|
02:56
ilbot3 joined,
nebuchadnezzar joined
07:20
domidumont joined
07:28
domidumont joined
11:46
dogbert17 joined
12:07
domidumont joined
14:26
fxsksn joined
15:55
Kaiepi joined
16:21
releasable6 joined
19:41
lizmat joined
20:56
FROGGS joined
20:59
MasterDuke joined
21:41
MasterDuke joined
|
|||
| MasterDuke | how do it tell if a routine gets inlined? | 21:50 | |
| mst | smash the stack in the body and read the backtrace? | 21:51 | |
| (sorry) | 21:52 | ||
| jnthn | MasterDuke: --profile and look at the call stack tab | 21:55 | |
| MasterDuke | jnthn++, mst-- | 21:56 | |
| mst | :D | 21:57 | |
| MasterDuke | mst: heh, made me think of the xkcd about the space bar and controlling a fan | ||
| jnthn: is there any information in any of the logs about why something doesn't get inlined? | 22:03 | ||
| for instance, Sequence.eager isn't inlined (with 245+k calls), but is only `multi method eager(::?CLASS:D:) { List.from-iterator(self.iterator).eager }` | 22:09 | ||
|
22:23
evalable6 joined
|
|||
| jnthn | MasterDuke: No, but if you compile MoarVM with the #define at the top of inline.c twiddled to a 1 instead, then it'll dump out reasons | 22:29 | |
| MasterDuke | thanks | 22:32 | |
| jnthn: i don't see a #define in inline.c. inline.h has `#define MVM_SPESH_MAX_INLINE_SIZE 384`, is that the one you're thinking of? | 22:34 | ||
| jnthn | MasterDuke: oh, could be in optimize.c | 22:39 | |
| MasterDuke | ah, i see it, thanks again | ||
| what's the number after the name? is that just some sort of id? | 22:42 | ||
| jnthn | Yeah, just the ID within the compilation unit. Useful to correlate with spesh log | 23:26 | |
|
23:26
ZofBot joined
|
|||