- Fixed some bugs in the divergence checking code and improved output formatting
- Fixed some bugs and formatting in the plotting code and tried a parallel version that turned out to actually be slower
- Found the issue with why Cholla didn’t seem to be updating. I thought setting the time between outputs to zero meant that it would output on every time step when actually it just forced the time step to be zero. Setting the compile time macro
OUTPUT_ALWAYSis what actually forces cholla to write out on every time step.
- Spent more time identifying bugs. The list of identified bugs thus far is:
- time step = 0.1 when
-Gis set which is required for NVCC to enable GPU debuggers. Unclear why at this time, might be a Cholla bug rather than an MHD bug
- X and Y velocity bug on left boundary. They both spike up really high on the first time step. Likely an issue with the boundary condition indexing
- Large errors in central region, likely some of the CT slopes or the magnetic field update.
- time step = 0.1 when
I spend quite a bit of time reading up on dynamos. I read some basics on them and have either read or am working on reading these papers. The dynamos of greatest interest to my research are the \( \alpha \) and \( \omega \) effects which together form the \( \alpha\omega \) dynamo. This dynamo is believed to be the primary effect that sustains the galactic magnetic fields. The \( \alpha \) effect is caused by helical fluid motion moving perpendicular to the magnetic field direction which causes the magnetic field to twist and turn into a shape resembling the letter \( \alpha \). The \( \omega \) effect is simply cause by differential rotation “winding up” the magnetic field. Together the two and take a begining magnetic field and stretch and fold it so that it returns to it’s original shape with an increased strength.
- Pakmer et al. 2018, Faraday rotation maps of disc galaxies
- Ntormousi et al. 2020, A dynamo amplifying the magnetic field of a Milky-Way-like galaxy
- Pariev & Colgate 2007, A Magnetic Alpha-Omega Dynamo in Active Galactic Nuclei Disks: I. The Hydrodynamics of Star-Disk Collisions and Keplerian Flow
- Pariev et al. 2007, A Magnetic Alpha-Omega Dynamo in Active Galactic Nuclei Disks: II. Magnetic Field Generation, Theories and Simulations
- NASA The Solar Dynamo
- Dynamo Theory by Andrew Gilbert
- Mean Field Theory Dynamo Solutions
I got interested in Rust as an HPC programming language so I did some poking around. Looks interesting but we’ll have to see how support goes in the next couple of years. If I was working on a high performance single node project I’d probably try it out; I’m not sure if it would work well yet for massively parallel systems though. Check out Costanzo et al. 2021 for an interesting performance comparison. Rust looks like it’s currently losing but other benchmarks show it being faster so it might just be a matter of which benchmark you pick.