Planting a Forest in Sky: Harnessing Parallelism in Skyrmion Racetrack Memory for Efficient Random Forest Data Placement

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/95557

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/95557

Title:	Planting a Forest in Sky: Harnessing Parallelism in Skyrmion Racetrack Memory for Efficient Random Forest Data Placement
Authors:	黃意勛;Huang, Yi-Hsun
Contributors:	資訊工程學系
Keywords:	斯格明子;賽道記憶體;隨機森林;高平行度;Skyrmion;Racetrack memory;Random forest;Parallelism
Date:	2024-07-23
Issue Date:	2024-10-09 17:00:48 (UTC+8)
Publisher:	國立中央大學
Abstract:	斯格明子賽道記憶體（Sky-RM）因其高存儲密度的潛力而備受關注，尤其是對於資料儲存需求不斷增長的現今尤為重要。Sky-RM 利用斯格明子的奈米尺度尺寸，與傳統記憶體技術相比顯著減少了物理空間需求。然而，由於其獨特的特性，包括位移、生成和消除，如果缺乏有效的演算法來管理這些操作，Sky-RM 的整體性能可能顯著低於傳統的動態隨機存取記憶體（DRAM）。因此，本論文致力於開發一種具有高並行性、低延遲、低能耗和高空間利用率的資料擺放方式。此外，我們將我們的方法應用到基於隨機森林的機器學習框架中，以檢驗是否實現了高平行度以及延遲和能耗的減少。;Skyrmion-based Racetrack Memory(Sky-RM) has gained attention due to its potential to offer high storage density, which is increasingly critical as the demand for data storage continues to grow. Sky-RM leverages the nanoscale size of skyrmions, allowing for a significant reduction in physical space requirements compared to traditional memory technologies. However, due to its unique characteristics, including shifting, generating, and eliminating, Sky-RM can exhibit significantly poorer overall performance compared to traditional DRAM if it lacks a robust algorithm to manage these operations effectively. Therefore, this paper is dedicated to develop a high parallelism, low latency, low energy consumption and high space utilization placement strategy. Furthermore, we integrate our method onto the random forest based machine learning framework to see whether high parallelism and reductions in latency and energy consumption have been achieved.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	119	View/Open

社群 sharing

Loading...