I am wondering about the performance behind color-space conversions for textures during reading.
If a project decides to go with a linear workflow, as in, when doing any sort of color manipulation, like lighting, they operate on within the linear color-space, is there any good reason to not store all texture assets in the linear color-space?
My understanding is that when loading textures which are stored in sRGB, the GPU can do the conversion to linear for you very quickly. Is that cost completely negligible? Why not just store all the textures in that form?
The only reason I can think against this is that you may want to store the textures in a color-space that looks like the one you intend to display to (if you're going to display to sRGB then you should store the textures in sRGB). In that case, why not master the texture as sRGB but then store them on disk already converted to linear colorspace?