WEKO3
アイテム
{"_buckets": {"deposit": "09a7d57d-f20c-436f-a47d-fa88f64a719c"}, "_deposit": {"created_by": 3, "id": "1151", "owners": [3], "pid": {"revision_id": 0, "type": "depid", "value": "1151"}, "status": "published"}, "_oai": {"id": "oai:uec.repo.nii.ac.jp:00001151", "sets": ["14"]}, "author_link": ["7125"], "control_number": "1151", "item_10006_date_granted_11": {"attribute_name": "学位授与年月日", "attribute_value_mlt": [{"subitem_dategranted": "2009-03-24"}]}, "item_10006_degree_grantor_9": {"attribute_name": "学位授与機関", "attribute_value_mlt": [{"subitem_degreegrantor": [{"subitem_degreegrantor_name": "電気通信大学"}]}]}, "item_10006_degree_name_8": {"attribute_name": "学位名", "attribute_value_mlt": [{"subitem_degreename": "博士(工学)"}]}, "item_10006_description_10": {"attribute_name": "学位授与年度", "attribute_value_mlt": [{"subitem_description": "2008", "subitem_description_type": "Other"}]}, "item_10006_description_7": {"attribute_name": "抄録", "attribute_value_mlt": [{"subitem_description": "GPU (Graphics Processing Unit) is suitable for parallel processing and has highperformance than existing CPU (Central Processing Unit), so GPU is now beingused for various applications. Many applications are getting speed-up using GPGPU(General-Purpose computation using GPUs), while the difficulty of GPU programmingis under problem.Application programmer must use special programming languages and librariesfor GPGPU programming to create programs running on GPUs, but using suchlanguages and libraries are difficult for programmers because it is necessary to learnGPU’s architecture, execution model of GPU, and programming language’s specification.Therefore it is a challenge of GPGPU to make GPGPU programmingeasy.In this paper, two approaches which make using GPGPU easy with commodityparallel programming environments are proposed.The first approach is a “parallel computation library” which hides parallel programmingfrom application programmers by offering parallelized computation kernels.This library has APIs similar to CPU’s libraries and computes using CPUsand GPUs. To get a high performance in anytime, this library divides the problemand executes them in parallel using CPUs and GPUs when problem size isbig enough and a balance between CPU’s performance and GPU’s performance issuitable, and uses an only CPU in other cases with appropriate balance suggestedby a tuning-script and a divide-module. “GPUPC GEMM Library” is a parallelmatrix-matrix multiplication library based on this approach. This library can get ahigh performance with GPGPU systems without parallel programming.The second approach is a “parallel programming environment” which offers a parallelprogramming environment like as existing parallel programming environmentsused on CPU. “OMPCUDA” is an OpenMP implementation for CUDA based onthis approach. OMPCUDA can execute parallel regions of OpenMP programs ona GPU, so that programmers can execute parallel programs on a GPU withoutlearning and writing GPU-specific parallel programs.Contributions of this study are proposal of new approaches which make usingGPGPU easy for many application programmers and evaluation of the effectivenessof approaches by implementing a library (GPUPC GEMM Library) and a system(OMPCUDA). Results of this study are significant from the perspective of indicatinghow easy-to-use environment for application programmers in computing systemshaving GPU should be.", "subitem_description_type": "Abstract"}]}, "item_creator": {"attribute_name": "著者", "attribute_type": "creator", "attribute_value_mlt": [{"creatorNames": [{"creatorName": "大島, 聡史", "creatorNameLang": "ja"}, {"creatorName": "オオシマ, サトシ", "creatorNameLang": "ja-Kana"}, {"creatorName": "Ohshima, Satoshi", "creatorNameLang": "en"}], "nameIdentifiers": [{"nameIdentifier": "7125", "nameIdentifierScheme": "WEKO"}]}]}, "item_files": {"attribute_name": "ファイル情報", "attribute_type": "file", "attribute_value_mlt": [{"accessrole": "open_date", "date": [{"dateType": "Available", "dateValue": "2016-09-16"}], "displaytype": "detail", "download_preview_message": "", "file_order": 0, "filename": "9000000354.pdf", "filesize": [{"value": "3.2 MB"}], "format": "application/pdf", "future_date_message": "", "is_thumbnail": false, "licensetype": "license_free", "mimetype": "application/pdf", "size": 