| dalek | arVM: 7193cda | timotimo++ | src/moar.c: oops. that was *not* the mutex ... |
00:09 | |
| arVM: d912f3c | timotimo++ | src/6model/6model.c: free debug names in STables |
|||
| arVM: 616acc8 | timotimo++ | src/6model/reprs/MVMCompUnit.c: free the string_heap_fast_table in CompUnit bodies |
|||
| arVM: 30d7a32 | timotimo++ | src/moar.c: free permroot descriptions. |
00:11 | ||
| arVM: 463fded | timotimo++ | src/core/threadcontext.c: free each TC's finalization queue in tc destruction |
00:16 | ||
| timotimo | in the current frame code, we're leaking frames that fell off the FSA size limit and got allocated with malloc instead. not sure what path should normally at some point reach those pointers and free them, but we're probably also missing cleanup for those that are "in" the FSA. though these would get destroyed whole with the FSA itself at teardown | 00:18 | |
| here i see a bunch of things leaking that got created from serialization_demand_object and actually have corresponding MVM_free in 6model_stable_gc_free. not sure why they don't end up getting freed | 00:21 | ||
| maybe we're not going through the whole gen2 to gc_free stuff on shutdown, or something like that? | |||
|
05:39
domidumont joined
05:44
domidumont joined
06:10
domidumont joined
07:21
cognominal joined
07:37
cognominal joined
08:24
brrt joined
08:48
zakharyas joined
09:24
zakharyas joined
11:48
brrt joined
|
|||
| brrt | good * #perl6 | 11:48 | |
| eh, #moarvm | |||
|
11:53
vendethiel- joined
|
|||
| jnthn | o/ brrt :) | 12:02 | |
| brrt | \o jnthn | 12:04 | |
| ehm, i was wondering how far along you are with the reframe branch | 12:05 | ||
| because, if you want to merge it, then, probably, you should just do it | |||
| jnthn | brrt: It builds NQP/Rakudo fine. Fails some spectests. | ||
| Some, some debugging remains. | |||
| brrt | right | ||
| jnthn | Plus the JIT thing. ;) | ||
| brrt | and disable the JIT for now, i was going to say | ||
| which, almost, works | |||
| something funny happening | 12:06 | ||
| jnthn | Well, I'll have to go debugging still, so there's time for getting the JIT to actually work :) | ||
| brrt | yeah, but there is my velocity, and there is yours | ||
| on the long list of to-do-things, i want better jit graph logging | 12:07 | ||
| rather than the adhoc logging we have now, better something like spesh logging | |||
| oh, i thought of a way to avoid mallocing the 'fake extop arguments' | 12:08 | ||
| we can put these in a 'data' section in the jit code | |||
| which is also where we can place (function) pointers, which would allow storage and linking | 12:10 | ||
| not only that, it's much more elegant that way... | 12:12 | ||
| do you think lifetime hole analysis is worth the complexity? | 12:16 | ||
|
12:47
harrow joined
|
|||
| jnthn | Sorry, was in a meeting... | 13:07 | |
| +1 on the data section stuff | |||
| LHA - don't know it's worth the complexity in the short term. | 13:08 | ||
| brrt | well, i don't really know either.... :-) | 13:14 | |
| hmmm | |||
|
14:14
colomon joined
15:15
colomon_ joined
15:21
colomon_ joined
16:30
domidumont joined
17:06
ggoebel114 joined
18:15
brrt joined
19:00
dalek joined
19:05
dalek joined
19:37
dalek joined
19:56
colomon joined
20:26
jnthn joined
20:33
dalek joined
20:38
colomon_ joined
20:40
dalek joined
|
|||