15 Nov
2011
15 Nov
'11
7:14 p.m.
On 15/11/11 5:12 PM, Eric Smith wrote:
Ethan Dicks wrote:
In that timeframe I fixed some bugs in Levco's Mandelbrot demo for their
Translink Transputer board for Macintosh NuBus, which on two T800's was
handily faster than the host Mac (a IIcx, probably). In fact, a single
T800 was quite a bit faster even for integer work.
I think I still have the code lying around, and the board.
--T
Build the worlds slowest Beowulf Cluster?
Sridhar Ayengar wrote:
I have one that probably is a contender for that
title.
Two machines each with 68010 @10MHz.
Not even close.
Possibly Ciarcia's multiprocessor 8751 system for Mandelbrot set
computations, in the October/November/December 1988 Byte magazines. It
was awesomely slow.
In that timeframe I was using a single Motorola DSP56001 to do
Mandelbrot set computations, and it was *much* faster than an array of
8051s. Somewhere I still have a printed listing of the assembly code.
The loop for a single-precision iteration was five instructions. The
double-precision loop was a lot longer. Originally I did this with a 386
PC, but later hooked it up to a Macintosh, and Dave Platt added plugin
support to MandelZot so I could use it without writing an entire custom
Macintosh application.
I see that Tristan did used the DSP56001 with an Amiga
host for
Mandelbrot set computations around 1990:
http://lorachnroll.blogspot.com/2010/03/old-hand-made-co-processor-for-my-a…