linux--perf

1、perf

       Perf 是用来进行软件性能分析的工具。通过它，应用程序可以利用 PMU，tracepoint 和内核中的特殊计数器来进行性能统计。它不但可以分析指定应用程序的性能问题 (per thread)，也可以用来分析内核的性能问题，当然也可以同时分析应用代码和内核，从而全面理解应用程序中的性能瓶颈。

2、perf安装

[root@VM_0_11_centos selinux]# yum -y install perf
Loaded plugins: fastestmirror, langpacks
Loading mirror speeds from cached hostfile
 * base: mirrors.aliyun.com

3、常用命令

1）perf list：查看当前系统支持的性能事件，可以列出所有能够触发 perf 采样点的事件。；

[root@VM_0_11_centos selinux]# perf list

List of pre-defined events (to be used in -e):

  alignment-faults                                   [Software event]
  bpf-output                                         [Software event]
  context-switches OR cs                             [Software event]
  cpu-clock                                          [Software event]
  cpu-migrations OR migrations                       [Software event]
  dummy                                              [Software event]
  emulation-faults                                   [Software event]
  major-faults                                       [Software event]
  minor-faults                                       [Software event]
  page-faults OR faults                              [Software event]
  task-clock                                         [Software event]

  msr/tsc/                                           [Kernel PMU event]

  rNNN                                               [Raw hardware event descrip
  cpu/t1=v1[,t2=v2,t3 ...]/modifier                  [Raw hardware event descrip
   (see 'man perf-list' on how to encode it)

  mem:<addr>[/len][:access]                          [Hardware breakpoint]

[root@VM_0_11_centos selinux]#

2）perf stat：对全局性能进行统计

//test.c 
#include <stdio.h>
#include <stdlib.h>
void longa() 
{ 
  int i,j; 
  for(i = 0; i < 1000000; i++) 
  j=i; //am I silly or crazy? I feel boring and desperate. 
} 
 
void foo2() 
{ 
  int i; 
  for(i=0 ; i < 10; i++) 
       longa(); 
} 
 
void foo1() 
{ 
  int i; 
  for(i = 0; i< 100; i++) 
     longa(); 
} 
 
int main(void) 
{ 
  foo1(); 
  foo2(); 
}

[root@VM_0_11_centos perf]# gcc -o test -g test.c
[root@VM_0_11_centos perf]# ll
total 16
-rwxr-xr-x 1 root root 9688 Feb 21 20:39 test
-rw-r--r-- 1 root root  356 Feb 21 20:39 test.c
[root@VM_0_11_centos perf]#

[root@VM_0_11_centos perf]# perf stat ./test

 Performance counter stats for './test':

            269.33 msec task-clock                #    0.982 CPUs utilized                                                                                                                      
                22      context-switches          #    0.082 K/sec                                                                                                                              
                 0      cpu-migrations            #    0.000 K/sec                                                                                                                              
               112      page-faults               #    0.416 K/sec                                                                                                                              
   <not supported>      cycles                                                                                                                                                                  
   <not supported>      instructions                                                                                                                                                            
   <not supported>      branches                                                                                                                                                                
   <not supported>      branch-misses                                                                                                                                                           

       0.274310192 seconds time elapsed

       0.269620000 seconds user
       0.000000000 seconds sys


[root@VM_0_11_centos perf]#
上面告诉我们，程序 test 是一个 CPU bound 型，因为 task-clock-msecs 接近 1。

3）perf top：可以实时查看当前系统进程函数占用率情况； 

#include <stdio.h>
#include <stdlib.h>

int main(void)
{
        long long i = 0;
        while(1) {
                i++;
        }
}

 

4） perf record：记录信息到perf.data（默认文件）；perf report：生成报告；

[root@VM_0_11_centos perf]# perf record – e cpu-clock ./test
Workload failed: No such file or directory
[root@VM_0_11_centos perf]# ll
total 28
-rw------- 1 root root 9920 Feb 21 20:52 perf.data
-rwxr-xr-x 1 root root 9688 Feb 21 20:51 test
-rw-r--r-- 1 root root  356 Feb 21 20:51 test.c
[root@VM_0_11_centos perf]# perf report