Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-6884
Main Title: On latency in GPU throughput microarchitectures
Author(s): Andersch, Michael
Lucas, Jan
Álvarez-Mesa, Mauricio
Juurlink, Ben
Type: Conference Object
Language Code: en
Abstract: Modern GPUs provide massive processing power (arithmetic throughput) as well as memory throughput. Presently, while it appears to be well understood how performance can be improved by increasing throughput, it is less clear what the effects of micro-architectural latencies are on the performance of throughput-oriented GPU architectures. In fact, little is publicly known about the values, behavior, and performance impact of microarchitecture latency components in modern GPUs. This work attempts to fill that gap by analyzing both the idle (static) as well as loaded (dynamic) latency behavior of GPU microarchitectural components. Our results show that GPUs are not as effective in latency hiding as commonly thought and based on that, we argue that latency should also be a GPU design consideration besides throughput.
URI: https://depositonce.tu-berlin.de//handle/11303/7706
http://dx.doi.org/10.14279/depositonce-6884
Issue Date: 2015
Date Available: 30-Apr-2018
DDC Class: 004 Datenverarbeitung; Informatik
Subject(s): GPU
microarchitecture
GPU architecture
latency
throughput
License: http://rightsstatements.org/vocab/InC/1.0/
Proceedings Title: 2015 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS 2015)
Publisher: IEEE
Publisher Place: Piscataway, NJ
Publisher DOI: 10.1109/ISPASS.2015.7095801
Page Start: 169
Page End: 170
ISBN: 978-1-4799-1957-4
Appears in Collections:FG Architektur eingebetteter Systeme » Publications

Files in This Item:
File Description SizeFormat 
andersch_etal_2015.pdf571.31 kBAdobe PDFThumbnail
View/Open


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