Linpack and Flops

2013-06-09

Linpack现在在国际上已经成为最流行的用于测试高性能计算机FLOPS的benchmark.

LINPACK was designed for supercomputers in use in the 1970s and early 1980s. LINPACK has been largely superceded by LAPACK
which has been designed to run efficiently on shared-memory, vector supercomputers.

LINPACK is a collection of Fortran subroutines that analyze and solve linear equations and linear least-squares probles. The
package solves linear systems whose matrices are general, banded,symmetric indefinite, symmetric positive definite, triangular,and tridiagonal square. In addition, the package computes the QR and singular value decompositions of rectangular matrices and applies them to least-squares problems. LINPACK uses column-oriented algorithms to increase efficiency by preserving locality of reference.

HPL - A Portable Implementation of the High-Performance Linpack Benchmark for Distributed-Memory Computers

1、intel提供的二进制：
wget -c http://registrationcenter-download.intel.com/akdlm/irc_nas/3058/l_lpk_p_11.0.3.008.tgz
源文件
wget -c http://www.netlib.org/benchmark/hpl/hpl-2.1.tar.gz

2、脚本实例，8核心xeon测了15分钟
#!/bin/bash

#export OMP_NUM_THREADS=2

echo "This is a SAMPLE run script. Change it to reflect the correct number"
echo "of CPUs/threads, problem input files, etc.."

datedate > lin_xeon32.txt

./xlinpack_xeon32 lininput_xeon32 >> lin_xeon32.txt

date >> lin_xeon32.txt
echo -n "Done: "
date

3、测试结果实例
Sun Jun 9 12:44:03 CST 2013
Intel(R) Optimized LINPACK Benchmark data

Current date/time: Sun Jun 9 12:44:03 2013

CPU frequency: 1.994 GHz
Number of CPUs: 2
Number of cores: 8
Number of threads: 8

Parameters are set to:

Number of tests: 9
Number of equations to solve (problem size) : 15000 14000 13000 12000 11000 10000 8000 6000 1000
Leading dimension of array : 15000 14008 13000 12008 11000 10008 8008 6008 1000
Number of trials to run : 1 2 2 2 2 2 2 3 4
Data alignment value (in Kbytes) : 4 4 4 4 4 4 4 4 4

Maximum memory requested that can be used=1800304096, at the size=15000

=================== Timing linear equation system solver ===================

Size LDA Align. Time(s) GFlops Residual Residual(norm) Check
15000 15000 4 108.253 20.7889 2.382393e-10 3.752309e-02 pass
14000 14008 4 88.835 20.5970 1.636866e-10 2.956007e-02 pass
14000 14008 4 88.677 20.6335 1.636866e-10 2.956007e-02 pass
13000 13000 4 65.647 22.3165 nan nan pass
13000 13000 4 65.166 22.4811 nan nan pass
12000 12008 4 51.527 22.3627 nan nan pass
12000 12008 4 51.664 22.3037 nan nan pass
11000 11000 4 40.907 21.6976 nan nan pass
11000 11000 4 41.213 21.5361 nan nan pass
10000 10008 4 35.834 18.6101 nan nan pass
10000 10008 4 36.096 18.4746 nan nan pass
8000 8008 4 18.612 18.3466 nan nan pass
8000 8008 4 18.475 18.4824 nan nan pass
6000 6008 4 8.102 17.7822 nan nan pass
6000 6008 4 8.078 17.8361 nan nan pass
6000 6008 4 8.046 17.9064 nan nan pass
1000 1000 4 0.076 8.7449 nan nan pass
1000 1000 4 0.077 8.6576 nan nan pass
1000 1000 4 0.077 8.6970 nan nan pass
1000 1000 4 0.077 8.7218 nan nan pass

Performance Summary (GFlops)

Size LDA Align. Average Maximal
15000 15000 4 20.7889 20.7889
14000 14008 4 20.6153 20.6335
13000 13000 4 22.3988 22.4811
12000 12008 4 22.3332 22.3627
11000 11000 4 21.6168 21.6976
10000 10008 4 18.5423 18.6101
8000 8008 4 18.4145 18.4824
6000 6008 4 17.8416 17.9064
1000 1000 4 8.7053 8.7449

Residual checks PASSED

End of tests

Sun Jun 9 13:00:04 CST 2013

4、参考网址
http://www.netlib.org/linpack/readme
http://www.top500.org/project/linpack/
http://www.netlib.org/benchmark/hpl/
http://www.netlib.org/utk/people/JackDongarra/faq-linpack.html
http://software.intel.com/en-us/articles/intel-math-kernel-library-linpack-download
http://www.codesky.net/article/201204/172385.html

分类：Linux | 标签： 、 |

相关日志

评论被关闭！