General-Purpose Computation and Visualization on the GPU Cluster

General-Purpose Computation and Visualization on the GPU Cluster PowerPoint PPT Presentation


  • 107 Views
  • Uploaded on
  • Presentation posted in: General

Download Presentation

General-Purpose Computation and Visualization on the GPU Cluster

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


1. General-Purpose Computation and Visualization on the GPU Cluster Zhe Fan Stony Brook University

2. Visualization Lab @ Stony Brook University Project Director: Prof. Arie Kaufman Visual Computing Cluster

3. Cluster Configuration 66 Nodes The first 34 nodes: HP xw8000 workstation Dual Intel Xeon 2.4 GHz nVidia GeForce FX 5800 Ultra The rest 32 nodes: HP xw8200 workstation Dual Intel Xeon 3.6 GHz nVidia Quadro FX 4500

4. Background Graphics Processing Unit (GPU) Data parallel computation High memory bandwidth High performance / cost ratio Becoming more general-purposed architecture General-Purpose Computation on GPU (GPGPU) Computation, simulation Volume rendering, volume processing Many others applications …

5. Background (cont.) New high-level programming languages BrookGPU RapidMind nVidia CUDA Microsoft Accelerator PeakStream

6. Motivations Current Status Mostly on single GPU Mostly in single module

7. Our Previous Work (2004)

8. ZippyGPU OO library toolkit for GPU cluster For development of computations, visualization, and systems combining multiple components of them

9. Multi-GPU Programming Model Tasks: Data partitioning Data communication Network data transfer Data transfer between GPU and main memory Requirements: Easy to program Efficient to execute

10. Approach: GA on GPU cluster Similar to Global Arrays (GA), combines shared-memory and distributed-memory programming models

16. Thank you!

  • Login