|
01:16
vendethiel joined
01:38
dalek joined
02:25
vendethiel joined
05:16
_longines joined
06:21
mojca joined
07:25
domidumont joined
07:29
domidumont joined
07:52
FROGGS joined
08:21
mojca joined
08:23
zakharyas joined
08:54
vendethiel joined
09:33
mojca joined
10:20
vendethiel joined
12:50
domidumont joined
12:51
domidumont joined
13:08
brrt joined
|
|||
| brrt | \o | 13:08 | |
| timotimo | o/ | 13:09 | |
| brrt | i fixed the json-fast bug fwiw :-) | 13:12 | |
| and somewhat-fixed the double-labelling bug | |||
| jnthn | brrt++ | 13:13 | |
| timotimo | i like the sound of that. so that's what the last commit was about? | 13:14 | |
| brrt | aye | 13:19 | |
| i'm still under the impression i'm not getting something about how to write a register allocator | |||
| jnthn | brrt: I'd probably try and find something I could nab from the literature :) | 13:27 | |
| brrt | hmm, yeah, the thing is | 13:30 | |
| the literature has steps, of which i don't fully understand the purpose | |||
| like discovering lifetime holes | 13:31 | ||
| why do that | |||
| (and processes things in reverse order. why do that, too?) | |||
| jnthn | Reverse order is normally 'cus it's doing some kind of backward-propagating data flow analysis or so | 13:32 | |
| brrt | this from the article which claims that data flow analysis is not necessary due to register allocation on SSA form | 13:33 | |
| jnthn | Link? | 13:34 | |
| brrt | i'm pondering writing a prototype in a language that is not C | ||
| lemessee | |||
| www.christianwimmer.at/Publications...mer10a.pdf | |||
| jnthn | "Lifetime holes occur because the control flow graph is reduced to" | 13:38 | |
| a list of blocks before register allocation | |||
| brrt | same in the expression compiler | 13:39 | |
| jnthn | The block order has a | 13:40 | |
| high impact on the quality and speed of linear scan: | |||
| ...I'm suspecting then that picking that order is part of the "fun" | |||
| brrt | hmmm | 13:41 | |
| oh, great | |||
| however | |||
| we're not dealing with loops (yet) | 13:42 | ||
|
13:51
colomon joined
|
|||
| brrt | it looks like a good strategy, dunnit? | 14:11 | |
| jnthn | I only had chance to scan through the paper very quickly (doing $dayjob stuffs) but it seems like a decent fix. | 14:12 | |
| *fit | |||
|
14:19
vendethiel joined
14:28
FROGGS joined
|
|||
| brrt | yeah. i want to keep complexity to a minimum | 14:28 | |
|
15:16
vendethiel joined
15:21
brrt joined
15:50
vendethiel joined
17:15
FROGGS joined
17:56
vendethiel joined
18:31
mojca joined
18:40
domidumont joined
19:02
vendethiel joined
|
|||
| dalek | arVM/east_asian_width: 200d39b | (Matthew Wilson)++ | tools/ucd2c.pl: add East_Asian_Width property |
20:01 | |
| diakopter | GitHub Desktop--)--)--)--)-- | 20:03 | |
| timotimo | huh? | 20:06 | |
| diakopter | -- | ||
| timotimo | what's bad about it? | 20:07 | |
| last time i tried it it seemed pretty neat | |||
| diakopter | it's neat | 20:14 | |
| I'm sure for some high percentage of users | |||
| but yes, I'm grousing about being that user who has a poor experience ;) | 20:15 | ||
|
20:45
geekosaur joined
|
|||
| timotimo | heh. | 21:08 | |
| well, at least it also gives you a powershell link with all git commands readily accessible on windows | |||
| i suppose a mac user wouldn't be impressed by that | |||
|
21:30
Ven joined
22:06
synopsebot6 joined
22:07
synopsebot6 joined
22:22
synopsebot6 joined
22:28
synopsebot6 joined
23:31
vendethiel joined
|
|||