SPEC Seal of Reviewal OMPM2001 Result
Copyright © 1999-2002 Standard Performance Evaluation Corporation
Fujitsu Limited
PRIMEPOWER HPC2500 ( 2.08GHz )
SPECompMpeak2001 = 36041      
SPECompMbase2001 = 34457      
SPEC license # HPG0003 Tested by: Fujitsu Limited Test site: Numazu, Japan Test date: Aug-2004 Hardware Avail: Jun-2004 Software Avail: Sep-2004
Benchmark Reference
Time
Base
Runtime
Base
Ratio
Peak
Runtime
Peak
Ratio
Graph Scale
310.wupwise_m 6000 164    36562       163    36818       310.wupwise_m base result bar (36562)
310.wupwise_m peak result bar (36818)
312.swim_m 6000 203    29580       203    29604       312.swim_m base result bar (29580)
312.swim_m peak result bar (29604)
314.mgrid_m 7300 224    32610       224    32610       314.mgrid_m base result bar (32610)
314.mgrid_m peak result bar (32610)
316.applu_m 4000 41.5  96479       41.5  96479       316.applu_m base result bar (96479)
316.applu_m peak result bar (96479)
318.galgel_m 5100 221    23061       221    23061       318.galgel_m base result bar (23061)
318.galgel_m peak result bar (23061)
320.equake_m 2600 98.7  26346       78.5  33105       320.equake_m base result bar (26346)
320.equake_m peak result bar (33105)
324.apsi_m 3400 56.7  59966       56.0  60664       324.apsi_m base result bar (59966)
324.apsi_m peak result bar (60664)
326.gafort_m 8700 256    34044       240    36226       326.gafort_m base result bar (34044)
326.gafort_m peak result bar (36226)
328.fma3d_m 4600 215    21426       206    22350       328.fma3d_m base result bar (21426)
328.fma3d_m peak result bar (22350)
330.art_m 6400 86.1  74331       86.1  74331       330.art_m base result bar (74331)
330.art_m peak result bar (74331)
332.ammp_m 7000 579    12094       502    13942       332.ammp_m base result bar (12094)
332.ammp_m peak result bar (13942)
SPECompMbase2001 34457        
  SPECompMpeak2001 36041        

Hardware
Hardware Vendor: Fujitsu Limited
Model Name: PRIMEPOWER HPC2500 ( 2.08GHz )
CPU: SPARC64 V
CPU MHz: 2080
FPU: Integrated
CPU(s) enabled: 32 cores, 32 chips, 1 core/chip
CPU(s) orderable: 8 to 128
Primary Cache: 128KBI+128KBD on chip
Secondary Cache: 4MB(I+D) on chip, per core
L3 Cache: None
Other Cache: None
Memory: 128GB
Disk Subsystem: 1x 73GB Ultra SCSI
Other Hardware: None
Software
OpenMP Threads: 31
Parallel: OpenMP
Operating System: Solaris 8
Fujitsu Parallelnavi 2.3 (NQS, CPU scheduler, largepage)
Compiler: Fujitsu Parallelnavi 2.4 (C & F90)
File System: ufs
System State: Multi-user
Notes / Tuning Information

 Baseline flags:
     C:    -KOMP,fast_GP2=2,V9,hardbarrier,largepage=1,prefetch=2,
            preex,a8,mfunc,ilfunc -O5 -x-
     F90:  -KOMP,fast_GP2=2,V8PLUS,hardbarrier,largepage=1,prefetch_line=4,
            prefetch_cache_level=3,mfunc=2 -O5 -Nautoobjstack

 Portability Flags:
     318.galgel_m:  -Am -Fixed -w
     328.fma3d_m:   -Am

 Extra Flags:
      330.art_m:    -DINTS_PER_CACHELINE=16 -DDBLS_PER_CACHELINE=8


 Peak Optimization Flags:

  310.wupwise_m: 
      F90: -KOMP,fast_GP2=2,V8PLUS,hardbarrier,largepage=1,prefetch_line=7,
           prefetch_cache_level=3,mfunc=2,ilfunc -O5 -Nautoobjstack
  312.swim_m: 
      F90: -KOMP,fast_GP2=2,V9,hardbarrier,largepage=1,prefetch_line=7,
           prefetch_cache_level=3,mfunc=2 -O5 -Nautoobjstack
  326.gafort_m: 
      F90: -KOMP,fast_GP2=2,V8PLUS,hardbarrier,largepage=1,prefetch_line=4,
           prefetch_cache_level=3,mfunc=2 -O5 -Nautoobjstack
           -Kunroll=4
  328.fma3d_m: 
      F90: -KOMP,fast_GP2=2,V8PLUS,hardbarrier,largepage=1,prefetch_line=6,
           prefetch_cache_level=3,mfunc=2 -O5 -Nautoobjstack
           -Kcommonpad=8,prefetch_infer
  332.ammp_m: 
      C:   -KOMP,fast_GP2=2,V8PLUS,hardbarrier,largepage=1,prefetch=2,preex
           ENV_OMP_NUM_THREADS = 30

 Alternate sources:
 
 Add critical region around update of linked list in parallel loop.
 Compulsory src.alt available as ompm-purdue1-20040324.tar.gz
 Used for 330.art_m base and peak.

 Peak sources:
 
 SPEC OMPL2001 source for 32bit systems modified for SPEC OMPM2001.
 Available as ompl src.alt in the SPEC OMP v3.0 release.
 Used for 312.swim_m, 320.equake_m, 324.apsi_m, 326.gafort_m, and 328.fma3d_m.

 System tunables:
     /etc/system:
         set shmsys:shminfo_shmmax=2147483648
         set shmsys:shminfo_shmmni=64
         set shmsys:shminfo_shmseg=64
         set autoup=172800
         set memscrub_period_sec=345600
     /etc/opt/FJSVpnrm/lpg.conf:
         JOB=96G,SHMSEGSIZE=2048M
     /etc/opt/FJSVpnrm/cpursc.conf:
         CPU_USE=0:0
     NQS:
         Per-process data size limit = UNLIMITED
         Per-process permanent file size limit = UNLIMITED
         Per-process memory size limit = 128 gigabytes
         Per-request memory size limit = 128 gigabytes
         Per-process number of cpus limit = 31
         Per-process stack size limit = 4 gigabytes
         Per-process CPU time limit = 2147483646.000
         Execution mode = SImplex
         Jobclass = 0

 Submitting the runspec to NQS:
     Run the qsub command with the following sh script.

         cd /spec/omp2001
         . ./shrc
         OMP_NUM_THREADS=31
         export OMP_NUM_THREADS
         runspec --config=Fujitsu --reportable --tune=all medium




For questions about this result, please contact the tester.
For other inquiries, please contact webmaster@spec.org
Copyright © 1999-2002 Standard Performance Evaluation Corporation

First published at SPEC.org on 16-Sep-2004

Generated on Fri Sep 17 10:39:12 2004 by SPEC OMP2001 HTML formatter v1.01