PSTSWM SP3-375 Algorithm Comparison

Performance Studies using

PSTSWM

IBM SP3-375 Winterhawk II Algorithm Comparison

(transpose FFT experiment B )

Date/Person: December 12, 1999 / P. Worley
Platform: IBM SP3 at Oak Ridge National Laboratory (bobcat.ccs.ornl.gov):
     62 4-way Winterhawk II SMP nodes (375 MHz POWER3 with 8MB L2 cache)
Environment: AIX 4.3.3;   PSSP 3.1.1
Code Version: 6.6.4
Compilation Options: mpxlf -O3 -qarch=auto -qtune=auto -qcache=auto
Link Options: -bmaxdata:0x70000000
Environment Variables: export MP_SHARED_MEMORY="yes"
Math Library: ESSL
Communication Library: MPI over the switch using user space for internode communication
MPI using shared memory for intranode communication
Number of Timesteps: 12
Results:

Transpose FFT (mpi) / Row Major Processor Ordering
Algorithm Comparison
  T21L16    T21L16    T42L16    T42L32    T42L16    T85L32 
  P=32x2     P=16x4     P=8x8     P=32x2     P=16x4     P=8x8  
  optimal algorithm   swtrans  logtrans  srtrans  swtrans  srtrans  srtrans 
  (alltoallv-min)/min     0.061    0.092    0.102    0.043    0.039    0.051 
  (generic-min)/min     0.498    0.421    0.084    0.174    0.162    0.000 

Transpose FFT (combined) / Row Major Processor Ordering
Communication Library Comparisons
  T21L16    T21L16    T42L16    T42L32    T42L16    T85L32 
  P=32x2     P=16x4     P=8x8     P=32x2     P=16x4     P=8x8  
  optimal library   mpi  mpi  mpi  mpi  mpi  mpi 
  (alltoallv-min)/min     0.061    0.092    0.102    0.043    0.039    0.051 
  (mpi-min)/min     0.000    0.000    0.000    0.000    0.000    0.000 

Patrick H. Worley / ( worleyph@ornl.gov)
Last Modified Monday, 15-Jul-2002 10:19:54 EDT.