G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of nesl. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....wonder, therefore, about the utility of being able to bound memory in terms of reachable space if determining the reachable space itself might be nondeterministic. As a partial solution to this problem, however, elsewhere we have shown bounds on the maximum reachable space of parallel programs [11]. We first consider the time bounds. Figure 4.8 shows the call graph for all the collector routines. The functions at the bottom (pushStack, popStack, and Allocate when the collector is on) take constant time since they only InitField Collect CopyLoc PushStack CollectOff CollectOn Write ....

Guy Blelloch and John Greiner. A provable time and space efficient implementation of nesl. In Proceedings of First International Conference on Functional Programming, May 1996.

....for NESL. Prins and Palmer [24] presented a different form of flattening using program transformations; this approach was further refined in [21,20] and here. The thread based execution model of nested parallelism has been shown to respect the step and work complexities of the source level metrics [9,5]. However, overheads and space requirements in the realization of this model require careful run time scheduling [4] fast synchronization [25] and granularity control (in the sense of [10] to make it practical. Blelloch [1] and Suciu and Tannen [34,33] have presented nested parallel languages ....

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In International Conference on Functional Programming, 1996.

....clause [10] also allows control to be separated from value under programmer control. Unlike strategies, moreover clauses are statically evaluated with a view to creating a static map of tasks to processors. Probably the most successful purely functional parallel functional language to date is NESL [1]. NESL s war cry is nested data parallelism, supported by built in operations on one dimensional arrays. NESL is not as general as GpH, but it is implemented on a variety of big parallel machines (Unix workstations, SP 2, CM5, Cray C90 and J90, MasPar MP2, and Intel Paragon) and delivers ....

Blelloch G.E. and Greiner J. "A Provable Time and Space Efficient Implementation of NESL", Proc. ICFP '96, Philadelphia, Pennsylvania, (1996), pp. 213--225.

....for NESL. Prins and Palmer [24] presented a different form of flattening using program transformations; this approach was further refined in [21,20] and here. The thread based execution model of nested parallelism has been shown to respect the step and work complexities of the source level metrics [9,5]. However, overheads and space requirements in the realization of this model require careful run time scheduling [4] fast synchronization [25] and granularity control (in the sense of [10] to make it practical. Blelloch [1] and Suciu and Tannen [34,33] have presented nested parallel languages ....

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In International Conference on Functional Programming, 1996.

.... Our design is novel in a number of respects: ffl Many researchers have proposed the use of static cost models to predict parallel performance, in particular Skillicorn [7] These are most effective where parallelism is limited to a small number of regular constructs, for example as in NESL [8] and FiSh [9] However, in general cost modelling is undecidable: program analysis results in a set of recurrence relations which must then be solved [10] In contrast, we use dynamic prototyping to try and establish typical behaviour. These implementation independent measures are then used with ....

....pmap 4 Example: arbitrary length integer matrix multiplication 4. 1 Overview For example, consider multiplying matrices of arbitrary length integers(ALIs) Such integers are represented as lists of integer digits, from least significant to most significant, left to right: e.g. 189652784532 = [2,3,5,4,8,7,2,5,6,9,8,1] to ease carry propagation. Addition and multiplication of ALIs may be defined as: add carry to ALI ) fun addc 0 t = t addc c [ c] addc c (h: t) h c) mod 10: addc ( h c ( h c) mod 10) div 10) t; fn : int int list int list ( add ALI to ALI with carry ) fun add c ....

Guy E. Blelloch and John Greiner. A Provable Time and Space Efficient Implementation of NESL. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

....theorem [35] None of these papers give a complete formulation of the underlying dynamic and static semantics of the language and thus, the proofs of correctness are necessarily ad hoc. Blelloch and Greiner give an abstract machine for evaluation of the parallel programming language NESL [10]. The goal of their work was to provide provable space and time bounds for an implementation of NESL. Their machine is based directly on the CESK machine [20] However, some details in their formulation, such as the representation of control information, are left implicit. 8 Summary and ....

