硬體輔助記憶體配置器之設計與效能分析

Design and Performance Evaluation of Hardware-Assisted Memory Allocator

林秀書(Hsiu-Shu Lin)

指導教授 ： 鄭福炯

大同大學/電機資訊學院/資訊工程學系所/碩士(2009年)

摘要

由於Java 應用程式及即時嵌入式系統的普及，應用程式對於動態記憶體管理的
需求越來越高，因此高效能記憶體配置器將越來越重要，而硬體之密度越來越
高，成本越來越低，及硬體可省電加速之特色，使得許多研究者注重於硬體輔
助記憶體配置器(hardware-assisted memory allocator 簡稱HAMA)之研究。但多
數的配置器並未考量尺寸盲點問題(size blind spot)及邊界盲點問題(boundary
blind spot)導致記憶體之使用率無法提高。
我們分析各種的硬體輔助記憶體配置器後提出一個硬體輔助記憶體配置
器，符合real-time system 對執行時間的限制，解決尺寸盲點問題及邊界盲點問
題，且不增加硬體花費可利於硬體實作。最後我們以Java 程式實做我們的及近
年來被提出的硬體輔助記憶體配置器，並使用SPECjvm98 效能評估程式量測
記憶體使用率之比率，結果顯示我們提出的硬體輔助記憶體配置器提高了記憶
體使用率。

關鍵字

記憶體管理 ； 記憶體配置 ； 記憶體回收 ； 爪哇語言

並列摘要

Modern programming languages such as Java and C++ highly demand efficient
dynamic memory management to allocate and deallocate blocks of memory
frequently. Furthermore, the density of logic on a chip is increasing dramatically
every year. Therefore, it is common to translate software to hardware to speed up the
execution time and reduce power consumption. Many researchers proposed
hardware-assisted memory allocators to address dynamic memory management.
However, None of those work addresses both the size blind spot and boundary blind
spot problems.
We propose a new hardware-assisted memory allocator to address the above
two problems and evaluate the performance with SPECjvm98. The experimental
results indicated that our hardware-assisted memory allocator effectively improving
the memory utilization.

並列關鍵字

memory management ； memory allocation ； garbage collection ； Java

參考文獻

Bloomington, Indiana, Dec. 1978

[2] Xiao Hui Sun, Jin Lin Wang, Xiao Chen, “An Improvement of TLSF

[3] Weixing JI, Feng Shi, Baojun Qiao, “A self-maintained memory module

supporting DMM,” ACM Compilers, architecture, and synthesis for embedded

[4] K.C. Knowlton, “A fast storage allocator,” Comm. ACM, Vol. 8, pp. 623–625, Oct.