, but at least expmod accounts for 72.4% of total run time 21:28
it looks like it's preventing the containing block from being jitted
it'd be trivial to jit that op 21:29
let's see how much faster that would make it
pff, what 21:31
expmod takes the result as an argument. i'd have to fix that first
damn, if we pull the repr_alloc_init out of the bigint op functions again later we can spesh them into optimized object initialization 21:34
don't actually wanna do it right now 21:37
i'd be interested to hear how far we are away from combining a spesh fact "known type" with a "call the allocate reprop" from inside an expr tree in the new jit 21:42
(paging dr brrt)
21:43 MasterDuke joined 22:21 rba joined
Geth MoarVM/fix_for_expmod: 214a0dadb4 | (Timo Paulssen)++ | 3rdparty/libtommath
fix for invmod
23:02
23:23 MasterDuke joined 23:33 TimToady joined 23:54 evalable6 joined