5 replies to this topic

[Sol_HSA]

I did a little research on different methods of rendering lots of cubes in OpenGL. Here's the writeup; http://iki.fi/sol/cubes.html

[Reedbeta]

Interesting. I would be curious to see how the results break down when rendering more complicated models (several hundred to several thousand polys) rather than cubes. I imagine it would change the tradeoffs around a bit but that the overall ranking of techniques from slow to fast would be the same. Storing all the matrices in a texture was an idea I wouldn't have thought of.

[rouncer]

Wavesurfing renders tonnes of cubes quick. look it up.

[Sol_HSA]

rouncer said:

Wavesurfing renders tonnes of cubes quick. look it up.

I wasn't trying to optimize the effect. I was trying to optimize the problem case. The objects themselves don't really matter either, I just happened to use a cube - I might replace that with a cube with rounded corners, for instance =)

[rouncer]

Only if you want to work in software, will be any worth to you. But ive noticed that it goes so much quicker than raytracing voxel volumes, if you werent restricted by hardware, youd displacement map that way.

I HATE MY VIDEO CARD.


this was made out of tiny little cubes. :)
you cant beat my graphix.

I admit i didnt wavesurf that, but what if its possible? id love to see it.

[Sol_HSA]

Sorry for updating an ancient thread. I revisited the instancing, implemented the new instanced_arrays extension and ran some benchmarks on both nvidia and ati, and also included some higher-poly models.

Results can be found at the bottom of http://iki.fi/sol/cubes.html