ecp-copa / cabanapic Goto Github PK

Add sample decks to show:

• How to custom init particles
• How to use an ES solver

We probably want to do it when the code is very simple.

(low priority)

• enable travis
• enable code coverage

From the LRF branch:

-- Configuring done
CMake Error at src/CMakeLists.txt:6 (add_library):
Target "CabanaPIC" links to target "Cabana::cabanacore" but the target was
not found. Perhaps a find_package() call is missing for an IMPORTED
target, or an ALIAS target is missing?

CMake Error at example/CMakeLists.txt:1 (add_executable):
Target "cbnpic" links to target "Cabana::cabanacore" but the target was not
found. Perhaps a find_package() call is missing for an IMPORTED target, or
an ALIAS target is missing?

-- Generating done

We need to extend the functionality so a user can:

• Select ES or EM field solver
• Select float/double precision
• Select 1d/2d/3d
• Support multiple species natively
• Support higher order shape functions

Am I correct with the assumption that the printed steps should ideally all show the same energy values?
I'm currently looking for different proxy-applications to compare tapered floating points with, and this project seemed well-structured to work for my case.

I'm just unsure of the printed results, as I don't know much behind the special physics involved. But from my limited understanding, the energies printed each step should ideally be identical at each step. This obviously is not possible with floating point arithmetic, but with higher precision, eg doubles instead of floats, the results should generally vary less between each step.

"2steam" should read "2stream"

The README.md states that one can choose the solver type by specifying the optional -DSOLVER_TYPE parameter.
When selecting "ES" as solver type cmake runs fine, but the make call throws errors:

~/Bachelor/proxies/CabanaPIC/example/example.cpp: In function ‘int main(int, char**)’:
~/Bachelor/proxies/CabanaPIC/example/example.cpp:269:26: error: ‘class Field_Solver<ES_Field_Solver>’ has no member named ‘dump_fields’
  269 |             field_solver.dump_fields(fpfd,fields, 0, 0, 0, dx,dy,dz,nx,ny,nz,num_ghosts );
      |                          ^~~~~~~~~~~
In file included from ~/Bachelor/proxies/CabanaPIC/example/example.cpp:11:
~/Bachelor/proxies/CabanaPIC/src/fields.h: In instantiation of ‘void dump_energies(field_solver_t&, field_array_t&, int, float, float, float, float, size_t, size_t, size_t, size_t) [with field_solver_t = Field_Solver<ES_Field_Solver>; field_array_t = Cabana::AoSoA<Cabana::MemberTypes<float, float, float, float, float, float, float, float, float>, Kokkos::HostSpace, 32>; size_t = long unsigned int]’:
~/Bachelor/proxies/CabanaPIC/example/example.cpp:264:102:   required from here
~/Bachelor/proxies/CabanaPIC/src/fields.h:722:40: error: no matching function for call to ‘Field_Solver<ES_Field_Solver>::e_energy(field_array_t&, float&, float&, float&, size_t&, size_t&, size_t&, size_t&)’
  722 |     real_t e_en = field_solver.e_energy(fields, px, py, pz, nx, ny, nz, ng);

Steps to reproduce (Release version 0.1.0):

cd CabanaPIC/
mkdir build
cd build
cmake -DCMAKE_PREFIX_PATH="~/Dependencies/Cabana/build/install" -DREQUIRE_HOST=ON -DSOLVER_TYPE="ES" -DCMAKE_CXX_STANDARD=20 ..
make

To have some more particle and field b.c.s.

It'd be nice to be able to easily switch different problems at compile time.

Normally it is okay (or even preferred) without it. But sometimes it is needed. Could this be something that the user can control?

The way energy is calculated currently disagrees with VPIC, so the tests are disabled

We need to:

1. Reconcile this difference
2. Add more tests

The way we update CBX in the push might accidentally be changing the fields:

https://github.com/ECP-copa/CabanaPIC/blob/master/src/push.h#L136

https://github.com/ECP-copa/CabanaPIC/blob/master/src/fields.h#L78
const float cj = 1; This is not the correct value for cj.
It should equal dt/eps0.

Currently we use atomics on the CPU and GPU. This will needlessly penalize CPU performance