|Name:||OpenMP 4.0 & Beyond|
|Time:||Monday, June 23, 2014
04:40 pm - 05:00 pm
CCL - Congress Center Leipzig
|Speaker:||Christian Terboven, RWTH Aachen University|
|Abstract:||With the increasing prevalence of multicore processors, shared-memory programming models are essential. OpenMP is a popular, portable, widely supported and easy-to-use shared-memory model. The latest version 4.0 of the OpenMP API is a major revision: places and binding policies allow to define a common way to express thread affinity, SIMD pragmas extend the thread-parallel execution to exploit data-parallel capabilities of current microarchitectures, cancellation of OpenMP constructs and task dependencies allow to parallelize new classes of algorithms with OpenMP and possibly the biggest addition is the support for compute accelerator devices.
This talk briefly reviews the achievements of OpenMP 4.0 and then describes the challenges it faces for being the node-focused parallel programming paradigm on future architectures. As the work on OpenMP versions 4.1 and 5.0 has already begun, this talk gives an overview about the direction of the OpenMP Language Committee.