Yeah, well, you just can't do it that way and be fast at the same time. I mean seriously fast. What you can do is decide how your comparison should be collated

Yeah, well, you just can't do it that way and be fast at the same time. I mean seriously fast. What you can do is decide how your comparison should be collated and preprocess your strings into a sort key form that is strictly big-endian binary (first byte has highest weight). Or little-endian I suppose, whichever works best for your hardware. Your sort key doesn't have to be the actual string.