|
github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today Set by moderator on 13 May 2018. |
|||
|
01:01
AlexDani` joined
01:29
zakharyas joined
01:57
ilbot3 joined
|
|||
| moderator | github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today | ||
|
05:31
domidumont joined
05:37
domidumont joined
06:22
domidumont joined
06:50
Ven`` joined
07:18
robertle joined
09:04
nwc10 joined
09:54
Ven`` joined
10:05
AlexDaniel joined
|
|||
| dogbert2_ | hmm, MoarVM panic: Collectable 0xb330df9c in fromspace accessed | 11:07 | |
| the code fails here, github.com/MoarVM/MoarVM/blob/mast...erp.c#L32, when running t/spec/integration/eval-and-threads.t | 11:15 | ||
| MVM_GC_DEBUG=1 and the nursery has been made smaller | 11:16 | ||
| jnthn, timotimo: should I report this ^^ or is it a known problem? gist.github.com/dogbert17/7b4c0161...0238fe8f29 | 11:29 | ||
|
11:56
Ven`` joined
12:03
Ven`` joined
|
|||
| jnthn | dogbert2_: Not known tome | 12:07 | |
| *to me | |||
| Please file an issue | |||
| dogbert2_ | jnthn: txh for looking, will create an issue | 12:20 | |
|
13:03
Ven`` joined
|
|||
| Geth | MoarVM/master: 5 commits pushed by (Bart Wiegmans)++, (Elizabeth Mattijsen)++ | 13:13 | |
|
13:15
domidumont joined
13:25
zakharyas joined
14:12
Ven`` joined
14:38
Ven`` joined
15:41
Voldenet joined
|
|||
| timotimo | jnthn: if "ensure_uninstrumented" for the instrumented profiler became a no-op, we'd be wasting some performance after profiling is ended, but we'd otherwise have to have some indirection somewhere for the bytecode we're running in a frame, and do some kind of lazy deletion for bytecode segments, too ... | 16:09 | |
| alternatively, at least in the profiler instrumentation case, we could go through all frames on the heap (need to consider ones that aren't on the stack) and just count how many profiling instructions come before the instruction pointer and subtract that amount when switching to the original bytecode | 16:11 | ||
| but that's also meh | |||
| jnthn | timotimo: hmm, yes :S | 16:13 | |
| Not to mention JIT | 16:14 | ||
| timotimo | aye | ||
| the profiler ops are, thankfully, completely harmless when called while profiling is turned off | |||
|
16:26
domidumont joined
|
|||
| [Coke] | (diff tool for windows) I am very happy with beyond compare for windows & mac | 16:29 | |
| yoleaux | 12 May 2018 03:13Z <Zoffix> [Coke]: are you in contact with Mark at TPF? The link to "his blog for April" in last grant update links to SO, not the blog: news.perlfoundation.org/2018/05/gra...l#comments (P.S.: I still have no idea who's *my* grant manager :|) | ||
| Geth | MoarVM: dd2d7f9edb | (Timo Paulssen)++ | 2 files don't uninstrument any more the instrumentation level barrier can be hit by one thread while another is currently running the bytecode, which causes it to reach a semi-random place in the bytecode and start interpreting values as instructions. |
17:24 | |
| lizmat | wow | 18:04 | |
| timotimo | hmm? | 18:06 | |
| lizmat | well, I just read the last commit message | 18:07 | |
| "start interpreting values as instructions" | |||
| timotimo | yeah | 18:08 | |
| that's where "const_iX NYI" came from | |||
| lizmat | aha, so would it make sense to bump MoarVM over this ? | ||
| timotimo | we could do that | ||
| lizmat | ok, I will then | 18:09 | |
| but after I get back from being afk for a few hours& | 18:14 | ||
| Geth | MoarVM: eebc4620e6 | (Timo Paulssen)++ | src/profiler/instrument.c profile: extract recursion loop for smaller stack frames dumping call graphs used to put 144 bytes onto the stack for every slice of recursion, now it'll deposit just 64 bytes for every slice and put a 128 byte frame on top to do most of the actual work. |
18:33 | |
| timotimo | it just so happens that the script i was randomly running to get profiler output has a little stack depth problem ... | 18:34 | |
|
20:15
committable6 joined,
squashable6 joined
|
|||
| lizmat | And another Perl 6 Weekly hits The Net: p6weekly.wordpress.com/2018/05/14/...ough-time/ | 22:23 | |
| timotimo | lizmat: the regex speed increase is actually for interpolating regexes into regexes | 22:25 | |
| lizmat | aaahh ok | 22:26 | |
| timotimo | oh, what did i fix in moarvm that was related to gc? | ||
| lizmat | eefd9a436285bd ? | 22:28 | |
| timotimo | ooh | 22:30 | |
| right, that just fixes a small bug in the gc tab in the profiler output, but that's good to have of course | |||
|
22:32
greppable6 joined
|
|||