Time measuring in C and linux

Question

I want to see how much is taken by the C program, so I wrote:

  #include<stdio.h>
  #include<stdlib.h>
  #include"memory.h"
  #include"memory_debug.h"
  #include<sys/times.h>
  #include<unistd.h>

  int (*deallocate_ptr)(memContainer *,void*);

  void (*merge_ptr)(node *);

  void* (*allocate_ptr)(memContainer *,unsigned long size);

  memContainer* (*init_ptr)(unsigned long );

  diagStruct* (*diagnose_ptr)(memContainer *);

  void (*finalize_ptr)(memContainer *);

  void (*printNode_ptr)(node *n);

  void (*printContainer_ptr)(memContainer *c);

  void info(memContainer *c)
  {
    struct tms *t;
    t=malloc(sizeof(struct tms));
    times(t);
    printf("user : %d\nsystem : %d\n %d",t->tms_utime,(int)t->tms_stime);
    diagnose_ptr(c);
    printf("\n");
    return ;
  }

but when I invoke this function I get 0 user time and 0 system time, even if I write:

for (i=0;i<100000;++i)
    for (j=0;j<10;++j)
    {}
info(c);

what am I doing wrong?

When you run `time a.out`, does that display the times you expect? — Douglas, Mar 24 '12 at 23:35
I suppose you don't care since your program exits after info, but if it does not, you have a memory leak. — ShinTakezou, Mar 24 '12 at 23:46

score 3 · Answer 1 · answered Mar 24 '12 at 23:34

3

The compiler probably optimizes away your for loops since they do nothing. Try incrementing a volatile variable.

If you only want to know the time, try running time ./app and it will print the cputime, wall clock time etc of the executed app.

answered Mar 24 '12 at 23:34

Erik Ekman

2,051
12
13

the thing is, i need to do those measures in the application, because the time needs to be compared in different points in code. – Andna Mar 24 '12 at 23:50

score 2 · Answer 2 · edited May 23 '17 at 12:11

The code could simply write a volatile variable at the start, put your 'work' in a function (in a separate file), then read the volatile after the 'work' and print something involving the volatile.

Or do some simple calculation with a part of the calculation buried in a function, or using a function return.

What platform (Operating system & Compiler) are you using?

I don't know what platform you are running on, but there are a few useful questions on stackoverflow about higher precision system clocks. High precision timing in userspace in Linux has several useful links and references.

Timing Methods in C++ Under Linux looked useful.

score 1 · Accepted Answer · answered Mar 25 '12 at 00:56

1

The below demo program outputs nonzero times:

#include<stdio.h>
#include<stdlib.h>
#include"memory.h"
#include<sys/times.h>
#include<unistd.h>
#include <iostream>
using namespace std;

int main()
{
    int x = 0;
    for (int i = 0; i < 1 << 30; i++)
        x++;

    struct tms t;
    times(&t);
    cout << t.tms_utime << endl;
    cout << t.tms_stime << endl;
    return x;
}

Output:

275
1

answered Mar 25 '12 at 00:56

Andrew Tomazos

66,139
40
186
319

It works, it seems that my prog is just running so short that values are so small. – Andna Mar 25 '12 at 07:58

Time measuring in C and linux

3 Answers3