OIDN animations

[epilectrolytics]

First attempt, 500 samples PathOCL, 6min/frame.

New attempt, 400 samples PGI+OIDN, 1.4min/frame.
There are some artefacts from slowing down 50% and blowing up to 3k resolution with DaVinci Resolve.

[lacilaci]

I guess this is without photonGI?

[epilectrolytics]

No, this was rendered with persistent PGI cache, bringing down render time from 6min to 1min 24s so that I could do a longer animation.

[lacilaci]

No, this was rendered with persistent PGI cache, bringing down render time from 6min to 1min 24s so that I could do a longer animation.

Excuse me what?! I did not expect pgi to be so stable

[B.Y.O.B.]

It's computed once and then only re-used. Only works if just the camera is moving.

[lacilaci]

I see, btw I was today rendering some interior previews that had rendertime faster than preprocessing. We need faster render startup.

[Sharlybg]

We need faster render startup.

opencl compilation is the killer in this case.

[FarbigeWelt]

If only camera is moving, why compiling openCL kernel each time? Why is kernel compilation not optimized as for view port render? It could save so much time for animations!

[B.Y.O.B.]

If only camera is moving, why compiling openCL kernel each time?

I'm pretty sure the kernels are cached and re-used if only the camera is moving.

[FarbigeWelt]

If only camera is moving, why compiling openCL kernel each time?

I'm pretty sure the kernels are cached and re-used if only the camera is moving.

I‘ll check it this evening.