ODK kickoff meeting – Description of WP3

ODK-kickoff-WP3.md

class: middle, center

WP3 – Component Architecture

by Luca De Feo

Sep 2, 2015. OpenDreamKit kickoff meeting.

view as a slideshow

---

Components

• separate modules inside a unique software,
• separate softwares interacting through
  • library calls,
  • APIs (e.g.: web APIs).

When combined together, they make up a full VRE.

Examples

• A math library: libGAP, libPari, Flint, ...
• A module: integers, rationals, number fields, finite fields, ...
• A full blown software: GAP, Pari/GP, Sagemath, Jupyter, ...

---

We want them

• Portable: mobile, desktops, cloud, ...
• Modular: easy install, buildi, test, distribute, remix, ...
• Flexible: personal computation, HPC, parallel platforms, ...
• Open:
  • open source, of course,
  • well documented,
  • easy to inspect: what algorithm is run? What are the alternatives?

---

The big picture

• OSS is great for research: powerful tools, open scrutiny, reproducibility, community, ...
• But there are still many areas where OSS is not so great:
  • Deep inspection: analyze algorithms & optimizations, compare frameworks;
  • Semantic interfaces: software that understands mathematics and communicates them;
  • Entry barriers: improve docs, UX, support for different platforms;
  • Credit: aknowledge contributions, peer reviews.

Let's improve these and bring about the next generation of mathematical software!

---

The tasks (1/2)

1. Portability (UVSQ):

• SageMath & co. on Windows (and other exotic platforms),
• continuous integration.

2. Interfaces between systems (UPSud):

• passing mathematical objects via SCSCP (OpenMath based RPC for computer algebra),
• automatic discovery of functionality, documentation, ...
• performance.

3. Modularisation and packaging (UVSQ):

• Virtualisation,
• Review community packaging & distribution in Python/SageMath/GAP, come up with a new solution,
• SageMath in Linux standard package systems.

4. Simulagora integration (Logilab):

• Deliver Simulagora VMs every 6 months.

---

The tasks (2/2)

5. Component architecture for High Performance Computing and Parallelism (UJF):

• Parallelism-friendly composition of components.

6. Document and modularise SageMathCloud’s codebase (UPSud):

• SMC is a moving target, evolving very quickly. Coordination with UWashington is top priority.

7. Improving the development workflow in mathematical software (UVSQ):

• Bug reporting, community patches, integration in VRE.

8. Python interface for OOMMF micromagnetic simulation library (Southampton):

• Demonstrator for Object Oriented MicroMagnetic Framework in Jupyter.