Location: Leeds University LIGHT Laboratories Conference Room (Level 4, LIGHT Labs), Clarendon Way, Leeds, LS2 9NL.

Agenda (draft):

1. Progress report (5 mins)

2. Report on survey (15 mins)

3. Summary of hackathon (80 mins)

• Overall (10 mins)
• Algo development via CIL (10 mins)
• Geometry (25 mins)
• Simulations (15mins)
• GE Signa (5 mins)

4. GPU (90 mins)

• Options: (Susan Thorne and Edo)
  • CUDA, OpenCL, OpenACC, OpenMP
  • Thrust, ArrayFire etc
• Where to add GPU support (SIRF, Gadgetron, STIR?) and how to avoid copying data from/to CPU memory (discussion led by Evgueni)
  • Proposal: SIRF should not know about GPU except possibly GPU configuration and tell objects when they’re being constructed to use GPU or CPU memory.
• Deployment: desktop vs cluster vs cloud (Casper, Ben)
  • How to use host GPU on Docker container or VM (if Nvidia and Linux host)