3200000.0, "url": {"label": "9000000354.pdf", "url": "https://uec.repo.nii.ac.jp/record/1151/files/9000000354.pdf"}, "version_id": "1e795335-b397-4618-a7d0-71c0f94a2707"}]}, "item_language": {"attribute_name": "言語", "attribute_value_mlt": [{"subitem_language": "jpn"}]}, "item_resource_type": {"attribute_name": "資源タイプ", "attribute_value_mlt": [{"resourcetype": "thesis", "resourceuri": "http://purl.org/coar/resource_type/c_46ec"}]}, "item_title": "汎用プログラミング環境を用いたGPGPUに関する研究", "item_titles": {"attribute_name": "タイトル", "attribute_value_mlt": [{"subitem_title": "汎用プログラミング環境を用いたGPGPUに関する研究", "subitem_title_language": "ja"}, {"subitem_title": "Study of GPGPU Using Commodity Programming Environment", "subitem_title_language": "en"}]}, "item_type_id": "10006", "owner": "3", "path": ["14"], "permalink_uri": "https://uec.repo.nii.ac.jp/records/1151", "pubdate": {"attribute_name": "PubDate", "attribute_value": "2009-03-24"}, "publish_date": "2009-03-24", "publish_status": "0", "recid": "1151", "relation": {}, "relation_version_is_last": true, "title": ["汎用プログラミング環境を用いたGPGPUに関する研究"], "weko_shared_id": -1}
汎用プログラミング環境を用いたGPGPUに関する研究
https://uec.repo.nii.ac.jp/records/1151
https://uec.repo.nii.ac.jp/records/115135639f8e-7952-4e84-b01c-b337269d8763
名前 / ファイル | ライセンス | アクション |
---|---|---|
9000000354.pdf (3.2 MB)
|
|
Item type | 学位論文 / Thesis or Dissertation(1) | |||||
---|---|---|---|---|---|---|
公開日 | 2009-03-24 | |||||
タイトル | ||||||
言語 | ja | |||||
タイトル | 汎用プログラミング環境を用いたGPGPUに関する研究 | |||||
タイトル | ||||||
言語 | en | |||||
タイトル | Study of GPGPU Using Commodity Programming Environment | |||||
言語 | ||||||
言語 | jpn | |||||
資源タイプ | ||||||
資源タイプ識別子 | http://purl.org/coar/resource_type/c_46ec | |||||
資源タイプ | thesis | |||||
著者 |
大島, 聡史
× 大島, 聡史 |
|||||
抄録 | ||||||
内容記述タイプ | Abstract | |||||
内容記述 | GPU (Graphics Processing Unit) is suitable for parallel processing and has highperformance than existing CPU (Central Processing Unit), so GPU is now beingused for various applications. Many applications are getting speed-up using GPGPU(General-Purpose computation using GPUs), while the difficulty of GPU programmingis under problem.Application programmer must use special programming languages and librariesfor GPGPU programming to create programs running on GPUs, but using suchlanguages and libraries are difficult for programmers because it is necessary to learnGPU’s architecture, execution model of GPU, and programming language’s specification.Therefore it is a challenge of GPGPU to make GPGPU programmingeasy.In this paper, two approaches which make using GPGPU easy with commodityparallel programming environments are proposed.The first approach is a “parallel computation library” which hides parallel programmingfrom application programmers by offering parallelized computation kernels.This library has APIs similar to CPU’s libraries and computes using CPUsand GPUs. To get a high performance in anytime, this library divides the problemand executes them in parallel using CPUs and GPUs when problem size isbig enough and a balance between CPU’s performance and GPU’s performance issuitable, and uses an only CPU in other cases with appropriate balance suggestedby a tuning-script and a divide-module. “GPUPC GEMM Library” is a parallelmatrix-matrix multiplication library based on this approach. This library can get ahigh performance with GPGPU systems without parallel programming.The second approach is a “parallel programming environment” which offers a parallelprogramming environment like as existing parallel programming environmentsused on CPU. “OMPCUDA” is an OpenMP implementation for CUDA based onthis approach. OMPCUDA can execute parallel regions of OpenMP programs ona GPU, so that programmers can execute parallel programs on a GPU withoutlearning and writing GPU-specific parallel programs.Contributions of this study are proposal of new approaches which make usingGPGPU easy for many application programmers and evaluation of the effectivenessof approaches by implementing a library (GPUPC GEMM Library) and a system(OMPCUDA). Results of this study are significant from the perspective of indicatinghow easy-to-use environment for application programmers in computing systemshaving GPU should be. | |||||
学位名 | ||||||
学位名 | 博士(工学) | |||||
学位授与機関 | ||||||
学位授与機関名 | 電気通信大学 | |||||
学位授与年度 | ||||||
内容記述タイプ | Other | |||||
内容記述 | 2008 | |||||
学位授与年月日 | ||||||
学位授与年月日 | 2009-03-24 |