I'm nearing the end of development of the first pass of my rigid body physics generator. I lazed about a bit, but mostly, I had problems with concepts of order in 3D space and as I would near the completion, a counter example would pop up and I'd be back to the beginning of the loop.
It's three passes. The first determines the simplest parts: volume and density, but all passes will use generally the same algorithm: loops along x and y, spans along z, but the second and third will need to do a bit more calculation (point mass determination, tensor creation, etc).
Now that I've thoroughly bored ya with the uninteresting leadin, the question is: would it be better to compute the two other passes, or save the span info to file and read it back in for the next two passes?
How many have tried a solution like this and what were the results?
The models will be as large as a carrier, maybe bigger and the spans will be 0.001m x 0.001m x Xm if that will give any idea of the computations being done.
NOTE: I designed this thing so i could easily recompile and use eight byte doubles and use 0.000001m x 0.000001m x Xm ( ).
It's three passes. The first determines the simplest parts: volume and density, but all passes will use generally the same algorithm: loops along x and y, spans along z, but the second and third will need to do a bit more calculation (point mass determination, tensor creation, etc).
Now that I've thoroughly bored ya with the uninteresting leadin, the question is: would it be better to compute the two other passes, or save the span info to file and read it back in for the next two passes?
How many have tried a solution like this and what were the results?
The models will be as large as a carrier, maybe bigger and the spans will be 0.001m x 0.001m x Xm if that will give any idea of the computations being done.
NOTE: I designed this thing so i could easily recompile and use eight byte doubles and use 0.000001m x 0.000001m x Xm ( ).