bglenden@colobus.cv.nrao.edu (Brian Glendenning) (01/31/91)
Hi. We're about to rewrite from the ground up a radio-astronomical
image processing system (AIPS). The present AIPS is about 10 years old
and consists of well over 500k lines of (mostly F77) code. The new
system will probably last as long and be at least as big.
C++ looks like it has many attractive features. We do have some
concerns that don't deal explicitly with the language features
however:
1. Is C++ here to stay: Is it going to be available on machines from
workstations to supercomputers (this seems to be the case now).
Will C++ be on the massively parallel machines that everyone says
are coming?
2. Portability: Are the C++ implementations close enough so that
different ports should be reasonable? Hints about what features to
avoid to allow the code to be reasonably portable? When will we
likely have a draft proposed C++ standard? An actual standard?
3. Efficiency: Although our "Math" libraries may well be in
fortran for speed, is the language otherwise efficient? e.g. will
the C code that cfront produces vectorize on a Cray or Convex?
Any other comments/experiences/hints would be appreciated,
particularly about large scientic systems. If possible, please respond
via email - I will provide a summary to the net when replies have
stopped coming in. Thank you.
Brian
--
Brian Glendenning - National Radio Astronomy Observatory
bglenden@nrao.edu bglenden@nrao.bitnet (804) 296-0286