D. Maskit. "A Message-Driven Programming System for Fine-Grain Multicomputers." California Institute of Technology Computer Science Department, CS-TR-94-03 (MS thesis) , Feb. 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... the Intel Paragon and the Cray T3D, which utilize two dimensional non torus and threedimensional torus meshes, respectively [7,12] They are also appropriate for research machines such as the Intel Touchstone Delta (two dimensional non torus) and the J and M Machines (three dimensional non torus) [9,13,14,15,17]. Many of these algorithms and techniques also apply to hypercubes (which can be viewed as multidimensional meshes with two nodes in each dimension) although that architecture is not specifically addressed. 4 Target Collective Communications The following collective communications routines are ....

D. Maskit. "A Message-Driven Programming System for Fine-Grain Multicomputers." California Institute of Technology Computer Science Department, CS-TR-94-03 (MS thesis) , Feb. 1994.

....specifics for the Graph Library that motivate the associated design decisions. 3.2.1 Low Latency RPC Mechanism As discussed in Section 3. 1, the Graph Library itself is an abstraction of a low latency, RPCbased programmingmodel found in programming systems such Message Driven C (MDC) [21]. 3.2. IMPLEMENTATION SPECIFICS 27 The RPC model used in the Graph Library allows the user to invoke a function with arguments on a remote computer: ffl f(len, args) comp: This statement causes the remote invocation of the function f with arguments args of length len on computer c. As with the ....

D. Maskit and S. Taylor. "A Message-Driven Programming System for Fine-Grain Multicomputers. " Software: Pract. and Exp., 24:953-980, 1994.

....were running programs within a few hours. After revising the chips in early 1992 to correct a few bugs, we built three J Machines: a 1024 node machine at MIT and 512 node machines at Caltech and Argonne. In parallel with the hardware development, several software systems were built for the machine [HCD89, Wal92, SD91, Mask94]. A die photo of the MDP chip with outlines identifying the major components is shown in Figure 1. The chip measures 1 x 1.5cm in a 1.2 m CMOS process. It was designed primarily using standard cells with hand placement for the data paths. The 32 bit integer CPU, at the lower right, measures 3.7 x ....

....by hardware on message arrival, synchronization was supported with presence bits on all memory locations and registers, and global (segmented) addressing was supported across the machine. These mechanisms were used to implement object oriented [HCD89] data flow [SD91] and message passing [Mask94] programming systems on the prototype hardware. In 1987 when the MDP was proposed, people built machines specialized for one model of computation: data flow machines, shared memory machines, message passing machines, and so on. It was generally accepted that a hardwired implementation of the ....

D. Maskit and S. Taylor, "A Message-driven Programming System for Fine-grain Multicomputers," Software - Practice and Experience. 24(10), pp 953-980, October 1994.

....have been emerging recently to ameliorate the problems of traditional message passing style programming. Among these are generic models such as Active Messages [24] and those aiming at better performance and or easing parallel programming, by means of messagedriven execution models. MDC [14] is a message driven programming system for the J machine [5] that introduces a code distribution scheme to cope with the shortage of memory on this machine. More closely related to the latency driven approach are Charm [9] and Cilk [2] Active Messages are suited for substantially more efficient ....

D. Maskit and S. Taylor. A message-driven programmingsystem for fine-grain multicomputers. Software---Practice and Experience, 24(10):953--980, October 1994.

....in a manner suited to the dependency checker. The Jade language is, in the end, so far removed from C that it is unclear why the authors have not produced entirely new constructs for describing data structures and program flow control. 3. 4 Call By Value Result (data flow) Message Driven C, MDC, [10] is an example of the class of language that could be characterised as Call By Value Result. Parallelism is supported by calling a function which accepts only values as arguments not references to shared data structures executes in parallel, and finally returns a result. These functions ....

Danial Maskit. A Message-Driven Programming System for Fine-Grain Multicomputers. MSc Thesis, Scalable Concurrent Programming Laboratory, CalTech, Pasedena, Ca., February 1994.

....Transmission of data between processes will be handled via explicit message passing. By default, all code in a program will be treated as distributed, and attempts to fetch code which is not physically local will result in implicit communication. The notion of function replication described in [5] will be supported by mapping the range of global memory containing specific functions into physically local memory at all nodes in a partition. This same technique of virtual to physical mapping will be used to support global variables. A process consists of a contiguous piece of virtual memory ....

Maskit, Daniel, A Message-Driven Programming System for Fine-Grain Multicomputers, Masters Thesis, California Institute of Technology, February, 1994.

....represents the lowest common denominator of available programming models. It can be implemented with a variety of methods, including pointer copying on shared memory machines and message passing on distributed systems; hardware implementations are also available in experimental architectures [6]. By focusing on a single communication concept, portable applications are easier to construct and maintain. The library is in use on a wide range of multicomputers, shared memory multiprocessors, and networked workstations, including the Cray T3D, Intel Paragon, IBM SP2, and SGI Power Challenge. ....

D. Maskit and S. Taylor, "A Message-Driven Programming System for FineGrain Multicomputers," Software - Practice and Experience, 24:953--980, 1994.

No context found.

MASKIT94. Daniel Maskit. A message-driven programming system for fine-grain multicomputers. MSc Thesis CalTech (February 1994).

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