- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
My program depending on MPI 2019 hangs in communication. I'm testing on 3 machines, two with RHEL7.6, one with RHEL7.
In testing, I got rid of external dependencies, and only used MKL examples. When testing pblas examples, all examples yield correct result. However, cluster_sparse_solver examples fail to run on multiple machines.
For local testing, i use cmdline: mpirun -n 3 ./cl_solver_unsym_c
For multi-node testing, the cmdline is: mpirun -n 3 -ppn 1 -hosts host1,host2,host3 ./cl_solver_unsym_c
When using FI_PROVIDER=sockets, I get the printings the program hangs in the callstack:
When using FI_PROVIDER=tcp, no print from MKL is observed and program hangs in the provided callstack.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
@yzhao sorry, but I can only provide help with the latest version of Intel MPI, not 2019. Please try to use the latest version available, 2021.14.1.

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page