H. Sinha. Mermera: Non-Coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Computer Science Department, Boston University, April 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to shared variables. This is more evident if the shared memory is not centralized but distributed among a number of processors, i.e. we have distributed shared memory (DSM) There has been a number of proposals and implementations of DSM systems providing di erent semantics, or consistency models [1, 2, 4, 6, 12]. The consistency memory models proposed in the literature can be broadly classi ed into strong and weak memory mod This work is partially supported by the CICYT under grant TEL99 0582 and the Comunidad Aut onoma de Madrid under grant CAM 07T 00112 1998. els. The strong memory models are ....

H. Sinha. Mermera: Non-Coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Computer Science Department, Boston University, April 1993.

....a SEQ view only needs to be compatible with the local orderings) Obviously, and accepting the fact that the future can not be predicted, any description by no means should allow that kind of sequences. However, axiomatic definitions whose read operations may report future are commonly used [1, 2, 7, 11, 12]. On the other hand, since operational descriptions only allow implementable sequences, they will rule out those sequences. By using the I O automata formalism, we can use an automaton to provide an operational definition (its wellformed traces will characterize the memory model behavior) ....

H.S. Sinha. Mermera: Non-Coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Computer Science Department, Boston University, April 1993. This article was processed using the L A T E X macro package with LLNCS style

....to shared variables. This is more evident if the shared memory is not centralized but distributed among a number of processors, i.e. we have distributed shared memory (DSM) There has been a number of proposals and implementations of DSM systems providing di erent semantics, or consistency models [1, 2, 4, 6, 12]. The consistency memory models proposed in the literature can be broadly classi ed into strong and weak memory mod This work is partially supported by the CICYT under grant TEL99 0582 and the Comunidad Aut onoma de Madrid under grant CAM 07T 00112 1998. els. The strong memory models are ....

H. Sinha. Mermera: Non-Coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Computer Science Department, Boston University, April 1993.

....and help. Arif Bhatti implemented the UDP version of Mermera, and with Sulaiman Mirdad, participated in its design. Figure 9 is 7 In the simplest configurations (e.g. low fan in fan out ATM switch) SONET frames may exert an aggregation effect. derived from Himanshu Sinha s Ph.D. thesis [29]. Hewlett Packard s Metrix Network Systems, Inc. provided us with free use of their NetMetrix software. ....

Himanshu Sinha. Mermera: Non-coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Boston University, Computer Science Department, 111 Cummington Street, Boston, MA 02215, May 1993.

....by using non coherence. We discuss how our implementation can be optimized in section 7. 2 Mermera s Memory Behavior In this section we briefly describe each of the memory behaviors supported by Mermera and explain how the behaviors are combined. A complete formal description can be found in [Sin93] Our system combines the behaviors of Coherent Memory [HS92] Pipelined RAM [LS88] Slow Memory [HA90] and Locally Consistent Memory [HS92] Our model consists of processes sharing a region of their address space. These processes may be running on different processors. The memories supported by ....

....of all writes by the same process is respected by all pro3 Defined in section 4. 4 Informally, a write x:w to location x is said to be observed by process i if the value returned by a read by process i is the value written by x:w or is a value that may have been influenced by x:w. 5 Sinha [Sin93] shows that coherent memory is equivalent to Lamport s sequentially consistent memory [Lam79] i.e. CO = SC) Coherent Non coherent Coherent (CO) Locally Consistent (LC) U U U U U U Weak Slow Pipelined RAM (PRAM) Causal Sequentially Consistent (SC) Dynamic Atomic (DA) Figure 1: ....

[Article contains additional citation context not shown here]

Himanshu Sinha. Mermera: Non-coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Boston University, Computer Science Department, 111 Cummington Street, Boston, MA 02215, May 1993.

....running communication intensive applications. Mermera is an example of a system that encourages applications to generate high traffic volumes, causing network congestion. It is a software shared memory system that has been developed to support parallel computing on a workstation network [HS93, Sin93] Processes that comprise a parallel program communicate through read and write operations to the shared memory provided by Mermera. Several different memory behaviors are supported, including one coherent 1 , and three non coherent (Pipelined Random Access Memory [LS88] Slow Memory [HA90] ....

Himanshu Sinha. Mermera: Non-coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Boston University, Computer Science Department, 111 Cummington Street, Boston, MA 02215, May 1993.

....used to prove the correctness of various program classes on different noncoherent memories. Most approaches establish the correctness of executions of certain programs, by showing that each such program induces only sequentially consistent computations on the shared memory system in question. In [29] we prove that asynchronous iterations [9] converge using slow memory. This is the first instance of a proof of correctness of an algorithm on noncoherent memories that results in executions that are not sequentially consistent. Different programs, even different parts of the same program, can ....

....programs, even different parts of the same program, can tolerate different levels of noncoherence. Therefore, it behooves a system to provide different behaviors to the programmer who can then mix and match the behaviors according to the needs of the different algorithms used in a program. In [21, 29] we present a formal description of mixed behaviors which permits executions where operations can correspond to slow, PRAM, causal and coherent memories. Attiya et al. 6] give a formalism in which one noncoherent behavior is combined with one noncoherent behavior. Other mixed coherence behaviors ....

[Article contains additional citation context not shown here]

H.S. Sinha. Mermera: Non-coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Boston University, (http://www.cs.bu.edu/techreports), May 1993.

No context found.

H.S. Sinha. Mermera: Non-Coherent Distributed Shared Memory for Parallel Computing. PhD thesis, Computer Science Department, Boston University, April 1993.

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