High-Performance Computing Project 1

COSC 420 - High-Performance Computing
Project 1

1 Description
Working in groups of two or three, you will solve (at least) three problems from the problem database
found at projecteuler.net. There are hundreds(!) of of problems to choose from, so you must also adhere to
the following restrictions:
1. No two groups may submit solutions to the same problem. When you assign yourselves to a group on
MyClasses, change the name of your group to reflect the three problems you chose (by ID from the
Project Euler website), and email the course instructor with your selection.
2. Your problems should show a range of difficulty (as rated by the Project Euler webpage): you should
solve problems in at least the 10th, 40th, and 60th percentiles (mouse over the yellow bar in the
“Difficulty” column to see the percentile). Note that you need to make an account and be signed in to
see the difficulty ratings!
3. Choose problems that allow you to take advantage of parallelism. In your project description and
documentation, describe your algorithmic approach to the problem, and analyze the speedup and
efficiency of a parallelized solution.
4. Choose problems on diverse topics. That is, they should not all be focused on number theory, graph
theory, strings, etc. If you are in doubt, consult with the instructor to see if you’re selection is diverse
enough.
Your solutions must use parallelism and parallel programming. Include with your code sbatch
scripts to take advantage of the HPC cluster. This means that you must work to develop an algorithm that
has a definite theoretical speedup over the sequential version. In your documentation, make this explicit,
with an explanation of your approach. Document which parallel programming constructs you use; try to take
the most advantage of the MPI tooling that you can. Try using custom MPI types, various point-to-point
communication, collectives, and topologies to help your program be more efficient.
Also include details of how your solutions evolved as you developed them. What went wrong? What are
the inherent difficulties in solving the problems, and how did you overcome them? If the problems are “too
hard” for simple brute force, what analytical tools did you employ to reduce your computational burden?
Design and implement separate programs to solve each of the three problems (but they may of course
share code for common, useful utilities that you write yourselves). The solution to the problem should be
printed out to standard output in a way that it can be checked by submitting it to the projecteuler.net
site.
Record both the time and number of operations performed by your solution, as appropriate for the problem; count an operation as being any single arithmetic operation on one of the data elements (comparison,
addition, multiplication, assignment, etc.). For the sake of getting the fastest solution possible, you may use
a flag or command-line parameter to turn the benchmarking features on and off.
Include a README file to describe each problem, its solution, and other required information. Include a
Makefile to compile the code. Be sure to include full and thorough documentation.
1
2 Submission
Submit a .zip file called Project1[LastName].zip (with [LastName] replaced with your own last name)
containing your source code, Makefile, documentation, and testing output, then upload it to the course
MyClasses submission page. Your submission must include checkable output (i.e. can be entered into the
Project Euler page to see if the solution is correct).
3 Bonus
If any bonus are completed, be sure to note it in the README file and provide appropriate output to
demonstrate its correctness.
1. Currently, the most recent problems (and consequently some of the least-solved) include problems
716-725. Your group may submit a solution to any of these, in addition to the three above, for a bonus
of 10 points.
2. Problem #696 is currently the least-solved problem on the site, only solved by 99 users! Any number of
groups may submit a solution to this problem, for a bonus of 10 points, similar to the above. This and
the problems listed above will be available to submit all semester! But your solution must be original,
and must take advantage of the HPC cluster.
3. For five bonus points, use GitHub, GitLab, BitBucket, or another public source-control platform to
host your group’s code and organize your joint efforts. Be sure to document that you did this in your
README.
2