Compact float4 array with CUDPP

Accelerated Computing CUDA CUDA Programming and Performance
1

I was wondering how to go at this since CUDPP only seems to support primitives such as integers and floats.
Would it be possible to simply enlarge the valid indicator array to match the size of the float4 array? Or could it happen that CUDPP actually changes the order of the elements?

EDIT:
In case anyone else wondered: Inflating the valid indicator array and passing the float4 array as a float array appears to work after a couple of tests.