Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-10-23

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:42 leedo joined #moarvm
01:38 TimToady joined #moarvm
01:56 ilbot3 joined #moarvm
01:56 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
05:06 brrt joined #moarvm
05:08 brrt so, adding a 'discard' node prior to the write barrier works
05:08 yoleaux 22 Oct 2017 23:27Z <samcv> brrt: can you add what you would like to the changelog for the expression engine? You'd be the best one to know what should be added to the changelog
05:09 brrt .ask samcv where should i write it? in the repo?
05:09 yoleaux brrt: I'll pass your message to samcv.
05:12 brrt oh, you just edit docs/Changelog
05:12 brrt i will
05:24 AlexDaniel` brrt++
05:26 Geth ¦ MoarVM/jit-expr-optimizer: 4 commits pushed by (Bart Wiegmans)++
05:26 Geth ¦ MoarVM/jit-expr-optimizer: 56dc3d981e | Add expression tree equivalence function
05:26 Geth ¦ MoarVM/jit-expr-optimizer: 544719d686 | [JIT] Simple tree-rewriting optimizer
05:26 Geth ¦ MoarVM/jit-expr-optimizer: f3031011f2 | [JIT] Add adhoc template application
05:26 Geth ¦ MoarVM/jit-expr-optimizer: 54259d3d37 | [JIT] Optimize multiple loads to a COPY
05:26 Geth ¦ MoarVM/jit-expr-optimizer: review: https://github.com/MoarVM/MoarVM/compare/f1d6abf0f6...54259d3d37
05:27 brrt only the last two commits are unique, rest is a rebase
05:30 brrt the tl;dr, we can optimize multiple loads to a copy, and this does seem to lead to less memory traffic
05:45 evalable6 joined #moarvm
05:48 Geth ¦ MoarVM: 1525165e41 | (Bart Wiegmans)++ | docs/ChangeLog
05:48 Geth ¦ MoarVM: Rearrange changelog entries for expression JIT
05:48 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/1525165e41
06:17 brrt joined #moarvm
06:31 brrt adhoc templates is one of those things that i should've done much sooner
06:48 nine brrt: I should have read Practical Lessons in Leadership 10 years ago. Try to beat that ;)
06:48 brrt what kind of book is that
06:49 brrt … by art petty and rich petro
06:49 brrt those names though
06:50 brrt also, good morning
06:51 nine It's...exactly what the title says. It's all those things that would have been so useful when I first became a team leader. Anyway, they are still useful now :)
06:51 nine And yes, good morning!
06:51 AlexDaniel` .oO( it's not a name if it's ≤10 characters! )
06:53 nine AlexDaniel`: http://www.kalzumeus.com/2010/06/17/falsehoods-programmers-believe-about-names/ ;)
06:57 AlexDaniel` there was another similarly good one: http://infiniteundo.com/post/25326999628/falsehoods-programmers-believe-about-time
06:58 AlexDaniel` although some of its falsehoods border with total stupidity
06:58 AlexDaniel` like, I really wonder if there are any programmers who don't know about february… hmm
07:02 domidumont joined #moarvm
07:08 domidumont joined #moarvm
07:29 brrt i think that's not about not knowing about it, it's about assuming the issue isn't very problematic
07:29 brrt eh
07:29 brrt the difference doesn't matter, i mean
07:31 rba joined #moarvm
07:46 committable6 joined #moarvm
07:48 committable6 joined #moarvm
07:49 zakharyas joined #moarvm
08:06 rba joined #moarvm
08:14 domidumont joined #moarvm
08:18 zakharyas joined #moarvm
08:41 rba joined #moarvm
09:11 domidumont joined #moarvm
09:12 brrt joined #moarvm
10:25 greppable6 joined #moarvm
10:32 samcv brrt, yeah in the repo. thanks!
10:32 brrt :-)
10:43 * samcv heads back to bed
10:49 brrt sleep well
10:52 timotimo hum.
