AM148: GPU Programming for Scientific Computations

This second course in scientific computing focuses on the use of parallel processing on GPUs with CUDA. Basic topics covered include the idea of parallelism and parallel architectures. The course then presents key parallel algorithms on GPUs such as scan, reduce, histogram and stencil, and compound algorithms. Applications to scientific computing are drawn from problems in linear algebra, curve fitting, FFTs, systems of ODEs and PDEs, and image processing. Finally, the course presents optimization strategies specific to GPUs. Basic knowledge of Unix, and C is assumed.

5 Credits

This class has not been taught recently.

Comments

Formerly AMS 0148

While the information on this web site is usually the most up to date, in the event of a discrepancy please contact your adviser to confirm which information is correct.