| W. T.-Y. Hsu and P.-C. Yew. An effective synchronization network for hot-spot accesses. ACM Transactions onComuter Systems, 10(3):167-- 189, August 1992. |
....to this problem, but such hardware is currently not available for most computing platforms. We present a software implementation for shared memorymultiprocessors. To compute GVT, we maintain a tournament tree, not unlike the hardware reduction networks or algorithms such as that described in [14] with processors at the leaves. Each node of the tree has a timestamp associated with it. The timestamp of a node signifies the minimum of the local virtual times and transient messages of the processors at the leaves in that node s subtree. The value at the root node is the GVT. Whenever the ....
W. T.-Y. Hsu and P.-C. Yew. An effective synchronization network for hot-spot accesses. ACM Transactions onComuter Systems, 10(3):167-- 189, August 1992.
....exacerbated by tree saturation [23] which even obstructs memory traffic to non hot spots locations. Significant progress has been made in reducing hot spot traffic, especially hot spot traffic due to synchronization. Techniques include separate synchronization networks (possibly with combining) [18] and hot spot free software algorithms that use distributed data structures [20] Furthermore, flow control mechanisms may be useful, especially when hot senders (processors with usually high request rates or a high favoritism to the hot spot) are a factor. Evaluating alternative techniques for ....
....systems can become unstable at favorite memory probabilities on the order of 1 to 2 under reasonable request rates. If the memory queues are of inadequate length, significantly lower favorite memory probabilities can cause instability. The techniques proposed in the synchronization literature [20, 18] (such as separate synchronization networks with combining or software algorithms using distributed data structures) will likely reduce the likelihood of hot spots. The simulated system does provide flow control in that the number of cycles before a source processing module submits a retry is a ....
W. T.-Y. Hsu and P.-C. Yew. An effective synchronization network for hot spot accesses. ACM Transactions on Computer Systems, 10(3):167--189, August 1992.
....at least one node in common with the forward path the node where the combining state is stored. Almasi and Gottlieb [1] give several examples of how such hardware combining can eliminate serial bottlenecks. Several alternative proposals for request combining have appeared in the literature [8, 11, 19, 21, 29, 32]. The primary focus of these efforts is on reducing the cost of the combining network. This is accomplished either by altering the topology of the combining network or by requiring the system software to reduce the amount of contention for shared data. This paper has two purposes. The first is to ....
....distinction is made between non combinable and potentially combinable requests (typically synchronization requests) and the interconnect dynamically determines the combining set of the potentially combinable requests. Two alternative techniques for IIC are presented by Tzeng [29] and Hsu and Yew [11]. Tzeng separates the interconnect into a routing section and a combining section. It is assumed that requests which may combine are distinguished from non combining requests by examination of the opcode. Such requests are directed to the combining section of the network. The combining section of ....
Hsu, W. T. and Yew, P.-C., "An Effective Synchronization Network for Hot-Spot Accesses," ACM Transactions on Computer Systems, vol. 10, pp. 167-189, August 1992.
....exacerbated by tree saturation [23] which even obstructs memory traffic to non hot spots locations. Significant progress has been made in reducing hot spot traffic, especially hot spot traffic due to synchronization. Techniques include separate synchronization networks (possibly with combining) [18] and hot spot free software algorithms that use distributed data structures [20] Furthermore, flow control mechanisms may be useful, especially when hot senders (processors with usually high request rates or a high favoritism to the hot spot) are a factor. Evaluating alternative techniques for ....
....If the memory queues are of inadequate length, significantly lower favorite memory probabilities can cause instability. The techniques proposed in the synchronization literature (such as separate synchronization networks with combining or software algorithms using distributed data structures [20, 18]) may well reduce the likelihood of hot spots. The simulated system does provide flow control in that the number of cycles before a source processing module submits a retry is a function of the number of retries previously sent [27] More sophisticated flow control mechanisms could be considered. ....
W. T.-Y. Hsu and P.-C. Yew. An effective synchronization network for hot spot accesses. ACM Transactions on Computer Systems, 10(3):167--189, August 1992.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC