In-Place Parallel Super Scalar Samplesort (IPSSSSo)

We present a sorting algorithm that works in-place, executes in parallel, is cache-optimal, avoids branch-mispredictions, and performs work O(n logn) for arbitrary inputs with high probability. We ran extensive experiments and show that our algorithm scales linearly in the number of cores on various multi-socket machines with 32 cores. On large inputs, we outperform our closest inplace competitor by a factor of 2.25 to 2.53 and our closest non-in-place competitor by a factor of 1.26 to 1.89. Even sequentially executed, we outperform our closest sequential competitor, BlockQuicksort, by a factor of 1.19 to 1.23 for large inputs. 1998 ACM Subject Classification F.2.2 Nonnumerical Algorithms and Problems