Michael Redenti

> I just tried running the Hasegawa-Wakatani example with `periodicX = true` (also using `hw:modified=false`, but I don't think that's relevant here). It did run, but very badly until I...

> Hi @mir1995 I think that the options are: > > * setting the INVERT_ZERO_DC flag, but understanding that the zonal flows will be missing. This may be important, depending...

> Hi @mir1995 I think that the options are: > > * setting the INVERT_ZERO_DC flag, but understanding that the zonal flows will be missing. This may be important, depending...

> @mir1995 Difficult... the zonal components evolve along with the turbulence, so are not something that can be calculated once and added together: The zonal flows take energy from the...

> @mir1995 Unfortunately I think solving that equation would require a modified 1D periodic solver, so doesn't get around the need to modify the cyclic inversion code. @bendudson ah! off...

Are there any updates regarding this? Do we finally have MPI debugging support in VScode that allows us to switch MPI process from the UI?

Cool, I will proceed that way. I see there is a already a NCCL building block that I will invoke before UCC. And will follow https://docs.nvidia.com/networking/display/sharpv300/using+nvidia+sharp+with+nvidia+nccl# to build the nccl...

@scotty110 Did you solve the issue, I have the same problem

> Also, how easy would it be with this design to add another backend, such as DuckDB? With this refactor, introducing a new SQL-based backend (e.g. DuckDB) requires implementing a...

> @mredenti Could you try fixing the unit tests. Before starting evaluating and testing this further, it'd be good to know that at least all unit tests pass. From a...