Citation Context

...communication necessary to keep the distributed memories consistent. It uses a lazy implementation [14] of release consistency [13] and multiple-writer protocols to reduce the impact of false sharing =-=[8]-=-. On the 100-Mbps ATM LAN, good speedups were achieved for Jacobi, TSP, Quicksort, and ILINK (a program from the genetic LINKAGE package [16]). TreadMarks achieved only a moderate speedup for a slight...

Citation Context

...action does not result in a large amount of communication. Various techniques have been used in TreadMarks to meet this challenge, including lazy release consistency [6] and multiple writer protocols =-=[3]-=-. This paper first describes the application programming interface provided by TreadMarks (Section 2), Next, we discuss the implementation challenges. (Section 3) and the techniques used to meet these...

Citation Context

... the number of messages exchanged in a software DSM should equal the number of messages exchanged in a message passing implementation of the same application. Therefore, Munin's write-shared protocol =-=[5]-=-, a software implementation of RC, buffers writes until a release, instead of pipelining them as in the DASH implementation. At the release, all writes going to the same destination are merged into a ...

Citation Context

...otations or explicit calls to access shared data [2, 11]. Another approach, called Shared Virtual Memory (SVM), uses the virtual memory hardware to detect access to data that is not available locally =-=[4, 13, 12]-=-. In most such systems, the granularity at which data is accessed and kept coherent is large, because it is related to the size of an application data structure or the size of a virtual page. We have ...

Citation Context

...ce [11], rather than letting arbitrary nodes access a global address space. Other systems have also restricted the DSM programming model to improve performance, reliability and programmability. Munin =-=[8]-=- lets programmers annotate variables with the access pattern they will have so as to choose an optimal consistency protocol for them. Linda [13] provides a tuple space programming model that may be im...

Citation Context

...n any system that uses some form of cache to speed up accesses. In particular, cache consistency protocols have been studied extensively in computer architecture [12], distributed shared memory (DSM) =-=[16, 5]-=-, network and distributed file systems [19, 13, 18], and distributed database systems [9, 10]. The consistency problems are slightly different in the four contexts. In computer architecture and DSM sy...

Citation Context

...d address translation hardware provides access control, though at memory page granularity. Nevertheless, it forms the basis of several distributed-shared-memory systems---for example, IVY [26], Munin =-=[4], and Trea-=-dMarks [17]. Though unimplemented by current commodity processors, additional, per-block access bits in a TLB entry could provide fine-grain access control. The "lock bits" in some IBM RISC ...

Citation Context

...et even for these machines there are efforts to implement a shared-memory programming model through runtime and kernel-level emulation of shared memory. Representative systems include Ivy [21], Munin =-=[8]-=- and others [27]. Most shared-memory machines, both hardware and software based, rely on data caching to exploit data locality and reduce communication. Cache coherence must be maintained for multiple...

Citation Context

...ss of work when an idle task is preempted [Litzkow88]. More closely-coupled systems can perform eager data transmission in periods when the network is lightly loaded. For example, release consistency =-=[Carter91]-=- mechanisms can pre-transmit updates to reduce the time to complete a release, which is often on the critical performance path. The choice of whether to compress data for network transmission can depe...