16.01.2015 Views

CS2013-final-report

CS2013-final-report

CS2013-final-report

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

• Grama, A., Gupta, A., Kumar V. (2003, 2nd edn.). Introduction to Parallel Computing. Harlow, England:<br />

Addison-Wesley.<br />

• To develop parallel programs C/C++ is used, MS Visual Studio, Intel Parallel Studio, cluster under MS HPC<br />

Server 2008.<br />

Why do you teach the course this way<br />

The goal of the course is to study the mathematical models, methods and technologies of parallel programming for<br />

multiprocessor systems. Learning the course is sufficient for a successful start to practice in the area of parallel<br />

programming.<br />

The course has a good reputation among students. The students that are studied this course were the winner in the<br />

track “Max Linpack performance” of the Cluster Student Competition at Supercomputing 2011. The course was a<br />

part of the proposal that was the winner of the contest “Curriculum Best Practices: Parallelism and Concurrence”<br />

of European Association “Informatics Europe” (2011) – see. http://www.informaticseurope.org/services/curriculum-award/105-curriculum-award-2011.html<br />

Body of Knowledge coverage<br />

KA Knowledge Unit Topics Covered Hours<br />

PD<br />

Parallelism<br />

Fundamentals<br />

Overview: Information dependencies analysis, decomposition,<br />

synchronization, message passing<br />

2<br />

PD<br />

Parallel<br />

Decomposition<br />

Data and task decomposition, Domain specific (geometric)<br />

decomposition 0.5<br />

PD<br />

Communication and<br />

Coordination<br />

Basic communication operations, Evaluating communication overhead<br />

2<br />

PD<br />

Parallel Algorithms,<br />

Analysis, and<br />

Programming<br />

Characteristics of parallel efficiency,<br />

Spectrum of parallel algorithms (Matrix calculation, System of linear<br />

equations, Sorting, Graph algorithms, Solving PDE, Optimization)<br />

OpenMP, MPI, MS VS, Intel Parallel Studio 10<br />

PD Parallel Architecture Structure of the MIMD class of Flynn’s taxonomy,<br />

SMP, Clusters, NUMA 1.5<br />

PD Parallel Performance Characteristics of parallel efficiency: speed-up, cost, scalability,<br />

isoefficiency. Theoretical prediction of parallel performance and<br />

comparing with computational results 2<br />

PD<br />

Formal Models and<br />

Semantics<br />

Information dependencies analysis, Evaluating characteristics of<br />

parallel efficiency (superlinear and linear speed-up, max possible<br />

speed up for a given problem, speed-up of a given parallel algorithm),<br />

Equivalent transformation of parallel programs 2<br />

Additional topics<br />

Isoefficiency,<br />

Redundancy of parallel computations,<br />

Classic illustrative parallel problems (readers-writers, dining philosophers, sleeping barbarian, etc.),<br />

Tools for development of parallel programs,<br />

Formal models based on Information dependencies analysis<br />

- 343 -

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!