There is also the fact that intermediate operation between floats are done in float precision and not double precision (as would be the case with C). It might b

There is also the fact that intermediate operation between floats are done in float precision and not double precision (as would be the case with C). It might be the behaviour one would expect but it can lead to serious degradations of a numerical computation.