EXA2PRO (Enhancing Programmability and boosting Performance Portability for Exascale Computing Systems) is a three-year project launched in May 2018 with €3.47 million in funding from the European Commission’s Horizon 2020 Future and Emerging Technologies for High-Performance Computing (FET-HPC) program. Coordinated by the Institute of Communications and Computer Systems (ICCS), EXA2PRO brings together seven partners*, including academic partners, a supercomputing center (Jülich) and an SME (Maxeler). The EXA2PRO advisory board will include key experts from the exascale/HPC computing field and relevant international initiatives, such as the US Exascale Computing Project. A summary of the project follows.
The EXA2PRO: Enhancing Programmability and boosting Performance Portability for Exascale Computing Systems H2020 project was officially lunched with a kick-off meeting at ICCS in May 2018. The project is developing a set of tools that addresses challenges imposed by the increasing heterogeneity and diversity of supercomputing centers and future exascale computing systems.
The EXA2PRO will deliver an integrated development environment leveraging high-level software abstractions that provide programming ease and productivity through multi-language parallel pattern exploitation, performance portability through effective code variants generation, increased memory efficiency at large-scale through sophisticated memory management tools and a runtime system that will provide multi-criteria scheduling and load balancing in heterogeneous computing architectures. Additionally, the EXA2PRO framework will integrate novel software engineering analysis and optimization tools that improve the design quality of the source code of applications that target supercomputing centers and enable the quantification of the economic impact of source code quality improvements.
The high-level API of the EXA2PRO programming environment will be based on an autotunable multi-backend skeleton programming framework than will offer portability, management of parallelism and heterogeneity support. A set of tools acting as “plug-ins” will access an annotated intermediate representation of the application’s source code to apply optimizations that will provide efficient resource management, fault tolerance and maintainable source code. The EXA2PRO runtime system will offer efficient resource management based on performance, load balancing, energy efficiency and fault tolerance criteria.
The components of the EXA2PRO programming environment will be either developed from scratch during the project lifetime or they will be based on HPC tools that will be enhanced and extended to meet specific exascale computing requirements. Several of these tools (e.g. SkePU, StarPU) have been used as critical components in HPC programming environments in past EC projects, such as the FP7 PEPPHER and EXCESS.
Four emerging applications from the high energy physics, material science, chemical processes for cleaner environment and energy storage will pave the way towards the effective exploitation of heterogeneity at the exascale level, by leveraging the EXA2PRO framework and Jülich’s supercomputing center infrastructures. The applications have been selected based on the following two criteria: the exploitation of future exascale systems and the impact on various domains in both the scientific community and the industry. Through the effective exploitation of the increased acceleration capabilities of heterogeneous systems, we expected significant improvements on the simulation and analysis of more complex problems, thus delivering significant impacts.
The EXA2PRO project will generate exploitable results in the form of a tool-chain that support diverse exascale heterogeneous supercomputing centres and concrete improvements in various exascale computing challenges. From technical perspective, the EXA2PRO will contribute to the transition of practical exascale computing, by addressing the challenges of programming in new heterogeneous architectures and having impact on standard bodies. From economic perspective, significant impact is expected on the domain of processes for cleaner environment and energy storage scientific community and industry.
*PARTNERS:
Greece: Institute of Communications and Computer Systems (ICCS); Centre for Research and Technology Hellas (CERTH)
France: Institut National de Recherche en Informatique et en Automatique (INRIA); Centre National de la Recherche Scientifique (CNRS)
Sweden: Linköping University
Germany: Forschungszentrum Jülich
UK: Maxeler Technologies Limited