Project description
Novel algorithms shepherd Big data on its journey through massively parallel systems
The increasing availability of huge amounts of data in fields as diverse as astronomy, protein engineering, climate science and finance offers an opportunity for a step change in understanding and predictive ability. However, to harness the opportunity hidden in all this data, we need increasingly fast, efficient and accurate numerical approaches compatible with high-performance computing systems. This parallel processing on steroids carries risks of propagating and increasing error through the division of labour and regrouping along the way to the final outcome. Handling of floating-point representations of data can be an important source of numerical errors or inaccuracy. The EU-funded Robust project will develop novel algorithms for either faster or more energy-efficient processing that will simultaneously boost the accuracy and reproducibility of the results of scientific computations.
Objective
Computations in parallel environments, like the emerging Exascale systems, are usually orchestrated by complex runtimes that employ various strategies to uniformly and efficiently distribute computations and data. However, these strategies, pursuing excellent performance scalability, may also impair numerical reliability (accuracy and reproducibility) of final results due to the dynamic and, thus, non-deterministic execution as well as non-associativity of floating-point operations. Additionally, scientific computations frequently rely upon only one working precision for computing problems with various complexities, which leads to the significant underutilization of the floating-point representation or the lack of accuracy. The Robust project aims to address the issue of reliable and sustainable scientific computations through developing robust, energy-efficient, and high performing algorithmic solutions for underlying numerical linear algebra solvers and libraries as well as applying these solutions in applications and kernels at scale. The fellow, Roman Iakymchuk, is an expert in numerical linear algebra and high-performance computing and will collaborate with the research team of Prof. Stef Graillat at the Sorbonne University, who are experts in numerical analysis and computer arithmetic. This unique collaboration and combination of skill sets are crucial to embed numerical reliability and sustainability in algorithmic solutions for linear algebra operations and solvers. The derivation of novel robust algorithmic solutions, which will lead to either faster or more energy-efficient execution, will also grant a user an opportunity to specify the expected output accuracy of computations while ensuring optimal intermediate precisions. This ambitious research project in conjunction with formal training and bespoke mentoring will enhance the fellow's academic profile, research experience, and broaden skill set in numerical analysis and computer arithmetic.
Fields of science
Keywords
Programme(s)
Funding Scheme
MSCA-IF - Marie Skłodowska-Curie Individual Fellowships (IF)Coordinator
75006 Paris
France