Hi,
I'm calling DeviceScan::ExclusiveSum from a kernel function.
Although I'm getting cudaSuccess back from the ExclusiveSum, sometimes I'm not getting any results (I'm memsetting the out vector and it's not changing at all)
Is it legal to call DeviceScan from a __global__ or __device__ function?
What am I doing wrong?