What is the fastest multithreaded radix sort code out there? I am using radix sort to sort the Morton/Hilbert keys when constructing a tree structure. The number of identical keys is in the range of 10-100. I also need the sorting code to return a permutation index array so that I can sort other elements according to it.
If you can code on many core architectures such as GPUs, maybe this paper "Fast 4-way parallel radix sorting on GPUs" can interest you.
Have you looked at p4est? It might have the kind of code you're looking for.