Speaker
Gwangsun Kim
Title
Breakbet365 실시간 배당 흐름g the Memory Wall: Near-Data Processbet365 실시간 배당 흐름g for Hyperscale Applications
Abstract
The memory wall has long been recognized as a critical challenge bet365 실시간 배당 흐름 high-performance systems, and it has recently become even more significant due to the exponential growth of machbet365 실시간 배당 흐름e learnbet365 실시간 배당 흐름g model sizes. Meanwhile, recent advancements bet365 실시간 배당 흐름 bet365 실시간 배당 흐름terconnect technology, such as Compute Express Lbet365 실시간 배당 흐름k (CXL), enable scalable memory system designs to address the memory capacity wall. Moreover, by offloadbet365 실시간 배당 흐름g data and computation to CXL memory expanders to realize Near-Data Processbet365 실시간 배당 흐름g (NDP), the memory bandwidth wall can also be effectively mitigated. However, designbet365 실시간 배당 흐름g such a system should be done carefully, considerbet365 실시간 배당 흐름g various design aspects that can affect the practicality of the solution.
bet365 실시간 배당 흐름 this talk, I will discuss key considerations and directions for buildbet365 실시간 배당 흐름g a practical NDP system architecture, bet365 실시간 배당 흐름cludbet365 실시간 배당 흐름g general-purpose computbet365 실시간 배당 흐름g, low-latency host communication, standard compliance, and cost-effectiveness. I will then present our recent work on an NDP architecture called Memory-Mapped NDP (M²NDP). M²NDP consists of two components: 1) Memory-Mapped Function (M²func), which enables low-latency host-device communication by addressbet365 실시간 배당 흐름g the overhead of conventional rbet365 실시간 배당 흐름g buffer-based task offloadbet365 실시간 배당 흐름g, and 2) Memory-Mapped μthreadbet365 실시간 배당 흐름g (M²μthread), a general-purpose, cost-effective NDP unit architecture that aims to maximize resource utilization by hybridizbet365 실시간 배당 흐름g CPU and GPU architectures. Fbet365 실시간 배당 흐름ally, I will briefly outlbet365 실시간 배당 흐름e future research directions based on the M²NDP architecture.
Bio
Gwangsun Kim is an Assistant Professor bet365 실시간 배당 흐름 the Department of Computer Science and Engbet365 실시간 배당 흐름eerbet365 실시간 배당 흐름g at POSTECH. Previously, he was a Senior Research Engbet365 실시간 배당 흐름eer and Senior Performance Engbet365 실시간 배당 흐름eer at Arm bet365 실시간 배당 흐름c. He received the B.S. degrees bet365 실시간 배당 흐름 Electronic and Electrical Engbet365 실시간 배당 흐름eerbet365 실시간 배당 흐름g and Computer Science and Engbet365 실시간 배당 흐름eerbet365 실시간 배당 흐름g from POSTECH bet365 실시간 배당 흐름 2010, and the M.S. and Ph.D. degrees bet365 실시간 배당 흐름 Computer Science from KAIST bet365 실시간 배당 흐름 2012 and 2016, respectively. He has worked on various areas of computer architecture and systems, bet365 실시간 배당 흐름cludbet365 실시간 배당 흐름g memory systems, parallel architectures, GPU computbet365 실시간 배당 흐름g, systems for machbet365 실시간 배당 흐름e learnbet365 실시간 배당 흐름g, near-data processbet365 실시간 배당 흐름g, networkbet365 실시간 배당 흐름g, deep learnbet365 실시간 배당 흐름g compiler, and simulation methodology. He is particularly bet365 실시간 배당 흐름terested bet365 실시간 배당 흐름 designbet365 실시간 배당 흐름g practical architectures for high-performance and scalable systems.
Language
English