. /../Lino Ray Tracer/ 1234567
written by Cryoburner on Jun 26, 2006 12:57
Oh good! That sounds fast.

I imagine a similar feature would be good for the realtime raytracer as well. : )
deep into the jungle
written by Ponche on Jun 26, 2006 13:07
Cryoburner said:
Oh good! That sounds fast.

I imagine a similar feature would be good for the realtime raytracer as well. : )
Yes, cause the most time is spent in preprocess.
written by Cryoburner on Jul 14, 2006 16:48
Ponche said:
I thought that some acceleration is needed! So i've implemented "uniform spatial subdivision". Improvement? The sphereflake at full detail and 600x600 took just 18 seconds! And one or two weeks ago i waited near 3 hours for it to render!! I'll be uploading it soon.
So, any progress on this?

I was thinking of creating a scene requiring an inhuman number of spheres and cylinders, so I thought it might be convenient to try this out. : )
deep into the jungle
written by Ponche on Jul 17, 2006 19:57
Yes, sorry!! Here it is. Try executing the example compiled application and check the speed up.

cd/zips/Ponche/lrt13 (56 Kb)
written by Cryoburner on Jul 18, 2006 04:50
Yes, that's very fast, although I never actually attempted to fully render sphereflake using the prior version. With the new version, it takes around 10 seconds to render at 400x400 on my Athlon64 3500+. At 800x800, with the camera dpp increased to 1024 accordingly, the same scene takes around 30 seconds to render. That's quite an optimization. : D

I also noticed that you increased the number of iterations for the 'sphereflake', so there's even more detail in this version. How many spheres are in that scene, anyway? : )
deep into the jungle
written by Ponche on Jul 18, 2006 10:20
Well, if i'm not wrong there should be exactly 7381 spheres and just one plane.
reading this thread
no members are reading this thread
. /../Lino Ray Tracer/ 1234567
20167, 15 queries, 0.061 s.this frame is part of the AnyNowhere network