Sustaining System-on-Chip Innovation by Enabling Open-Source Hardware
Abstract
Over the last decade, the system-on-chip (SoC) has emerged as the main advanced computing platform across many application domains, from data-center servers to mobile devices. Facing the slowdown of silicon dimensional scaling (also known as the end of MooreÕs Law), computer architects started combining increasingly heterogeneous components into a variety of SoC architectures to achieve high performance and energy efficiency. A state-of-the-art SoC architecture is realized by integrating generalpurpose processors, special-purpose processors, application-specific hardware accelerators, reconfigurable hardware blocks, and even analog/mixed-signal components into a single chip. The heterogeneity of these components is the key to the efficiency and performance of the overall system. Heterogeneity, however, increases design complexity in terms of hardware-software interactions, access to shared resources, and diminished regularity of the design. Open-source hardware (OSH) holds the promise of reducing the staggering costs of designing an SoC architecture, thus generating many opportunities for creative innovation in engineering new computing systems across a variety of application domains. The number of OSH projects is expected to grow steadily in the upcoming years, fueled by government funding programs, the activities of multi-institution non-profit organizations, and the many diverse contributions from academia and industry. To date, however, most OSH projects have been focused on the development of individual SoC components, such as a microprocessor core or an application-specific accelerator. While this is certainly useful, it leaves open a major research challenge, which can be formulated as the following question: How to enable the realization of a complete systemon- chip for a given target application through the efficient reuse and modular combination of a heterogeneous variety of open-source hardware components that are independently developed, possibly across different organizations for different purposes? The main goal of this project is to address this challenge. The proposed solution is based on the idea that the path towards achieving modularity and reusability of SoC components lies on the combination of the flexibility of latency-insensitive design mechanisms with the benefits of system-level design methods. The project will advance the scientific state-of-the-art in the design and programming of SoC architectures that combine many heterogeneous components. The main result of the proposed research is a complete system-level methodology for the design of individual reusable and modular OSH components as well as for the design of SoC architectures that are realized through the combination of these components. The methodology will be supported by a flexible computer-aided design (CAD) flow, which leverages a mix of commercial tools and open-source tools together with libraries of sample components, protocols and interfaces. The methodology and the supporting CAD flow will enable the rapid and inexpensive prototyping of individual components in the context of complete SoC architectures through their simulation with virtual platforms and their emulation with field-programmable gate array (FPGA) technologies. This will enable thinking at the system level while developing individual components. By raising the level of abstraction in the SoC design process to system-level design, the proposed effort promotes closer collaboration among hardware engineers and software programmers. By fostering the development of open-source hardware along the development of open-source software, the proposed effort contributes to sustain the innovation in advanced computing that modern societies have come to expect and that is currently threatened by the anticipated end of MooreÕs Law.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 01, 2019
- Source ID
- W911NF1910476
Entities
People
- Luca Carloni
Organizations
- Army Contracting Command
- Columbia University
- National Security Agency