“…I’d add a question of my own: what is the precision of the calculations? 16-bit, 32-bit? Integer, float? I think only somebody from the programmers can answer these types of questions. So, @sgospodarenko, could you forward those questions to somebody qualified to give the answer?..”
I posted this question on another forum.
“…So, I’m a programmer, and I’ve written my own software to do all this, but my response is generic, as I’m not a DxO or Adobe programmer. First, floating point representation internally is good, but not especially germane to this discussion. Same with 32-bit integers. It’s the difference between 8-bit and 16-bit that’s most significant to preserving both tone and color gradations…”
I hope, this is what you mean?
He also said:
“…Now, the color spaces really don’t care about such, and contrary to popular assumption, sRGB is not tied to 8-bit integer representation. But, converting early to sRGB is a different kind of clipping, one that discards color gradations in extreme colors, even if done to 16-bit data. If your images don’t have things like extreme blues from theatrical lights (my current nemesis), you’ll probably not notice because the “mainstream” colors will trot from working to output colorspace unmodified. But the cobalt-blue accent lights in my local theater gave me garish posterization when a considered mapping from camera->working color space was not done, even before the further compression into output color…”