uname -a, plus linpack output

[solipsistnation]

[root@toth linpack_202]# uname -a
Linux toth.wpi.edu 2.4.21-sgi300r3 #1 SMP Wed Apr 21 18:30:04 PDT 2004 ia64 ia64 ia64 GNU/Linux

[root@toth linpack_202]# cat lin_ipf.txt-2.4
Fri Jul  2 09:48:45 EDT 2004
Intel(R) Linpack data

Version: Intel(R) Linpack Version 2.0.2 for Intel(R) Itanium(R) Processor Family
Current date & time: Fri Jul  2 09:48:45 2004


Here is CPU, memory, OS, threads information

Intel(R) Architecture: Itanium(R) Processor Family
CPU version          : Intel(R) Itanium(R) 2 Processor
CPU frequency        : 1.500 GHz
Number of CPUs       : 16
Number of threads    : 16
Trying to determine L3 size, wait a minute, please...
Set L3 to 6144 KB. If that's not correct, run specific L3 binary.
OS                   : Linux Release 2.4.21-sgi300r3 Version #1 SMP Wed Apr 21 18:30:04 PDT 2004
Memory available     : 28167 MB
Machine epsilon      : 1.110223e-16

Parameters are set to:

Number of tests                             : 7
Number of equations to solve (problem size) : 40000 30000 24000 20000 15000 10000 8000
Leading dimension of array                  : 40016 30016 24016 20016 15000 10000 8016
Number of trials to run                     : 1     2     2     2     2     2     2
Data alignment value (in Kbytes)            : 4     4     4     4     4     4     4

Maximum memory requested that can be used=  524485888
============= Timing linear equation system solver =================

Size   LDA    Align. Time(s)    GFlops   Residual      Residual(norm)
40000  40016  4      752.068    56.7367  1.524397e-09 6.865275e+02
30000  30016  4      330.728    54.4308  9.646308e-10 5.792415e+02
30000  30016  4      330.590    54.4536  9.646308e-10 5.792415e+02
24000  24016  4      175.669    52.4687  5.971623e-10 4.482300e+02
24000  24016  4      175.362    52.5606  5.971623e-10 4.482300e+02
20000  20016  4      107.625    49.5621  3.701803e-10 3.334288e+02
20000  20016  4      107.662    49.5453  3.701803e-10 3.334288e+02
15000  15000  4      50.243     44.7912  2.946136e-10 3.538192e+02
15000  15000  4      50.264     44.7724  2.946136e-10 3.538192e+02
10000  10000  4      17.413     38.2981  1.017145e-10 1.832325e+02
10000  10000  4      17.484     38.1410  1.017145e-10 1.832325e+02
8000   8016   4      9.470      36.0581  7.384693e-11 1.662885e+02
8000   8016   4      9.484      36.0038  7.384693e-11 1.662885e+02

Performance Summary (GFlops)

Size   LDA    Align.  Average  Maximal
40000  40016  4       56.7367  56.7367
30000  30016  4       54.4422  54.4536
24000  24016  4       52.5147  52.5606
20000  20016  4       49.5537  49.5621
15000  15000  4       44.7818  44.7912
10000  10000  4       38.2195  38.2981
8000   8016   4       36.0310  36.0581

End of tests
Fri Jul  2 10:38:56 EDT 2004

Hint: 55GFlops is a LOT.

I have tried it with a 2.6 kernel, but it seems to work less well than the 2.4 kernel. There's a bit of a performance drop, which I found sort of disappointing. Some of it is that some of SGI's software doesn't appear to have been ported to the 2.6 kernel and doesn't work with it.

Anyway, there you go. uname -a output and a useful (if highly optimized) benchmark.