Bugs in NEW 2.4!

[Numsgil]

Yeah, I have a profiler, so I can find what calls take what amount of time, etc.  I didn't like VB watch much, so I went turbo powered and use Intel's VTune.

[Numsgil]

Just to beat a dead horse:

The simple, unoptimized algorithm is this:

Code: [Select]

For x = 1 to MaxRobs
  For y = 1 to Maxrobs
    check if bot x can see bot y.
Which is clearly O(n^2).

Computer Science is my minor.  I've taken everything but the algorithms course.  But don't think I don't know a thing or two about algorithms.  Especially in this regard, since I've put in 100s of hours to streamlining this vision process.

Using kd-trees, you could make this into O(log n).  Which is what I'm sorta looking into.  Unfortunately, kd-trees must be balanced to operate correctly, and rebalancing a kd-tree can take O(n log n) time.

If such is the case, other methods might prove better, such as fixed size space partitioning Carlo suggested a while ago.  Which is what I ws investingating in 2.4.  You can still see the code I commented out.  It's quite streamlined, and highly messy.  But fast.  You must admit it's fast.

Don't lecture me on theoretical computer science.  I am all on top of that.  Rather, it's the limitations and quirks of VB that are throwing me for loops.

[Botsareus]

Sounds to me like Greven is from cosmicrift...

[Endy]

Could someone please post the new sysvars flie up? It doesn't seem to be installing correctly for some reason...

Everything else is there, but the sysvars aren't making a showing.

[PurpleYouko]

I wonder if we could use some sort of external database for Bot X,Y coordinates then send a simple query to it that returns only those bots within a certain range.
Could work real well for collisions and vision.

I know some database queries can be slow though so it might not really help much.

[Numsgil]

Could someone please post the new sysvars flie up? It doesn't seem to be installing correctly for some reason...

Everything else is there, but the sysvars aren't making a showing.

No new sysvars.  Just use your old sysvars file.  I don't think there are anyway...

[Numsgil]

I wonder if we could use some sort of external database for Bot X,Y coordinates then send a simple query to it that returns only those bots within a certain range.
Could work real well for collisions and vision.

I know some database queries can be slow though so it might not really help much.

But you have to ask how the database is going to do that query search.  At some point, you have to deal with the physical task of finding the neighbors within a certain range.

[PurpleYouko]

Yes I know that. I guess it just depends which program is able to perform the query faster. To be quite honest I don't really know how fast a database can actually do it.

[Numsgil]

Back to Greven's initial problem:

Turns out that Translocation mutations are broken.  That was alot of the slowdown and problem with bots dying.  Each time a bot reproed it's kid was having like 300 mutations.  No wonder the bot was dying!

The only mutations I'm positive are working right are Point Mutations, Copy Error Mutations, Delta Mutations, and Insertions.  Even then there might be some minor errors.  But I think those are stable.

[Endy]

Turns out that Translocation mutations are broken. That was alot of the slowdown and problem with bots dying. Each time a bot reproed it's kid was having like 300 mutations. No wonder the bot was dying!

Yep, this was what I was seeing. Tons at once.

Had a major slowdown running Din and One in F1 mode. Not sure of where the problem lies...

On a positive note I saw that walls appeared to be working perfectly. I like how they don't appear to be oddball bots anymore, slowing down the program. I really like seeing different objects in the bots world, I think walls are a great start.

[Numsgil]

Hmm... I don't think I ever touched the walls code...

Now I'm going to have to give it a look see.

[Welwordion]

just tested 2.4 also was shocked when the bots diedlike flies ^^

[shvarz]

It's a beta.  Just to give you a taste :)

[Numsgil]

Try turning off mutations in the mutations panel.  That should stop them from dying.  If it doesn't, then it looks like I have some real bugs to hunt down.

[Endy]

Din seems to be relativly stable still there are some oddities with Tf but otherwise it's okay. Um... The One absolutly sucks now. It seems to move forward in a single direction doing nothing... I have no idea where the problem lies. I have trouble figuring out my own, let alone something like that. :)