26 Apr
2016
26 Apr
'16
8:59 p.m.
Seriously,
http://tinyurl.com/j46jg4p
http://www.computerhistory.org/collections/catalog/102743645
What in the world is this thing? Some kind of early parallel or vector
processor? External? How delightfully weird, except there is very little
online about it. Did any of you work with one of these? What kind of
instructions do they use? What is their bus width on the ...uh what? CPUs?
What kind of addressing modes & size(s) do these use? Why does the one
reference talk about the 64k RAM they had? Were these standalone machines
('cause it looks like it goes to a VAX-11 and "LSI-11") ?
This was before my time but fascinating, nonetheless.
-Swift
26 Apr
26 Apr
9:09 p.m.
On Tue, Apr 26, 2016 at 5:59 PM, Swift Griggs <swiftgriggs at gmail.com> wrote:
Seriously,
http://tinyurl.com/j46jg4p
http://www.computerhistory.org/collections/catalog/102743645
What in the world is this thing? Some kind of early parallel or vector
processor? External? How delightfully weird, except there is very little
online about it.
Try searching for ' cspi mini-map vector processor'
-- Charles
9:17 p.m.
On Tue, 26 Apr 2016, Charles Anthony wrote:
Try searching for ' cspi mini-map vector
processor'
Ah yes. The marketdroid blurb sums it up:
"Today, CSPI is recognized as a leading manufacturer of vector processors
which enhance a computer's ability to perform high-speed arithmetic.
Markets served include science and engineering applications in signal
processing, sonar and radar processing, image processing, seismic
processing and well logging, medical imaging (such at CT and MRI scanning,
and others."
-Swift
9:44 p.m.
On 04/26/2016 07:59 PM, Swift Griggs wrote:
Seriously,
http://tinyurl.com/j46jg4p
http://www.computerhistory.org/collections/catalog/102743645
What in the world is this thing? Some kind of early parallel or vector
processor?
We got a CSPI 6410 and had it hooked to a VAX 11/780. It
was probably a bigger brother to the machine you
link to. There was a big math library that came with it.
If you had really regular matrix operations, such as FFTs,
matrix multiplies and similar classic operations, it could
do them quite fast, in the several MFLOP range. The bigger
the matrix (as long as it fit in the memory of the unit) the
better, as the library just set up all the registers, loaded
the data and turned it loose. If you had a bunch of small
matrices, it was a lot less efficient, as it had the same
setup overhead for every task.
Yes, it is a vector processor, with a floating-point
multiplier and adder, some address arithmetic logic and a
sequencer.
Jon
2974
days inactive
2974
days old
3 comments
3 participants
participants (3)
-
charles.unix.pro＠gmail.com -
elson＠pico-systems.com -
swiftgriggs＠gmail.com