- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
If MKL libraries are linked to an MPI program by using the MKL link flag i.e. -mkl=cluster, then whether the problem size of mkl routine is divided between mpi processes alone? or between mpi processes as well as OpenMP threads also?
Thanks,
Sangamesh
Link Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Sangamesh,
If you use -mkl=cluster, it will be linking to cluster mkl lib, MPI version and wouldn't be using hybrid.
(Ref: https://software.intel.com/en-us/node/528338)
If you want hybrid parallelism, link the libs individually, and our MKL linkline advisor will give you the exact link line that you would want to use.
https://software.intel.com/en-us/articles/intel-mkl-link-line-advisor
Vipin
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page