Efficient FDK Algorithms on SIMD-accelerated Processors

• Item type: SIG Technical Reports(1)
• 公開日: 2020-07-23
• タイトル: Efficient FDK Algorithms on SIMD-accelerated Processors
• タイトル / 言語: en
• タイトル: Efficient FDK Algorithms on SIMD-accelerated Processors
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: アルゴリズム高速化
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: AIST-Tokyo Tech Real World Big-Data Computation Open Innovation Laboratory, National Institute of Advanced Industrial Science and Technology
• 著者所属: AIST-Tokyo Tech Real World Big-Data Computation Open Innovation Laboratory, National Institute of Advanced Industrial Science and Technology／RIKEN Center for Computational Science
• 著者所属: National Institute of Advanced Industrial Science and Technology
• 著者所属: National Institute of Advanced Industrial Science and Technology
• 著者所属: National Institute of Advanced Industrial Science and Technology
• 著者所属: Tokyo Institute of Technology／RIKEN Center for Computational Science
• 著者所属(英): en
• 著者所属(英): AIST-Tokyo Tech Real World Big-Data Computation Open Innovation Laboratory, National Institute of Advanced Industrial Science and Technology
• 著者所属(英): en
• 著者所属(英): AIST-Tokyo Tech Real World Big-Data Computation Open Innovation Laboratory, National Institute of Advanced Industrial Science and Technology / RIKEN Center for Computational Science
• 著者所属(英): en
• 著者所属(英): National Institute of Advanced Industrial Science and Technology
• 著者所属(英): en
• 著者所属(英): National Institute of Advanced Industrial Science and Technology
• 著者所属(英): en
• 著者所属(英): National Institute of Advanced Industrial Science and Technology
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology / RIKEN Center for Computational Science
• 著者名: Peng, Chen; Mohamed, Wahib; Shinichiro, Takizawa; Takahiro, Hirofuchi; Hirotaka, Ogawa; Satoshi, Matsuoka
• 著者名(英): Peng, Chen; Mohamed, Wahib; Shinichiro, Takizawa; Takahiro, Hirofuchi; Hirotaka, Ogawa; Satoshi, Matsuoka
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Computed Tomography (CT) is a widely used 3D imaging technology that requires compute-intense algorithms to generate volume data (or images). We propose a collection of novel back-projection algorithms that reduce the arithmetic computation and improve data locality. We also implement novel algorithms as efficient back-projection kernels that are performance portable over a wide range of CPUs. Unlike the conventional approaches that use OpenMP and target-specific SIMD intrinsics, we employ a high-level OpenCL implementation to generate the vectorized code and use the OpenCL local memory to prefetch the pixels at sub-pixel precision in a regular memory access fashion. Performance evaluation using a variety of Intel CPUs generations demonstrates that our back-projection implementation runs up to 10 times faster than the multi-threading optimized implementation.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Computed Tomography (CT) is a widely used 3D imaging technology that requires compute-intense algorithms to generate volume data (or images). We propose a collection of novel back-projection algorithms that reduce the arithmetic computation and improve data locality. We also implement novel algorithms as efficient back-projection kernels that are performance portable over a wide range of CPUs. Unlike the conventional approaches that use OpenMP and target-specific SIMD intrinsics, we employ a high-level OpenCL implementation to generate the vectorized code and use the OpenCL local memory to prefetch the pixels at sub-pixel precision in a regular memory access fashion. Performance evaluation using a variety of Intel CPUs generations demonstrates that our back-projection implementation runs up to 10 times faster than the multi-threading optimized implementation.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10463942
• 書誌情報: 研究報告ハイパフォーマンスコンピューティング（HPC）; 巻 2020-HPC-175, 号 6, p. 1-11, 発行日 2020-07-23
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8841
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会