Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-10108
For citation please use:
Main Title: A Quantitative Study of Locality in GPU Caches
Translated Title: Eine quantitative Untersuchung der Lokalität in GPU-Caches
Author(s): Lal, Sohan
Juurlink, Ben
Institution: Technical University of Berlin
Type: Conference Object
Language Code: en
Abstract: Traditionally, GPUs only had programmer-managed caches. The advent of hardware-managed caches accelerated the use of GPUs for general-purpose computing. However, as GPU caches are shared by thousands of threads, they are usually a victim of contention and can suffer from thrashing and high miss rate, in particular, for memory-divergent workloads. As data locality is crucial for performance, there have been several efforts focusing on exploiting data locality in GPUs. However, there is a lack of quantitative analysis of data locality and data reuse in GPUs. In this paper, we quantitatively study the data locality and its limits in GPUs. We observe that data locality is much higher than exploited by current GPUs. We show that, on the one hand, the low spatial utilization of cache lines justifies the use of demand-fetched caches. On the other hand, the much higher actual spatial utilization of cache lines shows the lost spatial locality and presents opportunities for further optimizing the cache design.
URI: https://depositonce.tu-berlin.de/handle/11303/11220
http://dx.doi.org/10.14279/depositonce-10108
Issue Date: 2020
Date Available: 26-May-2020
DDC Class: 004 Datenverarbeitung; Informatik
Subject(s): data locality
GPU caches
memory divergence
License: http://rightsstatements.org/vocab/InC/1.0/
Proceedings Title: International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation (SAMOS XX)
Publisher: Springer
Publisher Place: Berlin ; Heidelberg
Series: Lecture Notes in Computer Science
EISSN: 1611-3349
ISSN: 0302-9743
Appears in Collections:FG Architektur eingebetteter Systeme » Publications

Files in This Item:
lal_juurlink_samos2020.pdf

Accepted manuscript

Format: Adobe PDF | Size: 3.13 MB
DownloadShow Preview
Thumbnail

Item Export Bar

Items in DepositOnce are protected by copyright, with all rights reserved, unless otherwise indicated.