Massive Parallelization of Real-World Automotive Real-Time Software by GPGPU

Konferenz: ARCS 2017 - 30th International Conference on Architecture of Computing Systems
03.04.2017 - 06.04.2017 in Vienna, Austria

Tagungsband: ARCS 2017

Seiten: 8Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Hartmann, Christoph; Margull, Ulrich (Ingolstadt University of Applied Sciences, Esplanade 10, 85049 Ingolstadt, Germany)
Mader, Ralph (Continental Automotive GmbH, Siemensstr. 12, 93055 Regensburg, Germany)
Michel, Lothar; Ebert, Christos (Audi AG, Auto-Union-Str. 1, 85057 Ingolstadt, Germany)

General Purpose Computing on GPU (GPGPU) is a method that allows to reduce an applications Worst Case Execution Time (WCET) through massively parallel execution and thereby increasing its real-time capability. The integration of GPUs is relevant in automotive engineering, especially in advanced driver assistance systems.We herewith share the knowledge obtained from massively parallelizing automotive real-world applications in the powertrain domain. GPGPU in this sector is up to now an insufficiently explored field in literature. For this purpose we parallelized two CPU-intense real-time applications, achieving a WCET speedup of up to 6.7 under realistic circumstances. Furthermore, we present several powertrain related key issues regarding massively parallel approaches, identified through our investigations.