[Home]David Alves/Firebird

Robo Home | David Alves | Changes | Preferences | AllPages

Showing revision 24
Difference (from revision 24 to revision 24) (minor diff, author diff)
(The revisions are identical or unavailable.)
Firebird is based off of the Phoenix code, but uses DynamicClustering guns for both WaveSurfing and firing at the enemy. All other code is the same. Even the WaveSurfing class is the same, it just has a different value for the "enemy gun" variable. Coming soon to a rumble near you! --David Alves

Found out something interesting while testing... Firebird spends about a third the time of Phoenix in the movement method (~1150 vs ~3400 ms per round on my laptop), but spends almost ten times as long in the aiming method (~650 vs. ~65 ms per round on my laptop). Need to investigate why that's the case, I would expect it to be uniformly slower or faster... And yes, that does mean that Firebird is significantly faster overall. At least, it is until it turns on its flattener... --David Alves

Neat! Perhaps the movement / timing discrepancy is because a VisitCountStats system will access and sum all your buffers no matter how much data you actually have, while a DynamicClustering system crunches far fewer numbers if you only have a few pieces of data. And if it dodges bullets anywhere near as well as Phoenix, it probably doesn't have too many to deal with. =) Now to rederive the Lukious movement and we'll have a fully-DC-bot party going on in the top 10! -- Voidious

Awww ... I want to party, too :( -- Simonton

Oh come on, you'll probably crack the top 10 before I do with Lukious. =) -- Voidious

What's that you say? You want to have a TopTenDCParty? =) --David Alves

Actually, I just came to the sad realization that the first two bots to be knocked out of the top 10 would also be DC bots... -- Voidious

Adding 0.1 to the rumble now. Not expecting much, but it should hopefully be above 2000. Let the race to the top ten commence! --David Alves

I think it's pretty funny that this is so close to the rating of Lukious. And the Lukious gun rewrite left me with almost the exact same rating as the old gun. Are these just coincidence, or is there something specific in DC surfing that is holding us back? Hmm... -- Voidious

I dunno, I'll run a movement challenge. Maybe that will shed some light on things. --David Alves

0.2 is up now. Flattener disabled for this release. Surfing should be substantially improved. --David Alves

Just wondering, are you using Simonton's tree? Firebird seems pretty slow... -- Skilgannon

Yeah I am, but I do a lot of other processing. I'll see if it can be sped up in future versions. --David Alves

Yeah I did, but the tree isn't the problem. It's all the black magic I work on my scans after I extract them that's the problem. =) --David Alves

Incidentally, if anyone wants to look at a copy of my tree, I'd be willing to post it. Also, it is a well-known fact that "black magic" takes O(n^3 log n) time. --David Alves

Please don't read too much into the O(n^3 log n)... I made that number up as a joke. But yeah, n would be the number of scans retrieved. It's not the tree that's slow, and I'll see if I can speed it up. =) --David Alves

Funny that, it put me onto an interesting paper [MultiwayCuts]. -- Skilgannon


Robo Home | David Alves | Changes | Preferences | AllPages
Edit revision 24 of this page | View other revisions | View current revision
Edited November 7, 2007 8:09 EST by Skilgannon (diff)
Search: