But note if you go to a linear space, you need to bump up the precision of your numbers. 8 bits per channel doesn’t cut it anymore, you’ll want 12 or 16 bits pe

But note if you go to a linear space, you need to bump up the precision of your numbers. 8 bits per channel doesn’t cut it anymore, you’ll want 12 or 16 bits per channel (or even better, a floating point number). This might have a big effect on performance.