Login | Register
My pages Projects Community openCollabNet

Project home

2020-04-07: This site is going to be decommissioned and shut down on 2020-07-01. Please copy and archive any data you wish to keep before that date.

If you were registered and logged in, you could join this project.

Message from the owner(s)

Mostly individual scripts at this point instead of cohesive single tool.
Summary Genetic Algorithm Linpack Optimization
Category testing
License GNU General Public License
Owner(s) jkrauska


Develop a complete genetic algorithm tool set for determining optimal parameters for Linpack runs.


  • GA DNA decode in to Linpack HPL.dat translator
  • Fitness measurer, parent choser/crosser, population variance
  • Use the issue tracker to track features and enhancement requests. It is not just for bugs, pre-populate it with feature descriptions so that potential contributors know what you would like them to work on.

Current Project Assumptions

  • Use perl and bash scripts until functionally solid, then convert to single script.
  • Smaller N values are representative of larger N values

Related resources


For a myriad of reasons, clusters are often judged on their ability to solve dense systems of linear equations. Specifically the "Top 500" super computers in the world are judged using the High Performance Linpack benchmark or "HPL".

Overall Linpack performance can be thought of as a function many parameters:
  • cpu speed and instruction sets
  • memory capacity
  • system bus speeds
  • interconnect topologies, performance and design
  • linear algrebra library optimizations
  • compiler optimizations
  • communication stack and protocol optimizations
  • hpl linpack run parameters
Given that most cluster hardware is already in place, and the high likelihood that compile-time and communications options are generally  slow to change, this investigation focuses on the linpack run parameters.

The hpl benchmark contains several tunable parameters.
Linpack Tuning Document

To most cluster engineers (the authors included) the tuning explanations of the hpl parameters yield little clue as to the underlying effect of varying these parameters.  Not everyone can take a graduate mathematics course in advanced linear algebra in their free time.

Converting Linpack Parameters to Genetic Algorithm DNA string.

  • N           Problem Size
                 (fixed at 10,000 for “quick” fitness testing)
  • NB         Block Size: 0-1023 (10 bits)
  • PMAP     Process Mapping (row or column): 0-1 (1 bit)
  • P & Q     Grid Process Columns and Rows
                 (P x Q=Number of Procs)
                 For 64 Procs there are 7 possibilities (3 bits)
                 (1x64, 2x32, 4x16, 8x8, 16x4, 32x2, 64x1)
  • PFACT    Panel Factorization Method: 0-2 (2 bits)
  • NBMIN   Minimum Columns: 1-15 (4 bits)
  • NDIV      Number of Panel Divisions in Recursion: 0-7 (3 bits)
  • RFACT    Recursive Panel Factor: 0-2 (2 bits)
  • BCAST    Broadcast Method: 0-5 (3 bits)
  • DEPTH    Lookahead Depth: 1-3 (2 bits)
  • SWAP     Swap Algorithm: 0-2 (2 bits)
  • L1          Upper Right Transpose Method: 0-1 (1 bit)
  • U            Panel of Rows U Transpose Method: 0-1 (1 bit)
  • EQUIB     Equilibration Toggle: 0-1 (1 bit)
  • TOTAL     35 bits