Re: FORTRAN common blocks

[Russell Shaw <rjshaw@netspace.net.au>]

Richard James wrote:
> On Mon, 16 Jan 2006, Ricardo Biloti wrote:
>
> > >        One problem I have never been able to resolve - I accept that C is
> > > better for systems programming, but why on earth does anyone use it for
> > > scientific code?  Can somebody enlighten me?
> > Obviously, that assertion is a mistake. There are many scientists
> > writing scientific code in C. Few examples of scientific tools written
> > in C:
> >
> > Aztec (a parallel iterative library for solving linear systems)
> > FFTW
> > Gnu Linear Programming Kit
> > Gnu Scientific Library
> > NAG C Library
> > Open Optimization Library
> > PETSc, the Portable, Extensible Toolkit for Scientific computation
> > Seismic Unix
> > SLEPc, the Scalable Library for Eigenvalue Problem computation
> > SUNDIALS (SUite of Nonlinear and DIfferential/ALgebraic equation Solvers)
> > TAO (Toolkit for Advanced Optimization)
> >
> > Regards,
> > --
> > Ricardo Biloti
> > Department of Applied Mathematics
> > IMECC/UNICAMP

> My question was why do people do it?  The question of whether is one to
> which I already know the answer.

I haven't found anything i can't do in C easily. I couldn't imagine using
a language without pointers. Instead of copying 100x100 arrays by value as
a function parameter, i can just pass a pointer to it.

After looking at some papers on how to solve equations fast where the
matrices are sparse, it occured to me that the authors were misguided
by a pointerless language. Using pointers, i made a program to solve
problems with sparse matrices as one would "by hand", which eliminates
all kinds of possible numerical problems with "traditional" methods.