Solving Inverse Problems on GPUs

Andrea Borsic

Thayer School, Dartmouth College.

[Ed: GPUs are Graphics Processing Units, and are probably the future of scientific computing!]

The talk will cover aspects of GPU computing and will use Inverse Problems as an application example. Performance gains and algorithm design in accelerating computations with GPUs varies between dense linear algebra and sparse linear algebra operations. The application example will allow to discuss how GPUs perform differently in these two domains, and how this relates to hardware architecture and algorithm implementation. Part of the talk will cover design considerations related to the implementation of an Algebraic Multigrid Sparse Solver on GPUs and current results from ongoing work in this area.

Back to ACMS schedule