René Widera

The following example works fine if you do not use `static_assert` because due to a strange parsing stage in the compiler, the device functions are parsed but without setting `__HIP_DEVICE_COMPILE__`....

Sorry I have not carefully enough read the initial comment. Getting the warp size at compile time looks currently not possible.

> Please involve Fabian Koller in this discussion. He should knuw and can help! Why do you mention Fabian Koller (now with LogMeIn)? He wrote the first openPMD implementation, what...

Which operation do you perform with warpsize? Warp is no layer of alpaka therefore you can use it with #ifdef on nvidia hardware and on all other hardware write your...

Yes a simple example like `simpleAtomicIntrinsics` would be nice. We should also have the usage of the memory hierarchy within those examples.

It looks like the heap size is a compile time parameter and can not be choosen at runtime :-(

@ax3l There coming more changes, we should wait with the backport to the end of the next week. I think than we have all needed changes to support non cuda...

I have an older prototype to add support for CPU via alpaka: https://github.com/psychocoderHPC/mallocMC/commit/7c71dc0ac2e8ff4e7dec8d1795ca343304e78c47 The changes are currently not integrated in mallocMC.

for the record: @saipavankalyan also started to port mallocMC and reported linker issues with HIP-hcc https://github.com/ROCm-Developer-Tools/HIP/issues/1849

HIP prototype https://github.com/psychocoderHPC/mallocMC/tree/topic-hip-port Currently the cmake is not updated