|
00:17
benabik joined
00:56
FROGGS joined
01:20
flussenc1 joined
01:21
dagurval_ joined,
cxreg2 joined,
TimToady_ joined
01:39
flussence joined
01:41
btyler joined
01:49
ilbot3 joined
03:33
lizmat_ joined
03:35
lizmat__ joined
05:13
cxreg joined
06:49
lizmat joined
07:13
zakharyas joined
07:14
Ven joined
07:21
brrt joined
08:04
brrt joined
09:11
brrt joined
|
|||
| nwc10 | PyPy's second call for donations for STM is interesting: pypy.org/tmdonate2.html | 09:37 | |
| The current solution depends on having a huge address space available. Porting to any 32-bit architecture would quickly run into the limitation of a 2GB or 4GB of address space | |||
| and it's actually only for x86-64 Linux currently, using clang features | |||
| nowt wrong with that, but I was a bit surprised to discover how non-general it was | 09:38 | ||
|
10:22
brrt joined
10:41
brrt left
|
|||
| moritz | nwc10: also, when you need a huge address space, running under ulimit will be impossible | 13:23 | |
|
13:58
btyler joined
15:54
jnap joined
16:22
rurban_ joined
|
|||
| jnthn | I've been running/doing all my dev with spesh of late, and it feels fairly stable (as in, not seen it cause regressions). Think it's probably time to merge. :) | 16:43 | |
| So we get the daily runs and other feedback ahead of the monthly. | 16:44 | ||
| Can continue to do other experimental stuff in branch(es). | |||
| jnthn has a merge done locally | 16:49 | ||
|
16:51
dalek joined
|
|||
| dalek | Heuristic branch merge: pushed 130 commits to MoarVM by jnthn | 17:02 | |
| timotimo | i like the sound of that | 17:26 | |
| jnthn | Merged it | ||
| I'm working on multispec now. | |||
| timotimo | i saw a few commits land :) | 17:27 | |
| jnthn | Though will get some nom soon. | ||
| timotimo | made me a bit giddy with anticipation | ||
| i'm going to cook some noodles here now as well | |||
| jnthn | :) | 17:28 | |
| cognominal | what is multispec? | 17:33 | |
| jnthn | cognominal: Mechanism for avoiding calling onlystar protos. | 17:34 | |
| cognominal | bypassing them when possible? | 17:37 | |
| jnthn | Right | 17:38 | |
| nwc10 | jnthn: works on "my" machine | 17:39 | |
| jnthn | Basically, when we can find the palce to go directly in the cache. | ||
| nwc10 | (build with address sanitizer) | ||
|
17:39
rurban_ joined
|
|||
| nwc10 | s/build/built/ | 17:39 | |
| jnthn | nwc10: yay | ||
| timotimo | what other specs do we have besides boolification spec? | 17:47 | |
| jnthn | invocation | 17:48 | |
| container | |||
| timotimo | ah, right | ||
| jnthn | They all hang of the stable, meaning if you know the type you know what they're gonna be. | ||
| timotimo | yup, i remember that part | ||
| jnthn | *off | 17:50 | |
| hm, think my is_thunk thingy works out on Moar. | 17:58 | ||
| jnthn spectests while going to fetch nom | 17:59 | ||
| benabik | stable isn’t very stable if you keep changing it. | ||
|
18:08
FROGGS joined
|
|||
| TimToady | you just have to wait until it's Christmas, when the stable will contain all the usual animals | 18:10 | |
| cognominal | with jnthn, we already got one of the mages if this word is a cognate of magician. | 18:23 | |
| ho, you say Magi in english | 18:25 | ||
| timotimo | i don't know what "is_thunk" does :( | 18:40 | |
| dalek | arVM: 06f139e | jnthn++ | / (6 files): Add a way to flag a frame as a "thunk". This will allow us to ignore it in certain situations when considering caller and outer chains. |
19:06 | |
| timotimo | ah | 19:08 | |
| now i know \o/ | 19:13 | ||
| benabik | Now you just need red and blue lasers. | 19:14 | |
| dalek | arVM: 8e0419f | jnthn++ | / (6 files): Add ctx[outer|caller]skipthunks ops. |
||
| nwc10 | jnthn: multispec works on "my" machine | 20:20 | |
| jnthn | nice | 20:21 | |
| I'm now...uh...remembering how to PIR :/ | |||
| dalek | arVM/named_to_positional: b80ddf0 | (Timo Paulssen)++ | src/ (7 files): WIP nameds in the callsite |
20:39 | |
| jnthn | You can build all of Rakudo and NQP on Moar in less time than it takes to build CORE.setting on Parrot... | ||
| timotimo | gist.github.com/timo/1ebefa958416fa947096 <- got a hot tip for me, jnthn? | ||
| jnthn | ah shit... | ||
| it hangs loading CORE.setting :/ | 20:40 | ||
| timotimo | d'oh :( | ||
| jnthn | *sigh* | ||
|
20:40
jnap1 joined
|
|||
| jnthn | ah, think I found it... | 20:50 | |
| timotimo: Not really got much hints looking at the bt... | 20:53 | ||
| callsites[i]->arg_name = malloc((nameds / 2) * sizeof(MVMString)); | 20:54 | ||
| Should be MVMString * there | |||
| timotimo | oh! | ||
| right, because it's MVMString ** | |||
| jnthn | callsites[i]->arg_name | ||
| Wanna plural it I guess | |||
| timotimo | but MVMString is bigger than MVMString*, so it's not completely wrong | ||
| yeah, good idea | |||
| jnthn | callsites[i]->arg_name = 0; | ||
| s/0/NULL/ | |||
| MVM_gc_worklist_add(tc, worklist, frame->cur_args_callsite->arg_name[i]); | 20:55 | ||
| Missing & | |||
| But...you don't need to scan those. | |||
| timotimo | oh? | ||
| jnthn | They are anchored thanks to being in the string heap | 20:56 | |
| timotimo | oh! | ||
| that's good! | |||
| jnthn | You will need to do it for flattening case, though | ||
| Just like we scan ->args separately for that case. | |||
| But the issue in | |||
| MVM_gc_worklist_add(tc, worklist, frame->cur_args_callsite->arg_name[i]); | |||
| Is that we add the *address* of a collectable to the worklist | 20:57 | ||
| MVM_gc_worklist_add(tc, worklist, &frame->cur_args_callsite->arg_name[i]); | |||
| Would be correct. | |||
| So that's why you get the segfault. | |||
| timotimo | flattening case, i don't think i have anything for that yet, yes? | 20:58 | |
| jnthn | don't see it :) | 20:59 | |
|
21:11
btyler joined
22:06
JimmyZ joined
22:07
woosley joined
22:19
colomon joined
22:47
dalek joined
22:49
dalek joined
23:05
lizmat joined
23:08
lizmat joined,
woolfy joined
|
|||
| dalek | arVM: 994bd40 | jnthn++ | src/core/exceptions.c: Make backtrace generation ignore thunks. |
23:23 | |
|
23:38
lizmat joined,
woolfy joined
|
|||