Compiler and Runtime Based Parallelization and Optimization for GPUs

Student: Guray Ozen (Barcelona Supercomputing Center, Polytechnic University of Catalonia)
Advisor: Jesus Labarta (Barcelona Supercomputing Center, Polytechnic University of Catalonia), Eduard Ayguade (Barcelona Supercomputing Center, Polytechnic University of Catalonia)

Abstract: This thesis targets directive-based programming models to enhance their capability for GPU programming.  It introduces a new dialect model, which is a combination of OpenMP and OmpSs. The new model allows the use of multiple GPUs in conjunction with the heavily multithreaded capabilities in multicore processors automatically. The thesis also focuses on code transformation techniques and proposes the LazyNP method to support nested parallelism for irregular applications such as sparse matrix operations, graph and graphics algorithms. Finally, the thesis conducts a thorough exploration of loop scheduling methods on GPUs to find the advantage and disadvantages of them. It then proposes the concept of optimized dynamic loop scheduling.

The contributions of this thesis improve the programmability of GPUs. This has had an outstanding impact on the whole OpenMP and OpenACC language committee. Additionally, it includes contributions to widely used compilers such as Mercurium, Clang and PGI compilers.

Summary: pdf
Thesis Canvas: pdf



Back to Doctoral Showcase Archive Listing