D. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K. Schauser, R. Subramonian, T. von Eicken. LogP: a Practical Model of Parallel Computation. In Communications of the ACM, Vol.30, No.11, pp.79-85, November 1996.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

.... is a polynomial problem [9] while scheduling a fork join graph is NP complete [6] Extensions of the standard macro data ow model includes the one port model [13, 14] just as in this paper, each processor can communicate with at most another processor at a given time step) and the LogP model [7]. Scheduling fork graphs in the one port model [1] or in the LogP model [20] remains NP complete, even for an in nity of resources. Extensions of the result under the LogP model to allow message forwarding are dealt in [21] Collective communications on heterogeneous platforms Several papers deal ....

D. Culler, R. Karp, D. Patterson, A. Sahay, K. Schauser, E. Santos, R. Subramonian, and T. von Eicken. LogP: A practical model of parallel computation. Communications of the ACM, 39(11):7885, 1996.

....to be 32 bits and operate at 800 MHz. These parameters are similar to those used for the on chip router of the Alpha 21364 [22] Delay through a physical link is assumed to be equal to its length in number of tiles, with a minimum of one clock. Send and receive overhead of ten cycles is assumed [23]. Dimension order routing (DOR) is assumed on the mesh and true fully adaptive routing (TFAR) is assumed on the torus. Source routing is used for the generate topology to give the necessary flexibility to assign communications to di#erent routes and minimize contention. Deadlocks in the torus and ....

D.E. Culler et al. "LogP: A Practical Model of Parallel Computation". Communications of the ACM, 39(11):78--85, November 1996.

....clustering heuristics for reducing the relative communication overhead; see Gerasoulis Yang [9] Several models have been developed for modeling the communication and the parallelization overhead in these problems. In models with a ner communication representation (like in the LogP model [3]) the impact of the parallelization overhead is usually ignored. Recently, a new computational model called Malleable tasks (MT) has been proposed by Turek, Wolf Yu [19] as an alternative to the usual delay model. Under the MT model, the precedence task graph depicts a coarse grain vision of a ....

D. Culler, R.Karp, D.Patterson, A.Sahay, E. Santos, K. Schauser, R. Subramanian, and T. von Eicken [1996]. LogP: A practical model of parallel computation. Communications of the ACM 39, 78-85.

....are taken into account lead generally to harder algorithms than without communications: for large communication delays no constant approximation algorithm is known at this time. It is of course even more crucial for computational models with a ner communication representation like LogP [3]. Due to the intractability of the problem, the impact of the parallelization overhead is usually ignored. Recently, a new computational model called Malleable tasks (MT) has been proposed as a promising alternative to standard delay models. MT are computational units which may be themselves ....

D. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K. Schauser, R. Subramonian, and T. von Eicken. Logp: A practical model of parallel computation. Communications of the ACM, 39(11):78-85, 1996.

....model that predicts the execution time of an application given a set of parameters for a massively parallel distributed frame work. There are two important factors that dominate the execution time in parallel and distributed processing: the computation time and the communication time. 6] [8] and [16] present communication models for various distributedmemory architectures. Of these, we have chosen to base our analytical model on [8] We discuss this model further in Section 4. 10] and [11] propose models that predict the execution time of an application that incorporates both the ....

....are two important factors that dominate the execution time in parallel and distributed processing: the computation time and the communication time. 6] 8] and [16] present communication models for various distributedmemory architectures. Of these, we have chosen to base our analytical model on [8]. We discuss this model further in Section 4. 10] and [11] propose models that predict the execution time of an application that incorporates both the computation time and the communication time. However, 10] does not model the communication in sufficient detail. 11] has a reasonable model for ....

[Article contains additional citation context not shown here]

D. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K. Schauser, R.Subramonian, and T. Eicken. LogP: A practical model of parallel computation. Commun. ACM, 39(11), 1996.

....networks and processor computing power, two different models can be used. One model is a high level description of an architecture with only four parameters and the other model is a low level description with many more details and settings. The high level model is a description called LogP [3] [2]. The description contains four parameters. Latency is the time required to send data between any two nodes. Overhead is the processing time associated with setting up to send and receive data. Gap is the minimum wait time between sending messages. P is the relative processing speed of each ....

