How can I optimize vector normalization in a Molecular Modeling program?

[Andomar]

The optimizations I can think of:

VTYPE length;
length = v[0]*v[0] + v[1]*v[1] + v[2]*v[2];
if( length < 1.00001 && length > 0.99999 )
  return true;  // vector is already normal.
length = sqrt( length );
v[0] /= length;
v[1] /= length;
v[2] /= lenght;
return true;

The check for already-normalness will speed up 
things if the program tries to normalize vectors
more than once.  (You can't just temp == 0.0 
because floats aren't exact.)