PSTSWM Paragon Communication Protocol Performance

Performance Studies using

PSTSWM

Intel Paragon Protocol Performance

(distributed LT experiment A1 / recursive halving sum algorithm)

Date/Person: May 13, 1998 / P. Worley
Platform: Intel Paragon XP/S 150 MP at Oak Ridge National Laboratory:
     1024 MP nodes (3 50-MHz iPSC/860 processors per node)
Environment: Paragon OSF/1 Release 1.0.4 Server 1.4 R1_4_5
f77/Paragon Paragon Version R5.0.3
Code Version: 6.3
Compilation Options: if77 -O4 -Mnodepchk -Knoieee -Msafealloc
Math Library: KAI
Communication Library: MPI
Parallel Algorithm: halfsum
Partition: 1x8, 1x16, or 1x32
Results:

1x16 Processors / Problem T21L2
Runtime Statistics
min(mean-min)/min(median-min)/min(max-min)/min
  1.8165e-01   0.08   0.05   0.44 
Three Fastest
Protocols
1st2nd3rd
  c2   b0   c3 
       Number of Proctocols With
Runtimes Within X% of Min
1%5%25%
  1   10   24 

1x32 Processors / Problem T42L1
Runtime Statistics
min(mean-min)/min(median-min)/min(max-min)/min
  2.8047e-01   0.07   0.05   0.58 
Three Fastest
Protocols
1st2nd3rd
  c2   d2   c3 
       Number of Proctocols With
Runtimes Within X% of Min
1%5%25%
  1   13   24 

1x8 Processors / Problem T42L2
Runtime Statistics
min(mean-min)/min(median-min)/min(max-min)/min
  1.1822e+00   0.02   0.01   0.21 
Three Fastest
Protocols
1st2nd3rd
  b6   d6   b1 
       Number of Proctocols With
Runtimes Within X% of Min
1%5%25%
  14   24   25 

1x16 Processors / Problem T85L1
Runtime Statistics
min(mean-min)/min(median-min)/min(max-min)/min
  1.9619e+00   0.03   0.02   0.29 
Three Fastest
Protocols
1st2nd3rd
  b6   b5   b4 
       Number of Proctocols With
Runtimes Within X% of Min
1%5%25%
  8   24   24 

1x32 Processors / Problem T85L2
Runtime Statistics
min(mean-min)/min(median-min)/min(max-min)/min
  2.4029e+00   0.06   0.03   0.52 
Three Fastest
Protocols
1st2nd3rd
  b5   b6   b4 
       Number of Proctocols With
Runtimes Within X% of Min
1%5%25%
  4   16   24 

1x8 Processors / Problem T85L4
Runtime Statistics
min(mean-min)/min(median-min)/min(max-min)/min
  1.3376e+01   0.02   0.01   0.15 
Three Fastest
Protocols
1st2nd3rd
  b6   b4   b5 
       Number of Proctocols With
Runtimes Within X% of Min
1%5%25%
  13   24   25 

DISCUSSION

Partitions of the Paragon processor grid were used that match the processor subsets that the parallel algorithms would run on in a two dimensional data decomposition: 1x8, 1x16, and 1x32.
Patrick H. Worley / ( worleyph@ornl.gov)
Last Modified Monday, 15-Jul-2002 10:29:37 EDT.