I did a litte bit of research and it kept me wondering.. how about using spheres for voxels representation this time? Lets try it out and see how much the GPU can take.
Grid of 10^3 Spheres – there is no acceleration architecture implemented yet – Octree for spheres? that feels odd..
Oh, btw. That’s raytraced.. so far quite interactive.
It uses OpenCL on Radeon 5700 HD – only primary rays so far on ambient material. Give it some time.
Simple directional light lighting system implemented.
Secondary bounces in place (depth of bounces set to two) – little bit of hackery had to be done since there is problem with recursion in OpenCL.
Every third sphere is ambient.
Presentation done, no prize for the best poster, oh well..
Two more weeks.. 300 FPS when moves - Insane!
Small memory leak was driving me crazy; Debugging process was taking ages.. bored, frustrated and tired decided to play about with octrees depth and minimal/maximal number of elements in leaf.. Amazed by 300FPS discovered the stupid mistake of creating VBO each time the leaf is regenerated. Instant win; fix of memory leak and 10 times greater performance.
Who said that the mistakes are stupid?
1233560 cubes visible avg. FPS: 144.
Simple optimisation great performance boost! – Skip ALL the non-visible faces!
Progressive Grid of 256x256x256
Apparently I couldn’t resist. Last optimisation implemented.
Now seriously, report!
Perlin Noise in action
Since week all development stopped and now I am struggling with report. Just finished draft of first chapter.. it’s terrible.
Step forward :)
Last two nights were all about Octree, however this morning last bug has been hunted down.
Now I need to write a structure of report, good luck me
Oh and some metrics: