| japhb | I ... dunno? The last time I ran it, it worked on my install, which has a few non-core Perl 5 add-ons, but not like half of CPAN or anything. | 00:00 | |
| MasterDuke | it may be simpler just to start with 2018, ISTR some Configure.pl changes in 2016 or 2017 | 00:01 | |
| because of '.' no longer being in @INC and things like that | |||
| there's some command line option to start at a certain point in time? | 00:02 | ||
| ah.. github.com/MoarVM/MoarVM/issues/403 | 00:03 | ||
| japhb | I tried to make github.com/japhb/perl6-bench/blob/master/bench relatively easy to understand/hack on. | 00:06 | |
| timeall and analyze were necessarily less transparent. | |||
| :-/ | |||
|
06:39
domidumont joined
06:44
domidumont left
06:45
domidumont joined
07:15
robertle joined
07:31
brrt joined
|
|||
| brrt | good * | 07:32 | |
| yoleaux | 5 Jun 2018 21:24Z <samcv> brrt: it seems slower by the testing i've done | ||
| brrt | yeah, i'm not entirely sure that is a test that benefits that much by the JIT though :-) | ||
| samcv | brrt: 4.54 before after 4.60 for this thing | ||
| dd if=/dev/urandom bs=100 count=800000 | base64 > /tmp/stuff; perl6 -e 'use nqp; my $hash := nqp::hash(); my @file = "/tmp/stuff".IO.slurp.lines; my $t1 = now; for @file { nqp::bindkey($hash, $_, 1) };my $t2 = now; for @file { nqp::atkey($hash, $_) }; ' | |||
| and those times i ran it before 100 times, then ran it after the changes 100 times, then 100 times of before and after again | 07:33 | ||
| then averaged the scores | |||
| brrt | hmmm | ||
| samcv | also both before runs were slower than either after runs. | ||
| brrt | just for giggles, have you tried running that with MVM_JIT_DISABLE | 07:34 | |
| samcv | no i haven't. should i? | ||
| brrt | try it out | ||
| samcv | though actually | ||
| is it up to date with master? | |||
| brrt | not 100% sure it is | ||
| you've been productive the last few weeks :-) | 07:35 | ||
| samcv | that could be why | ||
| :-) | |||
| the speed difference may have been from github.com/MoarVM/MoarVM/commit/f0a837c41 | 07:36 | ||
| brrt | yeah, let me just check if the branch has that | 07:38 | |
| nope, it should have that commit | 07:40 | ||
| weird. and distressing | |||
| samcv | hm | ||
| brrt | it compiles less code and does less work | 07:45 | |
| samcv | brrt: i don't see that utf8 speed commit in that branch | 07:48 | |
|
08:02
lizmat joined
08:03
zakharyas joined
|
|||
| brrt | hang on.. there's a way to check it out | 08:07 | |
| jit-stack-walker should have it, for sure | 08:08 | ||
| samcv | hmm oh i think it wasn't fully updated | 08:09 | |
|
08:26
zakharyas left
08:28
zakharyas joined
|
|||
| samcv | ok i'm getting 4.38 after the changes and 4.35 before | 08:35 | |
| i gotta head to bed right now though | |||
| night o/ | 08:36 | ||
| brrt | good night | 08:40 | |
| (i'm still going to call that 'noise', though :-)) | |||
| .tell japhb perl6-bench doesn't work today anymore | 09:27 | ||
| yoleaux | brrt: I'll pass your message to japhb. | ||
|
09:34
brrt left
10:14
domidumont left
10:27
brrt joined
|
|||
| brrt | naive perl6 is still depressingly slow though... | 10:58 | |
| this while type-specialized nqp is pretty fast | 10:59 | ||
| m: say 1.034/0.993 | 11:15 | ||
| camelia | 1.041289 | ||
| brrt | meh, that's not so bad | ||
| this is on a hopelessly happy-case for the JIT, though, with explicitly-annotated ints | |||
|
11:26
lizmat left
11:49
lizmat joined
11:52
zakharyas left
12:35
brrt left
12:56
zakharyas joined
13:07
brrt joined
13:39
zakharyas left
13:40
zakharyas joined
|
|||
| brrt | in other 'OpenBSD hates JIT compilers' news: marc.info/?l=openbsd-cvs&m=152...17&w=2 | 13:52 | |
|
13:59
lizmat left
|
|||
| jnthn | o/ brrt | 14:06 | |
| Hope my review was of some use :) | |||
| brrt | it was, thanks | 14:09 | |
| i've figured out why we don't put `MVM_frame_effective_bytecode` in `frame.h` - at the point frame.h is included, `compile.h` (responsible for `MVMJitCode`) is not, so `MVMJitCode` isn't yet a complete type | 14:10 | ||
