|
github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm Set by AlexDaniel on 12 June 2018. |
|||
|
00:10
anatofuz joined
00:23
anatofuz left
00:41
lucasb left
00:53
anatofuz joined
01:45
anatofuz left
01:46
anatofuz joined
03:24
anatofuz left
03:27
anatofuz joined
03:32
anatofuz left
04:03
anatofuz joined
05:44
anatofuz left
05:45
anatofuz joined
06:17
anatofuz left
06:19
anatofuz joined
06:22
anatofuz left,
anatofuz joined
07:33
domidumont joined
|
|||
| nine | MasterDuke: well I didn't find any easy wins anymore. | 07:35 | |
|
07:56
sena_kun joined
08:22
domidumont left
08:59
anatofuz left
|
|||
| nine | Oh, t/06-telemetry/01-basic.t also has some tight timeouts! | 09:40 | |
| So I'm down to t/05-messages/10-warnings.t and t/02-rakudo/repl.t | 09:44 | ||
| AlexDaniel | again, it would be awesome to have a list of test files with tight timeouts | 10:55 | |
| nine | Well, are they actually tight? Note that I'm running in conditions where make test takes about 5 days instead of half a minute. | 11:34 | |
| lizmat | hmmm.. building moar gives me new warnings, I think: | 11:39 | |
| src/strings/utf8.c:173:13: warning: loop not vectorized: the optimizer was | |||
| unable to perform the requested transformation; the transformation might | |||
| be disabled or specified as part of an unsupported transformation ordering | |||
| [-Wpass-failed=transform-warning] | |||
| dogbert17 | five days to run make test! You must have made more changes than just MVMGC_DEBUG=3 ? | 11:54 | |
| Geth_ | MoarVM/gc_torture: 9d468afcaf | (Stefan Seifert)++ | 7 files Run the GC on every allocation and keep objects in the nursery for 32 GC runs |
11:57 | |
| nine | dogbert17: yes that ^^^ | ||
| dogbert17: turned out that running GC all the time also promotes objects to gen2 very quickly and then they won't be moved again. So GC_DEBUG=3 missed cases where that happened before the object was passed to code with missing MVMROOT | 11:58 | ||
| dogbert17 | nine: I might have to try that branch just for fun :) | 12:10 | |
|
12:10
lucasb joined
|
|||
| nine | Well it made me prefer gdb --args perl6 ... over perl6-gdb-m ... because the latter will take a couple of minutes just to show the additional help text ;) | 12:14 | |
| timotimo | d'oh :D | 12:17 | |
| we should be using --optimize=off for the help text display code :D | 12:18 | ||
| dogbert17 | hm, I can't see that branch | 12:28 | |
| timotimo | oh yeah, Geth_ is actually from another dimension, reporting what happens there to our dimension over here | 12:35 | |
| hence the _ | |||
| you'll have to git_ pull to get that branch | |||
|
12:40
anatofuz joined
|
|||
| dogbert17 | could it really be so easy ... | 12:44 | |
| and timotimo was right again :) | |||
| haha, this branch make valgrind look like a speed demon :) | 12:47 | ||
| nine | dogbert17: pro tip: run it _with_ valgrind to get those extra days where you don't have to worry about missing any output ;) | 12:54 | |
| dogbert17 | I got 'Unhandled exception: This representation (VMIter) does not support associative access (for type BOOTIter)' when trying to run t/05-messages/10-warnings.t | 12:55 | |
| I guess I made some kind of mistake | |||
| nine | You probably need to recompile the perl6-m executable | 12:58 | |
| The branch is based on fix_spesh_plugin_guard_set_threading_issue which adds fields to MVMInstance and MVMThreadContext | 12:59 | ||
| May also need to recompile dynext/* | 13:00 | ||
| dogbert17 | won't the recompilation take hours | 13:01 | |
| nine | Probably much longer :) Turn off GC_DEBUG and reduce MVM_KEEP_FROMSPACES for the recompilation and change them back later | 13:02 | |
| dogbert17 | ok, will give it a go | ||
| +++ Compiling blib/Perl6/Metamodel.moarvm | 13:03 | ||
| Segmentation fault (core dumped) | |||
| nine | Actually I have no idea how long a rakudo compile would take. I'd guess at least a week. A month wouldn't surprise me at all as there will be loads and loads of objects around that take their time getting to gen2 | ||
| did you do a full make clean? | 13:04 | ||
| dogbert17 | no, opps | 13:05 | |
| now things are looking better ... | |||
| nah, can't get it working, it SEGV's in nqp while building rakudo | 13:18 | ||
| nine | Could be that nqp needs a recompile, too, since nqp-m is also a binary. | 13:21 | |
| Sorry, don't remember what exactly I had to do. For obvious reasons I tried to do it as rarely as possible | 13:22 | ||
| dogbert17 | I can understand that :) | 13:24 | |
|
14:21
anatofuz left
14:23
anatofuz joined
14:58
anatofuz left
15:11
anatofuz joined
15:15
anatofuz left
15:58
Kaiepi left
16:58
cinch left
18:29
evalable6 left
18:34
evalable6 joined
19:49
AlexDaniel left,
AlexDaniel joined
19:50
AlexDaniel left,
AlexDaniel joined
20:05
sena_kun left
20:56
Kaiepi joined
22:46
anatofuz joined
23:03
anatofuz left
23:04
anatofuz joined
23:32
anatofuz left
23:56
anatofuz joined
|
|||