This is a trivial program that selects a series of numbers (1..n) to
fill an array with no repitition in a random order. 

In the first program (shuffle.c), uses an O(n^2) algorithm that stupidly
walks the array each time a new number is picked to see if the number
is already in the array. If it finds a repeat it tries again and keeps
repeating this step until a free number is found.

The second program (fast-shuffle.c) does a time/memory trade-off to
get an O(n) algorithm with an O(1) check time. This is done by 
creating a second array of equivilant size and tagging values as
used in order as they get put into the main array. The result? Each
time we pick a number, we know exactly where to check to see if it
has already been used. 

The Makefile provided compiles both programs -- no real magic there.
The programs have been compiled and tested under Linux although it 
should compile without hassle on any other OS including Windows. 

The timecards shell script allows you to run and collect data on
how much CPU time it takes to run both programs with various
numbers of cards (from 1000 to n-thousand). Simply run it with
a parameter of how many thousand to test with. The timecards.output
shows the output of "./timecards 50". Note that timecards.output 
will be overwritten each time you run timecards. If you want to
save the timecards.output file after each run, be sure to rename
it. results.gif is a graph of timecards.output. Note that this 
counts on the GNU time program being located at /usr/bin. You 
may need to change this program a little to make it work on your
system.

As you can see from results.gif, picking a good algorithm will
make a bigger difference than brute force processing speed. I
would have to run shuffle on a processor that is 100x faster to
even begin getting close to the performance of fast-shuffle. 

Enjoy.