iOrange

Well yeah I agree, the "pre-BC7" times are a bit weird - the spec didn't have any hard requirements for a bit-precise decoding, that 's why different implementations (even hardware...

Okay, I guess I see your point now. Our difference is rounding convention. I was having a "C-lang float to uint8" in mind, where it just drops any fractional part....

Yeah, so the 565 -> 888 conversion is just my carry-over code from my other (older) projects. I don't remember exactly where did I find that magic, I just know...

@richgel999 didn't expect an expert of your level to appear on my humble home project, truly an honour for me. And yes, I'm still on the fence on this issue....