Jan Stephan

My gut feeling is that we should mirror / mimic the atomic operations defined by the C++ standard. I guess that would require an implementation of `std::atomic_ref` at some point,...

Since alpaka started out as a layer above CUDA it inherited many of CUDA's designs. This is IMO another case where we should rethink this approach. For the SYCL back-end...

> As also discussed multiple times, other team members prefer to have names in the file headers to attribute people not writing papers and to guard against copying files from...

IMO we can leave out the year entirely. I'm basing this issue on the [REUSE](https://reuse.software/) recommendations by the Free Software Foundation Europe. Quote: > ### Which years do I include...

In alpaka we also gradually moved to option 2 over the past years. Personally I'm in favour of abandoning the year entirely.

We could set the Windows requirement to 3.24 and leave it at 3.18 for Linux / macOS.

There is [a long list of restrictions](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#extended-lambda-restrictions) that apply to these special lambdas. Some of them might come as a surprise to users. Are we okay with that?

How would you check for this? Sure, if two pointers are identical you can do a comparison. But what if this happens: ```c++ memcpy(ptr, ptr+offset, ...); ``` ?

Soooo... any news on this? :-)

Thanks for letting me know. For the beginning, `malloc_{device,host}` would be sufficient.