As arrived at in the tonescale thread here’s the values I’ve been using with Daniele’s formulation.
float n_r = 100.0f; // Normalized white in nits (what 1.0 should be)
float g = 1.15f; // Surround/contrast
float c = 0.18f; // Anchor point (mid-gray)
float c_d = 10.1f; // Approx. luminance of anchor point (nits)
float w_g = 0.14f; // Exposure (to make HDR brighter than SDR)
float t1 = 0.041f; // Shadow toe, glare compensation
float r_hit = 128.0f + 768.0f * ( _logf(n/n_r) / _logf(10000.0f/100.0f) ); // Scene-referred value "hitting the roof"
the value of c_d = 10.1 is such that 0.18 comes out at 0.1 (10.0 nits) to a few decimal places.
We could still tweak these more, if we feel a need to clean them up or further justify them. But for now, I’ve been keeping them static here.