SPEC® MPIM2007 Result

Copyright 2006-2010 Standard Performance Evaluation Corporation

IBM Corporation

IBM BladeCenter JS22 Express (4 GHz, 4x4 core)

MPI2007 license: 0005 Test date: Oct-2008
Test sponsor: IBM Corporation Hardware Availability: Nov-2008
Tested by: IBM Corporation Software Availability: Nov-2008
Benchmark results graph

Results Table

Benchmark Base Peak
Ranks Seconds Ratio Seconds Ratio Seconds Ratio Ranks Seconds Ratio Seconds Ratio Seconds Ratio
Results appear in the order in which they were run. Bold underlined text indicates a median measurement.
104.milc 32 860 1.82 871 1.80 869 1.80 32 860 1.82 871 1.80 869 1.80
107.leslie3d 32 1681 3.11 1719 3.04 1728 3.02 32 1692 3.08 1694 3.08 1686 3.10
113.GemsFDTD 32 1447 4.36 1445 4.36 1451 4.35 32 1447 4.36 1445 4.36 1451 4.35
115.fds4 32 873 2.23 898 2.17 873 2.24 32 863 2.26 859 2.27 863 2.26
121.pop2 32 1440 2.87 1436 2.88 1441 2.86 32 1440 2.87 1436 2.88 1441 2.86
122.tachyon 32 2056 1.36 2059 1.36 2055 1.36 32 2015 1.39 2022 1.38 2016 1.39
126.lammps 32 1183 2.46 1209 2.41 1194 2.44 32 1183 2.46 1209 2.41 1194 2.44
127.wrf2 32 2675 2.91 2670 2.92 2677 2.91 32 1774 4.39 1786 4.36 1771 4.40
128.GAPgeofem 32 667 3.10 668 3.09 667 3.10 32 667 3.10 668 3.09 667 3.10
129.tera_tf 32 2077 1.33 2077 1.33 2076 1.33 32 1510 1.83 1511 1.83 1501 1.84
130.socorro 32 1165 3.28 1166 3.27 1166 3.27 32 451 8.46 447 8.53 456 8.37
132.zeusmp2 32 1273 2.44 1325 2.34 1288 2.41 32 1273 2.44 1325 2.34 1288 2.41
137.lu 32 1607 2.29 1600 2.30 1604 2.29 32 1607 2.29 1600 2.30 1604 2.29
Hardware Summary
Type of System: Heterogeneous
Compute Nodes: IBM System JS22
IBM System JS22
Interconnects: InfiniBand
Ethernet
File Server Node: IBM System JS22
Head Node: IBM System JS22
Total Compute Nodes: 4
Total Chips: 8
Total Cores: 16
Total Threads: 32
Total Memory: 80 GB
Base Ranks Run: 32
Minimum Peak Ranks: 32
Maximum Peak Ranks: 32
Software Summary
C Compiler: IBM XL C/C++ Enterprise Edition V9 for AIX
Updated with the September 2008 Fix level
C++ Compiler: IBM XL C/C++ Enterprise Edition V9 for AIX
Updated with the September 2008 Fix level
Fortran Compiler: IBM XL Fortran Enterprise Edition V11.1 for AIX
Updated with the September 2008 Fix level
Base Pointers: 32-bit
Peak Pointers: 32/64-bit
MPI Library: IBM Parallel Environment for AIX, Version 5
Release 1
Other MPI Info: None
Pre-processors: None
Other Software: IBM Engineering and Scientific Subroutine Library
(ESSL) for AIX Version 4 Release 3 Updated with
PTF Set 3

Node Description: IBM System JS22

Hardware
Number of nodes: 1
Uses of the node: compute, head, fileserver
Vendor: IBM Corporation
Model: IBM System JS22
CPU Name: POWER6
CPU(s) orderable: 4 cores per blade
Chips enabled: 2
Cores enabled: 4
Cores per chip: 2
Threads per core: 2
CPU Characteristics:
CPU MHz: 4000
Primary Cache: 64 KB I + 64 KB D on chip per core
Secondary Cache: 4 MB I+D on chip per core
L3 Cache: None
Other Cache: None
Memory: 32 GB (4x8 GB) DDR2 500 MHz
Disk Subsystem: 1x146 GB SAS 15K RPM
Other Hardware: BladeCenter-H chassis
Voltaire 4X InfiniBand Pass-thru Module (P/N
43W4419)
Adapter: 4X InfiniBand DDR Expansion Card (CFFh) for IBM
BladeCenter (P/N 43W4423)
Number of Adapters: 1
Slot Type: PCIe x8 Gen2
Data Rate: 4x DDR 20Gbps
Ports Used: 1
Interconnect Type: InfiniBand
Software
Adapter: 4X InfiniBand DDR Expansion Card (CFFh) for IBM
BladeCenter (P/N 43W4423)
Adapter Driver: devices.pciex.b3157862.rte 6.1.2.0
Adapter Firmware: 2.3.0
Operating System: IBM AIX V6.1 with the 6100-02 Technology Level
Local File System: AIX/JFS2
Shared File System: NFSv3
System State: Multi-user
Other Software: None

