Horace framework redesign ideas

Horace framework redesign ideas

Overview of Horace and Herbert

• A scientific overview of Horace is given in the Horace paper.

• Brief overview of data classes and functionalities of Horace and Herbert

Current proposals

• AB's prototype distributed computing implementation

• AB's proposal for refactoring file operations

• AB's proposal for refactoring the sqw and dnd classes

PACE project objectives

• PACE project objectives

Taken from the PACE project brief (to be put online once sensitive information has been redacted).

Possible discussion topics

• What parts of Horace needs to be rewritten / refactored in order to accomplish the goals of PACE?
  • Should the rewrite stay with using Matlab, or should it use C++ or Python?
  • If Matlab is not used, should the Horace be rewritten entirely in another language?
  • What issues (re: maintenance, debugging, etc.) would be raised if we have a mix of Matlab and non-Matlab in the core of Horace?
• Given the objectives of PACE, how much developer effort should be spent on the Horace framework vs the other objectives (TobyFit, SpinW, Phonons, powders)?
• How important is it to support the use of Horace/PACE without needing licenses for:
  • Matlab toolkits (e.g. the Parallel Computing Toolkit)?
  • Matlab itself (no-license running)?
• How important is a parallel file system (where multiple nodes can read / write different parts of the same file on a shared network drive at the same time)?
  • How widespread is this capability on different machines or clusters?
  • The use case for this is whole-file element-by-element operations such as scaling all pixels or subtracting one sqw-file from another.

Rough estimates for developer time required for each task / work-package

Very, very approximate time estimates for various options for PACE

• Rough time estimates for developer time for PACE