The ALMA project
ALMA (Greek for ‘leap’) is an acronym for ALgorithm parallelization for Multicore Architectures. Driven by the technology restrictions in chip design, the end of Moore’s law and the quest for increasing computing performance, ALMA is a fundamental step forward in the necessary introduction of novel computing paradigms and methodologies.
The ALMA tool-chain will implement parallelization and optimization algorithms for a whole class of multi-cores. ALMA intends to show how various embedded reconfigurable multi-cores from Karlsruhe Institute of Technology and Recore Systems can be efficiently programmed using the same tool-chain and the same application code.
The project partners
The ALMA consortium brings together partners from industry and academia. The industry partners Recore Systems (The Netherlands) and Intracom Telecom (Greece) contribute their expertise in reconfigurable hardware technology for multi-core systems-on-chip software development tools and real world applications.
Five academic partners contribute their outstanding expertise in reconfigurable computing and compilation tools development: the project coordinator Karlsruhe Institute of Technology (KIT) (Germany), Université de Rennes I (France), University of Peloponnese (Greece), Technological Educational Institute of Mesolonghi (Greece) and the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation (Germany).
The focus: practical applications
To ascertain that programmers can directly apply the programming tool chains that the ALMA consortium develops in the course of the project, ALMA targets two cases in market domains with different, complementary requirements.
The first case is from the field of telecom, and intends to support the fast development of next generation of Point-to-Point / Point-to-Multipoint wireless communication systems. Multi-core architectures are a very good fit, since many performance-critical functions must be executed in parallel in order to meet real-time constraints. The call for short time-to-market at a minimum design effort requires high-quality models that produce optimized code, and release the designer from the code optimization burden through a set of tools that allow efficient optimization and parallelization of system-level models with minimum designer intervention.
The second application is in the area of multi-object detection and tracking. Multi-object tracking is often implemented on embedded systems with strong restrictions on power consumption and heating. The algorithms cope with a variable number of objects and must process data on each object within a small time frame. The computational load can vary heavily from image to image. For embedded systems offering low computational power, algorithms must be adapted and parallelized over multiple processors, often resulting in poor performance of the application. ALMA's tool supported, guided development, which hides the complexity of the architecture of the embedded system and the algorithm from the developer, should result in an enormous increase of the system performance and definitely to a reduction of time consuming implementation.
ALMA first results are in
Amsterdam, The Netherlands, January 19-21, 2015 - During the HiPEAC conference, the researchers in the ALMA consortium shared the first results on ALMA. View the poster....
This project is cofunded by the European Union under the 7th Framework Programme under grant agreement ICT- 287733. ALMA strengthens the position of the EU in the world market of multiprocessor-targeted software tool chains.