Why would using strchr before using strtok speed up runtime?

I have this code:

void
fill_array (unsigned int *iarray, char *string, int max)
{
  int ipos = 0;
  char *iholder;

  iholder = strtok (string, ",");
  while (iholder != NULL)
    {
      iarray[ipos++] = atoi (iholder);
      if (ipos == max)
    {
      return;
    }
      iholder = strtok (NULL, ",");
    }

}

It takes a string "1,2,3,4" for example, and enters the numbers into an array. I put this in a loop and got 3.3 seconds runtime.

With this code:

void
fill_array (unsigned int *iarray, char *string, int max)
{
  int ipos = 0;
  char *iholder;
  if (!strchr (string, ','))
    {
      iarray[0] = atoi (string);
      return;
    }
  ihol开发者_StackOverflow社区der = strtok (string, ",");
  while (iholder != NULL)
    {
      iarray[ipos++] = atoi (iholder);
      if (ipos == max)
    {
      return;
    }
      iholder = strtok (NULL, ",");
    }

}

It took about 1.4 seconds to execute.

The only difference is the strchr which I inserted just to see if it would run faster on single numbers, but it runs much faster on longer lists for some reason.

Can anyone explain why?

I am testing with this code:

int main ()
{
  unsigned int iarray[5];
  char str_test[] = "56,75,22,83";
  int i;
  struct timeval start;
  struct timeval end;

  gettimeofday (&start, NULL);


  for (i = 0; i < 10000000; i++)
    {
      fill_array (iarray, str_test, 5);
    }

  gettimeofday (&end, NULL);
  if (end.tv_usec - start.tv_usec < 0)
    {
      end.tv_usec += 1000000L;
      end.tv_sec -= 1;
    }

  printf ("Runtime: %ld s %03ld ms\n",
      end.tv_sec - start.tv_sec, (end.tv_usec - start.tv_usec) / 1000);

  return 0;
}

---

Presumably that first strchr is making your code drop out earlier so that it doesn't have to do as much processing?

---

Seems strange? Are you testing both codes on same machine? and how many times do you have taken the sample? and what is the code you are using to calculate time? please share that as well.