Generation and Verification of High-Performance Mathematical Software and Hardware

高性能数学软硬件的生成和验证

• 批准号: 536628-2018
• 负责人: Anand, Christopher
• 金额: $ 6.19万
• 依托单位: McMaster University
• 依托单位国家: 加拿大
• 项目类别: Collaborative Research and Development Grants
• 财政年份: 2020
• 资助国家: 加拿大
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=699284
• 关键词: Generation; Verification; Performance; Mathematical; Software

This project continues a decade-long collaboration between McMaster University and IBM Canada, inventing new software strategies to better utilize the z Mainframes which transact eCommerce and analytics and POWER supercomputers powering science. Coconut (COde CONstructing User Tool) is our experimental compiler platform, and comprises a series of Domain Specific Languages embedded in Haskell, together with symbolic code generation and several novel scheduling strategies. The work funded by this grant will allow us to finish work on randomly generating instruction schedules in order to not only produce very efficient code, but to understand the space of possible schedules, gain insight into existing scheduling strategies, and formulate new strategies. It will also allow us to fully develop a proposed processor enhancement which promises to cut Machine Learning times in half. The applications which will directly benefit from this work are Machine Learning and and Blockchain.

该项目延续了麦克马斯特大学和IBM加拿大之间长达十年的合作，发明了新的软件策略，以更好地利用z大型机进行电子商务和分析，并为科学提供动力。 Coconut（COde CONstructing User Tool）是我们的实验编译器平台，包括一系列嵌入在Haskell中的领域特定语言，以及符号代码生成和几种新颖的调度策略。 这项资助的工作将使我们能够完成随机生成指令调度的工作，不仅可以生成非常高效的代码，还可以了解可能的调度空间，深入了解现有的调度策略，并制定新的策略。 它还将使我们能够充分开发一种拟议的处理器增强功能，有望将机器学习时间缩短一半。 将直接受益于这项工作的应用程序是机器学习和区块链。