01:26 FROGGS__ joined 02:11 colomon joined 02:26 woosley joined 07:33 FROGGS joined
dalek arVM: f48f2eb | jnthn++ | / (5 files):
Implement file watchers.

On top of what libuv provides, and hopefully fairly portably. If we want recursive watching and so forth, better to implement that in Perl 6 rather than C.
10:34
10:42 brother joined 10:49 colomon joined
dalek arVM: 07b36b0 | jnthn++ | src/ (9 files):
Add infrastructure for cancellation.
11:26
arVM: 61c8cc7 | jnthn++ | src/io/timers.c:
Implement cancellation of timers.
nwc10 gosh, the MoarVM build is like the prototype ARM thermal envelope 12:03
12:36 btyler joined
jnthn nwc10: heh, that reference is obscure enough to leave on wondering if it's a compliment or not :P 12:38
12:38 dalek joined 13:41 LLamaRider joined 13:54 vincent21 joined
nwc10 I can't actually find a useful "citation", but I'm sure I did recently read the story about how they knew they couldn't afford to make the chip with a ceramic packaging 14:25
and they didn't have a way to calculate the power consumption, so they just played it safe
*very* safe as it turned out
14:41 vincent21 joined
timotimo i still don't get it :) 15:27
how does that apply?
15:29 colomon joined
FROGGS I understand it as that the build system is playing safe in a way that there is no guessing, but knowing or dying 15:33
timotimo oh, ok 15:34
FROGGS which would be the Perl 6 approach anyway... 15:35
15:41 bcode joined
nwc10 no, it was that jnthn wanted the build to be fast 16:07
he rather overshot "fast"
jnthn nwc10: How fast can you build a MoarVM? :) 16:11
nwc10 real 0m4.968s 16:12
user 0m8.739s
sys 0m3.147s
jnthn o.O
nwc10 from clean, with ccache
JimmyZ time make -j8
real 0m10.248s
user 0m28.807s
sys 0m10.310s
nwc10 "my" machine has 24 cores
jnthn core blimey...
JimmyZ 1 core here.. 16:13
FROGGS real0m13.927s
user0m37.892s
sys0m3.719s
nwc10 it's a production-grade server
FROGGS 2 cores with HT enabled
nwc10 it's not mine.
FROGGS and now recompile parrot *g* 16:14
</troll>
nwc10 not sure how many real, or if it's 12 with HT
FROGGS could be two six-cores with ht
(but could be anything else too)
nwc10 looks to be. dmesg suggests that it's one of these: ark.intel.com/products/47926/intel-...-intel-qpi 16:18
well, 2 of those
so, heck, actually it's old.
FROGGS we usually have these at work also 16:19
nwc10 equivalent parrot configure, build and install is 16:22
real 0m24.775s
user 2m58.766s
sys 0m25.024s
when it works
(looks like there is an intermittent parallel build failure - presumably a missing dependency 16:23
16:24 hoelzro joined
FROGGS I can't even build 6.3 fwiw 16:25
nwc10 I was building "Release 6.3.0 (supported release)"
moritz who supports it? :-) 16:36
rurban_ parrot does 16:53
what's the problem?
like: clean make with -j4 fails? 16:54
FROGGS rurban_: hmmm, I do not get the error anymore, it it failed like "Cannot find auto/expect.pm" during Configure 16:58
rurban_ that's a new probe, hmm. I'll check 16:59
FROGGS but I do not get that anymore for 6.1, 6.2, 6.3 and HEAD 17:00
rurban_ This probe came with 6.2.0 17:02
FROGGS rurban_: I'd just provide more information when I hit this problem again, okay? 17:05
jnthn: is that the right time to make hyperops multi threaded on moar? 17:15
rurban_ FROGGS: yes, please. sound like an interesting MANIFEST or git pull problem 17:17
FROGGS rurban_: k, will do
18:02 btyler joined 18:52 vincent21 left 19:07 brrt joined
brrt hey, #moarvm, i've just been informed that indeed i'll be building moar's JIT compiler for the summer 19:09
:-D
colomon \o/
GSoC?
