github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
brrt \o 10:21
I now think most of the issues we're having may be related to signs 10:22
dogbert11 hello brrt 10:35
timotimo that's a bad sign? 13:43
dogbert11 it's a sign of the times 13:46
brrt test case; trunc is dead 14:44
gist.github.com/bdw/0dd3ece19d8771...2fb7ce1216
timotimo nqp-m: gist.github.com/bdw/0dd3ece19d8771...2fb7ce1216 14:45
camelia ERROR
at <tmp>:12 (<ephemeral file>:<mainline>)
from gen/moar/stage2/NQPHLL.nqp:1685 (/home/camelia/rakudo-m-inst-1/share/nqp/lib/NQPHLL.moarvm:eval)
from gen/moar/stage2/NQPHLL.nqp:1928 (/home/camelia/rakudo-m-inst-1/share/nqp/lib/NQPHLL.moarvm:ev…
dogbert11 hmm, that doesn't look very good 15:57
dogbert11 tries to figure out when this happened ... 16:01
timotimo use bisectable with a "run" command to invoke nqp-m 16:03
dogbert11 is that the way to do it? 16:08
timotimo it's *a* way to do it :) 16:21
you could of course do it manually if you like
i wouldn't like it ;)
dogbert11 I'm trying manueal ATM and as you write it's not very fun :) 16:37
dogbert11 perhaps AlexDaniel will solve this quicker but here's what we have so far 16:45
nqp commit github.com/perl6/nqp/commit/8476ec...9948ca6c67 is broken 16:46
I got it wrong immediately github.com/perl6/nqp/commit/8476ec...9948ca6c67 is working but github.com/perl6/nqp/commit/4acf85...2272c3cc7f does not 16:47
AlexDaniel e: chdir ‘sandbox’; run ‘wget’, ‘gist.githubusercontent.com/bdw/0dd...trunc.nqp’ 16:52
dogbert11 between the two we have nine's recent work in nqp and MoarVM which doesn't lend itself to bisection
AlexDaniel c: HEAD chdir ‘sandbox’; run ‘wget’, ‘gist.githubusercontent.com/bdw/0dd...trunc.nqp’
where are the bots O_o
dogbert11 perhaps they ran away :)
AlexDaniel dogbert11: banned 16:54
dogbert11 how come 16:55
MVM_JIT_EXPR_DISABLE=1 seems to resolve the problem 16:56
AlexDaniel timotimo: btw do you know about this? gist.github.com/AlexDaniel/88b38af...3df31dd431 16:57
basically it bisects with builds from whateverable, but locally 16:58
so you can run any code you need
timotimo cool, thanks 16:59
bookmarked
AlexDaniel I wish somebody turned it into App::Bisect module, or something like that 17:00
timotimo CoreHackers may be the right namespace for that 17:02
AlexDaniel or that, ya 17:03
yoleaux 17:01Z <lizmat> AlexDaniel: I've documented all of my commits for this release (I think)
AlexDaniel timotimo: ah, also you'll need to read some comments in this file if you decide to use it github.com/perl6/whateverable/blob...on.pm6#L29 17:17
lizmat .ask samcv how hard would it be to create a "nqp::iterleft" op on an nqp::iterator object that would tell you how many elements can still be produced by the iterator 17:23
yoleaux lizmat: I'll pass your message to samcv.
lizmat that would allow us to make all Mappy iterators to easily become PredictiveIterators 17:24
samcv lizmat, for hashes specifically?
yoleaux 17:23Z <lizmat> samcv: how hard would it be to create a "nqp::iterleft" op on an nqp::iterator object that would tell you how many elements can still be produced by the iterator
timotimo how should it behave when the count is unknown? 17:25
we could have different "modes" of iterators in the future
lizmat return -1 ? 17:26
samcv lizmat, also i can do that for hashes
yes 17:27
lizmat I mean, if it can be done as a cheap operation...
(very) cheap
samcv well actually. hmm
lizmat if not, then just scratch the idea :-) 17:28
what happens if you do nqp::elems on an nqp hash ?
samcv we can quickly know how many are in the other buckets but i don't think we know where in the current bucket we are
timotimo for a hash it might have to count how often it has been iterated so far
samcv lizmat, that is very fast. it just returns the hash's count in the hash struct
timotimo because, yeah, you can't easily tell what you're pointing at at the moment
you can, however, just iterate the iterator to the finish line to figure out how long it takes, lol
lizmat samcv: so maybe an 'nqp::iterdone' op that produces the number values produced ? 17:30
but I guess you're not keeping that now, are you ?
timotimo well 17:31
the first entry in MVMIter is uint32 mode; there's a 32bit wide gap after that before the pointer to the target
and we only have two different modes at the moment anyway :)
oh
fourm odes
samcv yeah no we don't keep how far we've gone 17:32
timotimo in arrays we do, though, because that's how we go from one to the next there :)
"unsigned", is that a 32bit int? 17:33
lizmat gotcha... :-)
samcv: ok, so scratch the idea, it's probably not worth it
timotimo fwiw, just unbox_i on the iterator could give the number of iterations so far 17:34
so there doesn't even have to be a new op
lizmat hmmm... I don't think we use nqp::iterator on arrays in core 17:35
it's all based on atpos
timotimo we probably do if we "for @foo" in nqp code
lizmat m: my @a = ^10; use nqp; my $i := nqp::iterator(nqp::getattr(@a,List,q/$!reified/)); nqp::while($i,nqp::say(nqp::iterval(nqp::shift($i)))) 17:55
camelia This is not an iterator, it's a P6opaque (Scalar)
in block <unit> at <tmp> line 1
lizmat timotimo: ^^^ what am I doing wrong ?
