news@pucc-h (Usenet news) (04/25/84)
Call For Papers
SUPERCOMPUTER APPLICATIONS SYMPOSIUM
October 31 - November 1, 1984
West Lafayette, IN
Co-sponsored by:
Purdue University Computing Center
Purdue Center for Parallel and Vector Computing
Control Data
For the past two years, Control Data has co-sponsored an annual
applications symposium at one of its CYBER 205 customer sites. At
Colorado State University in 1982 and at Goddard Space Flight Center
in 1983, scientists, mathematicians, and engineers exchanged ideas on
methods and techniques for applying the CYBER 205 vector processor to
the solution of important computational problems. This year at Purdue
University the symposium will include papers describing parallel
processing methods using the CYBERPLUS multiple parallel processor.
Topics for papers include all aspects, both theoretical and applied,
of vector and parallel processing techniques using the CYBER 205 or
the CYBERPLUS. Theoretical papers should describe advances in general
numerical methods to take advantage of these two architectures.
Applied papers should describe the use of particular features of these
architectures to solve specific problems. All papers should be
addressed to an interdisciplinary audience with enough background
information to explain the topic to a general scientific audience.
Papers will be grouped in four general areas:
* Numerical Methods and Software Development
* Computational Fluid Dynamics and Weather
* Computational Physics and Chemistry
* Engineering and Petroleum Applications
Authors wishing to present a paper should submit five copies of a
two-page abstract to one of the conference organizers by June 29,
1984. Referees will evaluate the abstracts, and the conference
organizers will announce papers accepted for presentation around
August 1, 1984. Authors will be required to provide a full manuscript
at the beginning of the conference for publication in the conference
proceedings.
Conference Organizers
Robert Numrich
Control Data/HQN10V
8100 34th Avenue South
Minneapolis, MN 55440
Telephone: (612) 853-2690
Saul Rosen
Director, Computing Center
Purdue University
West Lafayette, IN 47907
Telephone: (317) 494-1787
ARPA: sxr@purdue
UUCP: ...!pur-ee!pucc-i:aa1ctk@ecsvax.UUCP (04/27/84)
What is a CYBERPLUS? Do they presently exist? Where are they?
jones@ut-ngp.UUCP (05/03/84)
[][][]
The CYBERPLUS is high speed scalar processor made of 15
independent functional units that can accept data every
machine cycle (1 machine cycle = 20 nanoseconds). It is
capable of 650 million instructions per second (this assumes
that you have an application that can use almost all of the
independent functional units at one time). The CYBERPLUS
has a floating point option, made up of three additional
functional units, capable of 65 MFLOPS with 64 bit floating
point numbers and 103 MFLOPS with 32 bit floating point
numbers.
Each functional unit can provide output to any or all of the
other functional units through a crossbar. There is a
register unit to provide high speed temporary storage but it
is not required when using a functional unit (you can do
memory to functional unit to memory operations).
Each CYBERPLUS processor can connect to other CYBERPLUS
processors through a number of ring interfaces. The ring
interface provides immediate transfer for the memory of one
CYBERPLUS processor to the memory of any one or combination
of other processors on the ring. The ring interface also
provides a indirect mode which places the data in a queue
state until the designated processor or processors are ready
to process the data.
Each CYBERPLUS has 256K of 64 bit high-performance data
memory. It is expandable to 512K 64 bit words. The program
instruction memory is made up of 4096 200-bit words. I
believe that the size of the program memory increases with
the floating point option. I can not remember by how much.
1K = 1024
William L. Jones
(The Bear)
{ihnp4,seismo,ctvax}!ut-sally!ut-ngp!jones
jones@utexas-11.arpajones@ut-ngp.UUCP (05/03/84)
[][][] One more inportant thing about CYBERPLUS, it is a trade mark of Control Data Coropration.