I am working with MPI_Comm_spawn to launch dynamic processes in my MPI application and I would like to use the process-to-core binding allocation strategy in processes spawned dynamically.

I already know that it is possible to set the "host" key in the MPI_Info parameter of the MPI_Comm_spawn call to spawn the process in a specific host. However, I would like to know how to bound a dynamic process to a specific processor core.

I tried passing to MPI_Comm_spawn the "host" info key as "compute-node-X binding:user=4", but it does not work and the process goes to any core randomly.

Thank you very much,

Regards,

Gonzalo.