Nick Shaw

It appears that because when all three channels are negative, `max(rgb)` is negative, the [`ach_shd` calculation](https://github.com/colour-science/aces-vwg-gamut-mapping-2020/blob/6b5d9b313564ad3a7668ffe2ce9bf8f908248990/model/GamutCompress.blink#L206) fails because `(1.0-ach)

When bypassing the shadow roll-off calculation in Blink, the results still don't match. This turns out to be because Nuke's `divide` merge mode is not a simple mathematical divide. It...

It is possible the issue I am seeing is not due to the wrapping, as it seems only to occur with images containing inf pixels.

It does indeed appear that inf pixels in the source result in NaN JMh values, which makes the iteration in the hue lookup fail, (`nan = something` are both false)...

The Daniele version was just a Desmos plot. The min and max were just "hard coded" into equations, rather than being parameters.

If you use the subtraction formulation then `r_hit_min` and `r_hit_max` are the min and max values for `r_hit`, which makes more logical sense.

> Need to verify how much changing r_hit from 896 to 768 affects the values resulting through the system tone scale. It actually affects it far less than you might...

SDR would not just have "least effect". It would be entirely unaffected, because that is only affected by `r_hit_min`.

Additionally, `REACH_GAMUT_TABLE` is probably not the right name, as this one is the cusp, but the table of M values at `limitJmax` gets called `REACH_CUSP_TABLE`.

The `flags` property of a NumPy array shows the storage order. ``` C_CONTIGUOUS : True F_CONTIGUOUS : False OWNDATA : False WRITEABLE : True ALIGNED : True WRITEBACKIFCOPY : False...