SYSINFO系统调用返回Linux上的错误负载平均值 [英] sysinfo system call returns wrong load average values on linux

问题描述

我有一些打印系统统计的C程序。

I have a C program that prints some system statistics.

#include <sys/sysinfo.h>
#include <stdio.h>

int main() {
  int days, hours, mins;
  struct sysinfo sys_info;

  if(sysinfo(&sys_info) != 0)
    perror("sysinfo");

  // Uptime
  days = sys_info.uptime / 86400;
  hours = (sys_info.uptime / 3600) - (days * 24);
  mins = (sys_info.uptime / 60) - (days * 1440) - (hours * 60);

  printf("Uptime: %ddays, %dhours, %dminutes, %ldseconds\n",
                      days, hours, mins, sys_info.uptime % 60);

  // Load Averages for 1,5 and 15 minutes
  printf("Load Avgs: 1min(%ld) 5min(%ld) 15min(%ld)\n",
          sys_info.loads[0], sys_info.loads[1], sys_info.loads[2]);

  printf("Total Ram: %lluk\tFree: %lluk\n",
                sys_info.totalram *(unsigned long long)sys_info.mem_unit / 1024,
                sys_info.freeram *(unsigned long long)sys_info.mem_unit/ 1024);


  // Number of processes currently running.
  printf("Number of processes: %d\n", sys_info.procs);

  return 0;
}

我的问题是，平均负载值比的/ proc / loadavg

My problem is that the load average values are different than /proc/loadavg

./a.out 
Uptime: 1days, 4hours, 1minutes, 16seconds
Load Avgs: 1min(13248) 5min(14880) 15min(11840)
Total Ram: 2052956k Free: 188104k
Number of processes: 265

为什么呢？我打印13248，但顶或猫的/ proc / loadavg给出0.24。有什么问题？

Why? I am printing 13248 but "top" or "cat /proc/loadavg" gives 0.24. What is the problem?

推荐答案

该调用不返回错误的值。注意，类型为整数，因此它不能被用来返回由运行时间打印的浮点数。

The call doesn't return the "wrong" values. Note that the type is integer, so it can't be intended to return the floating-point numbers printed by uptime.

请参阅this线程在如何跨preT返回值，以及如何转换为更熟悉的花车进行调查。

See this thread for an investigation in how to interpret the returned values, and how to convert to the more familiar floats.

我的猜测（读取链接之前）是，它只是重新presented由65,535放大（2 ¹⁶ ），这似乎是他们发现了什么，太。因此，通过65536.0分裂，或者是方式更清洁的（如在评论中指出）使用SI_LOAD_SHIFT不变，除以（浮点）（1 LT;＆LT; SI_LOAD_SHIFT）。

My guess (before reading the link) was that it's just represented as scaled up by 65,535 (2¹⁶), and that seems to be what they found, too. So divide by 65536.0, or to be way cleaner (as pointed out in a comment) use the SI_LOAD_SHIFT constant and divide by (float)(1 << SI_LOAD_SHIFT).

这篇关于SYSINFO系统调用返回Linux上的错误负载平均值的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！