HPX on nodes with different hardware architectures, possible?

[venartas]

Is it possible to run an HPX job that uses the HPX AGAS system across two nodes with different hardware architectures — for example, one x86 node and one ARM node (such as an NVIDIA DGX Spark)? Specifically, can I run an HPX component of type A on locality 1 (x86) and also request an instance of component A to be instantiated on locality 2 (ARM)?

[hkaiser]

That should be possible if you enable endian-aware serialization (cmake option: -DHPX_SERIALIZATION_WITH_SUPPORTS_ENDIANESS=On). However, we do not test this on a regular basis, so it might actually not work ;-) Please let us know.