Loading…
Back To Schedule
Friday, November 4 • 4:45pm - 5:30pm
Polly Loop Optimization and Accelerator Compilation

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Loop Optimization is important for high-performance computing but even
more for fast image processing, machine learning, and accelerator
programming. Over the last year the Polly loop optimization framework
has significantly evolved, with new support for data-layout
transformations, optimization of dense linear algebra kernels, and fully
automatic accelerator mapping support. Many of  these transformations
have been contributed by developers all over the world, including three
summer of code students. This BoF serves as a place for core developers
to gather, to discuss the current status of Polly, and to shape the 2016
development agenda of Polly. Hot topics are likely new automatic GPGPU
code generation facilities, recent improvements on correctness and
compile time, the new outer loop vectorization, and the recent addition
of @polly support in Julia. The Polly code base also relies heavily on
scalar evolution, value range analysis, and can serve as a basis
for performance, memory footprint, and data transfer modeling. We invite
Polly developers and all other interested developers.


Speakers
TG

Tobias Grosser

ETH Zurich


Friday November 4, 2016 4:45pm - 5:30pm PDT
3 - BoF (Rm LL21CD)