11-17-2019, 01:35 PM
sRGB only tells how to interpret the RGB coding in bytes of the JPEG/PNG. When Gimp converts to floating point, the data is usually put in "Linear light", so the Gamma-correction is removed. It is of course re-instated when you export to PNG/JPEG.
Note that internally, Gimp is always in FP-linear, the "precision" is only for storage (32-bit FP uses 4 times more RAM than 8-bit). This also means that unless you are in 32-bit FP there will be additional conversions, so 32-bit FP is normally faster than the other precisions.
Note that internally, Gimp is always in FP-linear, the "precision" is only for storage (32-bit FP uses 4 times more RAM than 8-bit). This also means that unless you are in 32-bit FP there will be additional conversions, so 32-bit FP is normally faster than the other precisions.