|
00:07
colomon joined
00:34
colomon joined
02:02
jnap1 joined
02:12
woosley joined
02:44
FROGGS_ joined
06:10
FROGGS joined
06:42
FROGGS joined
07:07
FROGGS joined
08:07
zakharyas joined
08:49
lizmat joined
08:52
lizmat_ joined
09:32
lizmat joined
|
|||
| nwc10 | jnthn++ # works on "my" machine again | 09:59 | |
|
10:01
lizmat joined
|
|||
| jnthn | nwc10: yay | 10:03 | |
|
10:24
lizmat joined
12:10
lizmat joined,
woolfy joined
12:14
woolfy left
12:33
woolfy joined
12:42
FROGGS joined
12:44
lizmat joined
12:46
woolfy joined
13:14
woolfy left
13:32
woolfy joined
13:33
woolfy left
13:55
btyler joined
14:31
jnap joined
14:36
jnap1 joined
15:27
cognominal joined
16:26
lizmat_ joined
16:27
woolfy1 joined
16:47
colomon joined
16:50
woolfy1 left
17:04
Guest1337 joined
19:52
btyler joined
20:58
brrt joined
|
|||
| brrt | hi #moarvm | 20:59 | |
| i came to lobby for pull request github.com/MoarVM/MoarVM/pull/85 | 21:00 | ||
| which i've updated :-) | 21:01 | ||
| jnthn | brrt: Thanks, had tasks today and evening meeting, but have some free time in just a moment :) | 21:07 | |
| brrt | no worries :-) | 21:09 | |
| jnthn | brrt: yeahl that looks bett4r | 21:19 | |
| whoa | |||
| brrt: yeah, that looks better | |||
| dalek | arVM: 57384a3 | (Bart Wiegmans)++ | src/gc/ (3 files): Compact the gen2 overflows list. Fixes issue #62. |
21:24 | |
| brrt | \o/ | 21:33 | |
| timotimo | if something in the overflows list has been killed, it's supposed to turn into a null pointer, right? | 21:57 | |
| jnthn | Should do. | 22:03 | |
| Mostly it's STables that end up there, which tend to be quite long-lived...though of course lots of throw-away mixins will produce such things. Well, may do if you've enough of a midlife crisis issue with 'em. :) | 22:04 | ||
| brrt | timotimo - i should note that in this version the live object beyond the size of the compacted list aren't really 'null'-ed | 22:05 | |
| which is ok because a): no algorithm should look beyond num_overflows and b): they are pointes to live objects | 22:06 | ||
| jnthn | uh, dead ones, no? :) | 22:07 | |
| brrt thinking | 22:12 | ||
| no, they're live pointers :-) | |||
| the algorithm copies pointers 'backwards' - the compaction is through a): skipping null pointers and b): setting the barrier to the live object cursor | 22:13 | ||
| so there are - in principle - live pointers in the space after the compacted area | 22:14 | ||
| but again, this doesn't matter because any algorithm that looks beyond the live cursor is already in error | |||
| and yes, i've been thinking too much about compacting :-) | 22:15 | ||
| timotimo | is compaction done every time an overflow object dies? | ||
| brrt | no, only after collection | 22:19 | |
| jnthn | brrt: oh, I see what you mean :) | 22:21 | |
| brrt: Sorry. $dayjob meeting had imperial stout... :) | |||
| timotimo | er, of course only after collection | 22:23 | |
| i'm asking because my moar-gdb.py plugin is just going from the 0th overflow to the num_overflows'th object in the overflows list and analyzes them | 22:24 | ||
| jnthn | timotimo: Then you're fine, I think | 22:29 | |
| timotimo: You'll just not see NULLs in ther enow | |||
| timotimo | good | ||
| brrt | :-) time for sleep | 22:54 | |
| jnthn | 'night, brrt++ | ||
| brrt | night jnthn | ||
|
22:54
brrt left
|
|||