github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
Set by moderator on 30 June 2013.
00:18 benabik joined 00:48 benabik joined 01:27 woosley joined 04:48 birdwindupbird joined 05:30 FROGGS[mobile] joined 05:46 FROGGS[mobile]2 joined 06:08 FROGGS joined 06:21 arnsholt joined 09:08 jlaire joined 09:13 yoleaux joined 09:27 itz joined, jnthn_ joined, pmichaud_ joined 09:31 tokuhirom_ joined 09:32 timotimo joined 09:40 woosley joined 10:04 not_gerd joined
not_gerd just submitted github.com/MoarVM/MoarVM/issues/41 10:04
jnthn just replied :)
nwc10 is it possible/is it sane/is it easy to get a github bot to announce new issues to channel? 10:05
not_gerd jnthn: great
jnthn not_gerd: Thanks for creating the issue, so we can track that one. :) 10:06
sorear nwc10: dalek used to have that functionality, but then github made incompatible changes to the issues API and dalek was never updated 10:15
10:52 not_gerd left 13:43 woosley1 joined 13:50 benabik joined 16:45 FROGGS joined 19:37 BinGOs_ joined 19:39 Util_ joined, rblackwe joined 19:43 BinGOs joined 19:54 BinGOs_ joined 19:56 TimToady_ joined 19:58 BinGOs joined 20:04 ssutch joined 21:34 Guest1337 joined
diakopter jnthn: u around? 21:55
jnthn diakopter: yeah 21:56
In the midst of debugging stuffs...
diakopter jnthn: do you know whether nqpj can cross-compile on moar?
jnthn Can we use nqp-jvm for the Moar build? 21:57
diakopter right, that's what I'm asking
setting up a new dev environment
not wanting to install parrot
well, setting up 2 new dev environments, really 21:58
one in windows; one in mac os x
lizmat: btw you convinced me to get a retina
how long has ActivePerl been non-zero-price for production/business use? 21:59
lizmat it's always good to have a retina if you want to look :-)
jnthn diakopter: No, we can't 'cus re-using the MAST -> bytecode compiler is done through a Parrot dynop.
diakopter oh yeah, that.
well, that's something to work on I could do I guess 22:00
(integrate compiler.c into moarvm/nqp)
jnthn: do you have a hint of where to start there? 22:01
(I'm guessing forking compiler.c to start)
jnthn I wrote compiler.c with a load of macros. In theory, we "just" need to define them.
For Moar.
diakopter right, I've done it before
you probably have forgotten :P 22:02
jnthn Then some kinda compile op, I guess.
Yes, probably ;-)
diakopter I made that tiny perl-ish expression/statement/assembly language not-quite-not-quite-perl
jnthn I guess it can get a hash of the various types it wants to know about rather than loads of operands.
And a tree.
ah, yes ;)
diakopter tested a couple of the edge cases of the bytecode compiler 22:03
jnthn Anyway, doing that will be one thing off the "what we need to bootstrap" roadmap
diakopter like number of constants, strings, code size, etc
yeah
22:03 cognominal joined
diakopter let's see, a hash of the types 22:04
(you said)
jnthn yeah
diakopter (and the mastree)
jnthn 'cus it needs to know MAST::Op, MAST::Label, etc.
diakopter (btw, I was poking fun at myself earlier because *I* had completely forgotten I'd redone the macros for compiler.c before) 22:05
jnthn: help me remember what compiler.c would need to do with the type objects (are they not accessible from the WHAT of the tree nodes)? 22:24
jnthn They're to identify what type of node we have. 22:25
diakopter theoretically you could look up their names and cache them the first time to get the pointers, but yeah 22:26
jnthn Look up...where? :) 22:29
diakopter I thought those nqp type objects stored their names somewhere 22:30
(on themselves)
jnthn Hm...guess in the meta-object...
But it's more robust to pass 'em in, I think.
diakopter k
23:05 Guest1337 joined
diakopter jnthn: how would the moarvm compile op help the nqpj running in the jvm 23:12
have the nqpj emit a program of the mast serialized?
(I can see how to do it that way) 23:16
sorear: any ideas?
jnthn diakopter: oh...I meant it'd help us get NQP bootstrapped on Moar, not that it'd allow JVM -> Moar 23:17
We'll have to build something that dumps and then loads the thing up in Moar to do that...
diakopter can't use the existing serialized stuff in bytecode?
jnthn I'm not sure I follow? 23:19
diakopter put the entire tree as bytecode constants
no? 23:30
jnthn It sounds like a lot of throwaway work, is all. 23:31
diakopter jnthn: I guess I'm not seeing all the needed work 23:32
jnthn Well, it means writing some way to serialize the MAST tree in NQP, then deserialize it in Moar to then turn it into a bytecode file, iiuc? 23:34
sorear diakopter: I'm sorry, I wasn't paying any attention to the convo
diakopter jnthn: right
jnthn diakopter: That feels like a reasonable amount of work to me, is all :) 23:35
Maybe I'm just tired. ;-)
'night 23:40
diakopter 'nite 23:42