bad news -- memory profiling

[Patrick Nichols <pnichols@graphics.csail.mit.edu>]

Hi Vitaly,

I tried running my server on the raw triangles you've given me, and it 
doesn't look like individual triangles will work as java objects. There 
are about 80,000 triangles in the basemap, and when I parse and load them 
I get out of memory errors after about 60,000 of them get processed. So, 
it looks like we'll have to substantially reduce the number of triangles 
in the basemap and merge them into spaces. 

The good news is that there are about 10,000 rooms on campus, so if we 
consider an upper reasonable limit to be, say, 50,000 total spaces 
(reserving 40K for the outside world), all you have to do is merge 
triangles to their neighbors. Convexity won't be an issue, given that the 
sum of any two triangles is still convex, so maybe you could just hand me 
a bunch of quads.

I have some other thoughts about the space names in your output; I'll 
think a bit more carefully and send something your way soon. For now, I'm 
taking a look at my data structures to see if I can tune them to use less 
memory.

-Patrick

-- 
* * *

Patrick James Nichols II
Graduate Student, MIT Computer Graphics Group
http://graphics.csail.mit.edu/~pnichols