D. E. Culler, R. M. Karp, D. A. Patterson, A. Sahay, E. E. Santos, K. E. Schauser, R. Subramonian, and T. von Eicken. LogP: A Practical Model of Parallel Computation. Communications of the ACM, 39(11):78--85, 1996.

....clustering heuristics for reducing the relative communication overhead; see Gerasoulis Yang [7] Several models have been developed for modeling the communication and the parallelization overhead in these problems. In models with a ner communication representation (like in the LogP model [3]) the impact of the parallelization overhead is usually ignored. Recently, a new computational model called Malleable tasks (MT) has been proposed by Turek, Wolf Yu [16] as an alternative to the usual delay model. Malleable tasks are computational units which may be themselves executed in ....

D. Culler, R.Karp, D.Patterson, A.Sahay, E. Santos, K. Schauser, R. Subramanian, and T. von Eicken [1996]. LogP: A practical model of parallel computation. Communications of the ACM 39, 78-85.

....(1) to allow a clear understanding of the e ect of network delays on the overall results, and (2) to avoid tying the results to a speci c network topology. This approach is also taken by Culler et al. in the LogP model, and stems from the lack of consensus as to the best network topology [11]. The page placement policy used in this work, rst touch after initialisation, was chosen after an investigation into the e ects of di erent page placement policies [41] The study also examined the interaction of the page placement policy with the contention reducing strategies proposed in this ....

David E. Culler, Richard M. Karp, David Patterson, Abhijit Sahay, Eunice E. Santos, Klaus Erik Schauser, Ramesh Subramonian, and Thorsten von Eicken. LogP: a practical model of parallel computation. Communications of the ACM, 39(11):78-85, November 1996.

....of basic computational operations executed by any processor in the local computation phase, and h is the maximum number of messages sent or received by any processor. The total execution time for the program is the sum of all the superstep times. An approach related to BSP is the LogP model [CKP93, CKP96]. LogP models the performance of point to point messages with three parameters: o (computation overhead of handling a message) g (time interval between consecutive message transmissions at a processor) and L (latency for transmitting a single message) The main difference between the two models ....

....on an SGI Challenge and an Intel Paragon. The claim that both efficiency and portability can be achieved by using the BSP model is supported by both theoretical and experimental results [Val90a, GLR99, KS99, McC93, Val90b, Val93, WG98] However, other general purpose models, such as LogP [CKP96], make similar claims. LogP models the performance of point to point messages with three parameters representing software overhead, network latency, and communication bandwidth. Under LogP, the programmer is not constrained by a superstep programming style. Although proponents of LogP argue that ....

David E. Culler, Richard M. Karp, David Patterson, Abhijit Sahay, Eunice E. Santos, Klaus Erik Schauser, Ramesh Subramonian, and Thorsten von Eicken. "LogP: A Practical Model of Parallel Computation. " Communications of the ACM, 39(11):78--85, November 1996.

....processor. The total execution time for the program is the sum of all the superstep times. The claim that both efficiency and portability can be achieved by using the BSP model is supported by both theoretical [1] and experimental [2] results. However, other general purpose models, such as LogP [3], make similar claims. LogP models the performance of point to point messages with three parameters representing software overhead, network latency, and communication bandwidth. Under LogP, the programmer is not constrained by a superstep programming style. Although proponents of LogP argue that ....

D. E. Culler, R. M. Karp, D. Patterson, A. Sahay, E. E. Santos, K. E. Schauser, R. Subramonian, and T. von Eicken. LogP: A practical model of parallel computation. Communications of the ACM, 39(11):78--85, November 1996.

....that two algorithms for the same problem are incomparable in the sense that one is faster on machine A, the other is faster on machine B. To overcome these problems, several proposals for so called parallel bridging models have been developed: for example, the BSP model [16] the LogP model [6], the CGM model [7] and the QSM model [1] A bridging model aims to meet the following goals: Its cost measure should guide the algorithm designer to develop efficient algorithms. It should be detailed enough to allow an accurate prediction of the algorithms performance. It ought to provide an ....

D. Culler, R. Karp, D. Patterson, A. Sahay, K. Schauser, E. Santos, R. Subramonian R., T. von Eicken. LogP: A practical model of parallel computation. C.ACM 39(11) (1996) 78--85.

