Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2015-09-03

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
02:58 ShimmerFairy joined #moarvm
04:27 virtualsue joined #moarvm
05:51 lizmat joined #moarvm
06:12 virtualsue joined #moarvm
06:35 xiaomiao joined #moarvm
06:45 FROGGS joined #moarvm
07:03 Ven joined #moarvm
07:29 ShimmerFairy joined #moarvm
07:33 lizmat joined #moarvm
08:51 Ven joined #moarvm
09:42 lizmat joined #moarvm
09:51 zakharyas joined #moarvm
10:13 Ven joined #moarvm
11:43 Ven joined #moarvm
12:04 lizmat joined #moarvm
12:44 lizmat joined #moarvm
13:03 zakharyas joined #moarvm
13:12 hoelzro good moarning #moarvm
13:12 nwc10 good *, hoelzro
13:14 timotimo heyo you two
13:15 hoelzro o/ nwc10 & timotimo!
13:17 hoelzro it turns out there's a bug in rakudo-m where if you create a Thread, it will never clean up the underlying ThreadContext.  I tried to add MVM_tc_destroy(thread->body.tc) to MVM_thread_join, but that results in segfaults and other fun.  Is there some other cleanup I need to be doing?
13:20 timotimo doesn't the interpreter have a list of all threads?
13:20 timotimo i think that's necessary for the gc to work properly
13:23 timotimo hm. i should get an overview of the things i've been doing (or rather trying to do) during spw
13:24 timotimo i tried to move forward with the lexical-to-local lowering and got baffled by things once more
13:24 timotimo i've made perl6-bench glr-compatible
13:24 timotimo what else? :\
13:24 hoelzro timotimo: does that thread still need a TC after the thread has been joined?
13:25 hoelzro and should the thread be removed from that instance thread listing?
13:25 timotimo probably needs to be kept, as the thread contexts all have their own nursery and gen2
13:25 hoelzro re: what you've been up to, I just remember the lexical -> local stuff
13:25 hoelzro ahhhh
13:25 hoelzro that's probably why it's mad
13:25 timotimo maybe we want to have newly created threads claim previous TCs
13:29 lizmat joined #moarvm
13:32 brrt joined #moarvm
13:33 brrt good *
13:34 timotimo brrt * good
13:34 brrt hi timotimo
13:37 hoelzro o/ brrt
13:37 brrt \o hoelzro
13:37 brrt are you at yapc::eu?
13:38 hoelzro timotimo: A joined thread can no longer allocate memory in its pool, right?
13:38 hoelzro brrt: I am not (
13:38 hoelzro =(
13:38 brrt shame
13:38 brrt it's really nice up here
13:38 brrt granada is awesome
13:38 brrt perl community is really alive this year
13:38 nwc10 Klima!
13:38 [Coke] I'mma let you finish, but SPW2015 was clearly the best european workshop of all time.
13:38 nwc10 however, the big blue room is a bit warmer than the venue
13:39 brrt i was clearly not at SPW2015
13:39 timotimo [Coke]: because of the kitties, right?
13:39 brrt nwc10: where are you now? we haven't talked all conference, and i'm leaving tomorrow morning
13:39 * [Coke] wishes he could have gone to all 3 conferences in a row. Ah well.
13:39 nwc10 brrt: second row in miyagawa's talk
13:39 [Coke] timotimo: Loved that cat, man.
13:39 brrt it was hard enough justifying one conference :-)
13:39 nwc10 I'll try to find you at coffee
13:39 brrt that's F1 right
13:39 brrt aye
13:39 [Coke] brrt: exactly. :)
13:40 timotimo brrt: may i ask you about the slides again?
13:40 brrt yes. if you have somewhere for me to put them, i will be happy to upload them
13:40 timotimo you have a google account, don't you?
13:41 timotimo we can use something like https://reep.io/ as well
13:42 timotimo https://rtccopy.com/ or this
13:42 timotimo https://rtccopy.com/#g6ZrBarS
13:43 brrt timotimo: https://github.com/bdw/slides/blob/master/Bart%20Wiegmans%20-%20YAPC::EU%202015-09-02.pdf
13:44 timotimo wow, github can display pdfs in-line
13:44 brrt nice
13:52 timotimo brrt: you're probably busy attending things rather than hacking on the jit, right?
