I spent most of a day playing with the
cuda-gdb debugger for CUDA code and it’s integration with VS Code. Currently it’s got some pretty significant limitations, it can’t show the contents of STL containers in host code or the values of variables in
__device__ functions plus it looks like the VS Code integration extension is dead.
The HLLD solver is DONE! All the tests pass, dual energy and passive scalars are supported and tested! I had to hammer out a handful of bugs, several of which were in the tests not the solver, and I completely redid how the solver handles dual energy and passive scalars now that I understand how they work better. I might have also found a bug in Enzo so I raised an issue on their github page. Next week I want to write some quick unit tests for the individual functions that the HLLD solver calls but then I’m off to work on the integrator.
- Read Minoshima & Miyoshi 2021 which is about a new, low dissipation, HLLD solver
Original HLLD paper: Miyoshi & Kusano 2005
Blog post on the HLLD Algorithm: HLLD Algorithm