Cray XE6 parallel tasks / parallel tasks per node

Question

In a job submission script (Cray XE6) what is the difference between 'number of parallel tasks' (-n) and 'number of parallel tasks per node' (-N)?

Answer 1

Im inexperienced in this but some googling gave:

It looks like a MPI parameter in an example:

# Launch the parallel job using aprun.
# Run the executable my_mpi_executable.x using total
# of 2048 parallel tasks, with 32 tasks assigned per node.
aprun -n 2048 -N 32 ./my_mpi_executable.x arg1 arg2

From Cray Brochure:

Each Cray XE6 blade includes four compute nodes for high scalability in a small footprint – up to 128 processor cores per blade or 3,072 processor cores per cabinet. Each compute node has two AMD Opteron™ 6300 Series processors (16 cores) coupled with its own memory and Gemini communication interface and is designed to efficiently run up to 32 MPI tasks. Alternately, it can be programmed to run OpenMP within a compute node and MPI between nodes.

From the info above, n can be up to 3072 for a full cabinet and N can be 32 for two opterons(6300 series).

Answer 2

A man aprun will solve all your questions for that.

-n means how many instances of your executable you want (for MPI, this means how many MPI processes you want).

-N means how many instances per node you want.

-d is the depth, i.e how many threads per instance.