書誌事項
- タイトル別名
-
- A study of improving practicality of framework for developing Agent Simulation using GPGPU
この論文をさがす
抄録
The purpose of this study is to improve practicality of a development framework for agent simulations. Recently, multi agent simulation (MAS) is necessary for fields of social science, such as study of epidemics of infectious diseases, analysis of traffic jam and understanding of social networks. These simulations require a large amount of time for processing because they have large scale agents and interactions. Some languages and environments including such as Repast-HPC, D-MASON and XAXIS, are developed that can execute simulations in parallel. They can execute simulations with short time. However, they require users to learn about parallel processing and how to describe those process with a programming language. In our previous research, we studied and developed framework that can execute simulations in parallel and can be written with Python that is easy to write and read program. The framework can speed up simulation by translating to GPGPU program from python program. However, it could not hold coherence of simulation executing between CPU and GPU, and was restricted speed up efficiency in some kind of simulation. In this paper, we will provide functions to hold coherence between CPU executing and GPU executing, and to optimize the data transfer between main memory and GPU memory and the execution in GPU. We also show some our practices to improve extendibility of the framework and to facilitate users’ understanding of the framework usage.
収録刊行物
-
- 法政大学大学院紀要. 情報科学研究科編
-
法政大学大学院紀要. 情報科学研究科編 16 1-6, 2021-03-24
法政大学大学院情報科学研究科
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1390009225533195520
-
- NII論文ID
- 120007119699
-
- NII書誌ID
- AA12746425
-
- HANDLE
- 10114/00023876
-
- ISSN
- 24321192
-
- 本文言語コード
- ja
-
- データソース種別
-
- JaLC
- IRDB
- CiNii Articles
-
- 抄録ライセンスフラグ
- 使用可