A DRAM Backend for The Impulse Memory System

نویسنده

  • Lixin Zhang
چکیده

T h e Im pulse A d ap tab le M e m o ry S ystem exp oses D R A M access patterns not seen in conven tion al m em ory system s. F or instance, it can generate 32 D R A M a ccesses ea ch o f w h ic h requests a fou r-byte w ord in 32 c yc le s . C on ven tion al D R A M b acken d s are o p tim ized for a ccesses that request fu ll cach e lines. T h e y m ay not b e able to handle sm aller a ccesses effectively . In this docum ent, w e describe and evaluate a D R A M b acken d that reduces the average D R A M access laten cy b y exp lo itin g the potential p aralle lism o f D R A M a ccesses in the Im pulse system . W e design the D R A M b acken d b y studying ea ch o f its im portant design options: D R A M organization , hot ro w p o licy , d yn am ic re­ ordering o f D R A M accesses, and in terleavin g o f D R A M banks. T h e experim ental results obtained from the execution-driven sim ulator P ain t [10] sh o w that, com pared to a conven tion al D R A M backend, the proposed b acken d can reduce the average D R A M access la ten cy b y up to 98 % , the average m em ory c y c le s b y up to 90 % , and the execu tio n tim e b y up to 80% . This effort was sponsored in part by the Defense Advanced Research Projects Agency (DARPA) and the Air Force Research Laboratory (AFRL) under agreement number F30602-98-1-0101 and DARPA Order Numbers F393/00-01 and F376/00. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official polices or endorsements, either express or implied, of DARPA, AFRL, or the US Government. 1 In tr o d u c t io n 3 2 O v e r v ie w o f T h e Im p u ls e M e m o r y S y s te m 3 2.1 H ardw are O r g a n iz a t io n ......................................................................................................................................... 4 2.2 R em ap p in g A lg o r i t h m s ......................................................................................................................................... 6 3 D R A M B a s ic s 7 3.1 Syn ch ro n ou s D R A M ............................................................................................................................................. 7 3.2 D ire ct R am bus D R A M ......................................................................................................................................... 8

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Impulse: Memory system support for scientific applications

Impulse is a new memory system architecture that adds two important features to a traditional memory controller. First, Impulse supports application-specific optimizations through configurable physical address remapping. By remapping physical addresses, applications control how their data is accessed and cached, improving their cache and bus utilization. Second, Impulse supports prefetching at ...

متن کامل

Impulse: Building a Smarter Memory Controller

Impulse is a new memory system architecture that adds two important features to a traditional memory controller. First, Impulse supports application-specific optimizations through configurable physical address remapping. By remapping physical addresses, applications control how their data is accessed and cached, improving their cache and bus utilization. Second, Impulse supports prefetching at ...

متن کامل

BlueCache: A Scalable Distributed Flash-based Key-value Store

A key-value store (KVS), such as memcached and Redis, is widely used as a caching layer to augment the slower persistent backend storage in data centers. DRAM-based KVS provides fast key-value access, but its scalability is limited by the cost, power and space needed by the machine cluster to support a large amount of DRAM. This paper offers a 10X to 100X cheaper solution based on flash storage...

متن کامل

Novel Memory Access Scheduling Algorithms for a Surveillance System

Abstract: The continuously growing functionality of digital video surveillance make the surveillance system integrate more streaming processors for serving more cameras to recoding their raw video streaming data. But the memory subsystem can not provide necessary bandwidth and become the bottleneck of whole system. Therein how to improve the performance of the accessing memory will become a maj...

متن کامل

The Hierarchical Multi-Bank DRAM: A High-Performance Architecture for Memory Integrated with Processors

A microprocessor integrated with DRAM on the same die has the potential to improve system performance by reducing the memory latency and improving the memory bandwidth. However, a high performance microprocessor will typically send more accesses than the DRAM can handle due to the long cycle time of the embedded DRAM, especially in applications with significant memory requirements. A multi-bank...

متن کامل

CAM-based retention-aware DRAM (CRA-DRAM) for refresh power reduction

As the main component for modern main memory system, DRAM stores data by capacitors, which must be refreshed periodically to keep the charges. As the size and speed of DRAM devices continue to increase, the overhead of refresh has caused a great power and performance dissipation. In this paper, we proposed a CAM (content-addressable memory)-based Retention-Aware DRAM (CRA-DRAM) system, a hardwa...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2000