Auto-tuning FFT using GPU

2011 Fiscal Year Final Research Report

• Project/Area Number: 22680002
• Research Category: Grant-in-Aid for Young Scientists (A)
• Allocation Type: Single-year Grants
• Research Field: Software
• Research Institution: Tokyo Institute of Technology
• Principal Investigator: NUKADA Akira 東京工業大学, 学術国際情報センター, 産学官連携研究員 (40545688)
• Project Period (FY): 2010 – 2011
• Keywords: ソフトウェア工学

Research Abstract

We have developed the NukadaFFT library, which is an auto-tuning FFT library for NVIDIA CUDA GPUs. It outperforms NVIDIA's CUFFT library in many cases. We also implemented a multi-GPU version for both single-node with multi-GPUs and multi-node, and achieved further speed-up.

Research Products

• [Journal Article] スーパーコンピュータTSUBAME 2. 0におけるLinpack性能1ペタフロップス超の達成 (2011)
  • Author(s): 遠藤敏夫, 額田彰, 松岡聡
  • Journal Title: 情報処理学会論文誌コンピューティングシステム Volume: Vol.4, No.4(ACS35) Pages: 169-179
  • Peer Reviewed
• [Journal Article] CUDAによる高速フーリエ変換,応用数理 (2010)
  • Author(s): 額田彰
  • Journal Title: 応用数理学会 Volume: 第20巻,第2号 Pages: 37-43
  • Peer Reviewed
• [Journal Article] 異種アクセラレータを持つTSUBAMEスーパーコンピュータのLinpack評価,応用数理 (2010)
  • Author(s): 遠藤敏夫, 額田彰, 松岡聡
  • Journal Title: 応用数理学会 Volume: 第20巻,第2号 Pages: 29-36
  • Peer Reviewed
• [Presentation] High Performance 3-D FFT using multiple CUDA GPUs (2012)
  • Author(s): Akira Nukada, Yutaka Maruyama, Satoshi Matsuoka
  • Organizer: In Proceedings of the Fifth Workshop on General Purpose Processing using Graphics Processing Units(GPGPU-5) in conjunction with ACM ASPLOS XVII
  • Place of Presentation: London, UK, ACM Press
  • Year and Date: 2012-03-03
• [Presentation] グリーンスパコンTSUBAME2. 0における電力危機対応運用 (2011)
  • Author(s): 遠藤敏夫,松岡聡,額田彰,長坂真路,四津匡康
  • Organizer: 情報処理学会研究報告
  • Place of Presentation: 札幌
  • Year and Date: 2011-11-28
• [Presentation] Peta-scale Phase-Field Simulation for Dendritic Solidification on the TSUBAME 2. 0 Supercomputer (2011)
  • Author(s): Takashi Shimokawabe, Takayuki Aoki, Tomohiro Takaki, Akinori Yamanaka, Akira Nukada, Toshio Endo, Naoya Maruyama, and Satoshi Matsuoka
  • Organizer: In Proc. of 2011 ACM/IEEE International Conference for High Performance, Networking, Storage, and Analysis(SC' 11)
  • Place of Presentation: Seattle, ACM Press
  • Year and Date: 2011-11-15
• [Presentation] Hamming Color Code for Dense and Robust One-shot 3D Scanning (2011)
  • Author(s): Shuntaro Yamazaki, Akira Nukada, Masaaki Mochimaru
  • Organizer: In Proc. of the 2011 British Machine Vision Conference
  • Place of Presentation: Dundee, Scotland, Springer
  • Year and Date: 2011-08-30
• [Presentation] Fast Fourier Transform for AMD GPUs (2011)
  • Author(s): Akira Nukada
  • Organizer: AMD Fusion Developer Summit 2011
  • Place of Presentation: Bellevue, WA
  • Year and Date: 2011-06-15
• [Presentation] スーパーコンピュータTSUBAME 2. 0におけるLinpack性能1ペタフロップス超の達成 (2011)
  • Author(s): 遠藤敏夫,額田彰,松岡聡
  • Organizer: 先進的計算基盤システムシンポジウム(SACSIS2011)
  • Place of Presentation: 秋葉原
  • Year and Date: 2011-05-27
• [Presentation] NVCR : A Transparent Checkpoint-Restart Library for NVIDIA CUDA (2011)
  • Author(s): Akira Nukada, Hiroyuki Takizawa, and Satoshi Matsuoka
  • Organizer: In Proc. of 20th Heterogeneity in Computing Workshop(HCW 2011), in conjunction with IPDPS 2011
  • Place of Presentation: Anchorage, AK, USA
  • Year and Date: 2011-05-16
• [Presentation] Low-overhead diskless checkpoint for hybrid computing systems (2010)
  • Author(s): Leonardo Bautista Gomez, Akira Nukada, Naoya Maruyama, Franck Cappello and Satoshi Matsuoka