11:59 Geth ¦ MoarVM: 5d2ee22e66 | (Timo Paulssen)++ | src/instrument/crossthreadwrite.c
11:59 Geth ¦ MoarVM: don't memcmp where it should strncmp in crossthreadwritelog
11:59 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/5d2ee22e66
12:06 lizmat MoarVM panic: Heap corruption detected: pointer 0x10ea11e50 to past fromspace  # whee!
12:16 lizmat MoarVM panic: Internal error: zeroed target thread ID in work pass
12:18 lizmat MoarVM panic: Heap corruption detected: pointer 0x1032e80a8 to past fromspace
12:18 lizmat just posting results of runs
12:19 brrt hmmm. results of what?
12:19 brrt *runs of what
12:19 lizmat https://github.com/rakudo/rakudo/issues/1202
12:25 lizmat Segmentation fault: 11
12:32 brrt joined #moarvm
12:37 timotimo is that with the crossthreadwrite log enabled? it might be explosive still
12:37 timotimo (and if you don't MVM_SPESH_BLOCKING it, it's even more explosive)
12:37 lizmat no, this is just running on HEAD
12:38 timotimo urgh
12:42 zakharyas joined #moarvm
12:43 zakharyas joined #moarvm
12:57 unicodable6 joined #moarvm
13:10 brrt joined #moarvm
13:32 brrt joined #moarvm
14:00 Zoffix joined #moarvm
14:01 Zoffix left #moarvm
14:01 Zoffix joined #moarvm
14:06 brrt joined #moarvm
14:15 ggoebel joined #moarvm
14:19 patrickz joined #moarvm
14:33 rba_ joined #moarvm
14:47 lizmat the code of https://github.com/rakudo/rakudo/issues/1202 now usually crashes in MoarVM space (and a lot less than before)
14:48 lizmat it now typicall always is: MoarVM panic: Heap corruption detected: pointer 0x1105d27d0 to past fromspace
15:07 brrt joined #moarvm
15:18 AlexDaniel` fwiw I tested it here also and sometimes it also hangs
15:19 AlexDaniel` in fact, most of the time it hangs
15:20 lizmat I've just had a hang in S07-hyperrace/race.t, test 29
15:20 AlexDaniel` hmm
15:20 lizmat I ran a sample of the process: https://gist.github.com/lizmat/d22cfb0c67f9866df3f5fe3b7187c15e
15:20 lizmat hope this is of help to somehow with moar gut internals knowledge
16:00 brrt1 joined #moarvm
16:02 AlexDaniel` fwiw changelog things: “seakable” → “seekable” and some lines have tabs
16:03 timotimo we should be able to tease out the exact location where things go wrong by lowering nursery size and crossing fingers
16:04 domidumont joined #moarvm
16:20 brrt joined #moarvm
16:21 dogbert17 joined #moarvm
16:35 lizmat timotimo: see my thoughts on that on #perl6-dev just now
16:35 timotimo does .elems sometimes cause autovivification?
16:36 moritz it shouldn't
16:36 moritz only writing accesses should autovivify
16:36 timotimo i mean for internally-used attributes, not for the list's storage
16:38 Zoffix It can cause reification
16:38 lizmat the thing is that .elems may cause changes in attributes of a List
16:38 Zoffix m: my @l := List.from-iterator: (gather { say "meow"; take 1; }).iterator; say "starting"; @l.elems.say; say "done"
16:38 camelia rakudo-moar 50324bb00: OUTPUT: «starting␤meow␤1␤done␤»
16:38 timotimo Zoffix: at least in the context of the ThreadPoolScheduler i don't think the lists actually have iterators involved in 'em?
16:38 lizmat because .elems may reify
16:39 Zoffix Ah. No idea :)
16:39 lizmat it's not about iterators
16:39 lizmat it's about .elems calling reify-until-lazy
16:39 lizmat and if fully reified, resetting $!todo
16:44 ggoebel joined #moarvm
16:50 rba joined #moarvm
16:50 zakharyas joined #moarvm
17:08 zakharyas joined #moarvm
18:53 Geth ¦ MoarVM: 6e59a07653 | (Samantha McVey)++ | docs/ChangeLog
18:53 Geth ¦ MoarVM: Add further entries to ChangeLog
18:53 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/commit/6e59a07653
19:55 lizmat .tell jnthn I wonder whether the difference between .hyper and .race is really whether results are buffered or not
19:55 yoleaux lizmat: I'll pass your message to jnthn.
