Overview of parallel computers, shared memory, message passing, MIMD and SIMD classifications. Understanding and use of message passing and synchronization facilities such as MPI. Study of parallel programming models such as master-slave, client-server, tast-farming, divide-and-conquer and pipelining. Performance analysis of parallel systems, execution time, time complexity, load balancing and scalability.

B524 involves a project. Credit not given for both B424 and B524.

Prerequisites: C243, M301.

Textbook: T. Rauber, G. Runger (2013): Parallel Programming for Multicore and Cluster Systems, 2nd edition, Springer.

Grading system:

• About 12 homework assignments: 20 points each
• About 12 quizzes: 5 points each
• 1 midterm exam: 50 points
• Final exam: 50 points
• B524 Graduate project: 50 points

Some ideas of projects for the graduate students.

Guidelines for assignments:

• The assignments will be posted on the course web page and on Canvas.
• The assignments are due at midnight of the due date.
• The programming assignments must be turned in on Canvas.
• No homework accepted after 1 week from the due date. A late homework loses 10% of the points per day, up to 50% of the points.
• Reasonable expectations concerning the program structure and clarity: functions should be commented and should not contain more than 20 lines of code. Multiple source files are expected when appropriate.

Programming Environment:

• OS: Linux, labs NS#207 and NS#209. Combinations will be provided by email. Some experiments will be performed on the Beowulf cluster in NS#207.
  IU computing server Karst: https://kb.iu.edu/d/bezu#access
• Compiler: g++. Examples of Makefile will be provided.
• Special libraries: MPI, pthread, openMP.
• Editor: according to preference: emacs, gedit, vi, pico.

Syllabus (Canvas)
Parallel Knoppix distribution 6.0
MPI resources
Pthreads documentation
OpenMP
Linux resources