Re: SANE on GS

[John_L._Graham@adbbs.antioch.edu (John L. Graham)]

First, a small but important point... SANE is precise, but not necessarily
accurate. The accuracy depends on your algorithm. If I told you Pi was 3.14
and your computer insisted it was 3.214159627, my answer is accurate, but not
very precise, while the other is precise, but not very accurate (a friend of
mine once memorized Pi to 200 places, we're still not sure why).

Second, speaking of the Applesoft ROM routines, while they aren't nearly as
precise as SANE (something like 7 places) they are relatively FAST. Someone
once wrote a program that would call the Applesoft ROM routines from Pascal
(Orca? TML?) and the head-to-head comparison showed about a 10-20 times
faster computation with the Applesoft vs. SANE calls. Soooo, if you could
live with 7 place precision it's like having a built-in FPE! Unfortunately,
the author never posted the code :( I'd love to get my hands on it.

I really like SANE when I'm tinkering with math, but for most applications I
hate it as it renders my code hopelessly slow. We bought an Innovative
Systems FPE, but found it gave unreliable answers and have since removed it.
What endlesly drive me nuts is compilers like AC/BASIC illustrate the GS is
capable of much higher speed math than SANE allowes without an excessive loss
in precision. In fact, I'm hip-deep in my new satellite tracking program
which is computationally the most demanding code I've ever worked on. For
some strange reason Apple failed to recognize high speed math as an important
feature on a computer. I dunno, maybe the integer math toolset is their
answer, but even that is relatively slow. Ultimately they seemed to have
taken the low road and threw a hardware answer at a software problem. Maybe
something's gotten lost in translation, but you'd think that something called
a computer would be designed to efficiently carry out computations *B^/

-John