CCRI: ENS: Collaborative Research: Modernizing gem5: Expanding the Reach of Computer Architecture Simulation

CCRI：ENS：协作研究：现代化 gem5：扩大计算机体系结构仿真的范围

• 批准号: 1925485
• 负责人: Matthew Sinclair
• 金额: $ 36.28万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-10-01 至 2023-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1925485&HistoricalAwards=false
• 关键词: CCRI; ENS; Collaborative; Research; Modernizing

Computer systems are increasingly important to society and the advancement of science. In recent years, technological trends have made it increasingly difficult to continue scaling their performance. As a result, research to further improve computer systems is vital. Traditionally, this research has relied on cycle-level software simulation of these computer systems to model the potential improvements of novel techniques. However, much of the software was written 15 years ago, and it is currently maintained by students and researchers. This project will make a concerted effort to improve the underlying simulation infrastructure, which will enable increased research output by the computer systems community. It will specifically focus on the gem5 infrastructure, which is widely used in the community (e.g., cited over 3000 times) and considered the de facto tool for simulating all components of the computing system.This project will enhance and rejuvenate the aging gem5 community infrastructure through improvements to the underlying gem5 components, community outreach and improved user's services, and developing new models for emerging devices required for evaluating important applications (e.g., machine learning, augmented and virtual reality, high performance computing, and system security). These improvements will increase the research output of gem5's broad user base and extend the ability to perform hardware-software co-design to the entire computer systems community. This project will broaden the already vibrant gem5 community through reducing the barrier to entry by (1) running two gem5 summer schools to introduce junior graduate students to gem5 and computer systems research, (2) simplifying gem5's interface, (3) improving the community's documentation and communication channels, and (4) supervising a community code shepherd to help researchers integrate their code with the community infrastructure. The investigators will also create a set of 'known-good' system configurations that will significantly increase the productivity of researchers performing detailed cycle-level research studies. These system configurations will be proxies for current state-of-the-art systems and will be publicly validated against real systems and workloads suggested by the community.This project will create a sustainable path for the future of the gem5 infrastructure by improving the code testing coverage and increasing the robustness of hardware device models. Additionally, community outreach efforts will broaden the community increasing the impact of gem5 and providing more contributors to continue development of this infrastructure after this project completes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

计算机系统对社会和科学进步越来越重要。 近年来，技术趋势使得继续扩展其性能变得越来越困难。 因此，进一步改进计算机系统的研究至关重要。 传统上，这项研究依赖于这些计算机系统的周期级软件仿真来模拟新技术的潜在改进。 然而，大部分软件是在15年前编写的，目前由学生和研究人员维护。 该项目将作出协调一致的努力，以改善基本的模拟基础设施，这将使计算机系统界能够增加研究成果。 它将特别关注在社区中广泛使用的gem 5基础设施（例如，被引用超过3000次），并被认为是模拟计算系统所有组件的事实上的工具。该项目将通过改进基本的gem 5组件、社区推广和改善用户服务，以及为评估重要应用所需的新兴设备开发新模型，机器学习、增强现实和虚拟现实、高性能计算和系统安全）。 这些改进将增加gem 5广泛用户基础的研究成果，并将执行硬件-软件协同设计的能力扩展到整个计算机系统社区。这个项目将通过以下方式来拓宽已经充满活力的gem 5社区：（1）开办两所gem 5暑期学校，向初级研究生介绍gem 5和计算机系统研究;（2）简化gem 5的界面;（3）改善社区的文档和沟通渠道，以及（4）监督社区代码管理员以帮助研究人员将他们的代码与社区基础设施集成。研究人员还将创建一组“已知良好”的系统配置，这将显著提高研究人员执行详细周期级研究的生产力。这些系统配置将成为当前最先进系统的代理，并将根据社区建议的真实的系统和工作负载进行公开验证。该项目将通过提高代码测试覆盖率和增加硬件设备模型的鲁棒性，为gem 5基础设施的未来创建一条可持续的道路。此外，社区外展工作将扩大社区，增加gem 5的影响，并提供更多的贡献者，以继续发展这一基础设施后，这个项目完成。这个奖项反映了NSF的法定使命，并已被认为是值得的支持，通过评估使用基金会的知识价值和更广泛的影响审查标准。

项目成果

Modeling Modern GPU Applications in gem5

在 gem5 中对现代 GPU 应用程序进行建模

• 发表时间: 2020
• 期刊: gem5 Users Workshop
• 作者: Roarty, Kyle;Sinclair, Matthew D.
• 通讯作者: Sinclair, Matthew D.

Enabling Reproducible and Agile Full-System Simulation

实现可重复且敏捷的全系统仿真

• 发表时间: 2021
• 期刊: 2021 IEEE International Symposium on Performance Analysis of Systems and Software
• 作者: Bobby R. Bruce, Ayaz Akram

Analyzing the Benefits of More Complex Cache Replacement Policies in Moderns GPU LLCs

分析现代 GPU LLC 中更复杂的缓存替换策略的好处

• 发表时间: 2023
• 期刊: 5th gem5 Users' Workshop
• 作者: Jarvis Jia;Matthew D. Sinclair
• 通讯作者: Matthew D. Sinclair

Improving the Speed of gem5’s GPU Regression Tests

提高 gem5 GPU 回归测试的速度

• 发表时间: 2023
• 期刊: 5th gem5 Users' Workshop
• 作者: James Braun;Matthew D. Sinclair
• 通讯作者: Matthew D. Sinclair

gem5 GPU Accuracy Profiler (GAP)

gem5 GPU 精度分析器 (GAP)

• 发表时间: 2022
• 期刊: 4th gem5 Users Workshop
• 作者: Charles Jamieson, Anushka Chandrashekar