CS 732 -- Project 3

• Sequential Version
  • Read in a file name from the the keyboard
  • Open the file and read the first number -- this will tell you how many numbers will be in the data file.
  • Allocate space for your array
  • Fill it up (from the file)
  • Start the stopwatch
  • Sort
  • Stop the stopwatch
  • Print out the computation time.

• Parallel Version
  • Implement the Bucket Sort
  • Time the process with the data in place (assume it starts on the nodes, and ends up on the nodes)

• For both the Sequential and the Parallel Versions
  • Compute the computation time for arrays of various sizes (stop when they begin to swap and thrash -- this will be different for each algorithm)
  • Compute speedup and efficiency for various number of processors.

• Here is a short program to generate data files if you wish to use it. generator.c

• Note:
  • We will add parallel sorting versions to this when we cover them later in the course.
• Submit all the appropriate material to the isntructor via email.

Notes Previous semesters had problems with

• Do not do any work on the master before the clock starts.
  (not even calculating the max and the min)
  • The master is only to open data file,
  • Read how mnay numbers there are (N)
  • Tell all the slaves how many numbers they will be getting.
  • have a for loop to read in N/num_proc elements and send to the next slave
  • Start the clock.
  • Send message to slaves to start
    (they can't start until this point)
  • Wait for message from all slaves that they are done.
  • Stop the clock
  • Output Wall Clock time
    (not CPU usage time)

• The slave should:
  • Receive how many numbers it will get
  • Allocate space
  • Receive those numbers
  • Wait for message to start
  • Do Parallel Bucket sort with slaves
    Yes this involves some exchange between slaves
  • When finished send message to master

• There is also to be a writeup turned in.

• Code:
  • Undocumentede code will not be graded
  • Code that is not indented properly will not be graded