Tpetra: Increasing number of threads leads to decrease of performance
Created by: freaklovesmango
That post refers a little to that one: #3827
I wanted to change Epetra to Tpetra, expecting a shorter execution time by increasing the number of threads while using just one MPI process, thus using shared memory functions via Tpetra/Kokkos.
Unfortunately, the results get worse by increasing the number of threads... and again, I wonder why. I am using Belos with Tpetra as the underlying library on a x86-64 system.