brrt yes 19:10
jnthn \o/
colomon brrt++
brrt tnx :-)
jnthn ++brrt ;
brrt++ too, for making it this far :)
.oO( I've just been informed I'll be mentoring for GSoC this summer... :D )
FROGGS \o/
brrt++
brrt yes that too :-D 19:11
and timotimo++ too
lizmat brrt++, jnthn++
and timotimo++ something to mention in the weekly :-)
brrt so i'm really excited of course 19:12
i've been reading up on dynasm
jnthn :) 19:13
brrt its basically a): an assembler-in-a-header-file and b): a preprocesor for that assembler
jnthn Yeah...afaict it basically turns the thing you write into instructions for the assembler... 19:14
brrt (actually, the assembler is split over both the header file and the preprocesor in a way)
and i think its probably best to focus on x64 19:15
FROGGS that would be state of the art, yes 19:16
jnthn Yes, that's my preferred focus too 19:18
brrt yes, and also - many more registers, 64 bit registers, etc
jnthn aye ;) 19:19
brrt and excellent documentation, i might add 19:20
(and dynasm fits well into a makefile-based-build) 19:21
moritz brrt: have you seen that blog post where a brainfuck JIT compiler is built? 19:22
timotimo oh, it is monday!
i totally forgot!
moritz blog.reverberate.org/2012/12/hello...-jits.html # this one
oh, and brrt++ jnthn++ timotimo++ 19:23
timotimo brrt++ jnthn++ :)
brrt yes i did moritz :-) it was on hacker news recently? i've the source code checked out 19:24
jnthn timotimo: Well, doing it now menas it's after my brunch merge
uh, branch
timotimo the asyncops branch? 19:25
jnthn yeah
well, async in Rakudo
moritz brrt: dunno, I remember it from when it came out, it was on HN and reddit then 19:26
19:28 zakharyas joined
brrt its a good article, if perhaps the only good documentation of dynasm 19:29
timotimo we only have async I/O for sockets thus far? 19:31
jnthn timotimo: Yeah, not files yet 19:34
timotimo: Though also there is timers, file watchers, and signals.
timotimo: So you can winch up your code...
timotimo yay :)
i must find another excuse to continue procrastinating that project :|
jnthn Damn, you were gonna be the way I knew if the signals thing worked :P 19:36
timotimo i could just as well build a few simple test applications first :)
i'll probably have something soon enough that we'll get it tested a bit for the next release :P
jnthn Also nice that on www.google-melange.com/gsoc/projec...e/gsoc2014 I see the LWP one too :) 19:37
timotimo yes! \o/
jnthn Hopefully we can get async variants of those in there somewhere too :)
timotimo those things are public, so i'm free to write about them on my blog? 19:38
i mean the gsoc applications being accepted 19:39
brrt moritz++ , FROGGS++ for the LWP project
jnthn yes, page just linked to is public
timotimo great!
FROGGS I don't see it yet :/
ahh, now I do :o) 19:40
timotimo i'll also link to the gists for crappy webserver and crappy LWP if that's all right
jnthn But they're SO crappy! 19:41
Make sure to say so :)
timotimo i will
where should i link for an example on timers? roast perhaps? 19:43
jnthn Dunno 19:44
brrt i see parrot indirectly got a student as well 19:45
good luck for them
timotimo indirectly? 19:46
jnthn timotimo: I dunno, just show Supply.merge(Supply.interval(2).map({ 'tick' }), Supply.interval(2, 1).map({ 'tock' })).tap(&say); sleep 10; :)
timotimo builds a fresh rakudo 19:47
moritz m: Supply.merge(Supply.interval(2).map({ 'tick' }), Supply.interval(2, 1).map({ 'tock' })).tap(&say); sleep 10; 19:48
camelia rakudo-moar 544ed9: OUTPUTĀ«(signal )Unhandled exception in code scheduled on thread 140555843340032ā¤Cannot call method 'more' on a null objectā¤ in block at src/gen/m-CORE.setting:18947ā¤ in block at src/gen/m-CORE.setting:18494ā¤ in method more at src/gen/m-CORE.setting:1ā€¦Ā»
timotimo :D
brrt brrt-to-the-future.blogspot.nl/ :-)
jnthn 544ed9? That's no HEAD...
moritz last build died with 19:50
Stage jast : Method 'jast' not found for invocant of class 'NQPMu' in (gen/jvm/stage2/QAST.nqp:3813)
jnthn moritz: oh, and all 3 backends have to build?
