The result we are seeing for sin pi (where pi=1p1) is 1.224606353822377e-16
which is the result produced by Intel's FSIN function, not from a software library. FSIN reduces its argument using a 66 bit approximation Pi to pi. Thus sin pi=sin(Pi-pi), and Pi-pi is not zero. If a prior library used a 64 bit approximation Pi to pi and did the same reduction, sin pi=sin(Pi-pi)=sin 0=0. Incidentally, FSIN returns its argument and sets an error bit if the argument is larger than 2^63. If software does not check this, it thinks sin(x)=x for large values of x. John ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
