M.I. Frank and M.K. Vernon. A Hybrid Shared Memory/Message Passing Parallel Machine. In Proceeedings of the International Conference on Parallel Processing. p. 232-237, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....from the shared memory bus. The BBN Butterfly [15] provided direct hardware support for block transfer. The Platinum system [6] implemented DSM using this mechanism. The Cray T3D also supports block transfer, but does not maintain cache coherence for data sent in this manner. Frank and Vernon [8] proposed a set of message passing primitives for DSM systems that allowed a possibly stale page to be sent between processes using their message passing primitives. Klaiber and Levy [12] 0 20 40 60 80 100 Jacobi FFT Execution Time Ratio DSM ONLY IMPLICIT DSM MPI EXPLICIT DSM MPI ....

M.I. Frank and M.K. Vernon. A Hybrid Shared Memory/Message Passing Parallel Machine. In Proceeedings of the International Conference on Parallel Processing. p. 232-237, 1993.

....We model shared memory systems by having a collection of ReadWrite objects. The abstract state of each such object consists of an integer or machine word. Other kinds of systems can be modeled as well, including hybrids of the message passing and shared memory paradigms, such as those described in [8, 10, 11, 15, 19]. The proposed research is to explore the theory and practice of this objectoriented model of concurrent systems, and to investigate how results in the general model apply to the subdomains of message passing and shared memory. The first step of this research is to elaborate the model. The second ....

Matthew I. Frank and Mary K. Vernon. A hybrid shared memory /message passing parallel machine. In C. Y. R. Chen and P. Bruce Berra, editors, ICPP '93, volume 1: Architecture, pages 232--6, St. Charles, IL, USA, 16--20 August 1993. Pennsylvania State University, CRC Press. 11

.... processor before it is actually used) and poststore (which is sender initiated and sends the data as soon as it is produced to potential consumer processors) Further, there have been several recent proposals to provide message passing style communication primitives in a shared memory machine [14, 17, 19]. We refer to a specific combination of memory model and coherence protocol, together with some explicit communication primitives as a memory system in this paper. All such latency reducing and tolerating mechanisms have simply one goal, namely, to make the parallel machine appear as close as ....

Matthew I. Frank and Mary K. Vernon. A hybrid Shared Memory/Message Passing parallel machine. In Proceedings of the 1993 International Conference on Parallel Processing, August 1993.

....pipelined prefetching, and by using forwarding, particularly with Forwarding Write. 5.2 Data Forwarding The Dash Deliver instruction [9] is an example of a sender initiated forwarding mechanism. Other forwarding mechanisms use directories to keep track of processors requesting forwarded data [5, 9, 38, 39, 40] or perform broadcast forwarding [41] Many of the directory based schemes implement receiver initiated forwarding because they rely on processors requesting data to add themselves to the directory. The action of data forwarding is similar, but not identical, to the update action in an update ....

Frank, M. I., and Vernon, M. K., "A Hybrid Shared Memory / Message Passing Parallel Machine", Proceedings of ICPP, 1993, pp. I-232-236.

....used. One potential disadvantage of software initiated schemes is the overhead introduced by prefetching and (1) related instructions; this overhead can have a significant impact on performance [1, 2] Although many mechanisms have been proposed that could be used to implement data forwarding [9, 10, 11, 12, 13, 14], most of these mechanisms have been introduced or evaluated for other purposes, such as reducing the latency of synchronization operations. Little performance information exists regarding the usefulness of these mechanisms for optimizing sharing accesses. Accordingly, the design and evaluation of ....

Frank, M. I., and Vernon, M. K., "A Hybrid Shared Memory / Message Passing Parallel Machine", Proceedings of the International Conference on Parallel Processing, 1993, pp. I-232-236.

.... processor before it is actually used) and poststore (which is sender initiated and sends the data as soon as it is produced to potential consumer processors) Further, there have been several recent proposals to provide message passing style communication primitives in a shared memory machine [16, 20, 22]. We refer to a specific combination of memory model and coherence protocol, together with some explicit communication primitives as a memory system in this paper. All such latency reducing and tolerating mechanisms have simply one goal, namely, to make the parallel machine appear as close as ....

Matthew I. Frank and Mary K. Vernon. A hybrid Shared Memory/Message Passing parallel machine. In Proceedings of the 1993 International Conference on Parallel Processing, August 1993.

....We model shared memory systems by having a collection of ReadWrite objects. The abstract state of each such object consists of an integer or machine word. Other kinds of systems can also be modeled, including hybrids of the message passing and shared memory paradigms such as those described in [42, 50, 54, 68, 75]. The proposed research is to investigate the use of compositional techniques in constructing transparently fault tolerant distributed object systems and proving properties of objects in such a system. In particular, the following points will be addressed: ffl Compositional reasoning techniques ....

Matthew I. Frank and Mary K. Vernon. A hybrid shared memory/message passing parallel machine. In C. Y. R. Chen and P. Bruce Berra, editors, ICPP '93, volume 1: Architecture, pages 2326, St. Charles, IL, USA, 1620 August 1993. Pennsylvania State University, CRC Press.

