🚀 go-pugleaf

This is the first of many installments summarizing what happened at
the 1987 Annual MUS meeting held in Braintree, MA from Tuesday,
April 27th until Thursday, April 30th.

Initially, I will summarize those things I personnally took notes on.
Any other persons who want to submit summaries, please do so.

Tuesday, Doug Rowan (VP Sales and Marketing) gave a talk on Masscomp.
Basically, the company is doing well. Machines in the field are
averaging 14 months MTBF with about 800 customer accounts. Recent
product annoucements include 3 CPU capability on the 5600 (not the
5700!) and favorable benchmark comparisions between the 5450 and
a Sun 3/160.

Later that day, Dave Cane (VP for R&D) gave a talk on near-term
(3 yr) developments. Here is some of the points I recorded.

1. Look for new Masscomp computers to use the VME bus as the
I/O bus. A path will be available for owners of the Multibus line
to upgrade to the VME-based archtecture. The MI bus will still
be used for all "high-speed" interconnects (e.g. memory, FPA,
VA, etc).

2. The 68030-chip will likely be the basic CPU for the VME-based line
of computers.

3. Multiprocessor capability will be available on all computers in
the VME-based line: from the smallest to the largest. The largest 
computer may be capable of having 12 processors on the bus with
the possibility of many VME busses to increase I/O bandwidth.

4. They are looking into offering a very low-cost product that
will be VME-based but have the CPU, memory and disk controller
on one VME board. Since the card count will be low and the package
small, this computer should be cheaper than the current 5300.

5. NFS should be available at the end of the calender year
(more on networking in another summary).

6. Diskless nodes may be supported. This might be useful for
creating "dedicated" computers that are basically data-acq stations
that take to a network and ship the data direct to disk on another machine.

7. Look for a new DACP product line that will give 10x performance over
the currect product line. Two products will likely be offered. One
will be a programable DACP, but much faster and the other will be an
ASIC (Application Specific Intergrated Circuit) system that will not
be programmable (i.e. code is not down-loaded at boot), but will be 
optimized for what many customers do with DACP now. This product will
also be faster and probably about the same cost as the current DACP.

8. Also, DACP modules may be developed to plug directly into VME instead
of the current STD bus.

9. Look for a better solution to the hardcopy from the graphics screen
problem (software) in the Summer 88 time frame.

10. A vectorizing FORTRAN compiler is being considered  as is more work
on ADA.

11. Work will be done on the VA-1 product line to being the increment
in performace more in line with expectations (i.e. the VA should be
much faster than the FPA for vector calculations when the vector
size is greater than some number like 60 or so [my opinion on the number
--sob]).

This is the end of summary 1. If you were there and you disagree with
my statements here, please drop me a line.

Stan Barber, President
Masscomp Users' Society
Moderator of comp.sys.masscomp