Sat 22 Jun 2019 16:00 - 16:30 at 106C - Session 5 Chair(s): Lenore Mullin

We present ALPyNA, an automatic loop parallelization framework for Python, which analyzes data dependences within nested loops and dynamically generates CUDA kernels for GPU execution. The ALPyNA system applies classical dependence analysis techniques to discover and exploit potential parallelism. The skeletal structure of the dependence graph is determined statically; this is combined with type and bounds information discovered at runtime, to auto-generate high-performance kernels for offload to GPU. We demonstrate speedups of up to 1000x relative to the native CPython interpreter across four array-intensive numerical Python benchmarks. Performance improvement is related to iteration domain sizes and the complexity of the dependence graph. Nevertheless, this approach promises to bring the benefits of manycore parallelism to end-user developers.

Sat 22 Jun
Times are displayed in time zone: (GMT-07:00) Tijuana, Baja California change

16:00 - 17:30: ARRAY 2019 - Session 5 at 106C
Chair(s): Lenore MullinSUNY Albany, USA
ARRAY-2019-papers16:00 - 16:30
Dejice Jacob, Jeremy SingerUniversity of Glasgow
ARRAY-2019-papers16:30 - 17:00
Henrik BarthelsRWTH Aachen, Paolo BientinesiUmeƄ University
ARRAY-2019-papers17:00 - 17:30
Martin KristienUniversity of Edinburgh, UK, Bruno BodinYale-NUS College, Michel SteuwerUniversity of Glasgow, Christophe DubachUniversity of Edinburgh