Nowadays, HPC systems frequently emerge as clusters of commodity processors  with attached accelerators. Moving from tedious low-level accelerator programming to increased development productivity, the directive-based programming models OpenACC and OpenMP are promising candidates. While OpenACC was completed about two years ago, OpenMP just recently added support for accelerator programming. We are happy to invite you to two talks from James Beyer from Cray, who was instrumental in creating OpenACC and is still driving the development of OpenACC as well as OpenMP for Accelerators.


Zeit:  Montag, 12.05.2014, 10.00 Uhr
Ort:   Seminarraum 2, Kopernikusstraße 6 (IT Center)
Thema: OpenACC 2.0++ and OpenMP 4.0++: The future of directive based programing on “accelerators”
Referent: Dr. James C. Beyer (Cray)
Slides: PDF
Abstract: OpenACC 2.0 addressed many important problems.  However, mistakes were made and features remained to be finished.  In most aspects OpenMP 4.0 followed along adopting features from OpenACC, however, there is still a large gap in capabilities.  In this talk I will present some of the important differences between the two directive sets and then discuss the features I believe OpenACC and OpenMP will be adding in the near future.  The goal of the OpenACC features is to provide a complete solution for all languages.  The goal of the OpenMP features is twofold, first to close the gap between OpenACC and OpenMP and second to add capabilities that were missed in the latest release.


Zeit:  Dienstag, 13.05.2014, 10.00 Uhr
Ort:   Seminarraum 2, Kopernikusstraße 6 (IT Center)
Thema: Transferring User-Defined Types in OpenACC
Referent: Dr. James C. Beyer (Cray)
Slides: PDF
Abstract:  A preeminent problem blocking the adoption of OpenACC by many programmers is support for user-defined types: classes and structures in C/C++ and derived types in Fortran. This problem is particularly challenging for data structures that involve pointer indirection, since transferring these data structures between the disjoint host and accelerator memories found on most modern accelerators requires deep-copy semantics. This talk will look at the mechanisms available in OpenACC 2.0 to allow the programmer to design transfer routines for OpenACC programs. Once these mechanisms have been explored, a new directive-based solution will be presented. Code examples will be used to compare the current state-of-the-art and the new proposed solution.

 

OpenACC and the Cray Compilation Environment (Webinar Slides, PDF)

 

  • Keine Stichwörter