Parallel Benchmark: Cleanup

[ax3l]

Clean leftovers from new MPI benchmark in #346.

• properly add auxiliary benchmark sources
• properly name for build with MPI only
• fix BlockSlicer API: no MPI needed
• includes
• add changelog

[ax3l]

hm, unless this shall be auxiliary as the documentation you wrote suggests (part of the library, not only part of the benchmark example...).

Maybe it's well-placed in auxiliary then. But then it should be included in our facade headers as well.

@franzpoeschel did you pass the whole communicator in case you want to determine e.g. a MPI grid position from it?

Maybe an openPMD::benchmark CMake target makes sense, outside of examples.

[franzpoeschel]

Passing the rank directly is probably the more sensible solution, yes.

[franzpoeschel]

part of the library, not only part of the benchmark example

I was intending for it to become part of the library to allow users to easily dissect a dataset into pieces. If we go with this, OneDimensionalBlockSlicer.cpp should probably be put back into the sources list again (possibly after moving it to auxiliary). If not, the complete class (i.e. OneDimensionalBlockSlicer) should probably go into the example file and be ridded from the API altogether otherwise. But this will result in users having to needlessly copy-paste the class if they want to use it..?

Maybe it's well-placed in auxiliary then.

So you suggest placing the whole parallel benchmark into auxiliary (headers as well as sources)? Sounds good for me.

[ax3l]

Yes, prospectively that would make a good place for that helper. I'll leave it in the dir for now. We also need to thing of the auxiliary headers for this workflow, they are currently not part of the public API but we still might want to add facade headers... dunno yet :)