| in another refactor, I want to inline that stuff into the spesh candidate anyway | |||
| but that's for another day | |||
| jnthn | OK | 14:11 | |
| brrt | i'm not feeling like figuring out a better ordering right now, although I suspect that we can include the spesh headers before core/frame.h | 14:12 | |
| jnthn | It's OK to move it, just wanted to know there was a reasonable reason :) | 14:14 | |
| Geth | MoarVM/jit-stack-walker: 7f39e47a56 | (Bart Wiegmans)++ | 2 files Fix some review things Leftover commented code and comments that had become stale |
14:19 | |
| brrt | anyway, if it's okay with you, i'll hit the 'merge' button then :-) | 14:20 | |
| jnthn | +1 | 14:23 | |
|
14:25
lizmat joined
|
|||
| Geth | MoarVM/master: 20 commits pushed by (Bart Wiegmans)++ review: github.com/MoarVM/MoarVM/compare/2...35aa7c4331 |
14:36 | |
| brrt | done | ||
| now all that is left is to blog about it :-) | |||
| nwc10 | and then beer? or have you been doing that already? | 14:37 | |
| brrt | no beer, not yet | 14:38 | |
| good point though | |||
| nwc10 | glad to be of *some* use | ||
| but not very much - can't even *buy* you the beer. At least, not this one this time. | |||
| brrt | hehe. Are you planning on going to TPC? | 14:41 | |
| i'm slightly afraid of the whole type punning situation in C now | 15:26 | ||
| jnthn | Cast your fears into the void... | 15:27 | |
| nwc10 | brrt: apparently I am: act.perlconference.org/tpc-2018-gla.../talk/7349 | 15:29 | |
| brrt | oh, i recall seeing that | 15:33 | |
| cool | |||
| jnthn: thing is, i was planning on doing the following evil trick | |||
| union { struct { MVM_VECTOR_DECL(MVMJitExprNode, nodes); }; struct { MVM_VECTOR_DECL(MVMuint32, links); } }; | |||
| this will give me a vector links and nodes, aliased to the same thing | 15:34 | ||
| i thought that was kind of cool | |||
| nwc10 | as found by one of work's sysamdins: img.pr0gramm.com/2018/06/06/61e7e5...ecefff.jpg | ||
| dogbert17 | brrt++ | ||
| brrt, there might be a bug lurking in your latest merge | 15:35 | ||
| got this: gist.github.com/dogbert17/f2a82fc0...6b559649a4 | |||
| brrt | tell me | ||
| i'll be surprised, I didn't add any additional free() | 15:36 | ||
| hmmmm | |||
| i'll try to reproduce | |||
| thanks dogbert17++ | |||
| nwc10 | ASAN is very excited | ||
| brrt | hehe, or so i hear | ||
| nwc10 | brrt: ASAN is excited by t/qast/01-qast.t | 15:37 | |
| same error trace as that valgrind report | |||
| brrt | ok. thanks. I'll debug | 15:40 | |
|
15:46
robertle left
|
|||
| brrt | use-after-free, i'm an idiot | 15:46 | |
| Geth | MoarVM: c58dadb1a8 | (Bart Wiegmans)++ | src/core/frame.c [frame.c] use-after-free fix Oops, didn't recognise ud == sr_data, so didn't see the use-after-free. |
15:47 | |
| brrt | thanks nwc10++, dogbert17++ | ||
| dogbert17 | brrt++, you made short work of that one :) | 15:50 | |
|
15:58
Redfoxmoon left
16:00
Redfoxmoon joined
16:01
Redfoxmoon left,
Redfoxmoon joined
16:25
Geth joined
16:45
robertle joined
16:57
AlexDaniel joined
17:06
zakharyas left
17:13
zakharyas joined
17:46
zakharyas left
18:04
brrt left
18:09
lizmat left
18:57
Merfont is now known as Kaiepi
19:12
domidumont joined
19:24
domidumont left
|
|||
| japhb | . | 19:45 | |
| yoleaux | 09:27Z <brrt> japhb: perl6-bench doesn't work today anymore | ||
| japhb | .tell brrt MasterDuke found it was just a couple lines that needed to be patched to get perl6-bench working with HEAD again. | 19:46 | |
| yoleaux | japhb: I'll pass your message to brrt. | ||
| japhb had totally forgotten he gave lizmat++ merge rights in the perl6-bench repo -- Thanks lizmat! | 19:47 | ||
| Kaiepi | i wasn to get into coding for moarvm at some point | ||
| it's been helping a lot with learning c | |||
|
20:19
zakharyas joined
20:35
TimToady joined
20:44
jimmyz joined
20:48
jimmyz left
20:57
zakharyas left
21:38
MasterDuke joined
22:08
MasterDuke left
22:23
lizmat joined
22:29
moritz left
22:30
moritz joined
22:36
MasterDuke joined
23:24
Redfoxmoon left
23:29
Redfoxmoon joined
|
|||
