Linux性能测试工具

在Linux Benchmark Suite Homepage网站上列举了诸多Linux性能测试工具，包括CPU/RAM/ROM/Cache/net等性能测试。

lmbench

测试工具

这个工具集中包含以下几种测试工具，我们主要使用到bw_mem工具进行Memory read和write测试。

1. Bandwidth benchmarks(带宽测试)

• Cached file read
• Memory copy (bcopy)
• Memory read
• Memory write
• Pipe
• TCP

2. Latency benchmarks(延时测试)

• Context switching.
• Networking: connection establishment, pipe, TCP, UDP, and RPC hot potato
• File system creates and deletes.
• Process creation.
• Signal handling
• System call overhead
• Memory read latency

3. Miscellanious

• Processor clock rate calculation

交叉编译

1. 从 How do I get LMbench? 或者从github下载源码；
2. 设置交叉工具链和编译FLAG
   修改src/Makefile中的CC和EXFLAGS为：

CC=/home/imx8x/prebuilt/toolchains/aarch64-imx8x-linux/bin/aarch64-poky-linux-gcc
EXFLAGS=-static -march=armv8-a -mfpu=neon -mfloat-abi=hard -mtune=cortex-a35 -funroll-loops

3. 编译
   使用make OS=arm-linux build编译。
   如果出现如下错误：

cd src && make
make[1]: Entering directory '/home/xxx/work2/util/lmbench3/src'
make[2]: Entering directory '/home/xxx/work2/util/lmbench3/src'
make[2]: *** No rule to make target '../SCCS/s.ChangeSet', needed by 'bk.ver'.  Stop.
make[2]: Leaving directory '/home/xxx/work2/util/lmbench3/src'
Makefile:117: recipe for target 'lmbench' failed
make[1]: *** [lmbench] Error 2
make[1]: Leaving directory '/home/xxx/work2/util/lmbench3/src'
Makefile:20: recipe for target 'build' failed
make: *** [build] Error 2

解决办法：

mkdir SCCS
cd SCCS
touch s.ChangeSet

上述编译好了之后，就可以在/bin/arm-linux/目录下生成可执行文件，关于各个可执行文件的用法，可以简要的参照：
lmbench1.0 manual pages
lmbench

memory性能测试
将bw_mem文件拷贝到板子上运行如下命令：

@android:/var # ./bw_mem 256M wr                                              
268.44 573.30
@android:/var # ./bw_mem 256M fwr                                             
268.44 3034.69
@android:/var # ./bw_mem 256M rd                                              
268.44 896.46
@android:/var # ./bw_mem 256M frd                                             
268.44 867.99

各个参数含义如下：

/*
* rd - 4 byte read, 32 byte stride
* wr - 4 byte write, 32 byte stride
* rdwr - 4 byte read followed by 4 byte write to same place, 32 byte stride
* cp - 4 byte read then 4 byte write to different place, 32 byte stride
* fwr - write every 4 byte word
* frd - read every 4 byte word
* fcp - copy every 4 byte word
* All tests do 512 byte chunks in a loop.                        
* XXX - do a 64bit version of this.
*/

输出结果为megabytes, megabytes_per_second。

参考资料
Lmbench-Ti
LMbench - Tools for Performance Analysis

STREAM

官方的说法是：

The STREAM benchmark is a simple synthetic benchmark program that measures sustainable memory bandwidth (in MB/s) and the corresponding computation rate for simple vector kernels.

翻译过来就是：
STREAM benchmark是一个简单的综合基准测试程序，用于测量可持续存储器带宽（以MB/s为单位）和简单矢量内核的相应计算速率。

交叉编译
1、下载源码

mkdir STREAM
cd STREAM
wget -r -R 'index*' -np -nH --cut-dirs=3 http://www.cs.virginia.edu/stream/FTP/Code/

2、指定交叉工具链：

CC=/home/xxx/work2/xxx/imx8x/prebuilt/toolchains/aarch64-imx8x-linux/bin/aarch64-poky-linux-gcc

在编译的时候，需要指定一些参数，这些参数对结果影响很大，建议直接查看stream.c源文件。

3、编译c语言版本的工具make stream_c.exe，生成stream_c.exe可执行文件。

4、执行结果为：

@android:/var # ./stream_c.exe                                                
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 134214 microseconds.
   (= 134214 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function    Best Rate MB/s  Avg time     Min time     Max time
Copy:            1731.3     0.106711     0.092416     0.125942
Scale:           1294.3     0.140618     0.123622     0.164197
Add:             1107.7     0.232000     0.216672     0.263167
Triad:            991.0     0.252958     0.242168     0.276101
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------

目前还不知道这个结果都是些什么，与上文的lmbench的bw_mem工具差异蛮的的，有待后续研究。

STREAM参考资料如下：
STREAM: Sustainable Memory Bandwidth in High Performance Computers
What is STREAM
stream gitlab