Someone famaliar with Cudpp?

How to call the prefix-scan in the lib cudpp? And how fast is it, say computing the prefix-sum of 16M integers? Someone can help me? I tried but failed lots of times…


There are examples included in the distribution:

There are some performance numbers in the paper (0.79ms for a prefix sum of 1M values):…_pub?pub_id=915

I think the performance has improved considerably since then.