I think that if you would like to optimize this code that much in C or assembler, you wouldn't convert the decimal strings to binary numbers at all.

That being said, I do not attack the original code, it is a nifty good idea.