Re: II emulator with math accelerator card emulation?

[pausch@saaf.se (Paul Schlyter)]

In article <5ad50d67.0407131428.47a2e1f3@posting.google.com>,
Rich J. <aiiadict@gmail.com> wrote:
> Are there any Apple II emulators out there
> with FastMath or any other sort of floating
> point/trig able math accelerator emulation?
> 
> even with a fast PC running a II emulator,
> applesoft floating point and trig functions
> take a long time.

I'm not aware of any such emulator.  The only emulator I know of
which emulates some other processor besides the 6502 is Apple II
Oasis, which also emulates a Z80 softcard - but that won't help
speeding up floating-point math.

Another add-on card for the Apple II was an 8088 card which enabled
you to run MS-DOS ver 1 on the Apple II.  However, it also came with
a patched version of Applesoft where floating-point math was some 10
times faster compared to the unpatched version of Applesoft.  If that
card was added to some emulator, the 8088 wouldn't need to be emulated
on a PC but could be run in native mode!  Well, almost: virtual-86
mode (a mode in the 386 and later which emulates the real mode of
an 8086/8088) would perhaps be preferable here -- anyway, the 8088
code would then run very fast.

But if you want fast f.p. math, isn't it grossly counterproductive to
do that in Applesoft on an Apple II emulator?  I mean, any modern PC
has a floating-point capability which exceeds even the supercomputers
of the time when the Apple II was introduced (there was only one
supercomputer back then, the Cray I: it ran at a clock speed of 80
MHz and had a floating-point computation speed approximately
equivalent to a 500 MHz Pentium III).


If you think the floating-point speed of Applesoft is too slow,
rewrite your application as a native PC application !!!!

One application I have which has survived through all these years was
originally written in Fortran-80 for Apple II CP/M.  It computed the
rise/set times for the Sun and Moon at one place during one calendar
year.  On the Z80 Softcard it took one full hour to compute the
rise/set times for each day in one year.  Later I rewrote this
application, first in Pascal and later in C.  The current version
takes only some 0.3 seconds to produce equivalent output, and then
I've added the rise/set times for six planets once a week too!

Interestingly, other pieces of PC software doing the same thing runs
much slower: one piece of software took some 10 seconds to produce
rise/set times for the Sun only during one year - but that software
was never written to run on old hardware with very slow f.p.
arithmetic.


If you want fast f.p. math, use modern computer hardware!  But if
you want to emulate the Apple II, live with the fact that f.p.
math on that computer was horribly slow by modern standards.

-- 
----------------------------------------------------------------
Paul Schlyter,  Grev Turegatan 40,  SE-114 38 Stockholm,  SWEDEN
e-mail:  pausch at stockholm dot bostream dot se
WWW:     http://www.stjarnhimlen.se/
         http://home.tiscali.se/pausch/