That kwazy guy, Henry Minsky writes, in part:
> Since I couldn't find a free-software fixed-point math library for Java,
> I wrote a simple one for the IAppli/CLDC platform.
>
> http://www.ai.mit.edu/people/hqm/imode/fplib/index.html
>
> .... A real
> numerical analyst would probably have a coronary looking at this
> implementation.
>.....
An EE friend of mine on a circuit simulation project was trying
to improve the performance of one of the algorithms. In this task,
she was dogged by her father's little guilt-trip: a physicist
working in CFD (Computational Fluid Dynamics; don't ask),
he had sniffed that any *real* numerical analyst in his field wouldn't
stoop to using double-precision floating point, since you could
do everything you needed with single precision, with a little
thought.
Upon hearing this story, our start-up's CTO exclaimed, "that's just
like what von Neumann said, only he said that anyone worth
his salt didn't even need floating point - you can do everything
in fixed-point arithmetic. With a little thought."
Of course, "a little thought" for John von Neumann was, like,
more CPU cycles than the rest of our little brains ever
crank out through their entire product lifespans. MY brain,
anyway.
The von Neumann story may be apocryphal. A five-minute
web-search turned up no confirmation.
> There are probably lots of errors. Don't use this code for anything
important
> without testing it.
Too late, Henry, I'm writing this in my hang-glider, diving
through fog, and my navigation system is running entirely
in a midlet. It's great! It leaves so much more time for
thumbty
SSH/TELNET CLIENT TIMEOUT - CONNECTION CLOSED.
[ Did you check the archives? http://www.appelsiini.net/keitai-l/ ]
Received on Wed Apr 4 04:57:50 2001