Fortran - Re: sin/sind

You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

Author

Re: sin/sind

glen herrmannsfeldt

2004-06-08, 3:58 pm

Kevin G. Rhoads wrote:

(snip

> Also note that using a sin2pi(x) does not eliminate the problem, it
> merely alleviates it. As x gets large, fractional(x) has less precision
> than x. This is just plain unavoidable. However, sin2pi(x) allows
> argument decimation by trivialities rather than requiring arcane
> manipulation to accurately decimate the input argument w/o humongous
> look up tables. Alleviation, not elimination.

Well, it eliminates the uncertainty in the argument reduction
due to the use of an irrational number. If one does argument
reduction with an approximate pi, as at least it was traditionally
done, there is a systematic error in addition to the
fraction bit loss problem.

-- glen

Sponsored Links