Course content

Optimizing algorithms for both single and multi-processors systems as well as accellerators such as GPUs. Choosing numerical algoriths, use of optimized libraries, compiler optimizations and program profiling. How to take advantage of PC-clusters and graphics cards for computationally large tasks which cannot be run on a single processor system will also be covered. Parallel programming assignments in, among others, MPI and CUDA are included in this course.

Learning outcome

Understanding parallel programing is becoming increasingly more important as PCs and desktops incorporate multicore and multichip multiprocessor technology. Knowledge: The goal of this course is hence: To give the students a good understanding of optimzing serial programs and algorithms within computational science. Skills: Develop the students programming skills for and future multi- and many-core processor systems.

Learning methods and activities

Lectures and programming/theory assignments.

The course will be given in English. All problem sets, including obligatory activity, work/reports counting towards your final grade, and exams will be given in English.

Compulsory assignments

• Excercises

Recommended previous knowledge

Operation systems and algorithms & data structures are recommended, but if you have a strong math background, that should offset any potential extra workload.

Course materials

Syllabus will be available at the start of the semester.

Credit reductions