[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Programming Performance on Symbolics
I spend most of my time working on user interfaces on the
Symbolics and by trial & error have found ways to do things
that seem efficient and useful. I do get useful information
over the SLUG-net and although I have not attended a national
SLUG conference yet. I am planning to atttend SLUG88.
However, I feel that there is a lot of very useful information
out there in the heads and software of Symbolics, Inc and SLUG
members, that is not available to me and other Symbolics users.
The courses that Symbolics offers donot provide this type of
information - they are a potpourri of available facilities and
donot address performance issues.
Thus, I am not aware that sufficient information is readily available
concerning optimization of performance on the Symbolics. What I
would like to see are:
1) an additional volume in the infamous 10
volume series, on the subject of performance. (dream on!)
2) a Symbolics course on performance and advanced window/flavor
techniques (see proposed subjects below)
3) tutorials and sessions at SLUG national meetings on
performance. Topics could include, as a start:
1) system configuration: hardware; physical memory, virtual
memory. networking, file serving with multiple nodes, ...
2) efficient use of Common Lisp functions: examples of usage
in appropriate situations, ... (for portable software)
3) efficient use of zeta-lisp, underlying Symbolics functions
to acheive maximum speed in e.g. managing large arays, how to
avoid consing, how to generate less gargage, which underlying
functions are faster than Common Lisp and/or Zeta Lisp,
.... (for non-portable software)
4) efficient use of Symbolics Genera & TV: windows and flavors;
e.g. fastest pop-up windows (small & large item lists), which
ways of displaying object in dynamic windows are fastest to
generate - fastest to access with mouse handlers, ....
5) perfomance hints on using Symbolics S- products (e.g.
S-Geometry). How to manage large numbers of objects, etc...
I would welcome a discussion on this issue on the SLUG network and
after appropriate input, etc., some concrete steps to be taken by
Symbolics, SLUG and SLUG members.
Will Taylor taylor%plu@io.arc.nasa.gov