timotimo forgot a decont 17:56
also
i'm not sure nqp::shift($i) returns the iterator itself
lizmat m: my @a = ^10; use nqp; my $i := nqp::iterator(nqp::getattr(@a,List,q/$!reified/)); nqp::while($i,nqp::stmts(nqp::shift($i),nqp::say(~nqp::iterval($i)))) 17:58
camelia 0
1
2
3
4
5
6
7
8
9
lizmat whee 17:59
AlexDaniel c: HEAD chdir ‘sandbox’; run ‘wget’, ‘gist.githubusercontent.com/bdw/0dd...trunc.nqp’ 18:00
committable6 AlexDaniel, gist.github.com/9fc76730444284fcad...221d5f5531
lizmat m: my @a = ^10 .reverse; use nqp; my $i := nqp::iterator(nqp::getattr(@a,List,q/$!reified/)); nqp::while($i,nqp::say(~nqp::shift($i)))
camelia 9
8
7
6
5
4
3
2
1
0
lizmat looks like nqp::shift already returns the next value
AlexDaniel c: HEAD run <nqp-m sandbox/trunc.nqp>
committable6 AlexDaniel, gist.github.com/accf0a11a05cfa166a...e6e6d09e72
timotimo cool 18:01
in that case, it's probably only necessary for hashes
or maybe even in that case you just need iterkey 18:02
AlexDaniel c: 2018.10 run <nqp-m sandbox/trunc.nqp> 18:02
committable6 AlexDaniel, ¦2018.10: «»
AlexDaniel bisect: old=2018.10 run <nqp-m sandbox/trunc.nqp>
bisectable6 AlexDaniel, Bisecting by exit code (old=2018.10 new=86ce4f4). Old exit code: 0 18:03
AlexDaniel hey that was easy?
bisectable6 AlexDaniel, bisect log: gist.github.com/65bb18fee0661dcfb5...18f124e6f8
AlexDaniel, (2018-10-30) github.com/rakudo/rakudo/commit/b8...08b1c37fe8
AlexDaniel c: b8176dc5bf^,b8176dc5bf run <nqp-m sandbox/trunc.nqp>
committable6 AlexDaniel, gist.github.com/92fc2e97799fc869f8...79930c606b
AlexDaniel who was looking for it? dogbert11? timotimo?
timotimo dogbert i think
AlexDaniel brrt?
.tell brrt bisected: colabti.org/irclogger/irclogger_lo...11-17#l196 18:04
yoleaux AlexDaniel: I'll pass your message to brrt.
AlexDaniel is there a ticket for it?
timotimo probably not yet
dogbert11 well that commit does quite a lot :) 18:06
timotimo yeah, we can't really bisect in between bumps yet 18:07
since whateverable builds rakudos up front, and it'd be immense to build them with every in-between commit up front
AlexDaniel no, I almost did it actually
the code started to get very hairy though, so I abandoned my attempt… 18:08
so yeah, it's not an issue and we can and should do it
someone just has to write some nice code for it :)
btw it's not M × N × R, it's just M + N 18:12
as in we don't have to try all combinations, just shift the *_VERSION a bit on bumps 18:13
github.com/perl6/whateverable/issues/40
dogbert11 holy shit, I think I found it 18:20
lizmat dogbert11++ 18:20
dogbert11 it seems to be this one github.com/MoarVM/MoarVM/commit/2c...2b4b131976
dogbert11 haven't done assembler since the 68000 days 18:21
dogbert11 something for nine or brrt or timotimo I guess 18:22
AlexDaniel I'm very tired today so I'll go to bed soon. See you all tomorrow! 18:33
lizmat good night, AlexDaniel++ 18:35
lizmat timotimo samcv did some benchmarks, but any different between an atpos based iterator and one based one nqp::iterator is lost in the noise 19:14
so I'm not going to pursue that atm
(as somehow it also breaks the install process when I do that change) 19:15
timotimo OK 19:34
nine dogbert11: how did you end up with that commit? With the expr JIT disabled brrt++'s testcase works fine. Same when I just comment out the expr JIT template for trunc_i16. 22:37
nine dogbert11: now the weird thing is that I think, my lego JIT implementation of trunc_i16 is just as wrong. It does exactly what the expr JIT template claims to do. I think, trunc_i16 should actually be exactly the same as extend_i16. We probably just emit a trunc_i16 in places where we really do use only the lower 16 bits. 22:40
dogbert11 hello nine 22:52
there are three files changed in that commit. I took the ones from the provious commit anf copied them to master. Real shady :) 22:54
*previous
I then wiped the rakudo install dir and rebuilt everything 22:56
probably not the best way to go about things 23:00

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!