Gelernter, D., Carriero, N., Chandran, S., Chang, S.: Parallel programming in linda. In: Proceedings of the International Conference on Parallel Programming. (1985) 255--263  Home/Search   Document Not in Database   Summary   Related Articles   Check

....memory performance found for parallel programming applications. Indeed, as we demonstrated, remote I O performance is equal to local I O performance. Perhaps more similar to the River environment is Linda, which provides a shared, globallyaddressable, tuple space to parallel programs [Car87, GCCC85] Applications can perform atomic actions on tuple space, inserting tuples, and then querying the space to find records with certain attributes. Because of the generality of this model, high performance in distributed environments is difficult to achieve [BKT92] Thus, while the distributed ....

David Gelernter, Nicholas Carriero, Sharat Chandran, and Silva Chang. Parallel programming in Linda. In 1985 International Conference on Parallel Processing, pages 255--263, 1985.

....CCS [28] A distinguishing feature is that tuples and operations over them are located at speci#c sites of a net and types are used to control access rights of processes over these sites. We start this section by summarizing the main features of Linda (the interested reader is referred to, e.g. [23, 16, 15] for more details) Then, we present the syntax of KLAIM (processes, types and nets) Most of the presentation of the untyped part of the language is borrowed from [18] There, we also outline the main features of the KLAIM type system without providing the actual syntax for types, the explicit ....

D. Gelernter, N. Carriero, S. Chandran et al., Parallel programming in Linda, Proc. IEEE Internat. Conf. on Parallel Programming, IEEE Computer Society Press, Silverspring, MD, 1985, pp. 255--263.

....Kenneth Manning, Pamela McCorduck, Joseph Traub, Hortense Calisher, Curtis Harneck, Wal Mee Ching and Mark Gilpin for the invaluable friendship and affection. 8 Finatly, my deepest gratitude to all mabets of tire e mtily: mmtltr, Hekn, Introduction In the parallel programming system Linda [4] [13], processes (called workers in this thesis) are uncoupled in time and space: they store and pick up logical tuples, units of data in Linda, in a shared memory like data structure referred to as the tupIe space. A typical Linda system consists of several workers and a tuple space. The tuple space ....

....Operations are executed as soon as the data needed are available. This chapter describes the Linda data structure and its operations, uses a simple example to explain how a Linda program runs, and introduces the notion of a Linda ke nel. More detailed descriptions of Linda can be found in [13] and [9] 2.1 Logical Tuples and Operations The basic data unit in Linda is a logical tuple, or tuple for short. A tuple contains a logical name followed by one or more ordered data elements, which can be either data values such as 1 , true , and John , or formals, which are typed variables ....

David Gelernter, Nicholas Carriero, Sarat Chandran, and Silva Chang. Parallel Programming in Linda. In Proceedings of International Conference on Parallel Processing, IEEE, 1985.

....Collections of (possibly identical) tuples are deposited in tuple spaces and hence shared between processes. Tuples are persistent objects and cannot be altered while they reside in a tuple space: in order to effect changes to a tuple it must be explicitly withdrawn, changed and then re inserted [Gelernter,85b] Two forms of tuples are defined by the tuple space model: passive and active. Passive tuples are of the form described above, containing a mixture of actual and or formal fields from the time of their creation through to that of their destruction. Active tuples differ by virtue of the fact ....

....in order to avoid confusion with a recent Lucent Technologies product of the same name. 93 require additional support services from their distributed systems platform, e.g. adaptive mobile applications. L2imbo incorporates a number of extensions to the original Linda model [Gelernter,85a] Gelernter,85b] which enable it to address the requirements of mobile environments that were identified in chapter 1. These extensions, some that are novel and others of which were inspired by the work surveyed in section 4.3, include: Tuple Space Model: multiple local, distributed and centralised tuple spaces ....

D. Gelernter, N. Carriero, S. Chandran and S. Chang, "Parallel Programming in Linda", Proceedings of the International Conference on Parallel Processing, August 1985, pp 255-263.

....components. Very often, network operating systems [12] do not provide high level interprocess communication facilities. In fact, building distributed applications even if feasible, becomes a hard work which requires an expert knowledge of the low level primitives of the network interface. Linda [2, 3, 4, 5, 6] is a high level parallel programming language intended for distributed programming. It consists of a reduced set of primitives over a virtual, associative shared memory, called tuple space, which can be embedded in any sequential programming language to transform it in a parallel language. ....

....an operational semantics of Linda. In Section 4, we discuss several alternatives to design a run time system for Linda in a computer network. In Section 5, we present the implementation developed. 2 Linda Linda is a reduced set of simple operations intended for distributed and parallel programming [2, 3, 4, 5, 6]. Linda is based on a tuple space memory which supports generative communication. 2.1 Tuple Space A tuple space is an associative memory which storage unit is the tuple. A tuple can be added, removed or read using the out, in or rd operations, respectively. Tuples are addressed via any collection ....

[Article contains additional citation context not shown here]

D. Gelernter, N. Carriero, S. Chandran, S. Chang. "Parallel programming in Linda". Proceedings of the International Conference on Parallel Processing, Aug. 85, 255-263.

....Klaim provides direct support for expressing and enforcing policies that control access to resources and authorize migration and execution of mobile processes. Klaim exploits a capability based type system as an aid to specify and enforce access control policies. Klaim consists of core Linda [6, 5, 2] with multiple located tuple spaces. A Klaim program, called a net, is structured as a collection of nodes. Each node has a name, and consists of a process component and a tuple space component. Sites are the concrete names of the nodes and are the main linguistic constructs to provide network ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proc. of the IEEE International Conference on Parallel Programming, pp. 255-263, IEEE Computer Society Press, 1985. 3

....CCS [28] A distinguishing feature is that tuples and operations over them are located at specific sites of a net and types are used to control access rights of processes over these sites. We start this section by summarizing the main features of Linda (the interested reader is referred to, e.g. [23, 16, 15] for more details) Then, we present the syntax of Klaim (processes, types and nets) 4 Most of the presentation of the untyped part of the language is borrowed from [18] There, we also outline the main features of the Klaim type system without providing the actual syntax for types, the ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proc. of the IEEE International Conference on Parallel Programming, pp. 255-263, IEEE Computer Society Press, 1985.

....provided in C itself (such as operator overloading and templates) and provide a rich library of operators useful in scientific engineering applications. 3.4.4 Other approaches Several additional languages are being used in parallel CSE applications. We will cite just a few examples. Linda[15] allows multiple parallel processes to communicate and coordinate by sharing a set of tuples: a tuple contains a sequence of data items. A process can deposit tuples in the shared tuple space, and request the system to retrieve tuples with specific characteristics. The language Cid[26] combines ....

David Gelernter, Nicholas Carriero, S. Chandran, , and Silva Chang. Parallel programming in Linda. In International Conference on Parallel Processing, pages 255--263, Aug 1985.

....is simply too small for the computation to amortize the communication overhead on larger systems. In Figure 7 and Figure 8 for larger square matrix multiplies, we begin to see an improvement in the linearity of speedup in all cases as problem size is increased. 7 Related Work PVM [3] and Linda [8] are examples of software systems that enable parallel processing on a cluster of workstations using existing LAN facilities. Interworkstation communication is accomplished through interface routines implemented over existing Unix interprocessor communication and networking facilities. These ....

D. Gelernter. Parallel programming in Linda. In Proceedings of International Conference on Parallel Processing, August 1985.

....small as the maximum of the times required for the individual function evaluations. Although linked lists are not used very often in parallel programming, note the future construct and its associated dataflow synchronization can be used in the context of other data structures. The Linda language [17] also folds synchronization into data accesses, although in the case of Linda, synchronization is done during associative access of a shared tuple space. 4.2 Object oriented MIMD Languages We describe HPC [13] and Java [18] These are both shared memory languages. 4.2.1 HPC HPC [13] is a ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in Linda. In Proceedings of the International Conference on Parallel Programming, pages 255--263, August 1985. 22

....hiding communication latency via asynchrony [30] Load balancing is provided via a distributed task queue [126] but the user can tailor load balancing as he or she desires to suit the needs of the application. Linda: Linda provides a shared, globally addressable, tuple space to parallel programs [29, 51]. Applications can perform atomic actions on tuple space, inserting tuples, and then querying the space to find records with certain attributes. Because of the generality of this model, high performance in distributed environments is difficult to achieve [11] There have been a large number of ....

David Gelernter, Nicholas Carriero, S. Chandran, and Silva Chang. Parallel programming in Linda. In D. Degroot, editor, 1985 International Conference on Parallel Processing, pages 255--263, 1985.

....of a set of operators, borrowed from Milner s CCS [30] for building processes. The distinguishing feature is that tuples and operations over them are located at specific sites of a net. We start this section by summarizing the main features of Linda (the interested reader is referred to, e.g. [22, 11, 10] for more details) Then, we present the syntax of Klaim. The process algebraic operators will be briefly presented in the subsection that contains the syntax of Klaim processes. 2.1 An overview of Linda Linda is a coordination language that relies on an asynchronous and associative ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proceedings of the Internatinal Conference on Parallel Programming, IEEE, pp. 255-263, 1985.

....policies for access control. The main guidelines for the design of Klaim and of its type system are: Klaim processes and types are network aware; networks and processes are di erent entities; access control policies are network coordination policies. Klaim consists of core Linda [17, 16, 11] with multiple located tuple spaces. A Klaim program, called a net, is structured as a collection of nodes. Each node has a name and consists of a process component and a tuple space component. Sites are the concrete names of the nodes and are the main linguistic constructs for providing ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proc. of the IEEE International Conference on Parallel Programming, pp. 255-263, IEEE Computer Society Press, 1985.

....applications. Klaim provides direct support for expressing and enforcing access control policies to resources and for authorizing migration and execution of mobile processes. Klaim exploits a capability based type system to specify and enforce access control policies. Klaim consists of core Linda [11, 10, 5] with multiple located tuple spaces. A Klaim program, called a net, is structured as a collection of nodes. Each node has a name, and consists of a process component and a tuple space component. Sites are the concrete names of the nodes and are the main linguistic constructs to provide network ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proc. of the IEEE International Conference on Parallel Programming, pp. 255-263, IEEE Computer Society Press, 1985.

....Klaim provides direct support for expressing and enforcing policies that control access to resources and authorize migration and execution of mobile processes. Klaim exploits a capability based type system as an aid to specify and enforce access control policies. Klaim consists of core Linda [15, 14, 9] with multiple located tuple spaces. A Klaim program, called a net, is structured as a collection of nodes. Each node has a name, and consists of a process component and a tuple space component. Sites are the concrete names of the nodes and are the main linguistic constructs to provide network ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proc. of the IEEE International Conference on Parallel Programming, pp. 255-263, IEEE Computer Society Press, 1985.

....to test and the list of fitness test cases, it simulates the controller for each of the test cases, and reports the fitness back to the master process. With N fc = 6, typically, this means that each task takes between 5 and 10 seconds. This is exactly the granularity which the Linda manual (Gelernter, 1985) specifies as optimal. We tested the parallel version of the GP software on a network using an R4000 Crimson, an R4000 Indigo, and three R3000 Indigos. The load balancing came naturally and the result was that controllers could be derived approximately three times faster than when the single R4000 ....

Gelernter, D. (1985). Parallel programming in linda. Technical Report 359, Yale University Department of Computer Science.

....section we first give an overview on the LiPS and Transis systems. The last subsection explains, why we need TRIPS in LiPS and shows, why Transis is a good choice for its implementation. 2. 1 LiPS Overview The LiPS system provides the application programmer with the tuple space paradigm [Gel85, GCCC85, GCS86, GCL86, BCGL87, Gel88, Gel86] of distributed computing, where the processes of a distributed application communicate through a shared memory that is laid out as a tuple space. This tuple space is replicated on several machines, each using a server process called the MessageServer. The ....

Gelernter D., Carriero N., Chang S., and Chandran S. Parallel Programming in Linda. IEEE Transactions on Computer, 1985.

....on more than 1000 machines within the next years. This paper presents some basic decisions taken when designing LiPS version 2.4. This version supports a software fault tolerant generative communication paradigm based on the tuple space, as introduced by the coordination language Linda [GCCC85] The next chapter contains an introduction to generative communication, a programming paradigm suited to implement distributed computations in networks of workstations. Then, an introduction to the terminology used for coping with faulttolerance is given along with the model of software failure ....

Gelernter D., Carriero N., Chang S., and Chandran S. Parallel Programming in Linda. IEEE Transactions on Computer, 1985.

....connected to the campus network at the University of Saarbrucken [BMS95, LMMS94, BLZ93, WD95] This paper presents some basic decisions taken when designing LiPS version 2.4. This version supports a fault tolerant generative communication paradigm based on the tuplespace, as introduced by the Linda[GCCC85] system. Fault tolerance is achieved by logging messages exchanged among processes, and keeping process checkpoints. Checkpoints are used to periodically save the state of individual processes while message logging keeps track of all communication activities among processes belonging to a ....

Gelernter D., Carriero N., Chang S., and Chandran S. Parallel Programming in Linda. IEEE Transactions on Computer, 1985.

....Closely coupled MPP systems are built by Convex. As to SCI connected distributed systems, there is a number of companies implementing hardware connections. The low cost is an important factor in the advance of the parallel processing. The emerging software platforms such as PVM [11] and Linda [12, 13] in the last few years are strong evidence. With the current network latency and bandwidth, solving a problem in parallel on a number of workstations is not very effective. The cost effectiveness of using existing workstations is so appealing that more and more applications are being ported on ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in linda. In Proc. Int. Conf. Parallel Processing, pages 255--263, 1985.

....requires to transmit data to another process, it terminates, yielding a value, and it is this value which is read by the receiving process. To access each other s data, it is necessary to share address spaces, though this may be used in a variety of ways (for example in an associative manner ( Gelernter 85] Carriero 86] Carriero 88] Live data structures are crucial in supporting languages which have nonstrict data structures, that is, data structures that can be accessed and manipulated before they are fully defined. This is a common requirement in parallel versions of lisp ( Halstead 85] ....

....creation of a broad corpus of knowledge about concurrent programming. The report analyses 19 toolsets of varying scope and maturity, using a set of classifications en route to eventual evaluation. The range of tools encompasses simple parallel programming languages (or constructs) such as Linda [Gelernter 85] and Strand [Foster 89] through to integrated environments consisting of an editor, debugger, performance monitors, and assorted program visualisers such as EXPRESS [Likudome] and TOPSYS [Bemmerl 88] The report falls far short of true comparative analysis of the offerings, but is a useful ....

D Gelernter et al, Parallel Programming in Linda, Proc Int Conf on Par Prog, St Charles IL, IEEE pp255-263, 1985.

....feature is that tuples and operations over them are located at specific sites of a net and types are used for controlling the access rights of processes relative to these sites. We start this section by summarizing the main features of Linda (the interested reader is referred to, e.g. [26, 17, 16] for more details) Then, we present the syntax of Klaim (processes, types and nets) Most of the presentation of the untyped part of the language is borrowed from [21] There, we also outline the main features of the Klaim type system without providing an actual syntax for types and the proof ....

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proceedings of the International Conference on Parallel Programming, IEEE, pp. 255-263, 1985.

....through the system with little cost. Further, none of these systems attempt to deal with the problem of slow producers, which is important in our environment. Perhaps more similar to the River environment is Linda, which provides a shared, globally addressable, tuple space to parallel programs [11, 22]. Applications can perform atomic actions on tuple space, inserting tuples, and then querying the space to find records with certain attributes. Because of the generality of this model, high performance in distributed environments is difficult to achieve [4] Thus, while the distributed aspects of ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in Linda. In D. Degroot, editor, 1985 International Conference on Parallel Processing, pages 255--263, 1985.

....hope that it will be accepted as the model, or proposeprogramming languages that can be efficiently mapped to a variety of competing models. Recent proposals by Valiant [41] and Steele [40] fall into the first category. Systems and notations such as Paralex, Par [23] UNITY [21] Linda [28], CODE, P 3 L [24] Prelude [42] and Phase Abstractions [35] fall into the second camp. In the case of Paralex, we inherit the properties of the data flow notation and keep further goals for architecture independence rather modest. Within the class of MIMD architectures, we strive for ....

D. Gelernter, N. Carriero, S. Chandran and S. Chang. Parallel Programming in Linda. In Proc. Int. Conf. Parallel Processing, St. Charles, Illinois, August 1985, pp. 255--263.

....Two related studies offer additional insight. Bal, Steiner and Tanenbaum (1989) focus on the language issue in distributed systems. Chang and Smith (1990) examined the features of parallel programming tools, but the notion of the conceptual model of computation is not addressed. The Linda language (Gelernter et al. 1985; Gelernter, 1989; Carriero and Gelernter, 1988a, 1988b) uses the concept of a tuple space for communication between concurrent processes. Processes use atomic operations to read, add or delete a tuple to from the tuple space. This model is powerful but it does not provide a high level parallel ....

D. Gelernter, N. Carriero, S. Chandran and S. Chang. Parallel Programming in Linda, in Proceedings of the International Conference on Parallel Processing, 1985, pp. 255263.

....in describing and implementing various properties of reactive systems [15, 4] Concurrent METATEM is an instance of an abstract model of computation, called the CMP Model. Lack of space prevents us from exploring this model in full, but we note its links with coordination languages, such as Linda [12], where heterogeneous networks of objects can be constructed, and give a brief overview of its features in 2. An important aspect of this model, and hence of Concurrent METATEM, is the separation of each individual object s definition into its interface definition and its internal definition, ....

....but not only is the identity of the sender always incorporated into a message, but also objects cannot manipulate their message queue as is intended in Concurrent METATEM. One, more fundamental, difference between BSP and our approach is that objects in BSP are message driven. The Linda model [12] has some similarities with this approach in that the shared data structures represented in the Linda tuple space can be seen as providing a broadcast mechanism for data. However, Linda is a general coordination language, while our computational model fixes much more than just the basic ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in Linda. In International Conference on Parallel Processing, August 1985.

....memory than to repeatedly reference them remotely. In addition, we would like to access objects directly on reference, without introducing extra levels of indirection. 3. 1 Previous Systems Several systems have already been developed for distributed parallel programming, including Ivy [4] Linda [3] and Tarmac [1] 5] Shared virtual memory systems such as Ivy move entire memory pages between processors, clearly inappropriate for a problem in which an object averages 100 bytes. Linda is based on the tuple space abstraction, and requires that all shared data be encoded as tuples. The ....

D. Gelernter, "Parallel Programming in Linda," Proceedings of the International Conference on Parallel Processing, pp. 255-263, Aug. 1985.

....processor. If these scheduling policies are employed, parallel programmers and interactive users can peacefully coexist on a NOW. 1 Introduction Rapid improvement in workstation performance has resulted in widespread interest in using networks of workstations (NOWs) for parallel processing [Gelernter 1985, Kronenberg et al. 1986, Carriero Gelernter 1989, Sunderam 1990, Blumrich et al. 1994] Because of the economies of scale, building parallel computers from mass produced hardware and software components can be more cost effective than building the system from scratch. Though massively parallel ....

Gelernter, D. Parallel Programming in Linda. In Proceeding of the International Conference on Parallel Processing, pp. 255--263, August 1985.

....through the system with little cost. Further, none of these systems attempt to deal with the problem of slow producers, which is important in our environment. Perhaps more similar to the River environment is Linda, which provides a shared, globally addressable, tuple space to parallel programs [10, 21]. Applications can perform atomic actions on tuple space, inserting tuples, and then querying the space to find records with certain attributes. Because of the generality of this model, high performance in distributed environments is difficult to achieve [4] Thus, while the distributed aspects ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in Linda. In D. Degroot, editor, 1985 International Conference on Parallel Processing, pages 255--263, 1985.

....communication arrangements. Once we have settled these issues we just need to look at where messages are going to then implement communication channels if that is indeed more efficient. As another type of object based system mention must also be made of the Linda parallel programming paradigm of [Gelerter et al. 1985] . With its central tuple space this provides another mechanism for communication between modules here called processors. The processors themselves can be programmed in any of a variety of languages but only interact with the tuple space through special Linda operators. This can be seen as a ....

D Gelerter, N Carriero, S Chandran, and S Chang. Parallel programming in Linda. In International Conference on Parallel Processing, August 1985.

....tuples exist in shared data objects called tuple spaces. Tuples can be dynamically deposited in and removed from a tuple space, though they can not be altered while resident in it. Changes can, however, be made to a tuple by withdrawing it from the tuple space, amending and then reinserting it [19]. Tuple spaces are shared between collections of processes, all of which have access to the tuples contained within. In classic distributed environments processes communicate across virtual channels described by bindings and formed from pairs of endpoints, c.f. Chorus ports and UNIX BSD 4.3 ....

D. Gelernter, N. Carriero, S. Chandran and S. Chang, Parallel Programming in Linda, in: Proc. International Conference on Parallel Processing (August 1985) 255-263.

....that parallel programming is difficult) Also importantly, compiler technology for data parallel languages is advancing by leaps and bounds, since it is possible to extract quite a bit of knowledge about the problem structure from the language constructs. Two parallel programming systems, Linda [7] and Distributed Memo [11] can represent data of arbitrary structure without mapping or distortion. Both have a globally accessible storage space for data items wherein items can be accessed by name. In both languages, the naming scheme is infinite. Therefore, one can store data of arbitrary ....

David Gelernter, Nicholas Carriero, S. Chandran, , and Silva Chang. Parallel programming in Linda. In International Conference on Parallel Processing, pages 255--263, Aug 1985.

....various thread ports Table 8.1: Paradigms for various parallel languages among processes through these shared variables can generates sufficient parallelism to model stream based programming. see [60] For example, the stream based quicksort can be written in PARLOG declaratively. 28] Linda [23,10] is a coordination language that provides a shared space (called tuple space) for tasks to communicate. Synchronization and dataflow relationship among tasks is enforced by tuple space operations implicitly. Linda puts little restriction on programming styles ( which actually depend on mother ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in linda. In International Conference on Parallel Processing, 1985.

....but not only is the identity of the sender always incorporated into a message, but also objects cannot manipulate their message queue as is intended in Concurrent METATEM. One, more fundamental, difference between BSP and our approach is that objects in BSP are message driven. The Linda model [17] has some similarities with this approach in that the shared data structures represented in the Linda tuple space can be seen as providing a broadcast mechanism for data. However, our computational model fixes much more than just the basic communication and distribution system. As mentioned ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in Linda. In International Conference on Parallel Processing, August 1985.

....CFT [56] PTRAN [60] 2. hardware specific : CFD [63, 54, 55] MPL [47] languages : DAP Fortran [55] 3. architecture class oriented languages MIMD based : Occam [35, 13] SIMD based : C [59, 58] Parallaxis [9, 10] Problem oriented languages explicitly parallel 4. task parallel : PVM [27] Linda [28], languages : Ada [2, 6, 12] 5. data parallel : Modula 2 [34] Vector C [46] languages : Actus [53, 55] Fortran D [26] HPF [43] implicitly parallel 6. functional : Haskell [37, 38] languages : extended ML [5] 7. data flow : VAL [49] Id [21, 51] languages : Sisal [48, 15] 8. equational : ASL ....

D. Gelernter, N. Carriero, Chandran S, and S. Chang. Parallel Programming in Linda. In Douglas Degroot, editor, Proceedings of the 1985 International Conference on Parallel Processing, pages 255--263. Computer Society Press, 1985.

....and the domains concept of MuPAD. The implementation of sponsoring is not possible in an easy way. Cache Data of Cluster 1 Data of Cluster 2 Memory of Cluster 1 Memory of Cluster 2 Memory Shared Figure 3: Macro parallelism implemented with a shared memory 4. 2 Tuple Space Tuple space [7, 8, 9] is a simple paradigm to parallelize sequential languages. A tuple space is a kind of a shared memory. There do not exist addresses in this memory. Instead data with specified properties can be read out. This paradigm is broadly used and accepted [1] and is used by the computeralgebra system ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in linda. In D. Degroot, editor, 1985 International Conference on Parallel Processing, pages 255--263, 1985.

....memory (tables, name registries) distributed log files, databases, and numerical iterative methods that produce values to be reduced by an associative and a commutative operation. 3 Associative segments provide an abstraction similar to the tuple space abstraction supported in Linda [GCCC85, Gel89] The following sections describe the semantics of both temporal and spatial consistency in Unify. 3.1 Temporal Consistency Temporal consistency methods in Unify fall into one of two categories: application aided consistency and automatic consistency. Application aided consistency ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel Programming in Linda. In Proceedings of the International Conference on Parallel Processing, August 1985.

....manages its resources. These systems mandate that the distributed kernel must execute on all the system nodes a limiting factor for many applications and environments. Distributed Programming Environments ################################# such as ISIS [3] Marionette [4] PVM [5] and Linda [6] consist of a software layer that executes above the individual operating systems of autonomous machines, and provides various distributed facilities. While these systems are more flexible than distributed operating systems, they too have disadvantages, e.g. limited applicability, unfamiliar ....

Gelernter, D., et al, "Parallel Programming in Linda", Proc. Intl. Conf. on Parallel Processing, IEEE, pp. 255-263, 1985.

....of failure tolerance in distributed applications will not diminish. Preparing for, detecting, and recovering from failures are error prone tasks unrelated to the goals of the programmer which should be provided by the programming environment. Experimental versions of environments like Linda [17, 49, 3], PVM [44, 16, 27] and the V kernel (e.g, 23] already offer support for failure tolerance. These systems provide this support at the cost of considerable implementation effort. Since reliable operation is seldomly the focus of work on programming environments, it is necessary to diminish the ....

Gelernter, D., Carriero, N., Chandran, S., and Chang, S. Parallel Programming in Linda. in: Proceedings of IEEE International Conference on Parallel Processing. 1985.

....distributed systems. The idea of viewing a collection of workstations on a network as a parallel multiprocessor is a popular one. There are numerous other projects that have been experimenting with different abstractions to provide on top of such a system. For the most part, systems such as Linda [15], Amber [12] nigen [2] AERO [3] and PVM [6] exist as library functions callable by applications and cannot be considered programming environments. On the other hand, systems such as apE [13] CODE [10] FrameWorks [18] and HeNCE [7] include a graphical notation for parallel computations but no ....

D. Gelernter, N. Carriero, S. Chandran and S. Chang. Parallel Programming in Linda. In Proc. Int. Conf. Parallel Processing, St. Charles, Illinois, August 1985, pp. 255--263.

....to be executed. Additionally, there is an always running send and receive thread controlled by the task manager. The communication between the threads can be synchronized through the access to the buffers of the task manager. The communication between the agent processes has been based on LINDA [11]. LINDA libraries provide us with a virtual shared memory (tuple space) This enables a simple implementation of the message passing concepts. Internally LINDA uses shared memory and sockets for the inter process comunication whenever appropriate. The varying communication costs are reflected in ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in linda. In International Conference on Parallel Processing, pages 255--263, Los Alamitos, Ca., USA, Aug. 1985. IEEE Computer Society Press.

....applications that access resources through this model. There are many plausible abstractions of distributed systems. Some examples of broad classes are: general message passing (e.g, Hoa78] distributed objects (e.g, DLA89] distributed or virtual shared memory (e. g, BZ91, BCZ91] Linda [GCCC85, ACG86] and languages (e.g, Tse90] In addition, several programming styles are often used on these paradigms, e.g, bag of tasks, master slave, producer consumer, macro pipelining, data parallel. These styles are some times incorporated as part of the abstraction for the sake of efficiency or ....

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in linda. In Proceedings of IEEE International Conference on Parallel Processing, 1985.

....are the values of name bindings defined in it. Evaluating an ALPHA form causes expressions in the ALPHA to be reduced to the value they represent. The parallel semantics of the language makes the reduction process a concurrent one. Symmetric Lisp is an excellent fit to Linda machines as well. [12] discusses the implementation on Linda of a preliminary form of Symmetric Lisp. Linda and Linda machines are described in [10] 5] Symmetric Lisp runs currently as a sequential interpreter written in Common Lisp. 8 Comparison to Other Work Very few programming languages support environments as ....

David Gelernter, Nick Carriero, Sarat Chandran, and Silvia Chang. Parallel programming in Linda. In International Conference on Parallel Processing, pages 255--263, August 1985.

No context found.

Gelernter, D., Carriero, N., Chandran, S., Chang, S.: Parallel programming in linda. In: Proceedings of the International Conference on Parallel Programming. (1985) 255--263

No context found.

D. Gelernter, N. Carriero, S. Chandran, and S. Chang. Parallel programming in Linda. In ICPP 1985, pages 255--263. IEEE.

No context found.

D. Gelernter, N. Carriero, S. Chandran, et al. Parallel Programming in Linda. Proceedings of the Internatinal Conference on Parallel Programming, IEEE, pp. 255-263, 1985.

No context found.

D. Gelernter, N. Carriero, S. Chandran and S. Chang, "Parallel Programming in Linda", Proceedings of the International Conference on Parallel Processing, August 1985, pages 255263.

No context found.

D. Gelernter, N. Carriero, S. Chandran and S. Chang, "Parallel Programming in Linda", Proceedings of the International Conference on Parallel Processing, pp 255-263, August 1985.

No context found.

Gelernter, D., Carriero, N., Chandran, S. and Chang, S. (1985) Parallel Programming in Linda. Proceedings of the International Conference on Parallel Processing, 255-263.

No context found.

Gelernter, D., Carriero, N., Chandran, S. and Chang, S. (1985b) Parallel Programming in Linda. Proceedings of the International Conference on Parallel Processing, 255-263.

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