|
00:32
dalek joined
01:19
ggoebel joined
01:30
TEttinger joined
01:46
dalek joined
01:47
JimmyZ joined,
avuserow joined
01:48
Util joined,
[Coke] joined,
sergot joined,
masak joined,
PerlJam joined,
psch joined
02:11
colomon joined
02:14
colomon_ joined
02:19
colomon_ joined
02:25
colomon_ joined
02:30
colomon_ joined
02:34
colomon_ joined
02:37
colomon joined
02:42
colomon joined
02:45
colomon_ joined
06:53
FROGGS joined
|
|||
| FROGGS | o/ | 06:53 | |
| nwc10 | \o | 06:54 | |
|
06:56
zakharyas joined
|
|||
| jnthn | o/ | 07:01 | |
| nwc10 | jnthn: never mind the 10 minute commute, is the curry good? :-) | 07:02 | |
|
07:02
rurban joined
|
|||
| timotimo | i hope jnth's teachings go well today :) | 10:52 | |
| jnthn | nwc10: No curry yet :) | 11:18 | |
| Last night was beer and burger | |||
| Maybe curry tonight :) | 11:19 | ||
| But certainly curry at end of week ocne I'm back home... | |||
| jnthn has a long list of curry recipes he wants to try cooking | |||
|
13:09
ggoebel joined
13:40
JimmyZ_ joined
14:04
_longines joined
14:56
brrt joined
|
|||
| brrt | \o | 14:57 | |
| nwc10 | o/ | 14:58 | |
| brrt | i just had a *horrible* road trip | ||
| nwc10 | oh :-( | ||
| tadzik | :( | 15:00 | |
| brrt | actually, train trip | ||
| dunno why i sad road trip | |||
| anyway, can you help me with something? i still need to represent the 3d array efficiently | 15:02 | ||
| oh, and another problem; i need to prune rulesets, and earlier = better | |||
| for pruning, i currently have no good solution other than the idea that it'll have something to do with topological sort | 15:06 | ||
| (early pruning, that is) | |||
| for the 3d array, i was thinking of having a linear array of 3 integers and using binary search | 15:07 | ||
| that's not the most space efficient way to do it, but it'll be very easy to implement | 15:10 | ||
| pruning (late) reduces 17003 rules to 1202 | 15:19 | ||
| timotimo | i put the "run" in "prune" | 15:50 | |
| [Coke] | You can't spell prune without a pun. | 15:56 | |
|
16:02
rurban joined
17:58
zakharyas joined
|
|||
| jnthn | .oO( I put in the prune and got runs... ) |
18:32 | |
|
18:42
FROGGS joined
19:36
zakharyas joined
19:49
brrt joined
|
|||
| brrt | hmm... i think i have a strategy for 'early' pruning | 19:51 | |
| what's the difference? oh, why not, i'll explain | |||
| (what's the use? currently the table construction algorithm is O(m*n^2), n = number of rules, m = number of IR tree node types | 19:52 | ||
| actually, that's wrong | |||
| n = number of rule*sets*, which are combinations-of-rules; given any tree node type (head) with i matching rules, we have 2^i combinations | 19:54 | ||
| but many of those 2^i combinations are 'invalid', i.e. are never generated | |||
| now, a simple way to figure out which sets *are* generated is by generating the entire table and picking all rulesets from those | 19:57 | ||
| but of course, by then i've already done my million-or-so iterations | |||
| so what is needed is a way to figure out all the rulesets generated by the grammar without generating the entire ruleset-table | 19:58 | ||
| first observation: the tiler table maps heads with specific children-states to new rulesets-states | 20:00 | ||
| hence, only rulesets with the same head can combine | |||
| hmm | 20:12 | ||
| dalek | arVM/even-moar-jit: e165986 | brrt++ | tools/tiler-table-generator.pl: Add rule pruning, making the table much smaller |
20:32 | |
| arVM/even-moar-jit: 499d4c2 | timotimo++ | / (12 files): add special support for nativecallinvoke to profiler |
20:44 | ||
| MoarVM/even-moar-jit: b0e6608 | timotimo++ | src/profiler/profile.c: | |||
| MoarVM/even-moar-jit: fix output of profiler wrt. native function names | |||
|
20:45
dalek joined
21:53
colomon_ joined
21:58
colomon joined
|
|||