UpDown: Programmable fine-grained Events for Scalable Performance on Irregular Applications

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Rajasukumar, Andronicus, Su, Jiya, Yuqing, Wang, Su, Tianshuo, Nourian, Marziyeh, Diaz, Jose M Monsalve, Zhang, Tianchi, Ding, Jianru, Wang, Wenyi, Zhang, Ziyi, Jeje, Moubarak, Hoffmann, Henry, Li, Yanjing, Chien, Andrew A.
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913452502548480
author Rajasukumar, Andronicus
Su, Jiya
Yuqing
Wang
Su, Tianshuo
Nourian, Marziyeh
Diaz, Jose M Monsalve
Zhang, Tianchi
Ding, Jianru
Wang, Wenyi
Zhang, Ziyi
Jeje, Moubarak
Hoffmann, Henry
Li, Yanjing
Chien, Andrew A.
author_facet Rajasukumar, Andronicus
Su, Jiya
Yuqing
Wang
Su, Tianshuo
Nourian, Marziyeh
Diaz, Jose M Monsalve
Zhang, Tianchi
Ding, Jianru
Wang, Wenyi
Zhang, Ziyi
Jeje, Moubarak
Hoffmann, Henry
Li, Yanjing
Chien, Andrew A.
contents Applications with irregular data structures, data-dependent control flows and fine-grained data transfers (e.g., real-world graph computations) perform poorly on cache-based systems. We propose the UpDown accelerator that supports fine-grained execution with novel architecture mechanisms - lightweight threading, event-driven scheduling, efficient ultra-short threads, and split-transaction DRAM access with software-controlled synchronization. These hardware primitives support software programmable events, enabling high performance on diverse data structures and algorithms. UpDown also supports scalable performance; hardware replication enables programs to scale up performance. Evaluation results show UpDown's flexibility and scalability enable it to outperform CPUs on graph mining and analytics computations by up to 116-195x geomean speedup and more than 4x speedup over prior accelerators. We show that UpDown generates high memory parallelism (~4.6x over CPU) required for memory intensive graph computations. We present measurements that attribute the performance of UpDown (23x architectural advantage) to its individual architectural mechanisms. Finally, we also analyze the area and power cost of UpDown's mechanisms for software programmability.
format Preprint
id arxiv_https___arxiv_org_abs_2407_20773
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle UpDown: Programmable fine-grained Events for Scalable Performance on Irregular Applications
Rajasukumar, Andronicus
Su, Jiya
Yuqing
Wang
Su, Tianshuo
Nourian, Marziyeh
Diaz, Jose M Monsalve
Zhang, Tianchi
Ding, Jianru
Wang, Wenyi
Zhang, Ziyi
Jeje, Moubarak
Hoffmann, Henry
Li, Yanjing
Chien, Andrew A.
Hardware Architecture
Applications with irregular data structures, data-dependent control flows and fine-grained data transfers (e.g., real-world graph computations) perform poorly on cache-based systems. We propose the UpDown accelerator that supports fine-grained execution with novel architecture mechanisms - lightweight threading, event-driven scheduling, efficient ultra-short threads, and split-transaction DRAM access with software-controlled synchronization. These hardware primitives support software programmable events, enabling high performance on diverse data structures and algorithms. UpDown also supports scalable performance; hardware replication enables programs to scale up performance. Evaluation results show UpDown's flexibility and scalability enable it to outperform CPUs on graph mining and analytics computations by up to 116-195x geomean speedup and more than 4x speedup over prior accelerators. We show that UpDown generates high memory parallelism (~4.6x over CPU) required for memory intensive graph computations. We present measurements that attribute the performance of UpDown (23x architectural advantage) to its individual architectural mechanisms. Finally, we also analyze the area and power cost of UpDown's mechanisms for software programmability.
title UpDown: Programmable fine-grained Events for Scalable Performance on Irregular Applications
topic Hardware Architecture
url https://arxiv.org/abs/2407.20773