13:53 brrt i'm actually working on the freeing-to-many-registers problem now
13:53 brrt (sorry tux, if you're reading it)
13:54 timotimo cool :)
13:56 brrt we're trying to free rbx
13:56 brrt ....
13:56 brrt wth
14:04 dalek MoarVM/even-moar-jit: 2bfd5dd | brrt++ | src/ (6 files):
14:04 dalek MoarVM/even-moar-jit: Introduce 'immortally' created values
14:04 dalek MoarVM/even-moar-jit:
14:04 dalek MoarVM/even-moar-jit: From the perspective of the register allocator, values residing
14:04 dalek MoarVM/even-moar-jit: in nonvolatile registers are 'immortal'. These registers should
14:04 dalek MoarVM/even-moar-jit: not be freed over overwritten.
14:04 dalek MoarVM/even-moar-jit:
14:04 dalek MoarVM/even-moar-jit: This does not fix all the bugs, but some of them.
14:04 dalek MoarVM/even-moar-jit: Also implement the handlers for pre-and-post branch logic,
14:04 dalek MoarVM/even-moar-jit: but we do not yet call them.
14:04 dalek MoarVM/even-moar-jit: review: https://github.com/MoarVM/MoarVM/commit/2bfd5dda3d
14:10 dalek MoarVM: 15eadcd | paultcochrane++ | src/spesh/candidate.c:
14:10 dalek MoarVM: Revert func_ptr free from commit 688eecd
14:10 dalek MoarVM:
14:10 dalek MoarVM: Unfortunately freeing the func_ptr wasn't the correct thing to do; the
14:10 dalek MoarVM: change actually caused a bad free of a variable.  Sorry about that!  This
14:10 dalek MoarVM: reverts the commit and thus removes a potentially bad free operation.
14:10 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/15eadcde04
14:10 dalek MoarVM: ac6ea4e | lizmat++ | src/spesh/candidate.c:
14:10 dalek MoarVM: Merge pull request #253 from paultcochrane/pr/revert-free-func_ptr
14:10 dalek MoarVM:
14:10 dalek MoarVM: Revert func_ptr free from commit 688eecd
14:10 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/ac6ea4e232
14:26 Ven joined #moarvm
14:39 rarara joined #moarvm
15:00 kjs_ joined #moarvm
15:25 brrt joined #moarvm
15:31 colomon joined #moarvm
15:47 dalek MoarVM/even-moar-jit: baa5b8a | brrt++ | src/jit/register.c:
15:47 dalek MoarVM/even-moar-jit: Always take register when putting a value there
15:47 dalek MoarVM/even-moar-jit:
15:47 dalek MoarVM/even-moar-jit: After all, the register is now in use by a live value
15:47 dalek MoarVM/even-moar-jit: review: https://github.com/MoarVM/MoarVM/commit/baa5b8a6f2
15:48 brrt battery is empty, that's all folks!
15:48 brrt joined #moarvm
15:49 brrt joined #moarvm
15:50 brrt joined #moarvm
15:50 brrt joined #moarvm
15:51 brrt joined #moarvm
15:51 brrt joined #moarvm
15:52 brrt joined #moarvm
15:52 brrt joined #moarvm
15:53 brrt joined #moarvm
15:53 brrt joined #moarvm
15:54 brrt joined #moarvm
15:55 brrt joined #moarvm
15:55 brrt joined #moarvm
15:56 brrt joined #moarvm
15:56 brrt joined #moarvm
15:58 brrt joined #moarvm
15:58 brrt joined #moarvm
15:59 brrt joined #moarvm
15:59 brrt joined #moarvm
16:00 brrt joined #moarvm
16:00 brrt joined #moarvm
16:01 brrt joined #moarvm
16:01 brrt joined #moarvm
16:02 brrt joined #moarvm
16:03 brrt joined #moarvm
16:03 brrt joined #moarvm
16:04 brrt joined #moarvm
16:22 lizmat joined #moarvm
16:34 kjs_ joined #moarvm
16:54 vendethiel joined #moarvm
16:55 hoelzro it seems that MVM_thread_cleanup_threads_list is never called to clean things up; when *should* it be called?
17:07 jnthn You generally can't clean up threads until 2 GC runs later.
