Post

Cholla Prep

Posted
By Robert Caddy 1 min read

MHD Debugging

I spent some more time bug hunting and found that the instability only shows up when a number is not exactly expressible in binary. After discussin this with Evan they pointed out that it could just be an artifact of turning a 3D algorithm into a 1D algorithm. So instead of continueing bug hunting I’m going to switch to implementing MHD into Cholla.

Miscellaneous Cholla Prep

Switching to working on Cholla requires quite a bit of prep so here’s what I did this week:

  • Cleaned up a few notes on the Hydro-Sandbox code so that it can be put away for awhile
  • Submit account request for the ORNL supercomputer system
  • Update my fork of Cholla
  • Figure out how to access POWER9/V100 nodes at Pitt’s H2P cluster
  • Update the modules I’m using on H2P and the POWER9 nodes
  • Get Cholla to compile on H2P
  • Get Cholla running on H2P in serial and parallel
  • Meet with Damon to discuss testing schemes

Near future plans

  • Check if I can get cholla to run with GPU-MPI on H2P
  • Check if Cholla’s regression test is working and fix it if it isn’t
  • Figure out how I want to do future tests in Cholla
  • Learn CUDA
  • Learn how Cholla works

Miscellaneous

MHD, Cholla
Constrained Transport Testing
This post is licensed under CC BY 4.0 by the author.