Intelligent Computing Memory Systems for Data-Intensive Applications

在大数据时代，对于许多数据密集型应用，如机器学习、视频处理、计算基因组学、大数据分析和网络处理等，其性能和能效不仅受到计算本身的限制，还受到数据通信的限制。因此，传统的通用和以计算为中心的平台不再能够满足不断增长的需求。 在这个提案中，我们的目标是设计下一代计算平台的数据密集型应用，与华为加拿大在其智能计算业务的合作伙伴关系（https：//e.huawei.com/ca/solutions/hic）。加拿大在细粒度可重构架构的架构设计和工具自动化方面有着悠久的创新历史。通过结合专门的硬件加速和以数据为中心的计算，我们建议开发一个智能计算存储系统，在那里我们将设计一个高性能和高能效的粗粒度可重构硬件加速器芯片，并将其集成到处理器和存储器端。此外，我们建议提供完整的系统支持，以便数据密集型应用程序可以轻松有效地利用整个系统并在它们所在的位置处理数据，包括传统的处理器，处理器端和存储器端加速器，以优化计算并最大限度地减少数据移动。 我们的研究将为大数据时代的下一代智能计算存储系统开发新的硬件架构、关键技术和可重用方法，以及相应的编译和建模工具。 大部分研究成果将是开源的，以造福加拿大更广泛的社区。我们的计划还将培养下一代高素质的专业人员，并为他们在高需求的信息和通信技术领域的职业生涯做好准备（https：//www.ic.gc.ca/eic/site/ict-tic.nsf/eng/h_it07229.html）。所有这些都将使加拿大在计算技术方面保持国际竞争力，而计算技术是经济增长和医疗保健改善的关键。

In the big data era, for many data-intensive applications, such as machine learning, video processing, computational genomics, big data analytics, and network processing, their performance and energy efficiency are limited by not only the computation itself, but also the data communication. As a result, traditional general-purpose and computing-centric platforms can no longer sustain the ever-increasing demand. In this proposal, we aim to design the next-generation computing platform for data-intensive applications, with the partnership of Huawei Canada in its business of intelligent computing (https://e.huawei.com/ca/solutions/hic). Canada has a long history of innovation in the architecture design and tool automation for fine-grained reconfigurable architectures. By incorporating both specialized hardware acceleration and data-centric computing, we propose to develop an intelligent computing memory system, where we will design a high performance and energy efficient coarse-grained reconfigurable hardware accelerator chip and integrate it onto both the processor and memory sides. Moreover, we propose to provide full system support, so that data-intensive applications can easily and efficiently utilize the entire system and process the data where they sit, including the conventional processor, processor-side and memory-side accelerators, to optimize the computation and minimize the data movement. Our proposed research will develop novel hardware architecture, critical technologies and reusable methodology, as well as corresponding compilation and modeling tools, for next-generation intelligent computing memory systems in the big data era. The majority of the research results will be open source to benefit a broader community in Canada. Our program will also train next-generation highly qualified professionals and prepare their career in the highly-demanded information and communication technology sector (https://www.ic.gc.ca/eic/site/ict-tic.nsf/eng/h_it07229.html). All these will keep Canada internationally competitive in computing technologies that are key to the economy growth and healthcare improvement.