I came across this paper and attempting to go through the implementation of randomized selection algorithm on page 8.
Step 1 of the algorithm suggest to compute $s = PrefixSum(n_i, p)$. Where $p$ is the number of processes and $n_i$ starts simply as a number of elements in each process. Page 4 has a definition of parallel prefix, but I do not understand how to apply it in the algorithm.
Example, let $n_1 = 5, p = 2$, what is $PrefixSum(5,2)$?