In my work, there are some special linear systems to be solve, such as,
I want to know how I can solve this kind of sparse matrix linear systems by cuSparse and if cuSparse has the ability to solve these problems.
Could you please give me some advices?
Thanks for your help!!!
This is probably not a fit for cusparse. Is submatrix A2 dense? Looks rectangular in your picture. Please provide more details. You might also want to look into cusolver, which has some iterative solvers, and least square methods.
I give a full description about my question.
In fact, in my work, there is a very very large sparse band matrix linear system to be solved,as following:
In general, the matrix A has a dimension of m×m (m≈100,000,000,000). So, this is a super large question that can’t be solved by only one device (Tesla A100 40GB) due to massive memry consumption.
Therefor, I decide to split this question into serval small subproblems, as following:
However, some issues block the realization of my idea. The most annoying one is how to split the hole linear system into subproblems and solve .them by cuSparse.
Do you meet this kind of issues or Nvidia have some solutions on this kind of super large question?
Please give some advice,Thanks