|
github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm Set by AlexDaniel on 12 June 2018. |
|||
|
00:14
ggoebel left
03:52
AlexDaniel left,
AlexDaniel joined
03:53
p6bannerbot sets mode: +v AlexDaniel
06:06
domidumont joined
06:07
p6bannerbot sets mode: +v domidumont
06:42
robertle joined
06:43
p6bannerbot sets mode: +v robertle
|
|||
| Geth | MoarVM: 8c2dd3ad22 | (Bart Wiegmans)++ | 2 files [JIT] Be more silent about internal failures Loudly complaining about internal failures tends to upset some people. |
07:01 | |
| nine | .tell brrt I tried to implement JITing of trunc_u16 and extend_u16. From my understanding it should just be mov TMP1, WORK[src]; mov TMP2w, TMP1w; mov WORK[dst], TMP2; and mov TMP1, WORK[src]; mov WORK[dst], TMP1; respectively but that causes endless looping on while $i < $uint16value { ... } | 07:13 | |
| yoleaux | nine: I'll pass your message to brrt. | ||
|
07:23
travis-ci joined,
p6bannerbot sets mode: +v travis-ci
|
|||
| travis-ci | MoarVM build passed. Bart Wiegmans '[JIT] Be more silent about internal failures | 07:23 | |
| travis-ci.org/MoarVM/MoarVM/builds/441515254 github.com/MoarVM/MoarVM/compare/5...2dd3ad2256 | |||
|
07:23
travis-ci left
07:48
brrt joined
07:49
p6bannerbot sets mode: +v brrt
08:29
patrickb joined,
p6bannerbot sets mode: +v patrickb
08:35
dogbert11 left
08:36
brrt left
10:01
brrt joined
10:02
p6bannerbot sets mode: +v brrt
|
|||
| brrt | \o | 10:21 | |
| yoleaux | 07:13Z <nine> brrt: I tried to implement JITing of trunc_u16 and extend_u16. From my understanding it should just be mov TMP1, WORK[src]; mov TMP2w, TMP1w; mov WORK[dst], TMP2; and mov TMP1, WORK[src]; mov WORK[dst], TMP1; respectively but that causes endless looping on while $i < $uint16value { ... } | ||
| brrt | .tell nine operations on the 'word' form of a x86 register do not clear the upper 48 bites | ||
| yoleaux | brrt: I'll pass your message to nine. | ||
|
10:27
domidumont left
10:36
ggoebel joined,
p6bannerbot sets mode: +v ggoebel
|
|||
| nine | brrt: oh my, then I have been misled by my own comment: github.com/MoarVM/MoarVM/blob/mast...dasc#L1258 | 10:59 | |
| yoleaux | 10:21Z <brrt> nine: operations on the 'word' form of a x86 register do not clear the upper 48 bites | ||
| nine | Should have looked at whose commit that was. | ||
| brrt | :-) | 11:04 | |
| I don't know where I read that from though | |||
| anyway, iirc, the upper 32 bits are cleared for a 32 bit operation | 11:05 | ||
| ... no perl devroom :-( | 11:14 | ||
|
11:45
ZzZombo joined,
p6bannerbot sets mode: +v ZzZombo
12:02
domidumont joined
12:03
p6bannerbot sets mode: +v domidumont
12:07
domidumont1 joined
12:08
p6bannerbot sets mode: +v domidumont1
12:10
domidumont left
12:12
dogbert2_ joined
12:13
p6bannerbot sets mode: +v dogbert2_
|
|||
| dogbert2_ | has jnthn gone up in smoke? | 12:18 | |
|
12:33
brrt left
|
|||
| diakopter | eh | 13:11 | |
|
13:12
leont joined,
p6bannerbot sets mode: +v leont
|
|||
| masak | dogbert2_: not unless it happened in the last 1.5 hours. | 13:13 | |
| dogbert2_: might it been that you are privmsg'ing with unreasonable expectations of speedy reply? ;) | |||
| I find viewing IRC as an asynchronous medium works best | |||
| jnthn is busying himself with using Perl 6 rather than implementing Perl 6 in recent days :) | 13:18 | ||
| yoleaux | 13 Oct 2018 11:37Z <Xliff> jnthn: Thanks for the help with Supplies. I have a final signal handler pattern set up. You can see that, here: | ||
