The program count allows to determine of the execution time for specific
operations.

There are one input parameters which you have to enter after starting the
program:

  ict  selects the operation (check the program)

The parameter n in the program can be used to determine the number of
executions. Currently it is set to n=10000 which yield 10^8 executions
of the respective operation

  n    is the number of outer loops (the inner loop parameter is set to 10000
       such that a choice of n=10000 implies 10^8 executions of the
       operations specified in the loop. 

For proper evaluation run the program with different compiler
optimizations, e.g., with the default optimization or with -O1 -O2 etc.

Modern compilers may show unexpected behavior. For instance in some cases
where the operations within a do loop do not depend on the loop index (and
are explicit) the compiler may execute the operation only once giving a
timing result basically equivalent to an empty loop.
Also compare as far as possible the actual result of the execution for
different optimization. The results may not be identical in which case
either the result from the default or from the optimized code should be 
rejected (depending on which one is correct)

On linux systems you can time the result with the command:

how:count> time ./count
     value for ict 

1 10000
      ict=      1
  1.10000002
0.214u 0.005s 0:07.46 2.8%      0+0k 0+0io 130pf+0w

Here the last line is created by the time command. the first entry is the
time for the program execution in seconds.