SPEC® MPIM2007 Result

Linux Networx

Linux Networx LS-1

SPECmpiM_peak2007 = Not Run

SPECmpiM_base2007 = 3.59

MPI2007 license:	021	Test date:	Sep-2007
Test sponsor:	Scali, Inc	Hardware Availability:	Apr-2007
Tested by:	Scali, Inc	Software Availability:	Aug-2007

Results Table

Benchmark	Base							Peak
Benchmark	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	545	2.87	544	2.88	545	2.87
107.leslie3d	32	1608	3.25	1610	3.24	1609	3.24
113.GemsFDTD	32	1057	5.97	1062	5.94	1057	5.97
115.fds4	32	662	2.95	662	2.95	662	2.95
121.pop2	32	939	4.39	942	4.38	941	4.39
122.tachyon	32	910	3.07	910	3.07	910	3.07
126.lammps	32	973	2.99	973	2.99	975	2.99
127.wrf2	32	1393	5.60	1394	5.59	1399	5.57
128.GAPgeofem	32	565	3.65	565	3.66	564	3.66
129.tera_tf	32	1007	2.75	1008	2.75	1008	2.75
130.socorro	32	803	4.76	802	4.76	803	4.76
132.zeusmp2	32	903	3.43	902	3.44	904	3.43
137.lu	32	1381	2.66	1375	2.67	1379	2.67

Hardware Summary
Type of System:	Homogenous
Compute Node:	Linux Networx LS-1
Interconnect:	InfiniBand
File Server Node:	Linux Networx Evolocity 1
Total Compute Nodes:	8
Total Chips:	16
Total Cores:	32
Total Threads:	32
Total Memory:	64 GB
Base Ranks Run:	32
Minimum Peak Ranks:	--
Maximum Peak Ranks:	--

Software Summary
C Compiler:	QLogic PathScale C Compiler 3.0
C++ Compiler:	QLogic PathScale C++ Compiler 3.0
Fortran Compiler:	QLogic PathScale Fortran Compiler 3.0
Base Pointers:	64-bit
Peak Pointers:	Not Applicable
MPI Library:	Scali MPI Connect 5.5
Other MPI Info:	IB Gold VAPI
Pre-processors:	None
Other Software:	None

Node Description: Linux Networx LS-1

Hardware
Number of nodes:	8
Uses of the node:	compute
Vendor:	Linux Networx, Inc.
Model:	LS-1
CPU Name:	Intel Xeon 5160
CPU(s) orderable:	1-2 chips
Chips enabled:	2
Cores enabled:	4
Cores per chip:	2
Threads per core:	1
CPU Characteristics:	1333 Mhz FSB
CPU MHz:	3000
Primary Cache:	32 KB I + 32 KB D on chip per core
Secondary Cache:	4 MB I+D on chip per chip
L3 Cache:	None
Other Cache:	None
Memory:	8 GB (8 x 1GB DIMMs)
Disk Subsystem:	250GB SAS hard drive
Other Hardware:	None
Adapter:	Mellanox MHGA28-XTC
Number of Adapters:	1
Slot Type:	PCIe x8
Data Rate:	InfiniBand 4x DDR
Ports Used:	1
Interconnect Type:	InfiniBand

Software
Adapter:	Mellanox MHGA28-XTC
Adapter Driver:	IBGD 1.8.2
Adapter Firmware:	5.1.4
Operating System:	SLES9 SP3
Local File System:	Not applicable
Shared File System:	GPFS
System State:	multi-user
Other Software:	None

Node Description: Linux Networx Evolocity 1

Hardware
Number of nodes:	8
Uses of the node:	file server
Vendor:	Linux Networx, Inc.
Model:	Evolocity 1
CPU Name:	AMD Opteron 248
CPU(s) orderable:	1-2 chips
Chips enabled:	2
Cores enabled:	2
Cores per chip:	1
Threads per core:	1
CPU Characteristics:	--
CPU MHz:	2200
Primary Cache:	64 KB I + 64 KB D on chip per core
Secondary Cache:	1 MB I+D on chip per core
L3 Cache:	None
Other Cache:	None
Memory:	8 GB (8 x 1GB DIMMs)
Disk Subsystem:	18 TB SAN interconnected by FC2
Other Hardware:	--
Adapter:	Mellanox MHXL-CF128-T
Number of Adapters:	1
Slot Type:	PCI-X
Data Rate:	InfiniBand 4x SDR
Ports Used:	1
Interconnect Type:	InfiniBand

Software
Adapter:	Mellanox MHXL-CF128-T
Adapter Driver:	IBGD 1.8.2
Adapter Firmware:	3.5.0
Operating System:	SLES9 SP3
Local File System:	Not applicable
Shared File System:	GPFS
System State:	multi-user
Other Software:	--

Interconnect Description: InfiniBand

Hardware
Vendor:	QLogic
Model:	QLogic Silverstorm 9120 Fabric Director
Switch Model:	9120
Number of Switches:	1
Number of Ports:	144
Data Rate:	InfiniBand 4x SDR and InfiniBand 4x DDR
Firmware:	4.0.0.5.5
Topology:	Single switch (star)
Primary Use:	MPI and filesystem traffic

Submit Notes

Scali MPI Connect's mpirun wrapper has been
used to submit the jobs. Description of switches:
  -aff manual:0x1:0x2:0x4:0x8: instruct the launcher to
     bind rank N..N+3 to the cores corresponding to the
     masks 1,2,4, and 8 respectively on each node.
  -npn 4: launch 4 processes per node.
  -rsh rsh: use rsh as method to connect to nodes.
  -mstdin none: do not connect the processes' STDIN
     to anything.
  -q: quiet mode, no output from launcher.
  -machinefile: file selecting the hosts to run on.
  -net smp,ib: prioritized list of networks used
     for communication between processes.

General Notes

Scali, Inc has executed the benchmark on Linux Networx's
Solution Center. We are grateful for the support from
Linux Networx and in particular Justin Wood in order to
finalize the submissions.

Base Compiler Invocation

C benchmarks:

/opt/scali/bin/mpicc -ccl pathcc

C++ benchmarks:

126.lammps:

/opt/scali/bin/mpicc -ccl pathCC

Fortran benchmarks:

/opt/scali/bin/mpif77 -ccl pathf90

Benchmarks using both Fortran and C:

/opt/scali/bin/mpicc -ccl pathcc /opt/scali/bin/mpif77 -ccl pathf90

Base Portability Flags

104.milc:	-DSPEC_MPI_LP64
115.fds4:	-DSPEC_MPI_LC_TRAILING_DOUBLE_UNDERSCORE -DSPEC_MPI_LP64
121.pop2:	-DSPEC_MPI_DOUBLE_UNDERSCORE -DSPEC_MPI_LP64
122.tachyon:	-DSPEC_MPI_LP64
127.wrf2:	-DF2CSTYLE -DSPEC_MPI_DOUBLE_UNDERSCORE -DSPEC_MPI_LINUX -DSPEC_MPI_LP64
128.GAPgeofem:	-DSPEC_MPI_LP64
130.socorro:	-fno-second-underscore -DSPEC_MPI_LP64
132.zeusmp2:	-DSPEC_MPI_LP64