izumi@amber.berkeley.edu (12/16/86)
* I am looking for array processors for IBM PC/AT (or PC, XT..). Does anybody make them? Any info, pointers are appreciated. Thanks in advance! Izumi Ohzawa izumi@amber.berkeley.edu .. ucbvax!amber!izumi
davej@entropy.ms.washington.edu (Dave Jenner) (12/19/86)
In article <1949@jade.BERKELEY.EDU>, izumi@amber.berkeley.edu writes: > I am looking for array processors for IBM PC/AT (or PC, XT..). > Does anybody make them? Any info, pointers are appreciated. > Thanks in advance! > Izumi Ohzawa > izumi@amber.berkeley.edu > .. ucbvax!amber!izumi Sky Computers, Inc. (714) 474-1940 (West Coast) (617) 454-6200 (National, MASS) Makes the 320PC 16-bit, fixed point array processor for the PC/XT (and AT) using the TI32010 signal processing chip. There is a (subset) C compiler to produce 320 native code as well as libraries. They also make (or are in the processes of making) the VORTEX-2000 floating point array processor for the AT. Dave Jenner Department of Astronomy, FM-20 University of Washington Seattle, WA 98195 (206) 543-6182 UUCP: ...uw-beaver!uw-entropy!phastvax!davej BITNET: davej@uwaphast INTERNET: davej@entropy.ms.washington.edu
steveb@zaphod.UUCP (Steve Brozosky) (12/20/86)
A company in San Diego named Marinco has an array processor board. It uses the 29116 chip and is running at 8 MIPS. The board works well but is quite expensive (I think around $5,000). Their number is 800)421 4808 in California.
wem@necntc.UUCP (Bill Meshach) (01/01/87)
If you want a "canned" array processor, this is not for you. But, if you want a powerful fixed point processor that you can program yourself, read on. It's a bit unconventional, but there's a very powerful board available from: Data Flow Imaging, Inc. 42 Charles St. Suite 2-C Westwood, N.J. 07675 (201) 666-7970 This board plugs into a PC, XT, or AT. It uses four uPD7281 Data Flow Processors. If you can use 17 bit (16 bit magnitude plus sign) fixed point math, and you can pipeline your algorithms, you should take a look. You can use floating point routines, but the hardware is fixed point. This thing will do a 16 X 16 matrix times another 16 X 16 matrix in about 3.8 milliseconds (17 bit fixed point numbers) -- not bad for under $1,000. The Data Flow Architecture is real neat!
jbn@glacier.STANFORD.EDU (John B. Nagle) (01/01/87)
That data flow engine might make a good graphics rendering unit. I would like to hear from anyone working in that direction, especially if they are interested in solid geometric modelling. John Nagle 415-856-0767