cuda::experimental::places::replicate_over_t

template<size_t... axes>
struct replicate_over_t

Designates where data will be stored (CPU memory vs.

on device 0 (first GPU), device 1 (second GPU), …)

This class uses a polymorphic design where all place types (host, managed, device, composite, future extensions) implement a common data_place_interface. The data_place class holds a shared_ptr to this interface and delegates operations to it. Tag selecting which grid axes a replicated data place replicates over (the remaining axes are SHARED: their fibers use one common instance)