Riding the third wave of workstation computing: The future of irregular symbolic computation

The third wave of computing is powered by the recent advent of chiplets: small integrated circuits which are packaged together to act as a single integrated circuit. In the mid 2000s, the second wave of computing introduced multi-core computers, where each "core" is able to perform computations independently. Second-wave computers typically offer 2-16 cores. Now, chiplets will allow 64 to 128 cores, or more, per computer. This third wave is motivated by the ending of Moore's law, which states that the number of transistors on an integrated circuit will double every two years. Increasing the number of transistors has been a main source of increased computing power over the past decades. But this is coming to an end. Chiplets, and their many cores, present a solution. However, the single biggest limitation to gains in computer power will be software's ability to effectively utilize this increasing number of cores. The objective of this RTI proposal is to support my research group with the hardware necessary to develop, test, experiment, and deploy scalable parallel algorithms for irregular parallelism and symbolic computation. In irregular parallelism, concurrency opportunities emerge dynamically as the algorithm progresses, and load-balancing and scheduling are a challenge. These irregular applications are not well-suited for execution on computer clusters as distributed computations. Moreover, in symbolic computation, applications are often limited by memory requirements and not computational power, again suggesting they are not well-suited to distributed computing. These applications instead require the tightly coupled resources of a single computer. Personal workstation computers are thus ideal execution environments for irregular and symbolic computation. I propose acquiring a computer with a 64-core AMD Epyc 9554P processor. This equipment will be crucial to the success of my research program, to be primarily funded by an NSERC Discovery grant. There is no financial overlap between this proposal and the Discovery grant. However, there is substantial conceptual overlap and, without the proposed equipment, our research will be severely delayed. The proposed equipment will provide state-of-the-art computing resources and hands-on training to all members of my research group. This will give them highly valuable skills in high-performance and parallel computing, alongside indirect skills including software development, operations, engineering, and testing. The outcome of this research will be algorithms, techniques, tools, and best practices which enable continued improvements in computing power and performance in workstation computers despite the end of Moore's law. This research thus supports applications in: data processing; modelling and simulation; graphics and video games; and computer-aided design and manufacturing. Scientists, engineers, and content creators will be able to tackle larger and more complex workloads than ever before.