Explanation, please!

[Douglas C. Schmidt]

Hi,

   Since I posted my original question there has been a great deal of
abstract discussion about the relative merits of the loop unrolling
scheme.  The topic has piqued my curiousity, so I when ahead and
implemented a short test program, included below, to test Duff's
device against the ``ordinary for loop w/index variable'' technique.
See for yourself....   

After some quick testing I found that gcc 1.26 -O on a Sun 3 and a
Sequent Balance was pretty heavily in favor of the regular (non-Duff)
loop.  Your mileage may vary.  I realize that there may be other
tests, and if anyone has a better version, I'd like to see it!


Doug Schmidt
----------------------------------------
#include <sys/time.h>
#include <sys/resource.h>

double Start_Timer();
double Return_Elapsed_Time();

#define MAX_NUM 100000
int array1[MAX_NUM ];
int array2[MAX_NUM ];
int *A = array1, *B = array2;

main(int argc, char *argv[]) {
   double Elapsed_Time;
   int    Count = argc > 1 ? atoi(argv[1]) : MAX_NUM;
   int i;
   
   for (i = 0;i < Count ;i++) {
      array1[i] = i + 1;
      array2[i] = i;
   }
      
   printf("Starting Duff's device timing...\n");
   Start_Timer();

   {
      int n = (Count + 7) / 8;

      switch(Count % 8) {
         case 0:  do { *A++ = *B++;
         case 7:       *A++ = *B++;
         case 6:       *A++ = *B++;
         case 5:       *A++ = *B++;
         case 4:       *A++ = *B++;
         case 3:       *A++ = *B++;
         case 2:       *A++ = *B++;
         case 1:       *A++ = *B++;
                  } while (--n > 0);
      }
   }

   Elapsed_Time = Return_Elapsed_Time(0.0 );
   printf("Elapsed time = %.3f\n",Elapsed_Time);

   for (i = 0;i < Count ;i++) {
      if (array1[i] != array2[i]) {
         printf("Yow, problems at location %d!\n",i);
         break;
      }
   }

   for (i = 0;i < Count ;i++) {
      array1[i] = i + 1;
      array2[i] = i;
   }

   printf("Starting ordinary copy timing...\n");
   Start_Timer();

   for (i = 0;i < Count ;i++) {
      array1[i] = array2[i];
   }

   Elapsed_Time = Return_Elapsed_Time(0.0 );
   printf("Elapsed time = %.3f\n",Elapsed_Time);

   for (i = 0;i < Count ;i++) {
      if (array1[i] != array2[i]) { 
         printf("Yow, problems at location %d!\n",i);
         break;
      }
   }

}      


/* no such thing as "negative time"! */
#define  ERROR_VALUE -1.0   

static struct rusage Old_Time;
static struct rusage New_Time;
static int    Timer_Set = 0;

#ifdef __STDC__
double Start_Timer(void)
#else
double Start_Timer()
#endif
{
   Timer_Set = 1;
   getrusage(RUSAGE_SELF,&Old_Time);        /* set starting process time */
   return(Old_Time.ru_utime.tv_sec + (Old_Time.ru_utime.tv_usec / 1000000.0));
}

/* returns process time since Last_Time (if parameter is not DEFAULT_TIME, */
/* i.e., (double) 0.0 ),otherwise, if parameter == DEFAULT_TIME then       */
/* the time since the Old_Time was set is returned.  Returns ERROR_VALUE   */
/* if Start_Timer() is not called first */
#ifdef __STDC__
double Return_Elapsed_Time(double Last_Time)
#else
double Return_Elapsed_Time(Last_Time) 
double Last_Time;
#endif
{
   if (!Timer_Set) {
      return(ERROR_VALUE);
   }   
   else {
    /* get process time */
      getrusage(RUSAGE_SELF,&New_Time);
      if (Last_Time == 0.0) {
         return((New_Time.ru_utime.tv_sec - Old_Time.ru_utime.tv_sec) + 
               ((New_Time.ru_utime.tv_usec - Old_Time.ru_utime.tv_usec) 
                / 1000000.0));
      }
      else {
         return((New_Time.ru_utime.tv_sec + 
                (New_Time.ru_utime.tv_usec / 1000000.0)) - Last_Time);
      }
   }
}



--
schmidt at bonnie.ics.uci.edu (ARPA)
"If our behavior is strict, we do not need fun." -Zippy th' Pinhead
"If our behavior is struct, we do not need defun." -Anon