Buffer cleanup

[biddisco]

This PR contains cleanups to buffer management that should improve default performance

1. Make buffer manager a singleton so that creation of communicator objects share the same buffer handler and reuse the same caches of buffers
2. Fix a missing freeAllBuffers is particleUpdate
3. replace the default kokkos views that for the storage of the buffers with unmanaged kokkos views that have stronger alignment (in this PR 1024 bytes) than the default kokkos views (normally 64 bytes)
4. make the default buffer handler the default, rather than the logging one

This PR also adds support for logging using the spdlog library and the fmt library for clean messages, this can be enabled or disabled at compile time using IPPL_LOG_LEVEL off, trace, debug, warn, error, critical etc etc by default it is off, to use it, one needs to find_package spdlog, fmt which are now provided in uenvs for alps

[biddisco]

This shows the performance with/without IPC enabled on alps. The buffer cleanup branch improves things, turning off ipc is still a major improvement

edit: erratum -the graph says heffte-gpu-enabled, but in fact it is disabled here.

[aaadelmann]

John very nice results! Can you also show this on Lumi and/or on Beverin?

[aaadelmann]

with GPU off you mean: -D Heffte_DISABLE_GPU_AWARE_MPI=OFF ?

[biddisco]

with GPU off you mean: -D Heffte_DISABLE_GPU_AWARE_MPI=OFF ?

yes. exactly this

[srikrrish]

@biddisco Could you please pull and merge the latest master as it contains a fix for the Jülich's CI. I can then add the label and check if the unit tests are passing.

[aaadelmann]

The last commit seams to break the compilation: https://ippl-bc4558.pages.jsc.fz-juelich.de/365538/allure-report/index.html#suites/0a150a27af0a5eb1faedcb6100255f56/782113b35712612a/

[srikrrish]

@biddisco Did you pull in the latest ci change in master in this branch? Otherwise the results won't be displayed

[biddisco]

John very nice results! Can you also show this on Lumi and/or on Beverin?

Based on the previous build fail - I will need to also add a specialization for HIP in the buffer creation before I run tests. I will get to it soon

[biddisco]

Further cleanups and reorganisation of the commits to add a new cmake option that allows you to turn on/off aligned buffers in case they are a problem on not yet supported platforms

IPPL_ENABLE_ALIGNED_COMMUNICATION_BUFFERS (default off)

Tested the branch on lumi without any problems. Alignment seems to make no difference, so it can be ignored

Meanwhile on beverin (AMD mi300A) we have hanging problems that appear unique to the alps amd cluster and are not connected to the PR.

[biddisco]

Here's a graph comparing master to cleanup branch
What you can see is that the cleanup makes a big difference when IPC is enabled (on daint, where we have problems), but when disabling IPC is the main thing to improve performance,