| nwc10 | people actually *use* Perl 6‽ | 13:23 | |
| diakopter | ikr | 13:24 | |
|
13:26
patrickb left
|
|||
| masak | nwc10: funny/ouch | 13:40 | |
| dogbert2_ | masak: just thought he'd been awfully silent the last few days. Was starting to suspect that he had escaped somewhere in order to do some analysis :) | 13:42 | |
| masak | the last few days, as in... the weekend? :P | ||
| dogbert2_ | indeed :) | ||
|
14:01
Sigyn left
14:04
Sigyn joined,
p6bannerbot sets mode: +v Sigyn
14:21
brrt joined,
p6bannerbot sets mode: +v brrt
|
|||
| brrt | hmm | 15:00 | |
| nine: any reason why we can't make everything go over nativeinvoke? | 15:01 | ||
|
15:01
zakharyas joined
15:02
p6bannerbot sets mode: +v zakharyas
|
|||
| brrt | or just because nativeinvoke can't handle everything just yet | 15:12 | |
|
15:25
fake_space_whale joined
15:26
p6bannerbot sets mode: +v fake_space_whale
|
|||
| nine | brrt: it doesn't handle all cases yet, most notably callbacks | 15:34 | |
| i.e. function pointers | |||
| brrt: btw. extend_u16 and trunc_u16 now work :) Thanks! Now there's just param_rp_u to implement | |||
|
15:34
brrt left
|
|||
| leont | brrt: your post on fork didn't mention pthread_atfork. Implementation detail or did you take a different approach? | 15:35 | |
| nine | How can we not have a known value for a literal?! | 15:51 | |
| Ah, because the known value goes through a coerce_iu and becomes unknown. Because our support for unsigned ints barely exists :/ | 15:52 | ||
|
16:03
zakharyas left
16:05
zakharyas joined
16:06
zakharyas left
16:08
zakharyas joined
16:09
p6bannerbot sets mode: +v zakharyas
16:46
domidumont1 left
16:58
domidumont joined
16:59
p6bannerbot sets mode: +v domidumont
17:03
robertle left
17:58
dogbert17 joined
17:59
p6bannerbot sets mode: +v dogbert17
18:01
robertle joined
18:02
p6bannerbot sets mode: +v robertle
18:04
zakharyas left
18:06
zakharyas joined,
p6bannerbot sets mode: +v zakharyas
18:49
domidumont left
18:50
brrt joined
18:51
p6bannerbot sets mode: +v brrt
|
|||
| dogbert17 | brrt: I might hace stumbled upon a JIT related problem | 18:58 | |
| *have | |||
| gist here: gist.github.com/dogbert17/10cad2d6...021a20560e | 18:59 | ||
| brrt | hmm | 19:05 | |
| dogbert2++ good catch | 19:09 | ||
| sorry, dogbert17++ | |||
| Geth | MoarVM: f79011cf91 | (Bart Wiegmans)++ | src/jit/compile.c [JIT] Initialize sequence number early So that if we destroy the code early, we can use it correctly. |
19:11 | |
| dogbert17 | now that's what I call fast turnaround :-) brrt++ | 19:13 | |
| brrt | some bugs are simple :-) | 19:16 | |
| nine: I'm kind of looking for an easy test case for nativeinvoke | |||
| and..... I'm not sure I'm very happy about having the NativeCall library make decisions based on MoarVM internals | 19:17 | ||
| Ideally, i'd just be invoke() and whatever NC implementation we had would use that | |||
| rather than have something like nativecallinvoke and nativeinvoke | |||
| (although, arguably, I'd be equally happy about sp_nativeinvoke) | 19:18 | ||
| i.e. some way in which we knew we could do better | |||
| then again, we do need something like nativecallbuild | |||
|
19:22
zakharyas left
20:31
brrt left
|
|||
| lizmat | And another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/10/15/...tom-for-6/ | 21:17 | |
|
21:22
robertle left
|
|||
| jnthn | lizmat++ | 22:10 | |
|
22:49
leont left
|
|||