Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-11-13

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

All times shown according to UTC.

Time Nick Message
02:55 ilbot3 joined #moarvm
02:55 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
03:03 MasterDuke joined #moarvm
03:08 releasable6 joined #moarvm
05:38 evalable6 joined #moarvm
06:17 domidumont joined #moarvm
06:17 domidumont1 joined #moarvm
06:25 domidumont joined #moarvm
06:43 geospeck joined #moarvm
07:12 geospeck joined #moarvm
08:00 brrt joined #moarvm
08:04 geospeck joined #moarvm
08:04 domidumont joined #moarvm
08:26 zakharyas joined #moarvm
08:33 zakharyas joined #moarvm
08:50 brrt good * #moarvm
08:52 brrt i've counted and if i'm packing even a bit loosely, i could easily encode most of the info that currently lives in the expr node info structure into the expr operand itself
08:52 brrt this would save memory, but more importantly, random memory accesses
08:53 brrt if i'm packing tightly then i've got room to spare
08:57 nwc10 good *, brrt
08:57 nwc10 sounds useful - callgrind would give you numbers, but would take a while to do so :-)
09:53 brrt true
09:53 brrt and the object is to have the optimizer or spesh-relevant info in a separate hash table since it's really sparse
09:54 brrt also, so that during modification i don't immediately have to the info array whenever i'm changing the nodes array
10:11 geospeck joined #moarvm
10:36 geospeck joined #moarvm
11:33 lizmat m: sub a() { await start { return } }; a  # hmmm, jnthn ?
11:33 camelia rakudo-moar 5f4b61b12: OUTPUT: «MoarVM panic: Internal error: Unwound entire stack and missed handler␤»
11:34 jnthn Already ticketed somewhere, but yeah, I'd not expect that to work out well
11:34 lizmat jnthn: I was wondering about the significance of "anon sub THREAD-ENTRY()" in Thread.pm
11:34 lizmat 1. why is that a sub and not just a block, and 2. why anon?
11:35 jnthn 1. because it having a name in the backtrace is really useful and maybe even relied upon in the backtrace heuristics, and 2. 'cus we only want the name on the sub, not to install it somewhere
11:37 lizmat ok, so it was not intended to catch any "return" from the started code ?
11:38 jnthn It couldn't
11:38 jnthn return is resolved purely lexically
11:39 lizmat TIL
11:40 lizmat BTW, is "sub a() { await start { return } }; a" worth an Issue ?
11:41 jnthn It's already filed somewhere
11:41 lizmat ack
11:43 lizmat jnthn: is it correct that no code gets executed after https://github.com/rakudo/rakudo/blob/master/src/core/Thread.pm#L31 ever ?
11:45 geospeck joined #moarvm
11:47 jnthn Not sure I follow
11:47 jnthn I mean, if you put code after it, it'll run
11:47 jnthn But only if the code() returns
11:47 jnthn Which for any thread pool thread, it never will
11:47 jnthn Because they sit in an infinite loop
11:47 jnthn Reading from the work queue
11:49 lizmat ah, duh
11:50 lizmat when I do a start block, it doesn't actually run its own thread, it just adds a task
11:50 lizmat duh
11:56 jnthn :)
11:58 brrt joined #moarvm
12:29 AlexDaniel joined #moarvm
12:57 domidumont joined #moarvm
13:41 greppable6 joined #moarvm
13:41 domidumont joined #moarvm
13:56 zakharyas joined #moarvm
14:13 brrt joined #moarvm
14:23 greppable6 joined #moarvm
14:26 brrt joined #moarvm
14:32 dogbert2 hmm, t/spec/S17-promise/stress.t has a tendency to hang when run with MVM_SPESH_BLOCKING=1
15:18 brrt joined #moarvm
16:14 brrt joined #moarvm
16:20 ugexe https://github.com/libuv/libuv/issues/1625 - any ideas? we'll be stuck to 1.14.2 (which requires a revert) otherwise
16:27 geospeck joined #moarvm
16:38 * brrt looking
16:39 brrt yay, i found a bug in expr_ops
16:41 brrt and the solution is… let's do the nonzero-field trick everywhere
17:06 zakharyas joined #moarvm
17:07 jnthn hm, not having much luck with https://github.com/MoarVM/MoarVM/issues/749
17:07 domidumont joined #moarvm
17:07 jnthn Thought I knew what might be going on but...no
17:07 jnthn I can't figure out where there's a moment it's unmarked
17:09 jnthn Will dig more tomorrow
17:09 ugexe oh i may have solved the libuv thing
17:11 nwc10 jnthn: that's "ok 62 - CLOSE phaser sees correct outer scope" ?
17:12 nwc10 ASAN refuses to make any comment :-(
17:16 jnthn Yeah
17:16 jnthn valgrind reliably triggers
17:17 jnthn Gotta head home now though...will look at it some more in the morning :) o/
17:17 nwc10 yes. a fresh head feels like the right pan
17:17 nwc10 plan
17:39 brrt joined #moarvm
18:17 zakharyas joined #moarvm
18:23 geospeck joined #moarvm
18:48 TimToady joined #moarvm
19:01 geospeck joined #moarvm
20:21 * lizmat wonders if it isn't time for a MoarVM version bump?
20:21 lizmat (looks like it's 13 commits behind)
20:21 * Zoffix bumps
20:22 jnthn That's an unlucky number. Let's wait for 14. ;)
20:22 lizmat cool
20:22 lizmat hehe
20:22 lizmat it's at 70, that's already 5x 14  :-)
20:28 releasable6 joined #moarvm
20:35 patrickz joined #moarvm
20:35 robertle joined #moarvm
20:48 AlexDaniel joined #moarvm
21:18 Geth ¦ MoarVM: ugexe++ created pull request #752: Revert "Use uv_fs_copyfile API"
21:18 Geth ¦ MoarVM: review: https://github.com/MoarVM/MoarVM/pull/752
21:46 zostay_ joined #moarvm
21:49 BinGOs_ joined #moarvm
21:53 leedo joined #moarvm
21:54 unicodable6 joined #moarvm
21:56 squashable6 joined #moarvm
22:07 ilmari[m] joined #moarvm
22:11 lizmat and another Perl 6 Weekly hits the Net: https://p6weekly.wordpress.com/2017/11/13/2017-46-spesh-explained/
22:17 MasterDuke joined #moarvm
22:21 benchable6 joined #moarvm
22:21 coverable6 joined #moarvm
22:22 quotable6 joined #moarvm
22:22 bisectable6 joined #moarvm
22:22 releasable6 joined #moarvm
22:22 committable6 joined #moarvm
22:22 nativecallable6 joined #moarvm
22:29 squashable6 joined #moarvm
22:30 Ven joined #moarvm
22:39 AlexDaniel` joined #moarvm
23:16 statisfiable6 joined #moarvm
23:27 statisfiable6 joined #moarvm
23:59 MasterDuke is MVMStringBody.num_graphs guaranteed to always be correct?

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