In CFD problems, as we decrease grid spacing, the number of grid points increases rapidly. If we use the approach of matrix generation than we need to declare very large matrices but, in principle, there is a limit to matrix size we can handle using FORTRAN or any other programming language. Infact, practically there is no such limit and people handle such matrices. I want to know the tricks and techniques for the same.

Please help.

Thanks.

Amit

in TDMA we always break up the problem into 1 -d irrespective of whether the problem is 2-d or 3-d. We never assemble a "global matrix" which FEM people do. Thus with finite volume technique, this problem generally do not come .

The matrices are generally very sparse, most of the entries are zero. There are software available for handling sparse matrices in an efficient way. See the online book by Youcef Saad titled "Iterative Methods for Sparse Linear Systems". If you cant find this on the net, then let me know so that I can mail it to you.

I am sometimes getting this time message .What does this mean Exactly? run-time error M6101: MATH - floating-point error: invalid Domain error******** Manas

Sounds like a data structure question. I suggest that you read about linked lists. This is the way to handle large unstructured matrices in fortran.

If you know the structure of your matrix beforehand, then you might be able to use an easier data structure, eg: tri-diagonal, penta-diagonal, whatever... if you don't then it's back to linked lists.

Neale.

Shouldn't be a problem if you deal with the nonzero elements only. You may look into techpubs.sgi.com; that may help.