github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
Set by moderator on 12 May 2018.
timotimo it's taken me like two weeks, but i've finally checked to see that turning off instrumentation is what's causing the strange crashes like "const_iX NYI" 01:04
it really is switching the bytecode from instrumented to original while another thread is executing from it
MasterDuke hm, easy fix? 01:06
timotimo all the profiler ops already behave themselves when profiling is actually turned off 01:09
so just keeping the instrumented bytecode would be at least safe, just bad for performance
01:57 ilbot3 joined
moderator github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
Geth MoarVM: bd9156c6a1 | (Timo Paulssen)++ | src/spesh/dump.c
put sizes of spesh graphs into the spesh log

since the size of a frame is important for inlining, it's interesting to know this.
02:49
MoarVM: 99971b5378 | (Timo Paulssen)++ | src/spesh/dump.c
fix format string parameters

why doesn't my gcc warn about this? no clue.
03:07
09:55 domidumont joined 10:12 robertle joined 10:14 domidumont joined
Geth MoarVM: bdw++ created pull request #858:
Moar JIT ops
11:29
MoarVM: d5830c8539 | (Bart Wiegmans)++ | tools/jit-comparify-asm.pl
[Tools] Make jit-comparify break on 'ret'

All JIT-compiled frames have only one 'ret' instruction ever, so anything after the first 'ret' instruction is the data section, which can be really confusing to disassemble. This is an option, but on by default.
11:33
MoarVM/jit-stack-walker: 13 commits pushed by (Bart Wiegmans)++
review: github.com/MoarVM/MoarVM/compare/1...aaf418b084
12:04
12:06 brrt joined
brrt I haven't found the bug on windows just yet... 12:06
12:45 zakharyas joined 13:39 domidumont joined 13:47 domidumont joined, bartolin joined 14:43 AlexDaniel joined
timotimo i'm leaning towards "uninstrumentation shall never happen", at least not in a multi-threaded context 16:14
22:31 lizmat joined 23:32 MasterDuke joined