1) Run HF on GPU for a 18p9 test-case to test workflow

2) Run HF on GPU for a prior validation run to double check results

Once HF is completed use the "cybershake script" to copy results to HPC and run BB subsequently

3) Stress test the GPU HF code on a large subduction simulation

4) Do a pre-test (full 18p9) run (including Alpine) to verify workflow / computation time.

5) Generate sources / VMs and get underway.




Requirements

  • Automated test to see if parameters specified in gm-sim version are being used in each individual task
  • Analyse CH usage to see if it is within expectations and that the CH estimation is working appropriately
  • Test with a large #realisations to make sure there is sufficient db performance


Timeline



1st April - Cybershake Starts (ideal case)

20th April - Cybershake must start by now.

31st May - CH runs out


Key Software tasks

  • Non-uniform grid generation (Jason)
  • NHM Source generation on the new workflow (? (James, Jason or Sung))
  • vs30 modifications (Viktor)
  • Workflow testing (as above - Jonney)


  • No labels