|
01:05
[Coke] joined
01:47
ilbot3 joined
01:48
TimToady joined
02:09
mst joined
05:20
brrt joined
05:50
lizmat joined
|
|||
| dalek | arVM/even-moar-jit: 84d8099 | brrt++ | docs/jit/register-allocator.org: More notes on register allocation |
06:09 | |
|
06:12
brrt joined
06:21
lizmat joined
|
|||
| dalek | arVM/even-moar-jit: cdb7856 | brrt++ | docs/jit/register-allocator.org: Remove redundant macro parameter |
06:28 | |
|
06:32
pyrimidi_ joined
06:44
brrt joined
|
|||
| dalek | Heuristic branch merge: pushed 22 commits to MoarVM/even-moar-jit by bdw | 06:47 | |
| brrt | oh, one more thing | 06:49 | |
| if nqp::say() could not be deprecated, that'd be awesome | |||
| i don't really care if its reduced to a broken call to puts | |||
|
06:50
domidumont joined
|
|||
| brrt | but it is a really useful as nobody uses it, and thus it can act as a test case | 06:50 | |
| arnsholt | I use nqp::say all the time, but only when debugging NQP code | 06:51 | |
| lizmat too :-) | |||
| brrt | my point exactly | 06:52 | |
| it's not in production code | |||
| arnsholt | What's the reason for the deprecation though? I missed that | ||
| brrt | string-based-io sucks | ||
| bytes based io is the future | |||
| for, well, excellent reason | 06:53 | ||
| s | |||
| arnsholt | Ah, right | ||
| brrt | but nqp::say() is string based io, so i'm asking for it to be exempted | ||
| noting that i don't really care if it is slow or broken | 06:54 | ||
| arnsholt | Yeah, that sounds reasonable | ||
| brrt | we might argue for a nqp::debug() to replace it | ||
| but why bother :-) | |||
|
06:54
domidumont joined
|
|||
| arnsholt | I can see the merit of renaming it debug, TBH | 06:56 | |
| nwc10 was wondering this - the functionality seems useful, but the name is (now) wrong | |||
|
07:04
kjs_ joined
07:25
zakharyas joined
08:08
kjs_ joined
08:26
lizmat joined
08:35
lizmat joined
08:43
domidumont joined
|
|||
| jnthn | Note that since we can choose exactly how we want to compile nqp::say, we could easily just say "oh, it compiles into an encode and a write to stdout" :) | 08:58 | |
| nqp ops aren't all direct mappings to individual Moar ops :) | |||
|
09:07
zakharyas joined
10:17
lizmat joined
10:47
zakharyas joined
10:51
lizmat joined
|
|||
| arnsholt | That's true | 11:31 | |
| But since names are an indicator of intended use, I think calling it debug or some such is probably a good idea | |||
| If not, some enterprising hacker is probably going to (rightly) confuse it with something intended for non-debug use | 11:32 | ||
| Especially since it has the same name as a normal Perl 6 function | |||
| jnthn | I'm not sure I see the harm if they do, tbh :) | 11:39 | |
| If it's just an encode + a write, it's doing pretty much what Perl 6's say would do anyway :) | |||
|
11:49
lizmat joined
12:13
lizmat joined
12:17
[Coke]_ joined
12:20
japhb joined
12:21
mtj_ joined
12:24
TheLemonMan joined
12:27
zakharyas joined
|
|||
| arnsholt | Yeah, that's true | 12:28 | |
| Depends on how hard you want to enforce the new and improved byte-oriented future, I guess =) | |||
| TheLemonMan | .tell brrt this might be interesting for you www.brendangregg.com/perf.html#JIT_Symbols :) | 12:59 | |
| timotimo | i think brrt was interested in nqp::say surviving because you can put that exact moar op into a frame and use it to trigger stuff like extra debugging | 13:02 | |
|
13:48
zakharyas joined
13:55
lizmat joined
13:57
lizmat joined
14:06
lizmat joined
14:24
lizmat joined
14:39
nebuchadnezzar joined,
arnsholt joined
14:45
tomboy65 joined
16:09
kjs_ joined
16:16
lizmat joined
16:53
domidumont joined
17:06
brrt joined
|
|||
| brrt | timotimo, jnthn: that's correct, i want it specifically to have something that is ordinarily never called | 17:06 | |
| and with a clear side-effect to boot :-) | |||
| also, TheLemonMan, that is interesting indeed; otoh there are already a bunch of standards (ELF? DWARF?) and integrations we could potentially like to support | 17:07 | ||
| doing that is never hard but it is always more or less tedious | 17:08 | ||
| timotimo | i've already told brrt about this, but it wasn't interesting the last time i suppose? :\ | 17:10 | |
| brrt | aw, didn't mean that | ||
| timotimo | :P | ||
| brrt | i must have forgotten, or, i must have missed it | ||
| i think the latter is a bit more likely :-P | |||
| timotimo | :) | 17:11 | |
| brrt | is it as hot in * as it is here? | 17:14 | |
| jnthn | Getting there | ||
| The next few days will be hottest | 17:15 | ||
| Then it'll drop off again | |||
| If the forecast is right :) | |||
| timotimo | it's getting uncomfortable here | ||
| jnthn | Making it not so bad is it seems the temperature will drop quite sharply on a night | 17:18 | |
| brrt is rather sad about it all | 17:21 | ||
| hmm | |||
| alright, let's try and get to work | 17:22 | ||
| i'm going to try and add register specs to the jit. we don't need them as much as i had thought, but still | 17:24 | ||
|
19:08
zakharyas joined
19:09
domidumont joined
19:15
kjs_ joined
|
|||
| timotimo | o/ | 19:23 | |
| i don't know what register specs are supposed to be | 19:26 | ||
| and i have no code yet to read to figure it out for myself :S | |||
| do we have control over frame pointer existence or nonexistence in moar and its jit, btw? | 19:32 | ||
|
20:34
domidumont joined,
lizmat joined
|
|||