Only thing I would like to see change still is bumping the minimum OpenPGL version.
After that this is ok to go into main for 4.0 (in two days from now).
@brecht bumping the version is…
It is mainly there to be able to validate the label, roughness and eta estimation. this will be useful when the BSDF models change (e.g., after Principle_v2).
Do you have a better idea or…
This is due to the fact that Blender/Cycles uses roughness^2 internally and squared roughness returns alpha.x * alpha.y so it is actually (roughness^2)^2 compared to the user input.
I…
I agree, I changed it to test in both cases for the ris_weight as well
@Sergey I switched to the safe checks and polished the code a little bet. It is ready for your final review now ;)
Done,
The other define # define RIS_INCOMING_RADIANCE will be gone after we bumped the Open PGL to 0.5.0
It is a bit hard to see the exact difference when images are put on the same canvas. It would be helpful to have them as individual images to help doing pixel-wise comparison.
For the…
I already talked with Ray. As soon as every dependency package updated to Open PGL 0.5 I would like to bump the minimal required version t 0.5 and get rid of all these defines.
This is an extension/variant of the standard RIS where I added some stuff. A detailed description of the method with all the insides will be part of the Siggraph2023 Course on path guiding, I can…