Why are you using floats rather than doubles? Try multiplying by 4, adding 0.125, using the floor(double) method, and dividing by 4 again. You may, however, introduce imprecision into your results like that. Note you will end up with a double by that technique.
Joined: Jul 03, 2009
I end up with something like this:
I am using float, because these are really simple numbers: the larger number used in the program could be 15! ;)
Multiply by 4, floor (not round), divide by 4 again. You'll get (original, *4, rounded, /4):
As for the imprecision that Campbell was talking about, I don't think that will happen since you're only working with multiples of 2. These can be represented 100% accurately in doubles (as long as neither the integer part becomes too large or the decimal part becomes too small).
You should still not use floats, even for small numbers. Where are you getting 0.13 from? That will not give the right answer at all.
You will get imprecision for numbers like 2.3, but presumably, as Rob suggests, the “floor”ing will get rid of that.
I think I should have said add 0.5, not 0.125. Sorry for that mistake.
I never said that flooring would get rid of precision problems (although it probably will). What I said is that because the decimal parts are all of a form of a*1/2 + b*1/4 + c*1/8, these can be represented in float / double with accurate precision, as long as the integer part doesn't cause it to be cut off.
As for float, because these numbers are so small, there will also be no problem with precision loss.
Joined: Oct 13, 2005
I think we actually agree, Rob. Only I am not expressing myself clearly.