github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
Set by moderator on 28 April 2018.
01:41 FROGGS_ joined 01:57 ilbot3 joined
moderator github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
02:03 Kaiepi joined 07:45 brrt joined
brrt \\o 07:45
samcv o/ 07:57
08:11 domidumont joined 08:17 domidumont joined
nine .tell jnthn So....what's the goal for pluggable-spesh exactly? Funny coincidence: the very day I saw the first pluggable-spesh commit I was already thinking about how unit tests for spesh would make development easier and how to not have to write them in C and if we access spesh from Perl 6 how cool it would be to be able to implement optimizations in Perl 6. 08:58
yoleaux 28 Apr 2018 18:26Z <Zoffix> nine: do you know if any recent work went in that would cause issues until you blow away .precomp dir? It happened with one of my programs after rakudo upgrade. Here's the second time it happens with my stresstest run where a bunch of tests fail and keep failing until you blow away .precomp dirs (declutter command in the gist): gist.github.com/zoffixznet/43e5db7...237262b856
nine: I'll pass your message to jnthn.
28 Apr 2018 18:34Z <brrt> nine: that is unfortunate, but after some thought, not unexpected, since I don't think that test exercises the JIT particularly hard
28 Apr 2018 21:25Z <Zoffix> nine: the problem I mentioned in earlier bot message now has an Issue with repro instructions for it: github.com/rakudo/rakudo/issues/17...-385205863
Geth MoarVM: ae36d60f92 | (Samantha McVey)++ | 3 files
Use <private-use-XXXX> and <surrogate-XXXX> for Uni names

Follow the Unicode recommendations for names. Previously we used the names that were given in UnicodeData.txt which are more descriptive e.g. <Non Private Use High Surrogate-XXXX> This now becomes <surrogate-XXXX>.
09:19
10:18 Kaiepi joined
jnthn .tell nine See my design notes at gist.github.com/jnthn/e51a06c6882f...0a3dd373e6 for what the pluggable spesh stuff is aimed at; let me know if you've any more questions :) 11:09
yoleaux 08:58Z <nine> jnthn: So....what's the goal for pluggable-spesh exactly? Funny coincidence: the very day I saw the first pluggable-spesh commit I was already thinking about how unit tests for spesh would make development easier and how to not have to write them in C and if we access spesh from Perl 6 how cool it would be to be able to implement optimizations in Perl 6.
jnthn: I'll pass your message to nine.
Geth MoarVM: b3e7dd36b2 | (Samantha McVey)++ | src/strings/unicode_ops.c
Make sure getstrfromname will resolve generated names

Should make sure <control-XXXX>, "UNIFIED IDEOGRAPH-XXXX" etc. will work with getstrfromname op. Some of these didn't work previously and now work, others would have stopped working with the previous commit. If anybody was using them this should make sure it doesn't break their code.
11:49
samcv jnthn: you around. we define strtoll and snprintf specially for MSVC in src/core/coerce.c but i would like to put it another location as well. where should it be put so other files can use it too? 11:58
moar.h is probably not the best place, though i guess we could include it from platform? not sure
Geth MoarVM: c702f4c66a | (Samantha McVey)++ | src/strings/unicode_ops.c
Use strtol instead of strtoll since MSVC doesn't define strtoll
12:03
jnthn No, moar.h is for folks embedding to import too, so we shouldn't put anything in there not prefixed with MVM_ 12:11
Or MVM for types
Something in platform and included per file that needs it is probably best 12:12
Though seems you found a way around it anyway :)
12:18 Kaiepi joined
samcv jnthn: made sure that "CJK UNIFIED IDEOGRAPH-XXXX" works. i made sure it doesn't allow spaces between the - and the number, and doesn't allow 0x or a + or minus sign since strtol silently handles those. though eventually i may want to make sure it restricts it to actual codepoints that are named that 12:27
so you can't do <control-3FAEE> etc
12:30 Ven`` joined
jnthn *nod* 12:43
Yeah, probably wise to do that later on
Well, or as soon as you want, but being strict about that will likely avoid some accident or other :) 12:44
samcv yeah 12:46
13:33 evalable6 joined 13:50 Ven`` joined
timotimo i haven't done any research whatsoever, but my inkling for profiler crashes is that maybe the instrumentation barrier when leaving profiled mode doesn't trigger for any of the threads other than the one actively leaving the code, and since we only check the barrier when entering code, not leaving (citation needed) we don't trigger until it's too late 14:24
so anyway, jstuder's slice op is getting jitted, except the example code has an sp_getstringfrom before that, which bails the template jit out (the exprjit can handle it, though. it doesn't reach that point, however) 15:00
so ... about that "spesh slot" arg type for emit c call ... :)
15:29 zakharyas joined 15:31 zakharyas joined 15:36 zakharyas joined
dogbert17 check this out, www.dyncall.org/news 16:17
www.dyncall.org/changelog#dc10 16:19
timotimo allocated space used for thunk now always W^X (req. for e.g. OpenBSD >= 6.0) 16:20
yay 16:21
16:21 zakharyas1 joined 16:36 zakharyas joined 16:49 Ven`` joined 17:06 zakharyas joined 19:14 Kaiepi joined 20:47 Kaiepi joined
Geth MoarVM: MasterDuke17++ created pull request #852:
Clean up src/jit/core_templates.expr a bit
21:37
22:40 Kaiepi joined