....differently in two BSP libraries. 1 Introduction It has been argued by many researchers that one of the reasons why parallel computing has not been very successful is the absence of a standard parallel computation model (PCM) suitable for designing and analyzing parallel algorithms (see, e.g. [20, 19, 16, 7]) Such a model must fulfill several crucial requirements. First, it should be simple enough to facilitate the design and analysis of parallel algorithms. Second, it should be accurate so that efficient algorithms developed in the framework of the model translate to efficient programs. Third, it ....

....contention, to distinguish it from contention within the network. Alternatively, one can use a staggered communication schedule in which processor i first communicates with processor (i 1) mod p, then with (i 2) mod p, and so on. The issue of endpoint contention was raised by the LogP model [7]. This model has a capacity constraint such that at most L=g messages can be in transit from or to any processor at any time, where L is the latency parameter of the LogP model and g is the gap. In [18] it is argued that the naive communication schedule can take p times as long as the staggered ....

D. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K. Schauser, R. Subramonian, and T. von Eicken. LogP: A Practical Model of Parallel Computation. Comm. of the ACM, 39(11):78--85, 1996.

No context found.

D. Culler, R. Karp, D. Patterson, E. Santos, A. Sahay, K. Schauser, R. Subramonian, and T. von Eicken. Logp: A practical model of parallel computation. Communications of the ACM, 39(11):78--85, 1997.

No context found.

D. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K. Schauser, R. Subramonian, T. von Eicken. LogP: a Practical Model of Parallel Computation. In Communications of the ACM, Vol.30, No.11, pp.79-85, November 1996.

No context found.

Culler, D., Karp, R., Patterson, D., Sahay,A.,Schauser, K.E., Santos, E., Subramonian, R., von Eicken, T.: LogP: a practical model of parallel computation. Commun. ACM 39 (1996) 78--85.

No context found.

D. E. Culler, R. M. Karp, D. Patterson, A. Sahay, E. E. Santos, K. E. Schauser, R. Subramonian, and T. von Eicken. LogP: A practical model of parallel computation. Communications of the ACM, 39(11):78-85, 1996.

No context found.

D.E. Culler et al., "LogP: A Practical Model of parallel computation", Communications of the ACM, vol. 39, pp. 78-85, 1996

No context found.

D. E. Culler, R. M. Karp, D. Patterson, A. Sahay, E. E. Santos, K. Erik Schauser, R. Subramonian, and T. von Eicken. Logp: a practical model of parallel computation. Communications of the ACM, 39(11):260--270, November 1996.

No context found.

D. E. Culler, R. M. Karp, D. Patterson, S. Sahay, E. E. Santos, K. E. Schauser, R. Subramonian, and T. von Eicken. LogP: A practical model of parallel computation. Comm. ACM, 39:78-85, 1996.

No context found.

David E. Culler, Richard M. Karp, David Patterson, Abhijit Sahay, Eunice E. Santos, Klaus Erik Schauser, Ramesh Subramonian, and Thorsten von Eicken. LogP : A practical model of parallel computation. Communications of the ACM, 39(11):78 - 85, 1996.

No context found.

D. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K. Schauser, R. Subramonian, and T. von Eicken. LogP: A practical model of parallel computation. Communications of the ACM, 39(11):78-85, November 1996.

No context found.

D. E. Culler, R. M. Karp, D. Patterson, A. Sahay, E. E. Santos, K. E. Schauser, R. Subramonian, and T. von Eicken. LogP: A practical model of parallel computation. Commun. ACM, 39(11):78-85, November 1996.

No context found.

D.E. Culler, R. Karp, D. Patterson, A. Sahay, E. Santos, K.E. Schauser, R. Subramonian, and T.V. Eicken. LogP: A practical model of parallel computation. Communications of the ACM, 39(11):78--85, November 1996.

No context found.

D. E. Culler, R. M. Karp, D. A. Patterson, A. Sahay, E. Santos, K. E. Schauser, R. Subramonian, and T. von Eicken. LogP: A Practical Model of Parallel Computation. Communications of the A CM, November 1996.

No context found.

David E. Culler, Rachard M. Karp, David Patterson, Abhijit Sahay, Eunice E. Santos, Klaus Erik Schauser, Ramesh Subramonian, and Thorsten von Eicken. LogP: A practical model of parallel computation. Communications of the ACM, 39(11):78--85, November 1996.

First 50 documents  Next 50

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