General Notes

 Blade[1] runs the following commands to compose the cluster:
      mkdev -c management -s infiniband -t icm
      /usr/sbin/mkiba -a 192.1.10.1 -m 255.255.255.0 -i ib0 -A iba0 -p 1 -P 0xFFFF -M 65532 -q 4000 -k off -Q 0x1E -S up
      startsrc -s ctcas
      preprpnode mpiblade1
      mkrpdomain mpiblades mpiblade1 mpiblade2 mpiblade3 mpiblade4
      startrpdomain mpiblades
      cd /usr/lpp/ppe.poe/samples/nrt
      make
      chmod 4755 nrt_api
      shutdown -rF
      su spec
      cd mpiblades.64ranks.load
      ../nrt_api -l

Node Description: IBM System JS22

Hardware
Number of nodes: 3
Uses of the node: compute
Vendor: IBM Corporation
Model: IBM System JS22
CPU Name: POWER6
CPU(s) orderable: 4 cores per blade
Chips enabled: 2
Cores enabled: 4
Cores per chip: 2
Threads per core: 2
CPU Characteristics:
CPU MHz: 4000
Primary Cache: 64 KB I + 64 KB D on chip per core
Secondary Cache: 4 MB I+D on chip per core
L3 Cache: None
Other Cache: None
Memory: 16 GB (4x4 GB) DDR2 667 MHz
Disk Subsystem: 1x146 GB SAS 15K RPM
Other Hardware: BladeCenter-H chassis
Voltaire 4X InfiniBand Pass-thru Module (P/N
43W4419)
Adapter: 4X InfiniBand DDR Expansion Card (CFFh) for IBM
BladeCenter (P/N 43W4423)
Number of Adapters: 1
Slot Type: PCIe x8 Gen2
Data Rate: 4x DDR 20Gbps
Ports Used: 1
Interconnect Type: InfiniBand
Software
Adapter: 4X InfiniBand DDR Expansion Card (CFFh) for IBM
BladeCenter (P/N 43W4423)
Adapter Driver: devices.pciex.b3157862.rte 6.1.2.0
Adapter Firmware: 2.3.0
Operating System: IBM AIX V6.1 with the 6100-02 Technology Level
Local File System: AIX/JFS2
Shared File System: NFSv3
System State: Multi-user
Other Software: None

General Notes

 Each blade runs the following commands to compose the cluster, where $CLUSTER_INDEX is 2-4 for Blade[2]-Blade[4]:
      mkdev -c management -s infiniband -t icm
      /usr/sbin/mkiba -a 192.1.10.$CLUSTER_INDEX -m 255.255.255.0 -i ib0 -A iba0 -p 1 -P 0xFFFF -M 65532 -q 4000 -k off -Q 0x1E -S up
      startsrc -s ctcas
      preprpnode mpiblade1
      cd /usr/lpp/ppe.poe/samples/nrt
      make
      chmod 4755 nrt_api
      shutdown -rF
      su spec
      cd mpiblades.64ranks.load
      ../nrt_api -l

Interconnect Description: InfiniBand

Hardware
Vendor: IBM Corporation
Model: 4x DDR InfiniBand
Switch Model: QLogic SilverStorm 9024
Number of Switches: 1
Number of Ports: 24
Data Rate: 4x DDR 20Gbps
Firmware: 4.2.1.1.1
Topology: single switch
Primary Use: MPI Communication

Interconnect Description: Ethernet

Hardware
Vendor: IBM Corporation
Model: 4-port Gigabit Ethernet
Switch Model: IBM BladeCenter 4-port Gigabit Ethernet switch
module (P/N 26K6483)
Number of Switches: 1
Number of Ports: 18
Data Rate: 1Gbps
Firmware: 1.08
Topology: single switch
Primary Use: File system

Compiler Invocation Notes

 Blade[1], with 32GB of memory and 32GB of paging space, was used to compile the benchmarks.

