O'KeefeHPS@sri-unix.UUCP (09/21/84)
From: O'Keefe HPS (on ERCC DEC-10)
PROLOG ACCELERATOR
Development of a Prolog machine based on the Abstract Prolog
Machine (APM) software architecture has been proposed to the
Alvey Directorate by a consortium of Cambridge Consultants
(CCL); Imperial College, London; and Logic Programming Associates.
This follows a design study for a Prolog machine which the group
completed for Alvey at the end of May.
If the proposal is approved, the group would develop a computer
which would implement the APM architecture (devised by Frank
McCabe of Logic Programming Associates) and would be offered as
an add-on to existing machines. Its effect would be that of a
Prolog "accelerator", enabling the system to which it is attached
to run Prolog at a much higher speed.
CCL is leading the consortium, with Paul Ruskin as project
manager.
{RAO'K: the APM is the target for Sigma Prolog, which is a
successor to micro-PROLOG. I believe that it still uses lists
whenever it possibly can, but has also introduced vectors. It
has hooks for some degree of coroutining by posting constraints
on variables. I have seen a document describing the APM, and my
main criticism is that itlooks as though it will need lots and
lots of memory. At the Prolog Implementors' Meeting held in July
the consortium claimed that the machine could be expected to
produce 30k LIPS and cost 2000 pounds.
These figures may be misprints in the minutes. I wasn't at the
meeting. It is worth pointing out that 10k LIPS or better should
be achievable on VAX 750s and the like without special hardware,
andwithout the difficulty of timesharing an addon.}