Automatic shadow biasing

[JMS55]

What problem does this solve or what need does it fill?

Shadow acne is an artifact that occurs when sampling the shadow map. The shadow map surface you intend to sample was discretized into the shadow map, which might not perfectly match the sample position.

The solution is to bias the sampling position away from the surface a little.

What solution would you like?

An automatic algorithm to compute a good shadowing bias.

State of the art is "Oriented Depth Bias" from FFXVI http://www.jp.square-enix.com/tech/library/pdf/2023_FFXVIShadowTechPaper.pdf#page=28.

What alternative(s) have you considered?

A constant shadow bias is already supported by our lighting system. It's hard to find a good value for this that doesn't result in objects appearing to disconnect from their shadows.

Additional context

An older method called "Receiver Plane Depth Bias" was tried in #10188.

[superdump]

It is not an alternative to receiver plane depth bias but rather an alternative to the manual depth and normal biases that we have today.

Receiver plane depth bias is about adjusting the depth bias being applied when comparing a shadow map sample that is offset from the screen fragment’s position in the shadow map for PCF. It or something like it is still needed even with oriented depth biasing I believe.