OMPL2001 Result: SGI SGI Altix 3000 (1500MHz, Itanium 2)

Benchmark	Reference Time	Base Runtime	Base Ratio	Peak Runtime	Peak Ratio
311.wupwise_l	9200	225	654427
313.swim_l	12500	171	1171495
315.mgrid_l	13500	368	586350
317.applu_l	13500	474	456023
321.equake_l	13000	1217	170902
325.apsi_l	10500	340	494730
327.gafort_l	11000	636	276645
329.fma3d_l	23500	853	440745
331.art_l	25000	231	1732791
SPECompLbase2001	536169
	SPECompLpeak2001	--

Benchmark

Reference
Time

Base
Runtime

Base
Ratio

Peak
Runtime

Peak
Ratio

311.wupwise_l

9200

225

654427

313.swim_l

12500

171

1171495

315.mgrid_l

13500

368

586350

317.applu_l

13500

474

456023

321.equake_l

13000

1217

170902

325.apsi_l

10500

340

494730

327.gafort_l

11000

636

276645

329.fma3d_l

23500

853

440745

331.art_l

25000

231

1732791

SPECompLbase2001

536169

SPECompLpeak2001

Hardware

Hardware Vendor:

SGI

Model Name:

SGI Altix 3000 (1500MHz, Itanium 2)

CPU:

Intel Itanium 2

CPU MHz:

1500

FPU:

Integrated

CPU(s) enabled:

128 cores, 128 chips, 1 core/chip

CPU(s) orderable:

4-256

Primary Cache:

16KBI + 16KBD (on chip) per core

Secondary Cache:

256KB (on chip) per core

L3 Cache:

6.0MB (on chip) per core

Other Cache:

N/A

Memory:

512 GB (16 x 1024 MB PC2700 DIMMs per 4 core module)

Disk Subsystem:

1 x 36 GB SCSI (Seagate Cheetah 15k rpm)

Other Hardware:

None

Software

OpenMP Threads:

128

Parallel:

OpenMP

Operating System:

SGI ProPack(TM) 3

Compiler:

Intel(R) Fortran Compiler for Linux 8.0 (Build 20040416)
Intel(R) C++ Compiler for Linux 8.0 (Build 20040416)

File System:

xfs

System State:

Single-user

Notes / Tuning Information
Baseline optimization flags: C programs: -openmp -O3 -ipo -ansi -ansi_alias (ONESTEP) Fortran programs: -openmp -O3 -ipo (ONESTEP) OpenMP runtime library libguide.a statically linked Extra Flags: 331.art_l: -DINTS_PER_CACHELINE=32 -DDBLS_PER_CACHELINE=16 Baseline user environment: OMP_NUM_THREADS=128 limit stacksize 256000 KMP_STACKSIZE 124M KMP_LIBRARY TURNAROUND OMP_DYNAMIC FALSE KMP_SCHEDULE static,balanced Alternate sources: Add critical region around update of linked list in parallel loop. Approved src.alt available as ompl-purdue1-20040324.tar.gz Used for 331.art_l base. For all benchmarks threads were bound to CPUs using the following submit command: dplace -x2 -cNTM1,0 $command, where NTM1 is the number of threads minus 1. This binds threads in order of creation, beginning with the master thread on cpu NTM1, the first slave thread on cpu NTM1-1, and so on. The -x2 flag instructs dplace to skip placement of the lightweight OpenMP monitor thread, which is created prior to the slave threads.

Notes / Tuning Information

 Baseline optimization flags: 
   C programs:       -openmp -O3 -ipo -ansi -ansi_alias (ONESTEP)
   Fortran programs: -openmp -O3 -ipo (ONESTEP)
   OpenMP runtime library libguide.a statically linked
 
 Extra Flags:
   331.art_l: -DINTS_PER_CACHELINE=32 -DDBLS_PER_CACHELINE=16

 Baseline user environment:
   OMP_NUM_THREADS=128
   limit stacksize 256000
   KMP_STACKSIZE 124M
   KMP_LIBRARY TURNAROUND 
   OMP_DYNAMIC FALSE
   KMP_SCHEDULE static,balanced

 Alternate sources:
 Add critical region around update of linked list in parallel loop.
 Approved src.alt available as ompl-purdue1-20040324.tar.gz
 Used for 331.art_l base.
 
 For all benchmarks threads were bound to CPUs using the following submit command:
 dplace -x2 -cNTM1,0 $command,
 where NTM1 is the number of threads minus 1.
 This binds threads in order of creation, beginning with the master
 thread on cpu NTM1, the first slave thread on cpu NTM1-1, and so on.
 The -x2 flag instructs dplace to skip placement of the lightweight
 OpenMP monitor thread, which is created prior to the slave threads.

First published at SPEC.org on 12-May-2004

Generated on Wed May 12 17:12:35 2004 by SPEC OMP2001 HTML formatter v1.01