Accommodating workload diversity in chip multiprocessors via
Accommodating workload diversity in chip multiprocessors via - polish dating hearts
Designing an efficient interconnection network for a chip multiprocessor (CMP) is a challenging problem.
In Section 3, we will describe a framework at the hardware level to support reconfigurable interconnects.In Section 5, we will present the experimental methodology used to evaluate different interconnection networks and the data collected for different benchmark suites.Finally, we will discuss some conclusions of the research presented in this paper.On the other hand, increasing number of cores in a CMP places a corresponding increasing demand on the bandwidth requirements of an interconnection network.Both these problems are depicted in Figure 1 which shows that the increasing delay of wires and the increasing number of cores that are utilized on a CMP result in more conflicting requests for a shared bus (Conflicting requests result when a request has to wait in a queue because the bus is currently not available.Our cost-effective reconfigurable interconnection network is implemented on a traditional shared bus interconnect with snoopy-based coherence, and it enables improved multicore performance.
The proposed interconnect architecture distributes the cores of the processor into clusters with reconfigurable logic between clusters to support workload-adaptive policies for inter-cluster communication.In this paper, we first motivate the need for workload-adaptive interconnection networks.Subsequently, we describe an interconnection network framework based on reconfigurable switches for use in medium-scale (up to 32 cores) shared memory multicore processors.This large increase in conflicts at the interconnect increases the resolution time for memory reference instructions and is one barrier to the high performance and throughput of multicore processors.One limiting factor to the efficiency of any interconnection network is that it is designed to serve a diversity of workloads.This data was collected for a multicore processor with a separate request bus and response bus.