|
00:25
colomon_ joined
02:05
colomon joined
02:52
colomon joined
06:32
domidumont joined
06:53
domidumont joined
06:58
domidumont joined
07:10
zakharyas joined
08:06
domidumont joined
09:32
Ven_ joined
10:11
domidumont joined
11:29
brrt joined
|
|||
| brrt | ehm, you may with reason not believe me | 11:41 | |
| jnthn | yeah, right | ||
| brrt | but i think i finally, really, truly know how to implement the register allocator 8-) | ||
| jnthn | \o/ | ||
| brrt | so now the issue is to make time for it | 11:42 | |
| one of the interesting bits is that i think i've discovered that linear scan (known as fast) is a subset of graph coloring | 11:43 | ||
| that is going to be a blogpost soonish | |||
| graph coloring is known as slow, because a): graph coloring is .... whatsitcalled? traveling-salesman-equivalent complexity | 11:44 | ||
| and linear scan is known as fast because linear | |||
| but in reality linear scan is just a special version of graph coloring | 11:45 | ||
| timotimo | NP-complete is the word you're looking for? | 11:46 | |
| brrt | i.. think so | ||
| yeah, thanks :-) | |||
| arnsholt | That is indeed NP-complete | 12:02 | |
| When I took a databases class some years ago, we spent quite a bit of time discussing serialization algorithms/strategies (that being rather important for databases) | 12:03 | ||
| brrt | uhuh | ||
| arnsholt | First we discussed locking, which has some obvious drawbacks, then we covered something called view-serializability (IIRC) | ||
| brrt | (fwiw, i'm not sure a live range in this model can still meanigfully live in two registers.... that just has to go then) | ||
| arnsholt | Think we spent like two lectures on it | ||
| brrt | do continue | 12:04 | |
| arnsholt | And then at the end they just casually drop, "but this turns out to be NP-complete" | ||
| >>>>>>.<<<<<<< | |||
| brrt | lol | ||
| yeah, sucks when that happens | |||
| that basically happened to my idea of 'let's find the ideal combination of renewable generators for a given area' | 12:05 | ||
| 'this is going to be easy, just fill them up with the most ideal spots first... oh wait' | |||
| arnsholt | Yeah, turns out a rather large fraction of interesting problems are NP-complete | ||
| brrt | or solved by union-find | 12:09 | |
| timotimo | when you get a job, you better find a union to join ... | 12:10 | |
| brrt | :-) | 12:45 | |
| okay, i'm off, i have to write thesis reports and stuff | |||
| see you later | |||
|
16:27
FROGGS joined
18:56
domidumont joined
|
|||