00:40 Kaiepi joined 01:12 Kaiepi joined 01:21 ZofBot joined, huggable joined 01:37 huggable joined, ZofBot joined 01:39 buggable joined 01:57 ilbot3 joined 02:27 Kaiepi joined
samcv timotimo, MasterDuke are there any release blockers atm for MoarVM? 03:07
MasterDuke don't think so 03:08
samcv: mostly off topic, but have you seen github.com/MoarVM/MoarVM/issues/745 ? 03:11
samcv no i haven't. thanks for bringing it up 03:21
Geth MoarVM: samcv self-assigned Code never reach in /src/strings/unicode_db.c github.com/MoarVM/MoarVM/issues/745
samcv++ created pull request #845: Use an array instead of a big switch for ShiftJIS decoding
05:12 domidumont joined 05:20 domidumont joined 05:25 nwc10 joined
06:28 domidumont joined
Geth MoarVM/spesh-limit-debug-log: 4b6e4ae69a | (Bart Wiegmans)++ | 6 files
[Spesh] Wrap debug printing

The intention is that MVM_spesh_debug_enabled(tc) can be made more specific than just having a spesh log file handle or not.
Specifically, setting MVM_SPESH_LIMIT in conjunction with MVM_SPESH_LOG will restrict the log to that particular frame.
06:32
MoarVM/spesh-limit-debug-log: 4186ace71b | (Bart Wiegmans)++ | src/spesh/optimize.c
[Spesh] Log graph of inlinee prior to inlining

Now that we can restrict the amount of logging done, no longer such a problem to log the entire graph.
07:30 zakharyas joined 07:40 zakharyas joined 10:10 zakharyas joined, brrt joined
brrt ohai #moarvm 10:10
samcv good * brrt 10:12
so for shiftjit we used to use two huge switches to convert back and forth, for valid indexes and codepoints. so since the indexes are contiguous except for 11 places, i made an array and then made a table with data of where the gaps were and the length of gaps so that's much faster to compile and posibly faster than having the compiler optimize large switch 10:14
but i'm unsure what to do for the reverse: because you have bsically random codeponits matching to random indexes
10:15 zakharyas joined
samcv looks like we save 28KB by getting rid of one of the big switches as well 10:17
jnthn If they're blocky enough then a 2-level table could work 10:20
samcv i mean they're really all over the place 10:21
sometimes there's a range of contiguous codepoints but most of it is just random since they don't map the same
idk. it's very long so it could be more contiguous than i think 10:22
Geth MoarVM: 0ad859add8 | (Samantha McVey)++ | 6 files
Use an array instead of a big switch for ShiftJIS decoding

Since we can easily map index->codepoint, turn the big switch into an array. There are only 11 gaps of the indexes, so we just check we aren't in a gap and apply an offset. Makes the shiftjis files compile twice as fast.
We still use a big switch for encoding because codepoints are not at all in order, so for now let the compiler handle this.
10:24
jnthn I'd left merging the PR for post-release, but I guess it's pretty safe. :-)
samcv jnthn: yeah. i tested it on every possibility with the old function 10:26
and iterated every choice and made sure output was identical. it should make compiling faster for those with slow computers
dunno how many there are. i know it took forever to compile shiftjis.o when i emulated MIPS :) 10:27
jnthn :-)
samcv++
brrt can i maybe also interest you in: github.com/MoarVM/MoarVM/pull/842 10:29
10:39 zakharyas joined 11:12 AlexDaniel joined 13:08 zakharyas joined 13:37 domidumont joined 15:14 domidumont1 joined 16:23 Kaiepi joined 17:04 domidumont joined 18:59 brrt joined
brrt \o 19:01
20:42 robertle joined
timotimo o/ 21:24
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/04/23/...u-station/ 21:26
samcv lizmat++ 21:29
timotimo lizmat++ 21:33
jnthn lizmat++ # thanks 21:34
23:12 Kaiepi joined