Guy E. Blelloch and John Greiner. A provable time and space efficient implementation of NESL. In ACM Conference on Functional Programming and Computer Architecture, pages 213--225, Philadelphia, PA, May 1996.

....code. Our design is novel in a number of respects: Many researchers have proposed the use of static cost models to predict parallel performance, in particular Skillicorn [3] These are most effective where parallelism is limited to a small number of regular constructs, for example as in NESL [4] and FiSh [5] However, in general cost modelling is undecidable: program analysis results in a set of recurrence relations which must then be solved [6] In contrast, we use dynamic prototyping to try and establish typical behaviour. These implementation independent performance models library ....

Guy E. Blelloch and John Greiner. A provable time and space efficient implementation of nesl. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

.... [9] and for region inference (see [24, 1] In areas like parallel programming, where efficiency is a paramount issue, good intermediate languages are even more critical to bridge the gap between high level languages (e.g. NESL) and efficient implementations on a variety of architectures (see [2, 3, 22]) However, in this area of computing intermediate languages (e.g. VCODE) have not made significant use of types, yet. This paper proposes a typed intermediate language S2 for vector languages, and shows how it may be used to extract useful information from programs written in the Nested Sequence ....

....useful guidelines for such refinement. There are also obvious extensions to the run time part of S2, e.g. recursive types. One must fill the gap between S2 and parallel machines (or already implemented intermediate languages, like VCODE) Moreover, translations should be efficient in the sense of [22, 3]. We have used a modified version of NSC with for loops rather than while loops. It should be possible to incorporate run time computational aspects in S2 using monads. In such extension it should be possible to translate more realistic languages. c arity and ML like definition of c c (f ; x) ....

Guy E. Blelloch and John Greiner. A provable time and space efficient implementation of NESL. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

.... [9] and for region inference (see [24, 1] In areas like parallel programming, where efficiency is a paramount issue, good intermediate languages are even more critical to bridge the gap between high level languages (e.g. NESL) and efficient implementations on a variety of architectures (see [2, 3, 22]) However in this area of computing, intermediate languages (e.g. VCODE) have not made significant use of types, yet. This paper proposes a typed intermediate language S2 for vector languages, and shows how it may be used to extract useful information from programs written in the Nested Sequence ....

....useful guidelines for such refinement. There are also obvious extensions to the run time part of S2, e.g. recursive types. One must fill the gap between S2 and parallel machines (or already implemented intermediate languages, like VCODE) Moreover, translations should be efficient in the sense of [22, 3]. We have used a modified version of NSC with for loops rather than whileloops. It should be possible to incorporate run time computational aspects in S2 using monads, and thus translate more realistic languages. One cannot expect that array bound checking can all be done at compile time. Indeed ....

G.E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

....over them. Our approach is to redevelop the theory of arrays to support the combinators while retaining constant time access. We illustrate the efficacy of this approach by implementing static estimates of shapes and parallel work, as compared to, say, the dynamically obtained estimates for Nesl [BG96] Efficient array access is based on direct access to the storage of array entries, typically in contiguous blocks of memory. List storage, however, is by allocating cons cells during evaluation, which introduces significant access costs. Our approach to arrays uses the syntax of lists, so that ....

....mechanisms for cost composition in a work and depth circuit complexity model, but severely restricts nesting since only mapping is allowed at outer levels, with folded computations limited to a small number of special cases. Blelloch and Greiner s profiling semantics for Core Nesl [BG96] essentially records work and space usage while fully evaluating programs. By contrast, our cost translation produces a program which contains just enough summary information about the source in order to calculate its costs, as a kind of abstract interpretation. Further, the Core Nesl profiling ....

G.E. Blelloch and J. Greiner. A provable time and space efficient implementation of Nesl. In ACM SIGPLAN '96 International Conference on Functional Programming, pages 213--225, 1996.

....programmer and the system implementer. Such cost models can be used to predict (or bound) execution times, memory utilization, or communication requirements for parallel programs running on real platforms. Other programming approaches that emphasize the use of a cost model include the NESL [9] and Cilk [10, 11] languages, which utilize a work depth cost model, and the Split C language [20] which utilizes the LogP cost model [21] In comparison to other approaches that provide shared memory communication in a synchronous environment, BOS does not require preallocation for each shared ....

G. E. Blelloch and J. Greiner, "A provable time and space efficient implementation of NESL," in Proceedings of the 1996 ACM SIGPLAN International Conference on Functional Programming, (Philadelphia, PA), pp. 213--225, 24--26 May 1996.

No context found.

Guy E. Blelloch and John Greiner. A provable time and space efficient implementation of NESL. In Proceedings of the ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

....time stamp of a value as the depth in the DAG at which it is computed, and then find upper bounds on the time stamps of the results to determine the depth of the computation. The model, as defined here, is basically the PSL (Parallel Speculative # Calculus) 23] augmented with arrays as in NESL [10]. Although the PSL only considered the pure # Calculus with arithmetic operations, the syntactic sugar we include affects work and depth by a constant factor only. In this paper we are actually assuming a slightly simplified model by considering only a first order language (it cannot pass ....

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In Proceedings of the ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

....wonder, therefore, about the utility of being able to bound memory in terms of reachable space if determining the reachable space itself might be nondeterministic. As a partial solution to this problem, however, elsewhere we have shown bounds on the maximum reachable space of parallel programs [4]. This work showed that if S 1 is the maximum reachable space of a serial implementation then the maximum reachable space of the (nondeterministic) parallel version is S 1 O(PD) where D is the depth of the computation (i.e. the length of the longest chain of dependency) Similar bounds hold for ....

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In Proceedings of the ACM SIGPLAN International Conference on Functional Programming, pages 213-- 225, May 1996.

....Blelloch, Gibbons and Matias [4] showed that for nested computations, the time bounds can be maintained while bounding the space by s1 O(pd) which for sufficient parallelism is just an additive factor over the sequential space. This was used to bound the space of the nesl programming language [5]. Narlikar and Blelloch [30] showed that this same bound can be achieved in a non preemptive manner (threads are only moved from a processor when synchronizing, forking or allocating memory) and gave experimental results showing the effectiveness of the technique. All this work, however, has been ....

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In Proc. International Conference on Functional Programming, May 1996.

....these techniques generate executions for a multithreaded computation on p processors that require no more than p Delta S 1 space. A scheduling algorithm that significantly improved this bound was recently proposed [2] and was used to prove time and space bounds for the implementation of NESL [4]. It generates a schedule that uses only S 1 O(p Delta D Delta log p) space on a standard p processor EREW PRAM, where D is the depth of the parallel computation (i.e. the longest sequence of dependencies or the critical path in the computation) This bound is lower than the previous bound ....

Guy E. Blelloch and John Greiner. A provable time and space efficient implementation of nesl. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

....of the dag based on this ordering. It is not hard to show that the transitive reduction of the dag resulting from executing a program in any nested parallel language will always be a (single source and sink) seriesparallel dag (as defined in Section 2) This has been shown for the Nesl language [BG96] Any series parallel dag G is (s; t) planar with counterclockwise edge priorities, where s is the single root source node and t is the single leaf sink node. Thus the P stack algorithm of Section 4.2 can be used to implement a pdf schedule on the dag. Nested parallel constructs include ....

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In Proc. ACM SIGPLAN International Conf. on Functional Programming, pages 213--225, May 1996.

....of a value as the depth in the DAG at which it is computed, and will then find upper bounds on the time stamps of the results in order to determine the depth of the computation. The model, as defined here, is basically the PSL (Parallel Speculative Calculus) 23] augmented with arrays as in NESL [11]. Although the PSL only considered the pure Calculus with arithmetic operations, the syntactic sugar we have included only affects work and depth by a constant factor. In this paper we are actually assuming a slightly simplified model by only considering a first order language (it cannot pass ....

G. Blelloch and J. Greiner. A provable time and space efficient implementation of NESL. In Proc. ACM SIGPLAN International Conference on Functional Programming, May 1996.

No context found.

G. E. Blelloch and J. Greiner. A provable time and space efficient implementation of nesl. In ACM SIGPLAN International Conference on Functional Programming, pages 213--225, May 1996.

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