Sunway supercomputer architecture towards exascale computing: analysis and practice

Abstract

In recent years, the improvements of system performance and energy efficiency for supercomputers have faced increasing challenges, which create more intensive demands on the architecture design for realizing exascale computing. This paper first analyzes the main requirements of exascale computing on the aspects of the parallel computing application and supercomputing center operation. Afterwards, a mapping scheme of "demands-challenges-architecture" is proposed. Then, the major challenges of exascale supercomputer, such as scalability, power consumption, data movement, programming and availability, are thoroughly analyzed, and the corresponding appropriate solutions are proposed. Moreover, this paper proposes the Sunway computer architecture towards exascale computing in which the many-core processor, network chipset and software system are all domestically-designed. The technology roadmap of Sunway supercomputer will hold the comprehensive design methods for the architecture, including the processor, interconnect network, assembly structure, power supply, cooling system, system software, parallel algorithm and application support, promising great advances for exascale supercomputing.

Exaptation within structure: the challenge of external resources and risk-taking for public R&D institutions

“…Contrary to the proposed sequential trajectory (Bouty, 2000), our research challenges the concept of a predetermined sequence, advocating instead for a three-step exaptationbased approach involving internal and external resources. We highlight the considerable challenge associated with achieving parity in supercomputing capabilities through commercially available components, as emphasized by Gao et al (2021). This recognition emanates from the identi cation of dynamic knowledge capabilities, particularly in the domain of people development, as expounded by Michailova and Zhan (2015).…”

“…The nal decision depends on why the interaction is happening, and whether the focus is on making money or being fair (Bouty, 2000). This elucidates the formidable challenge of achieving a state of supercomputing parity with commercially available components (Gao et al, 2021). IN fact, R&D units within Multinational Corporations (MNCs) manifest a distinctive dynamic knowledge capability, characterized by intricate interdependencies with the innovation potential of subsidiaries (Michailova & Zhan, 2015).…”

“…Entrepreneurial capabilities, encompassing dynamic and asset orchestration dimensions, empower companies to navigate uncharted futures, exempli ed by a robust early warning system enhancing responsiveness to disruptions (Teece and Leih, 2016). BSC offers distinctive opportunities for knowledge creation and innovation, notwithstanding challenges associated with managing the delicate balance between specialization and collaboration, linguistic intricacies, and navigating the evolving terrain of technology and research funding (Gao et al, 2021). Nevertheless, this study discerns a conspicuous void in the alignment of resources allocation, like recruitment investments, especially within small and medium-sized technology rms, particularly those bolstered by public funding, in relation to labor market dynamics (Arokiasamy et al, 2023).…”

“…Exaptation emerges, being an adaptation as pivotal mechanism, facilitating the repurposing and integration of pre-existing resources within novel contextual domains (Dew et al, 2004). But public R&D units encounter challenges in adhering to a sequential decision-making process, in uenced by factors such as familiarity, trust, and competitive dynamics (Bouty, 2000), so the pursuit of supercomputing parity with commercial counterparts poses a considerable challenge (Gao et al, 2021), since commercial counterparts exhibit dynamic knowledge capabilities with intricate interrelations (Michailova & Zhan, 2015). As an illustration, the implementation of a resilient early warning mechanism, founded upon entrepreneurial aptitudes, augments the capacity of entities to navigate ambiguous circumstances (Teece and Leih, 2016).…”

A ''New Ara'' for Vector Computing: An Open Source Highly Efficient RISC-V V 1.0 Vector Processor Design

“…The current leader is Fujitsu's Fugaku [1], [2] at the RIKEN research lab in Japan with 159 thousand nodes, each with a 48-core Fujitsu A64FX running at 2.2 GHz, supporting the Armv8.2-A Instruction Set Architecture (ISA) with Scalable Vector Extension (SVE) using 512 bit vectors. The Sunway TaihuLight [3] similarly builds on processors with vector extension-and while it is now ranked as number 4 (as of September 2021), it has been on the list since 2016. It has 26 thousand Sunway SW26010 CPUs, each chip running at 1.45 GHz with 4 clusters of 64 compute cores with a 256-bit vector unit each.…”