my (kinda) fast compiler back

(L) [2006/04/18] [fpsunflower] [my (kinda) fast compiler] Wayback!

Here's my O(n log n) compiler - it does a linear sort upfront, then builds recursively as usual, with 3 linear passes over the splits each time.


Keep in mind that this is java, so most of this is obfuscated by the lack of unions. ints are 32 bits and longs are 64 bits. >>> is an unsigned right shift.


First some methods to pack/unpack our splits:

(L) [2006/04/18] [Phantom] [my (kinda) fast compiler] Wayback!

Great to see more code here. Now people have an alternative for my bad code. [SMILEY Smile]

As soon as I have a decent NlogN compiler I'll post it on the forum too.
_________________
--------------------------------------------------------------

Whatever

(L) [2006/04/28] [Phantom] [my (kinda) fast compiler] Wayback!

Sunflower: What did you set EMPTY_BONUS to?
_________________
--------------------------------------------------------------

Whatever

(L) [2006/04/28] [tbp] [my (kinda) fast compiler] Wayback!

Note: i've tried varied bonus/malus to straddling or planar population, space cut solutions etc etc... but never got a better tree out of that.

Perhaps i haven't tried varied enough scenes or something. Still it surprised a me.

(L) [2006/04/28] [Phantom] [my (kinda) fast compiler] Wayback!

My compiler does not produce worse trees if I omit empty space bonusses completely.
_________________
--------------------------------------------------------------

Whatever