Is there any way that I want to remove duplicates in a array but maintaining the original relative order in array using cuda library? That means not sorting.

[zlwu92]

Moreover, if I have many of these arrays, and need to do the same thing for each array individually, can cub library help me to achieve this?

[gevtushenko]

Hello @zlwu92! Please, take a look at our cub::DeviceSelect::Unique.
Note, however, that our API considers unique elements in "runs" of keys. For instance:

in = [0, 0, 0, 0, 0, 0, 0, 0, 1, 1]
out = [0, 1] 

but

in = [0, 0, 0, 0, 0, 0, 0, 0, 1, 0]
out = [0, 1, 0] 

Regarding your second question, we don't have segmented nor batched version of cub::DeviceSelect::Unique at the moment.