PSTSWM Paragon Point-to-Point Communication Performance

Performance Studies using

PSTSWM


Intel Paragon SWAP Performance

(ordered swap of 8KB message using NX)

Date/Person: January 30, 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
Communication Library: NX
SWAP size: 1024 REAL*8 floating point values each direction
Message size: Largest - 1024 REAL*8 floating point values
Smallest - 1 REAL*8 floating point value
Processors: 0 and 1
Latency Definition:(T1024-T512)/512
Results:

ordered simple swap
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 75.19 52.42 48.1%
10 iter. 77.55 52.12 49.3%

ordered swap using nonblocking send
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 65.33 60.03 47.9%
10 iter. 71.42 59.61 52.0%
1 iter. w/overlap 59.71 67.71 49.4%
10 iter. w/overlap 64.89 65.49 51.9%

ordered swap using nonblocking receive
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 65.15 60.29 47.9%
10 iter. 71.23 59.54 51.8%
1 iter. w/overlap 67.20 67.85 55.7%
10 iter. w/overlap 69.70 64.93 55.2%

ordered swap using nonblocking send and receive
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 53.77 59.27 50.1%
10 iter. 71.76 59.97 52.5%
1 iter. w/overlap 59.34 78.70 57.0%
10 iter. w/overlap 63.08 73.43 56.5%

ordered swap using ready send
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 54.74 90.60 60.5%
10 iter. 55.59 90.67 61.5%
1 iter. w/overlap 66.52 68.28 55.4%
10 iter. w/overlap 72.14 65.49 57.7%

ordered swap using nonblocking ready send
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 54.56 90.65 60.4%
10 iter. 52.27 89.61 57.2%
1 iter. w/overlap 56.77 79.47 55.1%
10 iter. w/overlap 65.09 73.99 58.8%

synchronous
Data Statistics
unidirectional bandwidth estimated latency model error
(peak MByte/sec) (usec/msg) (max. rel. error)
1 iter. 60.77 76.83 57.0%
10 iter. 60.81 76.42 56.7%

DISCUSSION


Patrick H. Worley / ( worleyph@ornl.gov)
Last Modified Monday, 15-Jul-2002 10:29:20 EDT.
86062 accesses since 1/2/96.