EMAS Team termly Report
      -----------------------

     The summer has been a time of intense activity for the EMAS group as two
major pieces of hardware had to be commissioned.  Both involved substantial
software changes and both projects were completed within the hoped for
timescales  - timescales that had been agreed before it was realised that the
two events would become co-temporaneous! The team deserves the congratulation
of the rest of the ERCC for a notable and successful effort and it is
regrettable that hardware problems, particularly with the 2988, have prevented
the Users from realising the scale and success of these efforts.

     The DAP arrived in April with the usual sub minimum of documentation. An
elegant solution was devised for incorporating it into EMAS fully compatible
with the EMAS philosophy. Each DAP user is allocated the whole of a virtual DAP
which is connected into his address space like a virtual file. A virtual DAP,
which may be less than but not more than the size of the actual DAP, then
behaves exactly like a mapped file with the additional (and alarming!) property
that its contents change unpredictably with time. Multidap-programming is
possible where several users require DAPs but only when the sum of the virtual
DAPs is less than the total DAP store of 2 megabytes. These changes were
complete by June but the DAP Fortran was not ready by then. Testing gave
different answers from those obtained at QMC and some very severe and prolonged
debugging was needed before the fault was traced to a hardware error. By the
middle of August a DAP service was possible; the proposed restriction of DAP
work to background mode is not yet being enforced. Most DAP development was
done on split machines without interruption to service. It is regrettable that
a software problem that only occurred at service loads caused three system
closes or crashes in the early half of August. The full integration of the DAP
accounting into Director has yet to be accomplished but this should be complete
by the end of September. The assistance of Compiler group with the DAP and in
mounting DAP Fortran is gratefully acknowledged.

     The 2988 arrived in early July and was offering some sort of service on a
single OCP by early August. It was soon discovered that the hardware for
semaphoring a dual OCP worked differently on the 2988 from the 2980. Before
EMAS could run as a dual a substantial portion of the resident code had to be
rewritten. This hard and tedious job was completed in a few days by J.Maddock
and dual operation was achieved before the end of August. Another contribution
of note came from A.Gibbons who had to modify the file index areas of EMAS File
Systems to enable the 640Mb disc to be used. Further developments are needed to
obtain an easier IPL and to take advantage of the "domain" protection system.
However with so much achieved in a 9 week period the remaining tasks are
clearly only a matter of time.

     The depressing and ill-informed report by the internal working party on
languages caused much irritation in the group. These two projects were only
possible because of the power and diagnostics of IMP. Further the money for the
2988 did not run to any pre-delivery testing time for S-series hardware and so
the time was bought by IMP. The Scottish Home and Health Department also had a
working group looking for a suitable language to use for program development on
their 2966. Their working group seems more open minded than ERCC's and selected
IMP80 for all systems and non-numerical work. By providing IMP80 on VME2900 the
ERCC obtained an adequate amount of hands on time on the 2966 for preliminary
testing. This time would have cost several thousand pounds at commercial rates.
Critics of the cost of maintaining IMP should note that the sum thus raised
will provide about 5 years maintenance at the current manning level of less
than 0.1 persons per year!


                                        P.Stephens
                                        Sep 82