Submit Notes

The config file option 'submit' was used.
      submit = poe task_stride.2level.32+64rank 4 2 8 $ranks $command -procs $ranks -hostfile /spec/MapFiles/ib0hosts.8x.1-8

General Notes


 Environment settings:
     All ulimits set to unlimited
     ranks                = 32
     CWD                  = /spec/mpi2007
     MEMORY_AFFINITY      = MCM
     XLFRTEOPTS           = intrinthds=1
     MP_PGMMODEL          = spmd
     MP_MSG_API           = mpi
     MP_DEVTYPE           = ib
     MP_CLOCK_SOURCE      = AIX
     MP_STDINMODE         = none
     MP_SHARED_MEMORY     = yes
     MP_SINGLE_THREAD     = yes
     MP_EUILIB            = us
     NRT_WINDOW_COUNT     = 1
     MP_RESD              = no
     MP_PULSE             = 0
     ADAPTER_USE          = shared
     EUIDEVICE            = sn_single
     MP_CSS_INTERRUPT     = no
     MP_BUFFER_MEM        = 67108864
     MP_USE_BULK_XFER     = yes
     MP_BULK_MIN_MSG_SIZE = 8192
     MP_EAGER_LIMIT       = 65536
     MP_WAIT_MODE         = yield
     MP_INFOLEVEL         = 0
     MP_LABELIO           = no
     MP_STDOUTMODE        = unordered
     MP_PMDLOG            = no
     NRT_JOB_KEY          = 64

Compiler Invocation

C benchmarks:

 /usr/bin/mpcc_r 

C++ benchmarks:

126.lammps:  /usr/bin/mpCC_r 

Fortran benchmarks:

 /usr/bin/mpxlf95_r 

Benchmarks using both Fortran and C:

 /usr/bin/mpcc_r   /usr/bin/mpxlf95_r 

Portability Flags

107.leslie3d:  -qfixed 
115.fds4:  -DSPEC_MPI_LC_NO_TRAILING_UNDERSCORE   -qfixed 
121.pop2:  -DSPEC_MPI_AIX 
127.wrf2:  -DNOUNDERSCORE   -DSPEC_MPI_AIX 
130.socorro:  -DSPEC_NO_UNDERSCORE   -qcpluscmt 
132.zeusmp2:  -qfixed   -DSPEC_SINGLE_UNDERSCORE 
137.lu:  -qfixed 

Base Optimization Flags

C benchmarks:

 -bmaxdata:0x80000000   -O5   -D_ILS_MACROS   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K 

C++ benchmarks:

126.lammps:  -bmaxdata:0x80000000   -O5 

Fortran benchmarks:

 -bmaxdata:0x80000000   -O4   -qstrict   -qalias=nostd   -qhot=level=0   -qsave   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K 

Benchmarks using both Fortran and C:

 -bmaxdata:0x80000000   -O5   -D_ILS_MACROS   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K   -O4   -qstrict   -qalias=nostd   -qhot=level=0   -qsave 

Peak Optimization Flags

C benchmarks:

104.milc:  basepeak = yes 
122.tachyon:  -O5   -lessl   -D_ILS_MACROS   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K   -q64 

C++ benchmarks:

126.lammps:  basepeak = yes 

Fortran benchmarks:

107.leslie3d:  -O5   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K   -bmaxdata:0x70000000 
113.GemsFDTD:  basepeak = yes 
129.tera_tf:  -O5   -qessl   -lessl   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K 
137.lu:  basepeak = yes 

Benchmarks using both Fortran and C:

115.fds4:  -O5   -lessl   -D_ILS_MACROS   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K   -qstrict   -qalias=nostd   -qhot=level=0   -qsave   -q64 
121.pop2:  basepeak = yes 
127.wrf2:  -O5   -bmaxdata:0x80000000 
128.GAPgeofem:  basepeak = yes 
130.socorro:  -O5   -lessl   -D_ILS_MACROS   -bdatapsize:64K   -bstackpsize:64K   -btextpsize:64K   -qessl   -bmaxdata:0x80000000 
132.zeusmp2:  basepeak = yes 

Other Flags

C benchmarks:

 -w   -qsuppress=1500-036   -qipa=noobject   -qipa=threads 

C++ benchmarks:

126.lammps:  -w   -qsuppress=1500-036   -qipa=noobject   -qipa=threads 

Fortran benchmarks:

 -w   -qsuppress=1500-036   -qsuppress=cmpmsg   -qspillsize=32648 

Benchmarks using both Fortran and C:

 -w   -qsuppress=1500-036   -qipa=noobject   -qipa=threads   -qsuppress=cmpmsg   -qspillsize=32648 

The flags files that were used to format this result can be browsed at
http://www.spec.org/mpi2007/flags/MPI2007_flags.20081105.html,
http://www.spec.org/mpi2007/flags/IBM-XL.html,
http://www.spec.org/mpi2007/flags/IBM-AIX.html.

You can also download the XML flags sources by saving the following links:
http://www.spec.org/mpi2007/flags/MPI2007_flags.20081105.xml,
http://www.spec.org/mpi2007/flags/IBM-XL.xml,
http://www.spec.org/mpi2007/flags/IBM-AIX.xml.