Notes from October 20th, 2010 meeting

• Topic:
• Parallelizing CDAT's Core
  • Attendants
  • Charles Doutriaux
  • Jim Ahrens
  • Tom Maxwell
  • Claudio Silva
  • Huy Vo
  • Peer-Timo Bremer
  • Valerio Pascucci
  • John Patchett
  • Bob Drach
  • Emanuele Santos
  • Berk Geveci
• Summary
  • Some think we need to focus on use case 1 and 2 in the context of Paraview first and uses these to report quickly to sponsors and show great performance increase
  • Galen should be reporting soon with preliminary result on reading data in
  • There is a strong feeling that we need to think strong and hard on the design
  • seems that temporal/linear parallelization is the main thing we need to focus on for now
  • Should we come up with a vistrails/pythonic language/dialect ?
  • If we tweak the user scripting language, we need to come up with a translator
  • Need to keep the metadata around and pass them to Paraview
  • We should expect 50% of GUI-based users and 50% of scripting-based users
  • Everybody is going to look into more details at the CDAT specific use cases as well, see: Parallelizing CDAT Core Use Cases