  • Organizer: In Proc. of International Conference on High Performance Computing(HiPC 2010)
  • Place of Presentation: Goa, India
  • Year and Date: 2010-12-20
• [Presentation] Efficient PageRank on GPU Clusters (2010)
  • Author(s): Ali Cevehir, Cevdet Aykanat, Ata Turk, B. Barla Cambazoglu, Akira Nukada and Satoshi Matsuoka
  • Organizer: 情報処理学会研究報告
  • Place of Presentation: 札幌
  • Year and Date: 2010-12-17
• [Presentation] ヘテロ型スーパーコンピュータTSUBAME 2. 0のLinpackによる性能評価 (2010)
  • Author(s): 遠藤敏夫,額田彰,松岡聡
  • Organizer: 情報処理学会研究報告
  • Place of Presentation: 札幌
  • Year and Date: 2010-12-16
• [Presentation] GPUにおけるモデルに基づいた電力効率の最適化 (2010)
  • Author(s): 長坂仁,丸山直也,額田彰,遠藤敏夫,松岡聡
  • Organizer: 情報処理学会研究報告
  • Place of Presentation: 札幌
  • Year and Date: 2010-12-16
• [Presentation] An 80-Fold Speedup, 15. 0 TFlops, Full GPU Acceleration of Non-Hydrostatic Weather Model ASUCA Production Code (2010)
  • Author(s): Takashi Shimokawabe, Takayuki Aoki, Chiashi Muroi, Junichi Ishida, Kohei Kawano, Toshio Endo, Akira Nukada, Naoya Maruyama and Satoshi Matsuoka
  • Organizer: In Proc. of the 2010 ACM/IEEE conference on Supercomputing(SC' 10)
  • Place of Presentation: New Orleans, IEEE Press
  • Year and Date: 2010-11-17
• [Presentation] NukadaFFT : An Auto-Tuning FFT Library for CUDA GPUs (2010)
  • Author(s): Akira Nukada and Satoshi Matsuoka
  • Organizer: NVIDIA GPU Technology Conference 2010
  • Place of Presentation: Research Summit Poster, San Jose
  • Year and Date: 2010-09-22
• [Presentation] Statistical Power Modeling of GPU Kernels Using Performance Counters (2010)
  • Author(s): Hitoshi Nagasaka, Naoya Maruyama, Akira Nukada, Toshio Endo and Satoshi Matsuoka
  • Organizer: Proceedings of the First International Green Computing Conference(IGCC' 10)
  • Place of Presentation: Chicago
  • Year and Date: 2010-08-17
• [Presentation] High Performance Conjugate Gradient Solver on Multi-GPU Clusters Using Hypergraph Partitioning (2010)
  • Author(s): Ali Cevahir, Akira Nukada, and Satoshi Matsuoka
  • Organizer: Computer Science. Research and Development
  • Place of Presentation: Hamburg, Germany
  • Year and Date: 2010-05-31
• [Presentation] Fast Fourier Transform using CUDA GPUs (2010)
  • Author(s): Akira Nukada and Satoshi Matsuoka
  • Organizer: ETHZ-Tokyo Tech Workshop : Computing with GPUs, Cells, and Multicores
  • Place of Presentation: Zurich, Switzerland
  • Year and Date: 2010-05-11
• [Presentation] Linpack Evaluation on a Supercomputer with Heterogeneous Accelerators (2010)
  • Author(s): Toshio Endo, Akira Nukada, Satoshi Matsuoka, and Naoya Maruyama
  • Organizer: In Proceedings of 24th IEEE International Parallel & Distributed Processing Symposium(IPDPS 2010)
  • Place of Presentation: Atlanta
  • Year and Date: 2010-04-21
• [Presentation] A High-Performance Fault-Tolerant Software Framework for Memory on Commodity GPUs (2010)
  • Author(s): Naoya Maruyama, Akira Nukada, and Satoshi Matsuoka
  • Organizer: In Proceedings of 24th IEEE International Parallel and Distributed Processing Symposium
  • Place of Presentation: Atlanta
  • Year and Date: 2010-04-20
• [Book] Chapter 11 of "Software Automatic Tuning : From Concepts to the State-of-the-Art Results" (2010)
  • Author(s): Tamito Kajiyama, Akira Nukada, Reiji Suda, Hidehiko Hasegawa, and Akira Nishida
  • Total Pages: 175-192
• [Remarks] 下記URLにて本研究の成果の一部であるNukadaFFTライブラリソフトウェアを公開
  • URL: http://matsu-www.is.titech.ac.jp/~nukada/nufft/