Making static gravity more user friendly with custom gravity field flags

[evazlimen]

I have separated out the various hard coded static gravity fields with a switch statement that is determined by parameter file input. This way a user doesn't have to comment and uncomment things in the source code just to run a test.

• added the flag "custom_grav". Used to select the appropriate custom gravity field in static_grav.h. Goes in the parameter file.
• added a switch statment in static_grav.h to use the custom_grav flag.
  This is set up for the simple integrator, I wasn't sure if it was appropriate to also do so for VL and also wanted to be as minimally invasive as possible!
  Still have to do some cleaning up but wanted to see if I should be doing this differently.

[bcaddy]

This looks great, anything to remove the need to comment/uncomment code for proper functionality.

I know there's some other stuff that is initialized in parse_params. I moved most of those initializations in PR #334.

I'm not entirely sure what's causing some of the HIP builds to fail. Do you have access to an AMD system to test this on? If not I can try to hunt down the issue on one of the AMD systems I have access to.

Edit: I found the issue.

hipcc -g -O3 -std=c++17 -fPIE -DCUDA -DMPI_CHOLLA -DPRECISION=2 -DPPMC -DHLLD -DMHD -DVL -DDENSITY_FLOOR -DTEMPERATURE_FLOOR -DCPU_TIME -DOUTPUT -DHDF5  -DO_HIP -DGIT_HASH='"4725f356813745761925289ad2d7e90419fc4569"' -DMACRO_FLAGS='"-DCUDA -DMPI_CHOLLA -DPRECISION=2 -DPPMC -DHLLD -DMHD -DVL -DDENSITY_FLOOR -DTEMPERATURE_FLOOR -DCPU_TIME -DOUTPUT -DHDF5  -DO_HIP -DGIT_HASH='"4725f356813745761925289ad2d7e90419fc4569"'"' -Isrc -I/usr/lib/x86_64-linux-gnu/hdf5/serial/include -I/usr/lib/x86_64-linux-gnu/openmpi/include -c src/integrators/VL_2D_cuda.cu -o src/integrators/VL_2D_cuda.o
src/integrators/VL_3D_cuda.cu:303:3: error: too few arguments to function call, expected 29, have 28
  hipLaunchKernelGGL(Update_Conserved_Variables_3D, dim1dGrid, dim1dBlock, 0, 0, dev_conserved, Q_Lx, Q_Rx, Q_Ly, Q_Ry,
  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/opt/rocm-5.2.3/include/hip/amd_detail/amd_hip_runtime.h:251:46: note: expanded from macro 'hipLaunchKernelGGL'
#define hipLaunchKernelGGL(kernelName, ...)  hipLaunchKernelGGLInternal((kernelName), __VA_ARGS__)
                                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/opt/rocm-5.2.3/include/hip/amd_detail/amd_hip_runtime.h:248:89: note: expanded from macro 'hipLaunchKernelGGLInternal'
        kernelName<<<(numBlocks), (numThreads), (memPerBlock), (streamId)>>>(__VA_ARGS__);         \
        ~~~~~~~~~~                                                                      ^
src/integrators/../hydro/hydro_cuda.h:18:17: note: 'Update_Conserved_Variables_3D' declared here
__global__ void Update_Conserved_Variables_3D(Real *dev_conserved, Real *Q_Lx, Real *Q_Rx, Real *Q_Ly, Real *Q_Ry,
                ^

[evazlimen]

I changed the initialization of custom_grav and set it up for the VL integrator. I formatted with clang-17, not sure if I just have to wait for PR #330 to go through for the formatting check to pass?

[evaneschneider]

Awesome, yep -- we're doing a hack day over here so have a few PRs in the queue waiting on 310 and rebasing, but once #330 330 goes through the clang 17 formatting should work.

[evaneschneider]

FYI, I've pulled in the clang-17 PR, so this one should be good to go once it's updated.