Parallel benchmark on other systems

From Gerris

Jump to: navigation, search

This pages is dedicated to the storage of the results of the parallel benchmark the parallel Bénard–von Kármán Vortex Street example for various systems and implementations of MPI, with and without load-balancing. The benchmark test is described on the Parallel benchmark on multi-core CPUs page. The goal here is to collect results from as many systems as possible. If you have some time, feel free to run the benchmark on your own system and add your results.



System
  • Dual-Intel(R) Xeon(TM) CPU 3.06GHz + 2 virtual cores
  • Ubuntu 9.10 32-bits
  • Linux delauxs 2.6.31-20-generic
  • Gerris2D version 1.3.2 (100407-025453)
MPI versions

Open MPI: 1.3.2-3ubuntu1.1 (packages libopenmpi1.3, openmpi-common, libopenmpi-dev, openmpi-bin)

Results

#CPUs Relative speedup
1 1
2 (load-balanced) 1.49
4 (load-balanced) 1.34
2 (not load-balanced) 1.22
4 (not load-balanced) 1.29

System
  • Intel(R) Xeon(R) CPU X5650 2.67GHz
  • Ubuntu 10.10 64-bits
  • Linux 2.6.35-24-generic #42-Ubuntu SMP
  • Gerris2D version 1.3.2 (110108-043625)
MPI versions

Open MPI: 1.4.3

Results

#CPUs Relative speedup
1 1
2 (load-balanced) 1.83
4 (load-balanced) 3.08
2 (not load-balanced) 1.62
4 (not load-balanced) 2.11

MPICH2 Version: 1.2.1p1

Results

#CPUs Relative speedup
1 1
2 (load-balanced) 1.89
4 (load-balanced) 3.19
2 (not load-balanced) 1.63
4 (not load-balanced) 2.09
Personal tools
communication