A graphical computer system must give the user an opportunity to look
at an object from different angles. For that, we must be able, given t
wo points P1 and P2 in space and an angle PHI, to rotate an object by
the angle PHI about the line P1P2. This is usually done by rotating ve
rtices (or other points that represent an object) and connecting them.
Since an image can have lots of vertices, it is important to be able
to rotate each of them quickly. Therefore, we are interested in a rota
tion method that would consist of the smallest possible number of comp
utational steps. The usual method (see, eg., [1]) includes four nonari
thmetic operations (namely, computing sin PHI, cos PHI, and two square
roots), and seven multiplications of 4 x 4 matrices. Nonarithmetic op
erations are necessary because the known expression for a rotated poin
t includes sin PHI, cos PHI, and a square root. However, we can try to
reduce the number of arithmetic operations (+, -, X, :).In this paper
, we propose an algorithm that consists of only 30 arithmetic steps: i
t is less than 1/10 of the number of steps used in the traditional met
hod.