Reconfigurable High Performance Computing: Challenges and Possibilities

1
 Reconfigurable High Performance Computing 
Challenges and Possibilities

• Karen A. Tomko
• Electrical and Computer Engineering and Computer
  Science Department
• email Karen.Tomko_at_uc.edu
• http//www.ececs.uc.edu/ktomko

2
Outline of talk

• High Performance Reconfigurable Computing
• Overview of Reconfigurable Computing
• Floating Point in FPGA
• Reconfigurable High Performance Computing (HPC)
  Systems
• Design and Development Tools
• Taking advantage of Reconfigurable HPC
• An Example Reconfigurable Computing Application
• Challenges/Open Issues
• Lab overview
• People
• Areas of interest
• Current projects

3
Resources

• www.openfpga.org
• Reconfigurable Systems Summer Institute (RSSI),
  NCSA July 2005
• www.ncsa.uiuc.edu/Conferences/RSSI/index.html
• Getting started on the SRC white paper
• D. Caliga, V. Kindratenko, and D. Pointer,
  High-Performance Reconfigurable Computing
  Application Programming in C, January 10, 2006.
  http//www.ncsa.uiuc.edu/pointer/whitepapers.html
  
• Commercial Reconfigurable HPC System/Board/Softwar
  e Vendors (not necessarily a complete list)
• www.celoxica.com
• www.cray.com
• www.impulsec.com
• www.mitrion.com
• www.nallatech.com
• www.sgi.com
• www.srccomp.com
• www.starbridge.com
• www.stoneridgetechnology.com

4
ACL Lab People

• Director (ERC 546)
• Dr. Karen A Tomko 
• Student Researchers (ERC 516)
• Amr Ahmadain
• Olukayode Arole
• Mike Conrad
• Robert Culley
• Ashish Desai

• Arun Janarthanan
• Qingyuan Liu
• Nitin Shekhar
• Prakash Velayutham

5
Research Interests

• High Performance Computing
• Adaptive scientific simulation
• Reconfigurable High Performance Computing
• Partitioning Algorithms
• Reconfigurable Computing
• Low power design for FPGA
• Reconfigurable High Performance Computing
• Multi-technology FPGA
• Debugging Environments
• Performance Estimation/Benchmarking

6
Projects Performance Analysis and Enhancement of
Adaptive Multi-disciplinary Simulation

• Organizations/People
• Student(s) Qingyuan Liu, Prakash Velayutham
• Purpose Explore and improve the performance of
  complex parallel simulations.
• Development suite of performance analysis and
  algorithm development
• Our Initial Tasks
• Fine grained data-centric application profiling
• Adaptive batch scheduling

7
Performance Analysis and Enhancement of Adaptive
Multi-disciplinary Simulation (continued)

• Suggested Reading
• Fine Grained Profiling
• Amol Deshmukh, Qingyuan Liu, Karen Tomko, An
  approach for fine-grained profiling of mesh-based
  parallel programs, PDCS-04 conference
  proceedings, Sept 2004.
• Amol Deshmukh, An Approach For Fine-grained
  Profiling of Parallel Applications, Masters
  Thesis, University of Cincinnati, July 2004.
• Synthetic/Simulated Application
• Qingyuan Liu, Amol Deshmukh, Karen Tomko,
  Synthetic Simulation of Mesh-based Parallel
  Applications Driven by Fine-Grained Profiling,
  PMEO-PDS05 Workshop, IPDPS Conference, April
  2005.

8
Projects Reconfigurable High Performance
Computing

• Organizations/People
• Ohio Supercomputer Center, Springfield
• Current Student(s) Robert Culley, Ashish Desai
• Purpose reconfigurable computing for
  computationally intensive engineering/scientific
  simulation
• FPGA based accelerator for Finite-Difference
  Time-Domain electromagnetics simulation
• Our Primary Tasks
• Deploy FDTD E-field and H-field update
  calculation engine on Reconfigurable HPC System
• Performance analysis of resulting implementation

9
Reconfigurable High Performance Computing
(continued)

• Suggested Reading
• Floating-Point FPGA designs
• Pavle Belanovic and Miriam Leeser, A Library of
  Parameterized Floating-point Modules and Their
  Use, In Proceedings, International Conference on
  Field Programmable Logic and Applications,
  Montpelier, France, Aug. 2002.
• Ronald Scrofano, Ling Zhuo and Viktor K.
  Prasanna, Area-Efficient Evaluation of Arithmetic
  Expressions Using Deeply Pipelined Floating-Point
  Cores, in Proceedings of the 2005 International
  Conference on Engineering of Reconfigurable
  Systems and Algorithms, June 2005, Nevada, USA.
• Prior Work in the ASC Lab
• Sachin Gandhi, An FPGA Implementation of FDTD
  codes for Reconfigurable High Performance
  Computing, Masters Thesis, University of
  Cincinnati, Nov. 04.
• Shuguang Wu, FPGA Implementation of the FDTD
  Algorithm Using Local SRAM, Masters Thesis,
  University of Cincinnati, Nov. 05.

10
Low Power Design for FPGA

• Organizations/People
• Current Student(s) Arun Janarthanan, Kayode
  Arole
• Purpose
• Low power design and power analysis for FPGA
• Our Primary Tasks
• Low power FSM implementation for FPGA
• Hierarchical power profiler for FPGA applications

11
CAD Tools for MT-FPGA

• Organizations/People
• In collaboration with Prof. Beyette
• Current Student(s) Easwar Ramaswamy, Nitin
  Shekhar
• Purpose
• CAD tools and environment for testing and design
  exploration of the MT-FPGA
• Our Primary Tasks
• Automatic generation of VHDL simulation from XML
  description of the MT-FPGA
• Parameterized models of the MT-FPGA components