Future High Performance Computing Capabilities: Summary Report of the Advanced Scientific Computing Advisory Committee (ASCAC) Subcommittee

Multisector Dynamics: Advancing the Science of Complex Adaptive Human‐Earth Systems

“…There are however methodological, data availability, and computational gaps that are at present limiting the MSD community's ability to confront the complexity of human‐Earth systems and their feedbacks. There is a need for: (a) better integration with complexity science (Haimes, 2018; Meerow & Newell, 2015; Montuori, 2013), (b) improved modes of analysis for capturing uncertainties in how human systems shape dynamics (Axelrod, 2006; Filatova et al., 2016; Moallemi & de Haan, 2019; Polhill et al., 2016; Trutnevyte et al., 2019; Zellner, 2008), (c) computational advances that enhance representations of highly nonlinear and uncertain “state‐action” feedbacks (Bertsekas, 2019; Herman et al., 2020; Oikonomou et al., 2021; Powell, 2019), and (d) solutions to overcome computational scaling and scientific inference barriers to MSD research insights (Bergman et al., 2019; Hendrickson, 2020; McGovern & Allen, 2021; National Academies of Sciences & Medicine, 2016). Addressing these gaps will require deeper collaborations with the statistical, mathematical, and computational sciences.…”

Strategies for Democratization of Supercomputing: Availability, Accessibility and Usability of High Performance Computing for Education and Practice of Big Data Analytics

“…Given the explosive growth in quantity, diversity, complexity, granularity and acceleration of data generation, it has become impossible to meaningfully depend on desktops, servers or standalone computers to create competitive value, and an increasingly large number of disciplines have begun HPC evaluation and adoption processes, to ensure that they remain competitive in progressively data and computation intensive environments (Fiore et al, 2018). The already steep trend towards HPC engagement and adoption can be expected to become stronger with the advent of new technologies and the identification of new opportunities (Bergman et al, 2019). An investigation by the Council on Competitiveness discovered that the vast majority of US corporations with HPC capabilities had significant concerns about being able to hire persons with "sufficient HPC training", and that there are no easy solutions because "... there aren't enough faculty, researchers, educators, and professionals with the HPC skills and knowledge to fulfill the demand for talented individuals" (Lathrop, 2016).…”

“…The term "supercomputing" (or supercomputers) has been treated as being synonymous with HPC (or high performance computers), and has been parsimoniously described as a computer or a cluster of computers with far greater computing-memory-storage capabilities than a general computer, and as being "characterized by large amounts of memory and processing power " (George, 2020). So also, HPC has been varying defined as being a "combination of processing capability and storage capacity" that can efficiently create solutions for "difficult computational problems across a diverse range of scientific, engineering, and business fields" (Ezell and Atkinson, 2016), and also as being "massively parallel processing (MPP) computers" (Bergman et al, 2019). HPC can be classified as being homogeneous or heterogeneous, based on the use of similar or dissimilar processors (or memory, or similar HPC components) respectively, in its array of processors, such as homogeneous HPC with CPU arrays, and heterogeneous HPC with with CPU and GPU arrays (Gao and Zhang, 2016).…”

OpenSBLI: Automated code-generation for heterogeneous computing architectures applied to compressible fluid dynamics on structured grids

“…As previously mentioned, automated parallel abstractions can offer researchers an easy way to utilise modern hardware. In addition to x86-CPU and GPU-based architectures, ARMbased CPUs [47], Field Programmable Gate Arrays (FPGAs), and many-core accelerator cards such as the Intel Xeon Phi have all been cited as possible architectures of the future [48]. As the recent discontinuation of the Intel Xeon Phi product line [49] has highlighted however, it is not clear at this stage which, if any, of these architectures will ultimately prevail in the coming decades.…”