|
08:36
FROGGS[mobile] joined
09:33
FROGGS[mobile] joined
09:46
lue joined
11:01
carlin joined
12:48
brrt joined,
zakharyas joined
13:03
dalek joined
13:09
Util joined
13:58
dalek joined
16:11
zakharyas joined
17:02
brrt joined
17:12
rurban_ joined
18:33
njmurphy joined
18:58
FROGGS joined
|
|||
| timotimo | jnthn: how do you feel about recording the size of the inter-generational link list after each gc run in the profiler? | 19:59 | |
| jnthn | timotimo: I was thinking of implementing that. | ||
| timotimo | in that case, let me steal that from you, so that you can do something harder :) | ||
| jnthn | timotimo: :P | 20:00 | |
| I think that CORE.setting compilation suffers because it gets really long there | |||
| Having something recording it would be helpful in seeing if that's the case | |||
| timotimo | how do i handle a maybe or maybe not existing key in js? | 20:19 | |
| as in gc.gen2_roots // 0 | |||
| hoelzro | timotimo: roots = 'gen2_roots' in gc ? gc.gen2_roots : 0; | 20:20 | |
| (that's the best I can do) | |||
| timotimo | what'll gc.gen2_roots evaluate to if it doesn't exist? | 20:21 | |
| and what'll gc['gen2_roots'] evaluate to if it doesn't exist? | |||
| hoelzro | null | 20:22 | |
| gc.gen2roots is equivalent to gc['gen2_roots'] | |||
| timotimo | in that case, wouldn't +gc.gen2_roots work? | ||
| leedo_ | in Chrome it seems to evaluate to undefined | 20:24 | |
| timotimo | that'd be good enough | ||
| leedo_ | +undefined evaluates to NaN | 20:25 | |
| timotimo | :\ | ||
| that's not surprising i suppose | |||
| jnthn | .oO( NaNNaNNaNNaN WATMAN! ) |
20:26 | |
| timotimo | jnthn: i just run the new profiler output over rc-dragon-curve and it seems to be hovering just below 20k gen2_roots | ||
| anything in particular you'd like me to try? | |||
| jnthn | That doesn't sound excruciatingly bad... | ||
| timotimo | almost the exact same number for -e 'my @a; for (1 .. 1024_000) { push @a, 1 }; say +@a;' | 20:27 | |
| i'll go ahead and push the code, as it seems to work | 20:28 | ||
| rc-9-billion-names hovers just above 21k | |||
| or just below 22k actually | |||
| dalek | arVM: 23f2089 | (Timo Paulssen)++ | src/profiler/ (3 files): record size of gen2_roots list after each gc |
20:30 | |
|
20:33
dalek joined
|
|||
| timotimo | now if only there was a way to get an evaluatable profile from the core setting compilation ... | 21:01 | |
| FROGGS | what if we could enable just a few things that get recorded? | 21:06 | |
| timotimo | that'd be possible. perhaps through an environment variable? | 21:10 | |
| FROGGS | probable, yeah | 21:11 | |
| gnight | |||
| timotimo | the big thing is the call graph, but not having the call graph removes a crapton of data | ||