Evaluation of the FGCS Project
David H. D. Warren
Department of Computer Science
University of Bristol
problem with designing specialised experimental hardware is that any performance ad-
vantage that can be gained is likely to be rapidly overtaken by the ever continuing rapid
advance of commercially available machines, both sequentiual and parallel. ICOT's
PSI machines are now equalled if not bettered for Prolog and CCL performance by ad-
vanced RISC processors. And it seems very possible that commercial multiprocessors
such as Sequent Symmetry, the new Butterfly, and other recent machines could come
close to equaling the PIM performance if ICOT's software technology were ported to
those machines.
A subsidiary issue is whether it was necessary to target KL1 so much at distributed
memory hardware, with all the attendant problems of achieving good locality of com-
munication and good load balancing, rather than adopting a virtual shared memory
approach, for which scalable solutions are becoming increasingly well developed, in-
cluding ones supporting a quasi-UMA (uniform memory access) model (c.f. KSR-1
and the closely similar work on DDM that I have been involved in). In general, I feel
that ICOT perhaps devoted too great a proportion of its effort to developing hardware
and operating systems, and could perhaps have focussed its efforts more on the knowl-
edge processing software and applications which were central to the original conception
of the project.
This section of my report is rather long! Its length should be interpreted not so
much as a measure of criticism of ICOT's approach, which given the many constraints
they were operating under has been very productive I believe, but rather as a measure
of the complexity of the issues that I felt needed to be mentioned.
Overall Evaluation
The nature of the original FGCS announcement raised a lot of expectations that the
project could never have satisfied and certainly have not been satisfied. Unfortunately,
this makes it difficult for the project to be judged a success by the world at large,
which includes most of the media. However, I strongly believe that overall the project
has been a considerable success, and I think most fair-minded and properly informed
observers will share my view.
The project was a major success in galvanising worldwide activity and more im-
portantly for its scientific impact in stimulating worldwide research in new directions
inspired directly by the FGCS vision and ICOT's work. The project has also succeeded
in achieving its main concrete target of 100 megalips plus, an outstanding accomplish-
ment that shouldn't be diminished with the benefit of hindsight.
But above all, any research project such as FGCS should be judged in comparison
with comparable efforts by comparable institutions elsewhere. I believe the specific
- 147 -