PhotonGI cache

[lacilaci]

A test with lacilaci's scene.

hoooly sh!t

also note how more of sunlight is the cache giving to the scene compared to non-cached version.

[epilectrolytics]

A test with lacilaci's scene.

Well, well, well!

[B.Y.O.B.]

500 samples Bidir+Metropolis (pathdepth 16/16) for comparison.

[Racleborg]

Wow!

[Dade]

Side note: some of the Lacilaci's scene setting is a bit extreme at the moment; for instance, I have yet to implement Russian Roulette on photon paths, this mean the pre-processing code wastes a lot of time tracing photon paths carrying nearly zero light. RR greatly speedup this kind of stuff. Also using matte, instead of glossy, for the walls can speedup the rendering too.

[lacilaci]

Side note: some of the Lacilaci's scene setting is a bit extreme at the moment; for instance, I have yet to implement Russian Roulette on photon paths, this mean the pre-processing code wastes a lot of time tracing photon paths carrying nearly zero light. RR greatly speedup this kind of stuff. Also using matte, instead of glossy, for the walls can speedup the rendering too.

Speedup would be welcome.
Although using matte for walls might be faster, glossy with bump and mapped roughness would be more realistic.
However!
Maybe a rayswitch for GI rays could be a solution.
Maybe photonGI could internally calculate GI as if all materials are matte, but render as usual?

[Dade]

Maybe photonGI could internally calculate GI as if all materials are matte, but render as usual?

It is possible but it is a LOT of bias. I'm a bit more comfortable with the idea of leaving the option to choose (between glossy walls vs matte walls) to the end user.
Note: I was talking of pre-processing speed, for the rendering speed is all pre-computed so it doesn't matter.

[lacilaci]

Maybe photonGI could internally calculate GI as if all materials are matte, but render as usual?

It is possible but it is a LOT of bias. I'm a bit more comfortable with the idea of leaving the option to choose (between glossy walls vs matte walls) to the end user.
Note: I was talking of pre-processing speed, for the rendering speed is all pre-computed so it doesn't matter.

I understand. I think what I requested could be achieved in future with a rayswitch material.

[lacilaci]

500 samples Bidir+Metropolis (pathdepth 16/16) for comparison.

bidir is a little brighter though, higher bounce amount?

[Dade]

500 samples Bidir+Metropolis (pathdepth 16/16) for comparison.

bidir is a little brighter though, higher bounce amount?

BiDir max.depth 16/16 is (about) like path with 32.