| timotimo | we may get a noticable speed boost if we have the string constants start and end with " inside moarvm's C code that generates the datastructure :) | 00:05 | |
|
01:24
FROGGS_ joined
01:27
colomon joined
03:14
colomon joined
04:34
woolfy left
04:57
woosley joined
05:01
woosley joined
05:23
woosley joined
|
|||
| tadzik | hmm, so you do have to be in the nqp directory or not? | 05:28 | |
| when profiling | |||
|
05:40
woosley joined
|
|||
| TimToady | no, I just ran one in the spec directory | 05:47 | |
| er, perl-bench | |||
|
06:46
woosley joined
|
|||
| tadzik | hmm | 07:00 | |
|
07:04
woosley joined
|
|||
| jnthn | tadzik: In the very first code I pushed you did; then I pushed some stuff so it installs the template and looks it up from the install location. | 07:41 | |
| tadzik | jnthn: okay, I got it working now :) | ||
| jnthn: hmm. If I sum up all times in call_graph, plus all times in gcs, plus spesh_time, I don't end up with total_time. Did I forget about something? | 07:43 | ||
| jnthn | tadzik: You summed up exclusive times, I'm assuming? | 07:44 | |
| tadzik | jnthn: correct | ||
| jnthn: the sum is lower | |||
| 16461692 vs 17314798 of total_time | |||
| jnthn | A little lower, or a lot lower? | ||
| tadzik | so about 5% is lost | 07:45 | |
| profiler overhead? | |||
| jnthn | I think it's more that it doesn't calculate the total_time at the point we drop out of the last profiled call frame since we started, but rather at the point we are asked for the profile data. | 07:46 | |
| tadzik | close enough :) | 07:47 | |
| jnthn | I can probably do something about that, but need to worry about preparing my async talk some more first... | 07:48 | |
| tadzik | don't worry about it :) | ||
| if my end results ends up looking blatantly wrong, I'll get back to you | |||
| if it doesn't, I'll still get back to you though :P | 07:49 | ||
| jnthn: also, I noticed that things like lib/NativeCall.pm6.moarvm:-1 show up | 07:50 | ||
| how does that happen? | |||
| jnthn | That means "unknown line number" | ||
| The UI I wrote turns that into <unknown> | 07:51 | ||
| Ofter it is due to compiler-genreated fames, like the deserialization or load-time frame. | |||
| tadzik | right, but why is both NativeCall.pm6 and NativeCall.pm6.moarvm... oh, I see | ||
| jnthn | It fails to find the HLL filename and falls back to the VM one, iirc | 07:54 | |
|
08:49
Ven joined
|
|||
| sergot | o/ | 11:01 | |
|
11:23
lizmat joined
11:34
klaas-janstol joined
12:05
zakharyas joined
14:13
klaas-janstol joined
|
|||
| masak | "mentee" is a wonderful word. | 14:27 | |
| jnthn | :D | 14:28 | |
| jnthn is an apporvee of it | |||
| masak | hm, this ain't #perl6... :) | ||
| [Coke] | mentees: the frosh maker. | 14:45 | |
|
15:20
FROGGS joined
17:25
Ven joined
17:33
zakharyas joined
|
|||
| timotimo | oh the hue-mentee | 18:12 | |
|
18:23
Ven joined
18:35
cognome joined
19:05
Ven joined
19:26
cognome joined
19:58
Ven joined
20:04
zakharyas joined
20:05
colomon joined
20:10
lizmat joined
20:24
woolfy joined
20:56
woolfy left
21:46
colomon joined
22:33
flussence joined
22:36
colomon joined
23:12
btyler joined
|
|||