- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I tried to run cluster solver with one MPI process to solve a symmetric sparse linear system.
The software settings are:
- mkl 2022.2.0
- windows 10
- c++
The parameter are defined as follows
iparm[0] = 1; // no solver default
iparm[1] = 3; // parallel version of the nested dissection algorithm
iparm[5] = 1; // write solution to b
iparm[7] = 2; // max numbers of iterative refinement steps
iparm[9] = 8; // perturb pivot tolerance
iparm[10] = 1; // use nonsymmetric permutation and scaling MPS
iparm[12] = 1; // maximum weighted matching
iparm[27] = 0; // double precision
iparm[34] = 1; // C-style indexing
iparm[39] = 2; // distributed input
iparm[59] = 0; // cluster sparse solver mode
Since we write solution to b, I suppose the input for solution can be a dummy double. So I create a dummy double `ddum` and use `&ddum` as input for solution .
It works when I run with multiple processes. However, if I run with single process, it will overwrite the memory at the address of the solution. I check the length of the overwritten memory, it is of the size of the linear system (The example I test is 1188). I tried with a very small matrix (3x3), it did not give me any error.
Could anyone help me with the problem?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
Thanks for posting in Intel Communities.
We had tried with the sample example(C:\Program Files (x86)\Intel\oneAPI\mkl\latest\examples\examples_cluster_c.zip\c_mpi\cluster_sparse_solver\source\cl_solver_sym_sp_0_based_c.c) where it is using cluster_sparse_solver, we are able to run the code without any errors with single and multiple processes.
Could you please provide us with the complete sample reproducer code(data file if any) along with the steps if you had followed to investigate more on your issue?
Thanks & Regards,
Varsha
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thank you for your reply.
For the example you tested, you can try set ipam[5] = 1 (write solution to b) and print the x array. Cluster solver will write values to x as well.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
Thanks for your reply.
Could you please let us know what exact errors you are getting at the time of running the code on a single process?
And also, could you please let us know what compilation and running commands you are using for your code?
Thanks & Regards,
Varsha
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
We have not heard back from you. Could you please let us know what exact errors you are getting at the time of running the code on a single process?
And also, could you please let us know what compilation and running commands you are using for your code?
Thanks & Regards,
Varsha
![](/skins/images/7FC17B7B85029576C25F1E43CE255B51/responsive_peak/images/icon_anonymous_message.png)
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page