We are going to start a new project based on an existing C code which is not
in good structure, The resulting code will evolve for a relatively long
period of time and be very possiblly parallelized by using PVM in the future.

We are impressed with the C++ features in term of program development and
maintainance efficency described in literature. Could someone there give
me a perspective on C++ in scientific computation based on their experience
in C++? More specifically,

 (1) Is C++ a good choice for scientific computation since most of the
     variables are coupled (related)

 (2) when will the ANS stardard for C++ be published?

 (3) is c++ compiler avaliable on most of the different machines?
     for example, Sun Sparc station, Silicon graphics, IBM/RS600 and
     other main frame, Crary, Convex, etc.

 (4) How is the compatibility of the compilers on different machines?

Additional info for what we are doing:

Our group develops large-scale software to simulate geochemical systems.
The programs basically solve a set of time-dependent nonlinear partial
differential equations repeatively. Each program is tipically of 10,000-20,000
lines of code and CPU time intensive.

Your help is greatly appraciated.

Yhanks!

-- Qin