Text archives Help

---

I just committed an updated sphere intersection code that minimizes the surface acne bug roni reported (for spheres). It's turned off by default (need to change a #if to 1 if you want to try it) because it's not yet in sse and it causes a slight slow down which perhaps people don't want. What I did is to essentially cast a new ray that's closer to the hit point. This reduces the need for increasing epsilon. This iteration style fix would likely need to be done for other "complicated" primitives, such as cone and torus. Triangles, planes, and other simpler primitives don't experience as much numerical precision issues, and so they seem to work with the smaller epsilon.

I tried raising the T_EPSILON value, but that caused shadows to go missing. So just playing with T_EPSILON won't do (unless you are happy living with very small scenes). If you're ok with a small slowdown in exchange for greater robustness, then let me know and I'll create SSE and specialized versions of the more exact intersection test. Even better, could RTSL do it instead of me???

Thiago

Roni Choudhury wrote:

pinhole( -eye 14.7525 15.0702 8.86065 -lookat -3.9814e-08 -1.56811e-08 0.3 -up -0.00433055 -0.00206167 0.999988 -hfov 60 -vfov 60 -normalizeRays )

That's for a similar one (not exactly the same one I sent).

roni

James Bigler wrote:

Could you press 'c' and show us the camera parameters?

James

Roni Choudhury wrote:

I get the attached image by zooming out from the default scene.

Even without zooming out, strange speckles are apparent on the edges of the sphere.

roni

James Bigler wrote:

This is to let folks know that T_EPSILON was recently changed from 1e-3 to 1e-5. This might affect how some scenes behave. If you find problems (missed intersections, spotty shadows) please report them to the list.

Thanks,
James

---

• [Manta] T_EPSILON changed from 1e-3 to 1e-5, James Bigler, 02/07/2008
  • Re: [Manta] T_EPSILON changed from 1e-3 to 1e-5, Roni Choudhury, 02/09/2008
    • Re: [Manta] T_EPSILON changed from 1e-3 to 1e-5, Steven G. Parker, 02/09/2008
    • Re: [Manta] T_EPSILON changed from 1e-3 to 1e-5, James Bigler, 02/09/2008
      • Re: [Manta] T_EPSILON changed from 1e-3 to 1e-5, Roni Choudhury, 02/09/2008
        • Re: [Manta] T_EPSILON changed from 1e-3 to 1e-5, Thiago Ize, 02/11/2008

---

Archive powered by MHonArc 2.6.16.