Porting fortran files to cpp in Src/Extern/HYPRE

[OscarAntepara]

I have replaced AMReX_HABEC_2D.F90 with AMReX_Habec_2D_K.H and AMReX_Habec_2D.cpp
I have replaced AMReX_HABEC_3D.F90 with AMReX_Habec_3D_K.H and AMReX_Habec_3D.cpp

[WeiqunZhang]

Thanks for the PR! There are two issues need to be fixed. First, this breaks PETSc because Src/Extern/PETSc uses some files in Src/Extern/HYPRE. So you need to convert PETSc too. Second, it breaks when hypre uses long integer. For example, line 285 of AMReX_Habec_2D.cpp should use HYPRE_int instead of int.

[OscarAntepara]

There are still some issues that I have to check

[WeiqunZhang]

This 2d version of amrex_hpacoef is identical to the 3d version. Could we move them to a new file AMReX_Habec_K.H? Maybe move the following block in AMReX_HypreABecLap.cpp to AMReX_Habec_K.H too?

#if (AMREX_SPACEDIM == 2)                                                                                                                                                                                  
#include <AMReX_Habec_2D_K.H>
#elif (AMREX_SPACEDIM == 3)
#include <AMReX_Habec_3D_K.H>
#endif

[OscarAntepara]

sure!!

[WeiqunZhang]

Currently it does not compile with USE_CUDA=TRUE. Some kernels are not GPU ready yet. To make it fully work with GPU is beyond the scope of this PR. But at least we should make it compile, and we should disable GPU launch region for functions that are not ready for GPU. To disabe launch region, you can git grep LaunchSafeGuard for examples. You will need to use AMREX_HOST_DEVICE_PARALLEL_FOR_3D instead of AMREX_PARALLEL_FOR_3D, so that the launch region flag is respected.

[OscarAntepara]

Now it compiles with USE_CUDA=TRUE

[WeiqunZhang]

Should this be 4 instead of 5?

[OscarAntepara]

ohh yeah, now it is fixed! Thanks.

[WeiqunZhang]

Suggested change

	AMREX_HOST_DEVICE_PARALLEL_FOR_3D (box, i, j, k,
	{
	if (msk_arr(lo.x-1,j,k) > 0 && i==lo.x){
	Box tbx = box;
	tbx.setSmall(0,lo.x).setBig(0,lo.x); // 0: x-direction
	AMREX_HOST_DEVICE_PARALLEL_FOR_3D (tbx, i, j, k,
	{

The looping is on a slice only. Could you fix in other place too including 3D?