Is it possible to solve linear system of equations with parallel computing?

I am interested in solving big systems of equations with parallel computing. I have an old script that use the LU decomposition to solve it, but i was trying to speed it up using parallel computation. Do you have any suggestion?

Answer 1

[Sean de Wolski]

It sounds like you have multiple systems to solve, so I would suggest parallelizing over systems of equations. If you can provide more details, we might be able to help.

Answer 2

[Giovanni]

my problem is that the script i am using works for a K matrix that can be constant or not (for different loops over K*u=f the k matrix can change). In the case the K matrix is constant, the LU decomposition is perfect because I just evaluate once L and U and i solve several time the analysis. For that problem i don't care that much, so I am interested in solving K*u=f where K is a very big matrix. I am interested, because the computational cost, to use parallel processing also with different methods than the LU decomposition. Thanks

[Matt J]

For the constant K case, it might be better to gather up your different f into the columns of a matrix [f1,f2,...fn] and solve as

   K\[f1,f2,...fn]

If done this way, MATLAB's internal parallelization of mldivide will kick in.