MICROELECTRONIC SYSTEMS NEWS
FILENUMBER: 482
BEGIN_KEYWORDS
ASSESSMENT fpga conferences
END_KEYWORDS
DATE: october 1995
TITLE: An Assessment of Conferences on Field-Programmable Devices and Systems
An Assessment of Conferences on Field-Programmable Devices and Systems
(Contributed by Alan Hunsberger of the National Security Agency)
Enclosed are two abbreviated trip reports of conferences on
field-programmable devices and systems that occurred in 1995.
Full reports of FPGA'95, FCCM'95, PLDCon'95 and FPD'95
may be obtained via email by sending a request to Don Bouldin
at dbouldin@utk.edu.
* The Third International ACM Symposium on Field Programmable Gate Arrays
(or FPGA'95) was sponsored by the ACM Special Interest Group on Design
Automation (SIGDA) with support from Actel, Altera and Xilinx.
It was held from 12 to 14 February 1995 at the Monterey Marriott
in Monterey, California.
The conference began on Sunday evening with registration and a
reception. The two working days of the conference were broken into 8
one-hour formal sessions, each containing three presentations.
A banquet followed by a panel session was held on the
second evening.
The attendee list of 100 included:
42 from FPGA vendors or hardware companies such as Altera, AT&T,
Xilinx, Actel and Hewlett-Packard; 33 from academia including
several active research groups from the Univ. of Toronto, UCLA,
Univ. of Washington and UC Santa Cruz; 17 from CAD companies
such as Synopsys, Quickturn Design Systems, Cadence, Mentor Graphics and
NeoCAD; 5 from various other companies; and 2 from government agencies.
Sixteen of these were from outside the U.S. including Canada, France,
U.K., Belgium, and Australia.
Twenty-four papers were presented at the conference with 21 of these
being from academia and 3 from industry. About half of the papers
from academia were presented by graduate students, probably because
of the relative newness of the FPGA technology.
The topics of the presentations included:
General Purpose Architecture (ideas for better FPGA chip architectures),
Field Programmable Systems (techniques to put together a number
of FPGA chips to form a system), Applications (using FPGAs to do real work),
Logic Synthesis (theoretical discussions of techniques
for optimizing the implementation of circuits in FPGAs),
Architecture of Special Purpose Structures (ways to implement new
or better FPGA features),
Placement, Routing and Testing,
Multi-FPGA Partitioning (ways to break up a large
circuit into several/many FPGA chips), and
Applications and Bit-Serial Synthesis (more uses of FPGAs
to do real work).
Several interesting points were made during the conference.
a. An irony in FPGAS is that the people building them are full-custom
VLSI people who don't like or use CAD tools because the tools are not
very good for optimizing custom circuits; but the people who use FPGAs
have to use CAD tools to map their application circuitry onto the FPGA
chips.
b. Six areas are considered by Altera in developing an FPGA
architecture: price per gate, capacity, performance, time to market,
power requirements, and ease of use. The CAD tools and the hardware are
developed together (Altera gets an architecture idea, works on the CAD
tools, and then produces the final architecture). Two measures of
success are the die size and how well application circuitry fits on
the chip (25% of the architecture on the FPGA chip is added because of
poor CAD fitting results on test circuitry).
c. Eighty percent of design work today is through schematics. Synthesis
tools must know about the special resources on the FPGA chip and about
the resource limits that are critical to the design. CAD tools should
be good enough that the designer needs only to specify the intent of
the circuitry and not the implementation.
For additional information about FPGA'95, contact Pak Chan at
pak@cse.ucsc.edu. For information about FPGA'96 which will be
held in Feb. 11-13, 1996 at the Monterey Marriott, contact Jonathan Rose
at jonathan.rose@xilinx.com or access:
FPGA-96
* The IEEE Symposium on FPGAs for Custom Computing Machines (or
FCCM'95) was sponsored by the IEEE Computer Society Technical Committee
on Computer Architecture. This is the third such meeting, although this
year it was called a "symposium" rather than a "workshop".
It was held from 19 to 21 April 1995 (Wednesday through Friday) at
the Marriott at Napa Valley in Napa, California.
Two full days were needed for the presentation of papers.
Friday morning was allocated for nine vendors to
discuss and demo some of their latest activities.
One hundred fifteen people attended FCCM'95. I understand that this is
about the same number that attended the conference the previous two
years. Of the attendees, 93 were from the U.S. and 22 from other
countries including the U.K., Canada, France,
Germany, Japan, Spain, Switzerland and Taiwan. The number of attendees
from academia and industry were about the same. Six attendees were
from government-related organizations.
Only one industry attendee was from a major CAD tool developer (Data
I/O), the rest were from FPGA manufacturers and specialized hardware
product developers. The largest contingent was from Hewlett-Packard
(9); others included National Semiconductor (5), Xilinx (4), Annapolis
Micro Systems (4), and one or two from e.g., Altera, Atmel, DEC,
Pilkington, Quickturn, and Virtual Computer Systems. The largest
university contingents were from the University of California at
Berkeley and Virginia Tech (6 each) followed by Brigham Young, UCLA,
and UC at San Diego (3 each). Other academic attendees included those
from Michigan State, Brown, and Cal Tech.
Only about a dozen attendees had attended the FPGA'95 conference held
in Monterey this past February.
Twenty-six papers were presented at the conference: 22 were from academia and 4
were from industry. Topics included: Custom Computing Platforms,
Signal Transport, Run-Time Reconfiguration, Applications and Compiler Issues,
The Wednesday night open panel session was led by Duncan Buell and was
titled "Reconfigurable Computing Machines: Real or Fad". The session
was quite lively and comments from the various attendees included the
following:
- FCCMs are still a process in search of a problem, a "killer
application".
- We are now in a lull between playing with FCCMs and seeing how they
can be useful commercially.
- Reconfigurable units will eventually be included in general purpose
computers.
- We need better and cheaper tools to program FCCMs. There is a problem
as long as the tools are just like what you use to design hardware.
- Too much attention is being paid to the computation units and not
enough attention is being given to system issues.
For additional information about FCCM'95, contact Peter Athanas at athanas@vt.edu.
For information about FCCM'96 which will be held in April 17-19,1996 at the Napa
Valley Marriott, contact Jeffrey Arnold at jma@super.org or access:
FCCM-96.
dbouldin@utk.edu