Problem

Cub allows itself to place into a namespace via CUB_NS_PREFIX and CUB_NS_POSTFIX, such that multiple shared libraries can each utilize their own copy of it (and thus different versions can safely coexist). Static variables used for caching could otherwise cause problems (e.g., https://github.com/NVIDIA/cub/blob/main/cub/util_device.cuh#L212).

Thrust however depends on cub and

If we find a rust::* type in a C++ header per this list, include that in our generated cxx bindings.

In theory this is just a matter of contributing such types to known_types.rs. In practice it won't be that simple...

First step would be to add some tests to integration_tests.rs.

DeviceSegmentedReduce reduces each segment using the same kernel as DeviceReduce, and thus provides the same run-to-run consistency guarantee for non-associative types. The documentation for DeviceSegmentedReduce should be updated to reflect this.

E.g. this bit: https://github.com/NVIDIA/cub/blob/a8910accebe74ce043a13026f8e71d678cddd6c1/cub/device/device_reduce.cuh#L91-L95
needs to be ad