Skip to content

TRC-HPC/MPI_Overlap

Repository files navigation

MPI_Overlap

Build via: make

Invoke using: mpirun --hostfile hostfile --bind-to core --mca pml ucx -x UCX_IB_GID_INDEX=0 -x UCX_TLS=self,sm,ud,rc ./overlap_regdata #commtype #data_elements #neighbors

where #commtype is one of:

-t: two-sided non-blocking comm. + wait for completion after compute (overlap),

-tw: two-sided comm. with immediate wait and then compute,

-o: one-sided asynchronous comm. + wait for completion after compute (overlap),

-ow: one-sided comm. with immediate wait and then compute

#data_elements is the number of data elements to send between ranks (typically ~200000-600000)

and #neighbors is the average number of neighbors each rank will send and receive data from.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •