01:18 stmuk_ joined 03:59 dalek joined 05:22 brrt2 joined
brrt2 code.facebook.com/posts/1568350381...rformance/ 05:23
it is not very rich in detail unfortunately 05:31
one of the thingw i like about moar is that the responsibilities of spesh and the jit are cleanly separated 05:32
nwc10 we have 2 brrts now? You've managed to clone yourself? 05:49
Yay \o/ 05:50
05:58 domidumont joined 06:04 domidumont joined
brrt2 not sure if the other clone is currently very active 06:14
06:47 domidumont joined 06:54 domidumont joined 07:06 domidumont joined 07:09 domidumont joined 07:43 brrt2 joined 07:58 dalek joined
nine So it's a failover brrt? 10:24
10:42 btyler joined 10:45 brrt joined
jnthn timotimo: The threads form a linked list hanging off instance 10:55
timotimo: Linked list 'cus it makes handling thread death easier than an array :) 10:56
timotimo ah!
well teaches me for just looking into Instance
is that a sensible way to try to get profile data from all threads?
or should i also look into marking the thread stolen or something like that?
jnthn I think all threads *do* recording profiling data 10:57
timotimo yes, they do record
jnthn Or at least, can
timotimo is it safe to just grab the data while the threads are potentially still running, though? :) 10:58
jnthn Hmm 10:59
Good question :)
Potentially not :) 11:00
Also those threads may end up with "unfinished business" in their call stacks
Which may throw off timings etc.
12:36 zakharyas joined
stmuk_ some info added to github.com/MoarVM/MoarVM/issues/410 14:42
(result of bisecting rakudo & lldb thread backtrace)
and valgrind 14:49
timotimo acquired noms 15:00
16:43 zakharyas joined 16:58 FROGGS joined 17:03 domidumont joined 17:05 domidumont joined 17:54 domidumont joined 17:56 zakharyas joined 18:49 zakharyas joined 19:57 diakopter___ joined 20:08 domidumont joined 22:02 Ven joined 22:07 Ven joined 22:19 Ven_ joined 23:23 Ven joined 23:50 Ven joined