I think FROGGS++ fixed that but I guess we need to bump NQP_REVISION? 19:51
moritz jnthn: yes
FROGGS t;dalek> rakudo/nom: e974894 | (Tobias Leich)++ | tools/build/NQP_REVISION:
bump nqp rev for unbreak of pointies in if
moritz I guess it didn't try to rebuild after that 19:52
aye, it rebuilds at 8 minutes past the full hour
timotimo supplies based on timers (so Supply.interval) has been supported on JVM for a bit longer now, right? 20:04
TimToady does the sampler pick the first event in every interval, or a randome event in every interval? 20:05
*random
seems like the first event is going to be subject to timing coincidences if production is cyclic
jnthn TimToady: yeah... "last one" is another option. 20:06
TimToady that's also problematic
jnthn Yeah
First one is easiest to implement
TimToady how 'bout replace the last one based on decreasing probablility
jnthn oh wait, we meant opposite things by last :)
Hm :)
Yeah, that's not bad. 20:07
TimToady I think there's even an RC entry for that algo :)
jnthn TimToady: www.introtorx.com/content/v1.0.1062...tml#Sample is how they have it in Rx, fwiw 20:08
Not saying we have to go that way, just another data point
latest can make sense if your data is such that newer is always better, but you want to throttle the rate it comes at you. 20:09
lizmat are we speccing Supply.sample here? 20:11
with e.g. named param :last
TimToady discussing semantics, anyway
lizmat :first
:any
?
each of this accepting a number 20:12
jnthn I can see use cases either way, so maybe we do want to support more than one of them.
a number? :)
TimToady 5 last values?
jnthn Thing is, the adverbs seem mutually exclusive... 20:13
TimToady so maybe just a different method name
jnthn yeah
TimToady .sample-recent vs .sample-random
timotimo i must have been asleep for most of last week ... i don't really recall many things i could report except timers, async i/o, many supply methods (lizmat++) and the release 20:14
lizmat timotimo: Mouq's 17f4ee310239e1dd3d84c2205e479fdce8a7b966 20:15
timotimo oooh! good one! 20:16
lizmat lue's e63497d3e6efe4da929048c2a277cda70c12643d
jnthn Various assorted Rakudo threading bugs got fixed along the way too
lizmat Nami-Doc's e5725364c4479b6c8d0c3149e33350ca9d1a221f 20:17
Donald Hunter's 626fd391b8c3d820d2509ab560687cea1c81a10f 20:18
lizmat shuts up :-)
dalek arVM: 4a8dac6 | (Tobias Leich)++ | src/io/ (3 files):
stat the target of a symlink when we query for filesize

This will also follow symlinks after symlinks to get to an actual file.
20:27
timotimo brrt: please add another s to my last name on your first blog post :) 20:31
brrt sure :-) 20:32
Paulssen, is it? 20:33
timotimo yup 20:34
brrt changed it :-)
i'll try to be at the dutch perl workshop this friday, but i can't promise anything 20:39
jnthn brrt: OK. I'll see you there, if you can :) 20:40
brrt yes, if i can't be there, maybe we could meetup later that day?
utrecht is only 2 hours' train away from me
jnthn I'm staying that evening in Utrecht, I belive.
And have nothing to do but hang out with Perl people :) 20:41
So should work.
I dunno exactly what the post-conf plans are but we can figure a way to meet up.
brrt sure
i'm going to sleep in a few minutes, so i'll speak to you tomorrow :-)
bye o/ 20:42
20:42 brrt left
jnthn o/ 20:45
20:51 btyler joined
dalek arVM: a75ad37 | (Tobias Leich)++ | src/io/fileops.c:
return a value in all cases in follow_symlinks
21:42
arVM: 73f8fe1 | (Tobias Leich)++ | src/io/ (3 files):
rename follow_symlinks to make clear it stats
21:48