I'm trying to port the code of one Finite Element Analysis from Fortran77 in Ansi C.

The linear system is made of up more 10,000 lines: the matrix is really sparse. I'm wondering which algorith should be used to solve this linear system.

Among those problem I'm wondering if it is in general true (even for sparse matrix) that the solution of a linear system of N unknowns, requires an order N^2 to be solved. Or am I wrong?

Any idea or reply will been weel accepted.

Mastro

I think that in general the central point of view for solving sparse linear matrix is the way they are stored in memory.

The way exposed in the book is really efficent for memory but seems easy implementasble for solving only with iterative methods.

Are there any other easy and well documented ways for storing big (up 10e6) matrix in memory to solve with direct methods?

With best regards,
Francesco

mastro_it2000 at yahoo .it

P.S. Aren't there any other indexed like storing algorithms?