Communication Behavior

The following figures provide a preview of the communication patterns of various applications.
They are dynamically generated from live performance data in PADS.

An application's message volume and counts are obtained using mpiP, a lightweight profiling library for MPI applications. It is used to collect basic communication characteristics (point-to-point and collective calls) as well as the communication topology.



AMG2013 - 3D_laplace

AMG2013 - default

BoxLibMiniAMR - inputs_2d

BoxLibMiniAMR - inputs_3d

Exp_CNS_NoSpec - inputs_3d

GTC - 50msteps

GTC - 128p-50msteps

GTC - 256p-50msteps

GTC - 512p-50msteps

GTC - 1024p-50msteps

HPCG - default

HPL - default

LSMS - Fe16

LSMS - Fe16

LULESH - sm_8x38

LULESH - default

MCB - 4x4

MCB - 16x16

MCB - 8x8

MOCFE - MOCFEbone-default

MultiGrid_C - inputs.3d_n_cells_128

MultiGrid_C - inputs.3d_n_cells_256

MultiGrid_C - inputs.3d_n_cells_512

NEKBONE - communication_suite

NEKBONE - mgrid

Nek5000 - eddy-8p

Nek5000 - vortex-8p

Nek5000 - eddy-64p

Nek5000 - vortex-64p

XGC - input_setup_neutral-default

XGC - input_setup_neutral-planemajor

XGC - input_setup_neutral-interplanemajor-poldecomp-true

XGC - input_setup_neutral-planemajor-poldecomp-true

XGC - input_setup_neutral-interplanemajor-poldecomp-10ts-SPECTRAL0

miniAMR - two-spheres-16p

miniAMR - sphere-diagonal-27p

miniAMR - expanding-sphere-64p

miniFE - 1056x1024x1024

pcluster_accel - tnnc_gsmnp_vertical_profiles-1.7GB

pcluster_accel - tnnc_gsmnp_vertical_profiles-1.7GB-64p
.