Lab 8
                             
1. Execute the instruction  "more /proc/cpuinfo" to discover things about your processor 
  such as the clock speed.
 
2. Read about the Intel rdtsc (ReaD Time Stamp Counter) instruction on wikipedia.org

2. Make sure you understand how "longlong" ints are handled on our 32 bit Linux machines.
   Compile and run the following program and note the format specifier "%lu".

   cis lclient19:~/class/oct21>more onefile.c
   #include <stdio.h>
   unsigned long long rdtsc();
   void main() {
      unsigned long long mylong;
      mylong = rdtsc();
      printf("Unsigned long long is %llu\n", mylong);
   }
   unsigned long long rdtsc() {
      return 123456789101112ull;
   }

3. Separate onefile.c into main.c and rdtsc.c, create rdtsc.s, and make sure you
   understand the listing.

cis lclient01:~/2012/rdtsc>more rdtsc.s
           .file   "rdtsc.c"
           .text
   .globl rdtsc
           .type   rdtsc, @function
   rdtsc:
           pushl   %ebp
           movl    %esp, %ebp
           movl    $ 2045822408, %eax
           movl    $28744, %edx
           popl    %ebp
           ret
           .size   rdtsc, . rdtsc
           .ident  "GCC: (GNU) 4.4.5 20101112 (Red Hat 4.4.5 2)"
           .section        .note.GNU stack,"",@progbits
cis lclient01:~/2012/rdtsc>

4. Strip rdtsc.s of the unnecessary instructions and insert the rdtsc instruction. Where does
   rdtsc leave the 64 bit counter? Where does a C function return a 64 bit unsigned integer?
   
5. Compile and execute the following main program with your new rdtsc.s function. The output
   should look something like the following:

   cis lclient01:~/2012/rdtsc>more testrdtsc.c
   #include <stdio.h>
   unsigned long long rdtsc();
   void main() {
      unsigned long long a, b, c, d;
      a = rdtsc();
      b = rdtsc();
      c = rdtsc();
      d = rdtsc();
      printf("b a = %llu\n", b a);
      printf("c b = %llu\n", c b);
      printf("d c = %llu\n", d c);
   }
   cis lclient01:~/2012/rdtsc>gcc testrdtsc.c rdtsc.s
   cis lclient01:~/2012/rdtsc>./a.out
   b a = 104
   c b = 112
   d c = 104
   cis lclient01:~/2012/rdtsc>

    
6. Estimate the number of clock cycles needed to execute the rdtsc instruction. Describe in
   writing the way you obtained your estimate. A good answer would use several ways. Show
   the output from your test program. (Hints   a function with 10 rdtsc instructions or a 
   function with no rdtsc instructions?) How many clock cycles are required to call and
   return from a function.
   
For the remaining questions, use the following main program. It is generally a good idea to
call the code you are timing at least two or three times. In many cases, the first measurement
is larger as you fill caches of various sorts (e.g. physical memory, level 2 cache, level 1 cache).
Show the run times and try to explain them.

   #include <stdio.h>
   unsigned long long rdtsc();
   void test();
   void main() {
      unsigned long long a, b, c, d;
      a = rdtsc();
      test();
      b = rdtsc();
      test();
      c = rdtsc();
      test();
      d = rdtsc();
      printf("b a = %llu\n", b a);
      printf("c b = %llu\n", c b);
      printf("d c = %llu\n", d c);
   }
   
7. How long does it take to print a constant string?

   #include <stdio.h>
   void test() {
      printf("Hello World\n");
   }
   

8. How long does it take to set the elements of a 4 by 4 matrix.

   #define SIZE 4
   int matrix[SIZE][SIZE];
   void test() {
      int i, j;
      for (i = 0; i < SIZE; i++) {
         for (j = 0; j < SIZE; j++) 
            matrix[i][j] = i + j;
      }
   }
   

9. Repeat part 8 setting SIZE equal to 16, 64, 256, and 1024.

10. Repeat part 9 using optimization level 3 (gcc  o3 main.c test.c rdtsc.s).

11. Repeat parts 8, 9, and 10 using a column scan (rather and a row scan) of matrix, e.g.

   #define SIZE 4
   int matrix[SIZE][SIZE];
   void test() {
      int i, j;
      for (i = 0; i < SIZE; i++) {
         for (j = 0; j < SIZE; j++) 
            matrix[j][i] = i + j;
      }
   }
   
Note that running the code is trivial. The grading is determened by the insight shown by 
your answers.