19:55 lizmat .tell jnthn I could see .race internally working as a Supply, emitting values from several threads as they become available
19:55 yoleaux lizmat: I'll pass your message to jnthn.
19:57 lizmat .tell jnthn instead of .pushing to an IterationBuffer until the end of the batch
19:57 yoleaux lizmat: I'll pass your message to jnthn.
20:08 timotimo lizmat: performance is much better if you submit big batches of data compared to individual pushes (for example to a supply)
20:09 lizmat timotimo: yeah, but if you have a serializer at the end of the chain that does post-processing, you would like the values to appear gradually, rather than in blocks
20:09 lizmat perhaps, not always I guess
20:11 timotimo samcv: i think the "left-over calloc in NFA" is a fix to a feature that's also new in this version; an oversight that was more or less quickly fixed?
20:11 samcv ah ok i can remove it then
20:11 timotimo in fact, it's the commit right before it: 054ad4e and the original work in 3059ba2829
20:12 samcv thanks
20:12 timotimo using FSA in the NFA might be noteworthy, but probably not
20:12 timotimo starting to use FSA in the VMArray impl will have a much bigger impact, for example
20:12 timotimo thanks for doing the release work :)
21:11 cognominal m: say $/.^name
21:11 camelia rakudo-moar e4a5bb17c: OUTPUT: «Nil␤»
21:12 cognominal m:'a' ~~ /a/ say $/.^name; say $/.^role
21:13 evalable6 cognominal, rakudo-moar e4a5bb17c: OUTPUT: «(exit code 1) ===SORRY!=== Error while compiling /tmp/cNH0Xg9Mx1␤Two ter…»
21:13 evalable6 cognominal, Full output: https://gist.github.com/d7c4524130f2d246ea6dbbeff546f314
21:13 cognominal m:'a' ~~ /a/; say $/.^name; say $/.^role
21:13 evalable6 cognominal, rakudo-moar e4a5bb17c: OUTPUT: «(exit code 1) Match␤No such method 'role' for invocant of type 'Perl6::Metamodel::ClassHOW'. Did you mean any of these?␤    none␤    note␤    one␤    roles␤␤  in block <unit> at /tmp/UlHPkcm3Zc line 1␤»
21:13 cognominal m:'a' ~~ /a/; say $/.^name;
21:13 evalable6 cognominal, rakudo-moar e4a5bb17c: OUTPUT: «Match»
21:13 cognominal m:'a' ~~ /a/; say $/.^roles;
21:13 evalable6 cognominal, rakudo-moar e4a5bb17c: OUTPUT: «(exit code 1) Unexpected named argument 'transitive' passed␤  in block <unit> at /tmp/ZXYhBGq0yJ line 1␤»
21:14 cognominal m:'a' ~~ /a/; say $/.^roles;
21:14 evalable6 cognominal, rakudo-moar e4a5bb17c: OUTPUT: «(exit code 1) Unexpected named argument 'transitive' passed␤  in block <unit> at /tmp/vJi3P3aNh7 line 1␤»
21:16 cognominal m: say $/.^roles;
21:16 camelia rakudo-moar e4a5bb17c: OUTPUT: «()␤»
21:16 cognominal m:'a' ~~ /a/; say $/.^roles;
21:16 evalable6 cognominal, rakudo-moar e4a5bb17c: OUTPUT: «(exit code 1) Unexpected named argument 'transitive' passed␤  in block <unit> at /tmp/3z9nwNPSjz line 1␤»
21:38 lizmat and another Perl 6 Weekly hits the Net: https://p6weekly.wordpress.com/2017/10/23/2017-43-hyper-lands-racing/
21:54 evalable6 joined #moarvm
22:59 Voldenet joined #moarvm
22:59 Voldenet joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary