Hillis, W. D. (1985). The Connection Machine. MIT Press, Cambridge, Ma.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....VFSR has no corresponding attribute and because the GA runs were simulated sequentially. However, a generation performance plot would be a more accurate estimate of the power of the genetic algorithms, especially if the GA simulations were run on parallel hardware such as the Connection Machine [20]. Then all function evaluations (performed in one generation) could be evaluated in parallel. It should be noted that parallel processing of GA can lead to spurious correlations and premature convergence [21] It should be noted in this context that VFSR also lends itself well to parallelization. ....

W.D. Hillis, The Connection Machine, The MIT Press, Cambridge, MA, (1985).

.... in other population based algorithms can be easily adapted to the ACO structure (e.g. migration and di usion models adopted in the eld of parallel genetic algorithms, see for example reviews in [16, 38] Early experiments with parallel versions of AS for the TSP on the Connection Machine CM 2 [63] adopted the approach of attributing a single processing unit to each ant [4] Experimental results showed that communication overhead can be a major problem with this approach on ne grained parallel machines, since ants end up spending most of their time communicating to other ants the modi ....

W. D. Hillis. The Connection Machine. MIT Press, 1982.

.... are an approach to graph rewriting that generalises some existing formalisms: Alan Bawden proposed a framework in 1986, initially by the name Connection graphs , as a programming language for distributed computation based on his experience with the massively parallel Connection Machine at MIT [Baw86,Hil85]. His PhD thesis develops the idea further (now calling them Linear graph grammars , influenced by the link with linear logic described below) and uses them as a key formalism in the design of a compiler, where the linear graph grammars serve as the basis both for an intermediate representation ....

W. D. Hillis. The Connection Machine. MIT Press, Cambridge MA, 1985.

....at a node, it is immediately forwarded to another node; in other words, a packet is treated like a hot potato . Hot potato routing algorithms have been observed to work well in practice [4] and have been used in parallel machines such as the HEP multiprocessor [25] the Connection machine [13], and the Caltech Mosaic C [24] as well as in high speed communication networks [19] Hot potato routing algorithms are well suited for optical networks since it is di#cult to bu#er optical messages [1, 26] 1 L 0 1 2 3 4 5 6 0 1 2 3 Butterfly Mesh L 1 . 0 2 Leveled Network ....

W. D. Hillis. The Connection Machine. MIT press, 1985. 18

....Currently, all the parallel implementations of ACO algorithms use ant level parallelism. These implementations are brie y reviewed in the following. The rst parallel versions of an ACO algorithm was Bolondi and Bondanza s implementation of AS for the TSP on the Connection Machine CM 2 [34]. The approach taken was that of attributing a single processing unit to each ant [2] Unfortunately, experimental results showed that communication overhead can be a major problem with this approach on ne grained parallel machines, since ants spend most of their time communicating to other ants ....

W. D. Hillis. The Connection Machine. MIT Press, 1982.

....and adequate. Fortunately, dramatic strides are being made in computer architecture at just the time that connectionist the oretical models are being explored. These fields are not unrelated. Connectionist models I241141 served as initial inspiration to designer of new generation hardware (e.g. 101) though many parallel architectural ideas were already being explored in the pursuit of greater speed. This followed the realization that we were approaching asymptotes for speeds possible with serial uniprocessors. I believe that developing appropriate hardware will prove to be the easiest ....

....Nonetheless, still there is a potentially large set of cases where weights will have to be generated by hand, or by yet to be discovered learning methods. Clearly, every concept cannot be connected to every other directly. This would require n s connections for n concepts, where n is at least 10c. Some solutions have been suggested (e.g. the microfeature ideas in [37] but none seems easy to program. 2.5 Fault Tolerance Since a large number of nodes (or modules) have a bearing on a single connectionist d.ecision (e.g. lexical selection or prepositional phrase attachment) then not all ....

Hillis, D., The Connection Machine, Cambridge, MA: MIT Press, 1985.

.... schedule multiple, simultaneous data movements across the network, between the functional units and the The tight integration with the processor pipeline is a key design aspect of the Raw architecture [13] which earlier register mapped network architectures including the Connection Machine CM2 [3] and Warp [1] lack. network, as well as between the register set and the network. A typical DSA instruction such as fma 4, 4, N1, W2 route N1 S1, W2 E2 consists of two parts. The fma operation is a oating point multiply and add compound instruction. It multiplies the values arriving on ....

Daniel W. Hillis. The Connection Machine. MIT Press, Cambridge, MA, 1985.

....node. Hot potato routing algorithms can be realized easily in hardware, since there is no need for bu ering or queuing. Moreover, hot potato routing algorithms have been observed to work well in practice. Hot potato algorithms have been used in the HEP multiprocessor [24] the Connection Machine [15], the CalTech Mosaic C [23] and in high speed communication networks [20] Hot potato routing algorithm are also well suited for use in optical switching networks [1, 14] Our algorithm is greedy : whenever possible packets are sent closer to their destination. Greedy algorithms are adaptive: ....

W. D. Hillis. The Connection Machine. MIT press, 1985.

....of classifying diverse load balancing algorithms. A number of well known dynamic load balancing algorithms for different systems and applications have been selected for this purpose. A small description is presented for each algorithm, followed by a detailed classification. The Central Algorithm [3], 4] Firstly, the average workload is computed and broadcasted to every processor in the system. Then, the processors are classified into 3 classes: idle, overloaded, and the others. The algorithm tries to match each overloaded with an idle peer. The Rendez Vous Algorithm [5] 6] It allows ....

Hillis , W.D.: The Connection Machine. MIT press, 1985.

....for 3D) Each process communicates with its parents, its children and its neighbors to compute the solution to the N body problem according to the above lgorithm. If a processor is assigned to each process and all processors operate in a SIMD style, this architecture is connection machine like [12] in the sense that there are many equal processors operating in SIMD style which communicate via many connections to neighbors . The difference between the connection machine ud this network is that this network is not a hypercube; instead, its connections are derived from the structure of the ....

....maps the pyramid of processes to three different processor configurations: A pyramid of processors corresponding to the processes pyramid graph, where each vertex is assigned a process and each arc a communication wire . An hypercube of processors ,i.e. a connection machine like axchitecture [12]. smaller axray of larger processors , e.g. 2 x 2 x 2 processors. The performance of the BH and the GR algorithms are compared by calculating their communication and computation times in these configurations The calculation is done for several typical values of N, the total number of particles, ....

D. Hillis, The Connection Machine. MIT Press, 1985. 51

....to a computational structure in the form of a pyramid like graph, where each vertex is a process, and to a SIMD computational algorithm. A pyramid of processors corresponding to the processes pyramid graph, where each vertex is assigned a process and each arc a communica tion wire . [12]. A smaller array of larger processors , e.g. 2 x 2 x 2 processors. Computation and communication time are computed for the first two cases and estimates are derived for the third case. The main qualitative conclusion is that large simulations (about 1M par ticle) can be performed on any of ....

....for 3D) Each process communicates with its prents, its children and its neighbors to compute the solution to the N body problem according to the above algorithm. If a processor is assigned to each process and all processors operate in a SIMD style, this rchitecture is connection machine like [12] in the sense that there ae many equal processors operating in SIMD style which communicate via many connections to neighbors . The difference between the connection machine and this network is that this network is not a hypercuhe; instead, its connections are derived from the structure of the ....

[Article contains additional citation context not shown here]

D. Hillis, The Connection Machine. MIT Press, 1985. 51

....which permit fast parallel solutions to these problems. Then, we describe each problem and present its solution. In each case, we provide an estimate of the running times for the sample problems on the current version of the Connection Machine. 1. 1 The Connection Machine The Connection Machine [Hillis85] is a powerful fine grained parallel machine having between 16K and 64K processors, operating under a single instruction stream broadcast to all processors (figure 1) It is a Single Instruction Multiple Data (SIMD) machine, because all processors execute the same control stream. Each of the ....

....integer values so that 5 the mesh connections in the Connection can be used for brush fire propaga tion, and the EMST depends on the Voronoi Diagram. All other examples assume real arithmetic when necessary. The parallel computing environment at the MIT AI Lab consists of a Connection Machine [Hillis85] with 16K processors, with a Symbolics 3650 Lisp Machine as host. Connection Machine programs utilize Lisp syntax, in a language called Lisp [Lasser86] Statements in Lisp programs are compiled and manipulated in the same fashion as Lisp statements, contributing signif icantly to the ease of ....

[Article contains additional citation context not shown here]

D. Hillis, The Connection Machine, MIT Press, Cambridge, 1985.

....parallel processing have produced a diverse selection of architectures and programming styles. At the architectural level a broad distinction can be made between parallel machines that execute in single instruction stream, multiple data stream (SIMD) mode (an example is the Thinking Machines CM 2 [1] ) and those which execute in multiple instruction stream, multiple data stream (MIMD) mode (examples are the Sequent Symmetry [2] and Intel iPSC 2 [3] A further distinction can be made within the MIMD category based on the characteristics of the machine s address space(s) If the address ....

W.D. Hillis, The Connection Machine, MIT Press, 1985.

....for shared memory MIMD machines seem to be rather scarce. Many installations use simple thread packages, that allow multiple threads that share the whole address space. On the other hand, there do exist a number of SIMD languages that e ectively allow data sharing through pointers or array indexes [40, 41, 42, 43]. The main limitation in these languages is that all the parallel computations are identical, and must proceed in lockstep. This is relaxed in some recent languages such as HPF [44] where the computations are only loosely synchronous. Finally, some parallel languages are tailored very closely to ....

W. D. Hillis, The Connection Machine. MIT Press, 1985.

....is assumed to take place via circuit switched connections. During every communication cycle processors exchange data, with the (source, destination) pairs forming a bijection between the set of processors and itself. This model idealizes the behaviour of a SIMD system (e.g. the Connection Machine [1]) fetching data which have been skewed to randomize placement and avoid conflicts. Although permutation requests present a worst case scenario, effective routing of simultaneous circuits is crucial in systems having less stringent communication patterns, such as the Intel iPSC 2 which uses ....

W. D. Hillis, The Connection Machine, MIT Press, Cambridge MA, (1985).

....noisy, otherwise is equal to 0.5 (for more details about the relationship between and Ai see [16] 4 Simulation results 4. 1 Implementation on a Connection Machine CM2 In this section, we briefly describe the architecture of the Connection Machine, a more detailed description can be found in [13]. The Connection Machine is a single instruction multiple data (SIMD) parallel computer with 8K to 64K processors. Each processor is a 1 bit serial processor, with 32K bytes of local memory and a 7MHz clock. For a given application, the user can dynamically define a particular geometry for the set ....

W. D. Hillis. The Connection Machine. Cambridge, MA, 1985.

.... processor or MPP and the CLIP series of computers developed by Duff and his colleagues represent the classic embodiment of von Neumann s original automaton [5, 6, 7, 8, 9] A more general class of cellular array computers are pyramids and Thinking Machines Corporation s Connection Machines [10, 11, 12]. In an abstract sense, the vari ous versions of Connection Machines are universal cellular automatons with an additional mechanism added for non local communication. Many operations performed by these cellular array machines can be expressed in terms of simple elementary operations. These ....

W. Hillis, The Connection Machine. Cambridge, MA: The MIT Press, 1985.

.... on several occasions with minor variations to solve different problems: Alan Bawden proposed a framework in 1986, initially by the name Connection graphs , as a programming language for distributed computation based on his experience with the massively parallel Connection Machine at MIT [Baw86,Hil85]. His PhD thesis develops the idea further (now calling them Linear graph grammars , influenced by the link with linear logic described below) and is a key formalism used in the design of a compiler, where the linear graph grammars serve as both an intermediate representation used in compiler ....

W. D. Hillis. The Connection Machine. MIT Press, Cambridge MA, 1985.

....motivated by theory, called Lamping an autodidactic engineer, but the first such real engineer was Bawden. 3. 2 Implementation experience In the mid 1980s, Bawden was working on the seemingly unrelated problem of programming massively parallel SIMD computers, such as the Connection Machine [Hil85] In order to control communications network contention problems 17 in such machines, he was working with programming languages in which communications bottlenecks are prevented by the simple expedient of outlawing the ability to make copies of references. Exploring this space of computational ....

W. Daniel Hillis. The Connection Machine. MIT Press, 1985.

....neural hardware, and seek to support more general classes of ANNs [6] 13] 18] Although some neurocomputers could potentially support dynamic topologies more directly in hardware, rather than in software, they currently do not. Of course, general parallel machines, like the Connection Machine [9] and the CRAY [1] can simulate the desired dynamics in software, but these machines are not optimized for neural computation. LIT supports general classes of ANNs and dynamic topologies in an efficient parallel hardware implementation. LIT redesigns the original network into a hierarchical, ....

Hillis, W. Daniel. (1985).The Connection Machine. Cambridge, Mass.: MIT Press.

No context found.

Hillis, W. D. (1985). The Connection Machine. MIT Press, Cambridge, Ma.

No context found.

W. D. Hillis. The Connection Machine. MIT Press, Cambridge, Ma., 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT Press, Cambridge, Ma., 1985.

No context found.

W. Daniel Hillis. The Connection Machine. MIT Press, 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT press, 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT press, 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT press, 1985.

No context found.

Hillis, W. D. (1985) The Connection Machine. MIT Press, Cambridge, MA.

No context found.

D. Hillis. The Connection Machine. MIT Press, Cambridge, MA, 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT press, 1985.

No context found.

Hillis, W. D. The Connection Machine, MIT Press (Cambridge, MA, 1985).

No context found.

W. D. Hillis, The Connection Machine, Cambridge, The MIT Press, 1985.

No context found.

W.D. Hillis, The Connection Machine, MIT Press, Cambridge, MA, 1985.

No context found.

W. D. Hillis, The Connection Machine, The MIT Press, 1985.

No context found.

W. D. Hillis, The Connection Machine, MIT Press, 1985.

No context found.

W.D. Hillis, "The Connection Machine", MIT Press, 1985.

No context found.

D. W. Hillis. The Connection Machine. The MIT Press, Cambridge, MA, London, England, 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT press, 1985.

No context found.

Danny Hillis. The Connection Machine. MIT Press, Cambridge, Massachusetts, 1985.

No context found.

W. D. Hillis. The Connection Machine. MIT press, 1985.

No context found.

W.D. Hillis, The Connection Machine (The MIT Press, Cambridge, MA, 1985).

No context found.

Hillis, W. D. The Connection Machine. The MIT Press, Cambridge, 1985.

No context found.

W. D. Hillis, The Connection Machine, MIT Press, Cambridge, Mass., 1985.

No context found.

W. D. Hillis, The Connection Machine, (MIT Press, Cambridge, MA, 1985).

No context found.

Hillis, W.D., The Connection Machine, MIT Press, Cambridge, Mass., 1985.

No context found.

D. Hillis, The Connection Machine, IV[IT Press, Cambridge, 1985.

No context found.

Hillis, W. D (1985) The Connection Machine. MIT Press, Cambridge, Massachusetts, USA.

No context found.

W.D. Hillis. The Connection Machine. The MIT Press, Cambridge, MA, 1985.

No context found.

W.D. Hillis, The connection machine, The MIT Press, 1985.

No context found.

Hillis, W. D. 1985: The Connection Machine, MIT Press, Cambridge.

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