Spencer Bryngelson

The debug cases are failing on this case (perhaps among others, they only run a fraction of the cases in CI): ``` m_mpi_proxy.fpp:1289: @:ALLOCATE_GLOBAL(ib_buff_send(0:-1 + gp_layers * (m + 2*gp_layers...

If this works, we will need to test that it works on NVIDIA and AMD GPUs in single precision mode.

@aricer123 CodeCov fails randomly all the time, and it isn't because of the 'diff hit'. So don't worry about that. I'm cloning your repo now to have a look. The...

you have a slight problem here in `precision_select` module: ```f90 integer, parameter :: wp = double_precision #ifdef MFC_MPI integer, parameter :: mpi_p = MPI_DOUBLE_PRECISION ``` notice that one has to...

Requesting some amount of review from @wilfonba and @henryleberre

Single precision doesn't work because you have many double precision intrinsics left, like `dlog`, `dexp`, `dabs`. You also have things like `1d-15` or `5.0d12` hanging around in the code. These...

> @sbryngelson incorporating the patch file seems to be failing tests, i simply incorporated the changes along with the fixes to m_precision_select. Did this happen on your system? What could...

It's actually somewhat suspicious (maybe?) (for the test suite, I suppose) that going down to single precision only barely causes tests to fail: ```console Failed test tests/D79C3E6F: 1D -> bc=-1...

Double precision does not build with my patch. The reason contradicts @henryleberre's assumption I think. Using `5e-1` has no way of knowing what precision to use. By default it goes...

Note that my patch does not cover all the `d` exponential commands in the code. My regex did not do all the work. For example, in `m_phase_chage` there is still...