samcv what is the best way to learn how to contribute or at least understand MoarVM's JIT code? 01:17
ugexe so libuv 1.16.0 seems to not build on any windows configuration ive tried 03:41
i sent a PR to fix the problem, so getppid will have to wait until that lands
second libuv bug ive found through moarvm :P 03:42
samcv nice! 05:47
brrt good * #moarvm 09:05
lizmat_ good *, brrt 09:23
lizmat brrt: "[02:17:54] <samcv>what is the best way to learn how to contribute or at least understand MoarVM's JIT code?"
lizmat would also like some guidance there :-) 09:24
brrt oh, wow, ehm, hmmm 09:25
there's some documentation....
what part would you like to know about?
anyway, i have to get back to you at some later point 09:27
lizmat brrt: that's ok, I have to be afk in a bit as well :-) 09:35
and then later for the rest of the day
jnthn Given the JIT takes a spesh graph as input, and a lot of the opts are performed by that stage, learning about spesh is also a worthwhile endeavor for those looking to work on performance things. I suggest going through my recent 4-part blog series side by side with the code. :) 10:23
nwc10 "recent"? You appear to be IRCing from the future 10:35
or is "next week" the new "this week" ?
jnthn haha 10:42
D'oh, I finished drafting part 4 last night, then left proof-reading/posting until this evening. :)
brrt what jnthn said is also very, very true 12:44
interestingly, the JIT (and especially the expression JIT) takes a very different approach to its internal operations than does spesh 12:45
this is more interesting to C hackers than to anyone else, but whatever
arguably though, i'd say that spesh is more advanced, in a way, but also more difficult to use 12:53
… for me anyway 12:54
(and with use, i mean 'to hack on')
nine I think spesh hacking requires fuller and often detailed knowledge about how the different parts of moarvm fit together and work. 13:11
jnthn Wow, they've started digging a trench right outside of my office door. 13:12
nine
.oO(What do you mean by "leaving the office"?)
jnthn They've been digging all the way along the street, right next to the buildings. I didn't yet figure out what they're doing. 13:13
nwc10 beer pipeline?
ilmari for lunch?:
jnthn My fantasy is "laying optical fibers for super-fast internet"
Well, or what nwc10 said works very nicely also :P 13:14
brrt nine: i agree, yes
nine I'm sure the trench can accomodate both
brrt
.oO( ip-over-beer )
jnthn I semi-agree, but the JIT has to also be involved in deopt and inlines :)
brrt true
nine Yet I've been able to to quite a few things with and to the JIT without ever worrying about those :) 13:15
brrt anyway, i want to start hacking on having the stack walker instead of writing 'here am i' labels at every turn
and the thing i'm dreading most is to hack our build system to support it
because i'm fairly sure the stack walker assembly i've cooked up works 13:16
jnthn Finally, finished and proof read: 6guts.wordpress.com/2017/11/09/moa...nt-guards/ 21:28
moritz jnthn++ 21:46
MasterDuke jnthn: "space the both spesh and", i think the "the" should be "that"? 22:37
and great series of articles 22:45
japhb jnthn: "Provided once is found" 23:13
jnthn MasterDuke, japhb: Thanks, fixed those :) 23:17
japhb Good post, too. 23:21
jnthn :)

Powered by Cro and the Raku® Programming Language.

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

Thank you!