Many of our functions assume that the trace is an MPI trace. A lot of them also assume that there is messaging information.
Specifically, we use the "process" column to groupby and aggregate metrics, plot the timeline, and calculate time profile. For a multi-threaded program, some of our readers do not create a process column at all.
These lead to issues with OpenMP and Pthreads programs that are multi-threaded but only run in a single node.
Many of our functions assume that the trace is an MPI trace. A lot of them also assume that there is messaging information.
Specifically, we use the "process" column to groupby and aggregate metrics, plot the timeline, and calculate time profile. For a multi-threaded program, some of our readers do not create a process column at all.
These lead to issues with OpenMP and Pthreads programs that are multi-threaded but only run in a single node.