[mb.pejvak]

hi friends;
Coefficient matrix in NS equations (Turbulence model equations and ...) are extremely large, and if it is defined as A(E,E) (E: number of cells), computer can not compile this huge data. one way is using the methods were introduced in dealing with sparse matrix [1] (CDS,CRS, ... ).
now I want to know whether or not specific methods have been introduced which are suitable for CFD particularly in unstructured Grid?

thank you so much in advance.

[1] Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods", Ch. 4, available in: http://www.siam.org/books

[cdegroot]

Yes, sparse matrix representations are absolutely mandatory for CFD problems. I have developed in a number of CFD codes and certainly we store our matrices in a sparse format, e.g. Yale sparse matrix format.

[mb.pejvak]

Originally Posted by cdegroot

Yes, sparse matrix representations are absolutely mandatory for CFD problems. I have developed in a number of CFD codes and certainly we store our matrices in a sparse format, e.g. Yale sparse matrix format.

Dear cdegroot;
Thank you so much. do you know which method is more suitable for CFD, particularly in unstructured grid which does not have a regular data structure or is not a banded matrix?

[cdegroot]

Originally Posted by mb.pejvak

Dear cdegroot;
Thank you so much. do you know which method is more suitable for CFD, particularly in unstructured grid which does not have a regular data structure or is not a banded matrix?

https://en.wikipedia.org/wiki/Sparse_matrix

I typically use the Yale format, but I am sure others are suitable.

[ztdep]

any sparse matrix storage form is suitable for the cfd computation.
for example, coordiniate form, csr form, csc form, and the other.
you can download the sparskit from
http://www-users.cs.umn.edu/~saad/software/
to find transformation between different matrix form.