17:07 jnthn 'cus that's how long it takes for their nursery to be redundant.
17:08 jnthn There is logic in orchestrate.c that arranges merging of gen2 into a live thread.
17:08 jnthn That's about the point when a ThreadContext can go away
17:09 [Coke] jnthn: hio!
17:09 jnthn o/
17:11 timotimo cool, i didn't know about that
17:11 timotimo so on brrt's latest commit i can get a little ways into the first file in the nqp build
17:11 jnthn brrt++ # talk
17:11 timotimo but after it enters a jitted frame, it segfaults, and the stack gets a bit trashed apparently
17:11 * jnthn read the slides
17:11 timotimo i imagine from the slides there was quite a bit more in the audio track than in the video track
17:12 * jnthn is sad to not have been able to do his MoarVM talk at YAPC::EU this year
17:12 jnthn But, health first...
17:12 lizmat jnthn++  # good plan
17:13 jnthn I'll be gradually inserting myself back into hacking over the coming days, as energy levels allow (so, probably just a little bit at first)
17:15 kjs_ joined #moarvm
17:21 timotimo yes, health very first, indeed
17:25 jnthn Aye
17:26 jnthn Though I suspect my first task ("screw it just merge glr already") will be an easy one. :)
17:28 TEttinger joined #moarvm
17:37 ggoebel joined #moarvm
17:41 [Coke] jnthn: yes, please do that. :)
17:41 FROGGS joined #moarvm
17:45 FROGGS o/
17:49 _itz_ joined #moarvm
18:14 colomon joined #moarvm
18:18 hoelzro jnthn: thanks for weighing in!  so I have to wait for a thread's state to be MVM_thread_stage_destroyed for a run before I can clean it up?
18:18 hoelzro ex. GC 1: state = destroyed; GC 2: nothing; GC 3: can clean up thread ?
18:27 lizmat joined #moarvm
18:34 jnthn hoelzro: I think by destroyed it can go
18:35 jnthn hoelzro: There's a clearing_nursery before that iirc
18:35 jnthn There's an enum of all these states somewhere :)
18:35 hoelzro =)
18:36 hoelzro jnthn: I tried putting a MVM_tc_destroy here: https://github.com/MoarVM/MoarVM/blob/master/src/gc/orchestrate.c#L88, but that also segfaulted
18:36 hoelzro I must be missing something else
18:36 colomon joined #moarvm
18:38 hoelzro also, that "will be cleaned up (removed from the lists) shortly" - any idea where that happens?
18:39 jnthn orchestrate.c iirc
18:40 * hoelzro .oO( iir.c )
18:51 colomon joined #moarvm
18:56 jnthn .oO( ir.c )
18:59 dalek joined #moarvm
19:01 colomon_ joined #moarvm
19:48 brrt joined #moarvm
19:48 brrt good evening or any other day time
19:56 lizmat joined #moarvm
20:02 brrt joined #moarvm
20:16 _itz_ brrt: excellent talk on jit and I would be interested in any links you have for eco issues
20:17 _itz_ jnthn++ # much hard work and rest deserved
20:35 brrt joined #moarvm
20:35 brrt thank you _itz_
20:35 brrt were you at yapc::eu or are you talking about the slides?
20:51 Peter_R joined #moarvm
21:04 nwc10 good UGT, brrt
21:55 _itz_ brrt: I am at YAPC and saw the talk
22:00 brrt joined #moarvm
22:01 brrt oh cool
22:01 brrt did i meet you afterwards
22:01 brrt goot UGT nwc10
22:01 brrt did you stay long still?
22:01 brrt (irc connection has been spotty in hotels, to say the least)
22:03 _itz_ no we didn't meet .. may try and talk tomorrow
22:07 brrt only early in the morning, because i'm leaving just after the first talk
22:07 brrt shame. i'll very very very likely will be going to nine's talk
22:09 brrt ok, my current theory is... what
22:10 brrt my code is weird, and bad
22:10 brrt and i feel bad
22:44 _itz joined #moarvm
22:50 kjs_ joined #moarvm
22:56 FROGGS joined #moarvm
22:58 mj41 joined #moarvm
23:03 lizmat joined #moarvm
23:12 mj41_nb joined #moarvm
23:14 mj41_nbx joined #moarvm
23:56 colomon joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary