Evaluation of the FGCS Project
David H. D. Warren
Department of Computer Science
University of Bristol
sary in an advanced research project.
The research leaders of ICOT are scientists of very high calibre much respected by
their international colleagues. Most of them have been with the project for its du-
ration and have provided continuity of direction. Most of the other ICOT staff have
been working on three-year assignments from the companies. The resulting lack of
long-term continuity of ICOT staff, and the fact, as I understand it, that ICOT could
not hand-pick the majority of its staff, are additional handicaps to pursuing advanced
research not shared by comparable institutions such as ECRC, MCC and SICS.
Major Technical Achievements
The FGCS project has produced many significant technical achievements. Some of
the major accomplishments which are of particular interest to me and which I would
highlight are the following.
First, ICOT has achieved its foremost concrete objective of building a parallel infer-
ence machine with a performance exceeding 100 megalips. Given the state of the art at
the time the project was announced, when Prolog performance was at best 40,000 lips
and large-scale parallel machines hardly existed, this achievement is quite remarkable
and should not be underestimated. Although KL1 lips are not quite as powerful as
Prolog lips, ICOT's achievement still represents a leap forward by more than three
orders of magnitude.
On the language side, I consider GHC to be a most significant contribution. It
embodies, in my opinion, the most elegant encapsulation of the committed-choice lan-
guage (CCL) concept, simplifying and clarifying what was introduced by Parlog and
Concurrent Prolog.
In its parallel implementations of KL1, ICOT has significantly advanced the imple-
mentation technology for CCLs. My own group has drawn on this work, and on the
key idea of GHC, in our implementation of Andorra-I.
Although I have some reservations about that ICOT has committed itself entirely
to CCLs and the concurrent logic paradigm, it cannot be denied that ICOT's PIM
machine and operating system PIMOS are a powerful demonstration of what is possi-
ble in terms of building a machine and operating system entirely based on a CCL. It
strikes me as something of an heroic feat, akin to climbing Everest or putting a man
on the Moon, which opens our minds to future possibilities while perhaps not bringing
immediate economic benefit.
As part of its programme for producing demonstrations of KL1 and PIM, ICOT
- 144 -