It would be interesting to see the performance of TBB malloc included in the mix. For our application (parallel graph evaluation) we found the TBB allocator out

It would be interesting to see the performance of TBB malloc included in the mix. For our application (parallel graph evaluation) we found the TBB allocator outperformed jemalloc. The margin was not huge, but enough to make it worthwhile using it. However a colleague at a company with somewhat similar software found jemalloc outperformed TBB, so your mileage definitely varies depending on application. It can also happen that these allocators use more memory than regular malloc since they hold onto local pools, so that may be a concern for some use cases.