Indigo Renderer Forum

skip further subdivisions levels attempts if no quads nor tris where modified during the last level. is this a safe condition ?

Usually when I run into ram issues because of displacements I just roughly estimate right subdivs by checking the displacement map res in megapixel. Basically a 2048x2048 map would have 4.19 MP for example, so it wouldn't show more detail with more than 4.2 million quads. So a 1 mil plane with a subdiv of 1 would be already pretty much there etc. So I think the displacement map res vs. poly count on object is a good condition.

However not so sure what happens after it triangulates the quads. It shouldn't give more detail after that, right?

I simply use the Pixel Threshold to limit the subdivision, its only important to know that the Pixel Threshold is applied on internal resolution, before Super Sampling!

pixel threshold is a must and working magnificently. It is sadly ineffective in spherical mode, the functionality being skipped/ignored.

I can manage the RAM well but as it is with subdivisions conditions being ignored with the spherical camera the poles are looking ugly downstairs. now that's maybe because I didn't test the resulting map in conditions but stared at its renderingninstead. can it be so ?

kklors, I'll ignore you if you talk about texels again to me in the next days good tip of course.

now that's maybe because I didn't test the resulting map in conditions but stared at its renderingninstead. can it be so ?

This seems to be true partially. View-dependednt subdivisions seem to apply with the spherical camera, but the distance ought to be modified by some logarithm I'm guessing. There must be an optimum view-dependent algorithm specific to the spherical camera mode.

On a different note, swapping camera types may render the view-dependent optimisation strictly inoperating in some instances.

I hope I clarified the report somewhat.