I have been using this system on a windows 98 machine for some months and
have found it works very well. The compiler has proved to be very robust
particularly when dealing with modules that make use of F95 features such as
pointer components initialised to NULL() and ELEMENTAL procedures. Fplus has
managed handle the advanced f95 very well and it is months now since I
managed to find any bugs. Those I did find in an earlier version were fixed
fairly quickly but then NASoftware is run by some old friends and colleagues
and is based only a mile or so down the road from here so I had a very direct
line of contact.
The software management built into the compile process is very helpful for
building modules and testing them. It keeps pretty good track of dependencies
and recompiles modules automatically most of the time when changes have been
made that could effect a USEing program. Compilation is not as fast as some
other systems I have used but since the link=>.exe generation is largely
integral if the whole check dependencies-compile-link-go sequences actually
compare not that badly. The .exe code produced is not breathtakingly fast
although the polyhedron comparisons indicate that it is not too desparately
out of line except perhaps when running F77 style codes.
F77 is a weakness (or a strenghth depending on your point of view). Fplus is
fairly strict about supporting bog standard f77 and very few of the most
universal non-standard extension. It would be a bit of a pain to get a large
VAX Fortran 77 program to run on fplus unless the original authors has been
very standard conscious. If used for teaching or serious redevelopment in F95
this could be an advantage rather than a problem. However, users with large
amounts of existing none too standard F77 could have more difficulty with
Fplus than with some of the other PC Fortran 95 products.
The Salford FTN95 for instance adds F95 to their previously much extended
FTN77 processor and does a great job of running F77 codes that previously
ran on FTN77. Sometimes these actually run better on FTN95. The Salford
compiler unfortunately is less robust when processing highly F95 style new
code.
The Fplus debugger is a bit crude and far from bug free itself. Its way of
displaying components of derived type variables and arrays is clumsy and
sometimes it proved quicker to revert to the old fashioned approach of
sticking loads of prints in the code and debug it in the time honoured way.
Fplus also has the interesting addition that it is actively tracking the
standards process and already implements a number of the TRs that are
proposed to form part of F2000. It also builds in intrinsic support for the
String standard ISO 1539-2:1994.
In spite Fplus' greater reliability for F95 code, my Department mainly
supports the Salford product for Fortran on our NT network because of its
backwards compatibility with our users' F77 code. However, I use Fplus most
of the time myself because of its better F95 support.
--
Lawrie Schonfelder
Director, Computing Services Dept.
The University of Liverpool, UK, L69 7ZF
Phone: 44(151)794 3716, Fax: 44(151)794 3759
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|