Automatically Parallelizing Compilation Framework for Massively Parallel Architectures

Event Sponsor: 
Argonne Leadership Computing Facility Seminar
Start Date: 
Apr 7 2017 - 10:30am
Building 240/Room 4301
Argonne National Laboratory
Kavitha Madhu
Speaker(s) Title: 
Indian Institute of Science, Bangalore
Hal Finkel

Effective parallelism expression in applications for varied target architectures and runtime platforms is hard. This presentation covers a target-agnostic intermediate representation that expresses concurrency without implying it directly. The IR also supports data distribution patterns and a frontend that automatically generates parallel IR from sequential code. The usefulness of this abstraction is proven using performance evaluation metrics and also the abstraction's ease of use in static analysis tools such as WCET analyzers.

Miscellaneous Information: 

Conference room is located on the fourth floor. Go to the right after you get off of the elevator. The room will be on your left side.

Please click below to add this event to your calendar.