|
|
eclipse newsletter 3.0 |
From the Adonis pipeline, eclipse has now grown to become a multi-purpose tool to program original data reduction algorithms which are too greedy in CPU and memory/disk use to be implemented within standard astronomical software packages. It was the case for SOFI/ISAAC, which were in need of a command to support the jittering data reduction. Algorithms in this domain easily reach a billion pixel operations for a single data set processing. It means that without dedicated routines in low-level C, it is impossible to reach reasonable amounts of processing time with a decent-sized workstation, especially when the process takes place in a telescope during the night as it is the case for the VLT dataflow system.
eclipse is now officially part of the tools which will help us build parts of the VLT data flow system, as a complementary data reduction engine to help MIDAS cope with the enormous data flow produced by the VLT. It will mainly be restricted to image processing tasks, since this is the area where we are in deep need of fast processing. Spectroscopy is completely covered by MIDAS already in pipeline mode, it will not be necessary to go to dedicated procedure development for these modes.
Ah, yes eclipse is still written in C, in ANSI C and only with ANSI C. As far as I know, this language may be obsolete by now but is still the only standard offered on any Unix workstation. It is convenient to program, offers all functionalities we need to implement efficient pixel-crunching in a portable way. Moving to C++ for example is out of question because of portability reasons. Example? The same code compiled with the same compiler (g++) on similar but different platforms (Solaris and HPUX in this case) is not compatible on fundamental issues such as memory management. No ANSI C++ standard is stable, and last but not least... the language grammar is not complete -- mostly because of operator overloading. Interested readers are referred to C++??: A critique of C++ (3rd edition). Because of its lack of direct memory access capabilities, Java is an interesting tool but for image processing, where we are deeply in need of low-level but portable access to the machine. What will be the language of the future for portable image processing applications? Who knows?...
Parsing through the list of sites accessing these web pages, one can
see a growing interest for the eclipse library from astronomical
sites all over the world. Indeed, there have been recorded downloads
from nearly all european observatories. American users are harder to
track since there is now an unmonitored download site for eclipse
in Tucson, Arizona. Lately, I learned (almost by accident) that the
Subaru telescope in Hawaii makes use of eclipse together with
other packages (IRAF and Skycat) through a CORBA environment. See
the following paper for more references about this subject:
The Distributed Analysis System Hierarchy (DASH) for the SUBARU Telescope.
In the future, I would greatly appreciate receiving more feedback about
what people do with this software package.
I hope eclipse users will tolerate the multiple changes which took place between version 2.0 and 3.0, especially the fact that the tar file grew up from 500k to about 800k due to the addition of the SFL library. Script writers will have to modify certainly some command calls since some changes affect both command and option names. Many thanks anyway for all eclipse users, for your support, for your help in reporting bugs and suggesting modifications. Again, I would strongly appreciate being kept informed of where and how is eclipse used, at least for my own information!
N. Devillard
Wed May 27 17:48:33 MET DST 1998