concurrent data structures, with a signi•cantly simpler design KEYWORDS concurrent data structures; parallel programs; processing-in-memory; near-memory computing 1 NEAR-MEMORY COMPUTING „e performance gap between memory and CPU has grown ex-ponentially Memory vendors have focused mainly on improv-
Data Structures + Hardware 6 • Tight integration between algorithmic design and hardware characteristics • Memory becomes an active component in managing data • Managing data structures in PIM • Old work: pointer chasing for sequential data structures • Our work: concurrent data structures
near-memory compute units, and between the host threads and near-memory compute units is required to meet high concurrency and correctness guarantees In summary, this paper makes the following contributions: •We propose hybrid data structures, which are cache-conscious concurrent data structures designed for new NMP architectures
the-art concurrent data structures This paper makes the following contributions: •We define a generic near-data-processing (NDP) architecture that is well-suited for concurrent data structures •We implement actual software kernels of the NDP-based concurrent data structures on a cycle-accurate full-system
Abstract of \Concurrent Data Structures with Near-Memory Processing: Software-Hardware Co-Design" by Jiwon Choe, Ph D , Brown University, May 2022 In recent years, there has been a renewed interest in near-memory processing (NMP) architectures as a workaround for the performance and energy issues of frequent and irregular memory access,
and software simulators con?rm our theoretical ?ndings, and they show that the LPM approach can be applied in diverse computing platforms and can effectivelyguide performance optimization of memory systems Index Terms—Memory wall, memory stall time, ef?ciency, performance optimization, layered performance matching (LPM), memory
How to Design Data Structures for PIM? Zhiyu Liu, Irina Calciu, Maurice Herlihy, and Onur Mutlu, "Concurrent Data Structures for Near-Memory Computing" Proceedings of the29th ACM Symposium on Parallelism in Algorithms and Architectures(SPAA), Washington, DC, USA, July 2017 [Slides (pptx) ( pdf )] 12
Avoid data movement by doing computation in memory • Old idea Near Memory Computing: Architecture 5 Goals: PIM Concurrent Data Structures 1
on Distributed Memory Computer Systems operation to be sufficiently close to the minimum designing concurrent data structures for parallel systems