René Widera

> * Also the distribution of jobs is different from Free Runner. We have few strong runners, so it makes more sense to run many configurations in a single job...

For an unknown reason, the OpenACC test shows access to invalid memory and HIP requires over 60 minutes to compile the atomic test.

> We need to revisit this once accessors no longer live inside `alpaka::experimental`. Where do they live now? Are they not experimental anymore?

> > IMO this should be only used for debug mode. > > Do we want this in hard debug mode (`-DCMAKE_BUILD_TYPE=Debug`) or just in some sort of verbose mode?...

It is not about what is shipped with the distribution. The question is which CMake versions you mostly find on HPC systems. I would only vote for increasing the version...

> How is this handled with our current CMake requirement? There are also no POWER binaries available on cmake.org. I guess the Summit / Taurus admins have installed CMake 3.18...

> Ubuntu 22.04 will make CMake 3.22 available. I propose that we switch to this version, too (which would affect alpaka 0.10). Any objections @psychocoderHPC @SimeonEhrig? We should not increase...

The feature that `AtomicRefStlLock` is save between different grids is needed to allow a atomic's on the hierarchy `Grids{}`. You are right that there is no need to have a...

I would keep the issue open but as you said `AtomicStdLibLock` is only a fallback and I do not see that it is at the moment an issue.

I used already a plain clang for HIP as prototype with hard coded path to the libs. https://github.com/psychocoderHPC/picongpu/commit/be71be2a8e2233dc97abe98358ecbf0a21b97220 This branch is was for CRAY to test PIConGPU without hipcc. Note:...