Deprecating the Program Counter - Implicitly Parallel Programming for the Multi- and Many-Core Future
- Date
- Dec 1, 2015
- Time
- 2:30 PM - 3:30 PM
- Speaker
- Sebastian Ertel
- Affiliation
- Institut für Technische Informatik, Professur Compilerbau
- Language
- en
- Main Topic
- Informatik
- Other Topics
- Informatik
- Description
- The rise of multi- and many-core architectures in the last decade is bound to continue. For example, researchers at leading chip manufacturers envision multi-processor systems with thousands of cores in the near future. In contrast, the productivity in developing programs that make use of all these cores is decreasing. This is primarily due to the fact that most widely-used programming languages were originally designed to execute on a single-processor machine. Concurrent programming models, such as threads and locks, were added to these languages to utilize multiple cores at once but require expert knowledge to avoid new hazards such as deadlocks and data races. In the first part of this status talk, we will review prevalent and novel programming models, languages, compilers and runtime systems to evaluate their potential to close this productivity gap. We start with the most familiar imperative languages, such as C/C++ and Java, where deriving a parallel execution is known to be challenging due to prevalence of shared state. We proceeds towards a more functional language design where the implicit extraction of parallelism from a program is known to be easy due to the absence of state. In the second part, we will introduce a new programming model and language that allows to exploit the best of both worlds: stateful computations and implicit parallel execution. The remainder of the presentation will investigate the potential of the proposed approach to address the productivity gap for modern multi-processor systems and conclude with an outlook onto future work in the context of my dissertation. Diese Veranstaltung wird unterstützt von Professur Compilerbau.
Last modified: Dec 1, 2015, 8:51:35 AM
Location
TUD Andreas-Pfitzmann-Bau (Computer Science) (APB 1004 (Ratssaal))Nöthnitzer Straße4601069Dresden
- Homepage
- https://navigator.tu-dresden.de/etplan/apb/00
Organizer
TUD InformatikNöthnitzer Straße4601069Dresden
- Phone
- +49 (0) 351 463-38465
- Fax
- +49 (0) 351 463-38221
- Homepage
- http://www.inf.tu-dresden.de
Legend
- Biology
- Chemistry
- Civil Eng., Architecture
- Computer Science
- Economics
- Electrical and Computer Eng.
- Environmental Sciences
- for Pupils
- Law
- Linguistics, Literature and Culture
- Materials
- Mathematics
- Mechanical Engineering
- Medicine
- Physics
- Psychology
- Society, Philosophy, Education
- Spin-off/Transfer
- Traffic
- Training
- Welcome
