Optimizing NEURON Simulation Environment Using Remote Memory Access with Recursive Doubling on Distributed Memory Systems

Algorithm 1

Recursive doubling algorithm for target calculation.

Input: Step number, Rank of processor

Output: target

    Start

    flag = Rank/((step − 1))

    if flag is even

        if IsPowerOfTwo(step)

              target rank + (2(step − 1))

        else

              diff step − previous power of two from rank

        if diff is positive number

              target Rank + (2(diff + previous power of two − 1))

        else

              target Rank + 2(Previous power of two)

    else

        if IsPowerOfTwo(step)

              target Rank − ((step − 1))

        else

        diff step − previous power of two from Rank

        if diff is positive

              target Rank − ((diff + previous power of two − 1))

        else

              (Previous power of two)

    Return target value

    End