R
Richard
I raised this 2 years ago but was not clear enough. PowerCalc (which is
part of the Power Toys suite) has a serious bug in the evaluation of
Trig functions such as Sin(x). Specifically it fails to evaluate
Sin(pi/2 + n*pi) for n ~ 30, 56, 104 and 392 at 32, 64, 128 and 512 bit
precisions respectively. The answers of course should be 1.0 but are
reported as numbers much greater than one. Clearly the program is
simply summing the power series for Sin without taking mod(x,2pi)
first. For the last two years I have used this as an example of poor
programming to an undergraduate maths for science that class I teach.
Amusingly the class probably includes some Gates Scholars.
Richard Ansorge.
part of the Power Toys suite) has a serious bug in the evaluation of
Trig functions such as Sin(x). Specifically it fails to evaluate
Sin(pi/2 + n*pi) for n ~ 30, 56, 104 and 392 at 32, 64, 128 and 512 bit
precisions respectively. The answers of course should be 1.0 but are
reported as numbers much greater than one. Clearly the program is
simply summing the power series for Sin without taking mod(x,2pi)
first. For the last two years I have used this as an example of poor
programming to an undergraduate maths for science that class I teach.
Amusingly the class probably includes some Gates Scholars.
Richard Ansorge.