You're search is not being benefited by the parallel execution, because your search is entirely too trivial. Every parallel program loses some efficiency because it has extra overhead to run it. That is what you're seeing here.

What you should do is up the ante here - make the array one that is on the heap and large - you can malloc it or just make it a global array for this test. (local arrays can't be this big) - say 1,000,000 random integers.

Now divide up the work so each parallel thread or process, sorts 200,000 of those integers. Do that test a few times with MPi and then repeat it with just a single process sorting the same random int's. (If you don't reseed the random seed generator, it will always give you the same random number sequence).

That would be a much better test.

One good way to get to know what's going on with j, is to add a print statement to the inner loop. Something that will give you a chance to study just what is going on:

And ALWAYS use [ code] tags around C code, on any forum - otherwise it becomes mangled by the forum software, and very hard to study.

printf("i is %d and j is %d\nPress Enter when ready for the next loop\n",i,j); 
getchar();

Add those two lines right inside the inner for loop, and run the program. Watch the values of i and j, and how they are now coordinated.

I promise you'll learn something good.

And welcome to D.I.C!

This is C, rather than C++, but if you study it, you'll see how the logic uses the index, and some subtraction, to count up the frequency of each letter.


You will need to save this with a dot c extension to compile it with the C compiler.

#include <stdio.h>
#include <string.h>

int main(void) {
   int i,len;
   int frequency[26];  
   char text[]={"Out in the West Texas town of El Paso,\nI fell in love with a Mexican girl.\n"};

   printf("%s\n",text);
   //get the length of text string
   len=strlen(text);
   printf("len: %d\n\n",len);
   
   for(i=0;i<26;i++) {   //set all frequency array elements to 0
      frequency[i]=0;
   }
   printf("\n\n");
   //get frequency of letters
   for(i=0;i<len;i++) {
      if(text[i]>='a' && text[i]<='z') {        //count up lowercase letters
         frequency[text[i]-'a']++;
         printf("frequency of %c: %d\n",text[i], frequency[text[i]-'a']);
         
      }else if(text[i]>='A' && text[i]<='Z') {    //counts up uppercase letters
         frequency[text[i]-'A']++;
         printf("frequency of %c: %d\n",text[i], frequency[text[i]-'A']);
      }
      if(i%8==0) {
         printf("Hit enter to continue\n");
         getchar();
      }

   }
   printf("\n\n");
   for(i=0;i<26;i+=2) {                  //show the frequencies of each letter
      printf("%c: %3d   %c: %3d \n",i+'a',frequency[i],i+1+'a',frequency[i+1]);
   }
   printf("\n");
   return 0;
}

Every time you enter some data, there is a newline generated as you hit Enter. With numbers, that's not a problem, because they're bigger. With a letter (char) however, that's a real problem, because a newline '\n' IS a char - AND it will drive scanf() crazy when it gets a char and a newline, instead of a number.

So some tricks to stop it:

1) With char entry using scanf() put a space before the % format specifier:
scanf(" %c", &myCharVariable); //note the space before the %

2) With char entry before a number entry, you can add this:

getchar(); //eats up one newline, without stopping your program

Put it right after the scanf() for the char.

Use those two tricks and you should be fine - but there are more, and more industrial strength tools for doing this, as well (like deleting ALL the input from the input buffer), but this is probably all you'll need.

Don't you think you got a LITTLE carried away with your development?

Get one function's basics working correctly, before you get working on the next function. Testing each, as you go. You'll have to back up now, by REM'ing out /* */ sections of code to find this error. The good thing is it probably won't take as long as it seems, and you probably won't make this rookie move, anytime soon.

No one knows your code as well as you do, and I don't think anyone can look at your code (cold), without running it, and clearly see the details here. Your time to practice debugging code, has arrived, imo. Borland has a good debugger, and remember, the bugs can maybe run, but they can't hide!

Good luck, and remember not to develop your programs like this.