Success story: as Ampere project ends, the RESPECT project opens doors for next steps for critical embedded systems
The AMPERE project asserted its presence in the 2023 ISC High Performance Conference, one of the fields most important conferences, which seeks to foster the growth of the global HPC community of technology providers and users. In addition to sponsoring the 18th Workshop on Virtualization in High Performance Cloud Computing, the AMPERE project is being acknowledged as a source of technology transfer for the RESPECT (Reliable Heterogeneous Parallelism for Embedded Critical Systmes) project in its poster presentation.
RESPECT, coordinated by the Barcelona Supercomputing Center and comprised of business mentors BOSCH and AirBus, builds on the AMPERE- produced Task Dependency Graph for multi-criteria analysis. It expands on AMPERE's results to rise to technology readiness level 6 by consolidating the taskgraph framework, extending and testing features for resilience and consolidating and extending application domains (automotive and space).
RESPECT seeks to develop DePE, a unified framework that facilitates the development of advanced critical software targeting parallel and heterogeneous systems considering high performance and crticial demands.
Respect, began in October 2022 and is conducting research on four different fronts:
- Parallel Programming Models: to abstract the complexities of the underlying processing architecture while providing enough flexibility to easily exploit the system
- Performance optimizations: to minimize the impact of parallel execution in terms of contention and overhead, and to address heterogeneity and interoperability with other pogramming models
- Resilience: to develop techniques for fault detection, like replication and N-version programming, and fault recovery, like coars grained checkpointing
- Reliability: to provide functional and timing correctness and to provide reliable parallel runtimes.
Applications of the RESPECT project include spacecraft applications for image and radar processing, data and image compression, signal processing and machine learning and automotive applications for the evaluation of heterogenous parallel programming models. These use cases will allow the project to further demonstrate the DePE technology in a relevant environment.
The project is funded under the Department of Research and Universities of the Generalitat de Catalunya and will last for 18 months.