Where the scheduling algorithms currently in TBB shine are in Shared Multi-Processor (SMP) environments where multiple CPUs share a common memory system. These architectures facilitate the sharing of work among the cores. Whether those cores are distributed singly across a board or ganged together in various organizations upon single chips will determine various performance characteristics of interest for blocking (sizing problems to available cache resources), but all SMP architectures will benefit from the basic scheduling, which tries to keep data local as much as possible and to parcel out the work so that all the cores stay busy.