....1 SW [7, 26] A second advantage of software protocols is that they lead to a natural unification of message passing with shared memory. A shared memory program can send messages when a user or compiler wants to bypass shared memory overheads and can determine in advance how data will be used [5, 10]. A third advantage is the opportunity to develop language specific memory models, such as Loosely Coherent Memory [14] Thus far we have argued the benefits of software coherence protocols without specifying whether the software runs in the system (in kernel mode) or at user level. In fact, the ....

Matthew I. Frank and Mary K. Vernon. A Hybrid Shared Memory/Message Passing Parallel Machine. In Proceedings of the 1993 International Conference on Parallel Processing (Vol. I Architecture), pages 232--236, August 1993.

....systems such as Alewife [ABC 95] FLASH [KOH 94] and Start NG [CAA 95] support both models, but do not allow the user to combine their features arbitrarily. For example, none of these systems allows data to be sent in a message without being renamed at the receiver. Frank and Vernon [FV93] propose extensions for a shared memory machine that send data without renaming. However, their message passing and shared memory primitives interact in a fixed way and do not allow arbitrary coherence policies. The Tempest mechanisms may be implemented in hardware, but Tempest protocols are by ....

Matthew I. Frank and Mary K. Vernon. A hybrid shared memory/message passing parallel machine. In Proceedings of the 1993 International Conference on Parallel Processing (Vol. I Architecture), pages 232--236, August 1993.

.... has begun focusing on supporting multiple paradigms in a single computer by integrating shared memory and message passing and allowing a compiler to select the model appropriate for a program or data structure [24] Frank and Vernon proposed messagepassing mechanisms for shared memory systems [12]. The MIT Alewife system handles some cache coherence events in software and allows shared memory programs to send explicit messages. Their preliminary results show that some run time operations, for example, task creation, are more efficiently implemented with explicit messages than shared memory ....

Matthew I. Frank and Mary K. Vernon. A Hybrid Shared Memory/ Message Passing Parallel Machine. In Proceedings of the 1993 International Conference on Parallel Processing (Vol. I Architecture), pages 232--236, August 1993.

....from the shared memory bus. The BBN Butterfly [15] provided direct hardware support for block transfer. The Platinum system [6] implemented DSM using this mechanism. The Cray T3D also supports block transfer, but does not maintain cache coherence for data sent in this manner. Frank and Vernon [8] proposed a set of message passing primitives for DSM systems that allowed a possibly stale page to be sent between processes using their message passing primitives. Klaiber and Levy [12] examined the inherent costs associated with shared memory and message passing and argued that future ....

M.I. Frank and M.K. Vernon. A Hybrid Shared Memory/Message Passing Parallel Machine. In Proceeedings of the International Conference on Parallel Processing. p. 232237, 1993.

....the memory latency of sharing accesses since data can be sent to destination processors caches immediately after being produced; data prefetching can only approach this level of (3) latency tolerance. Although many mechanisms have been proposed that could be used to implement data forwarding [14, 15, 16, 17, 18, 19], most of these mechanisms have been introduced or evaluated for other purposes, such as reducing the latency of synchronization operations. Little performance information exists regarding the usefulness of these mechanisms for optimizing interprocessor data communication. Accordingly, the design ....

Frank, M. I., and Vernon, M. K. A hybrid shared memory / message passing parallel machine. Proc. 1993 International Conference on Parallel Processing. 1993, pp. I-232-236.

....that comes with the shared memory paradigm. So the key here is to provide the flexibility in the memory system to derive the performance advantage of message passing without sacrificing the programming advantage of shared memory. This issue has been the focus of several recent research endeavors [7, 11, 16, 19, 26]. The Beehive memory system provides this flexibility along two orthogonal dimensions. Along one dimension we exploit knowledge of data sharing patterns in the applications to decide whom to send the data to (i.e the future sharer set ) This builds on our earlier work [24] wherein we proposed a ....

M. I. Frank and M. K. Vernon. A Hybrid Shared Memory/Message Passing Parallel Machine. In Proceedings of the 1993 International Conference on Parallel Processing, 1993.

....the memory latency of sharing accesses since data can be sent to destination processors caches immediately after being produced; data prefetching can only approach this (3) level of latency tolerance. Although many mechanisms have been proposed that could be used to implement data forwarding [14, 15, 16, 17, 18, 19], most of these mechanisms have been introduced or evaluated for other purposes, such as reducing the latency of synchronization operations. Little performance information exists regarding the usefulness of these mechanisms for optimizing interprocessor data communication. Accordingly, the design ....

Frank, M. I., and Vernon, M. K. A hybrid shared memory / message passing parallel machine. Proc. 1993 International Conference on Parallel Processing. 1993, pp. I-232-236.

....The system must also implement the processor side support described in the previous section. Since these mechanisms comprise the bulk of the complexity of a software extended system, it is important to note that the benefits of these mechanisms extend far beyond the implementation of shared memory [42, 29]: integrating message passing and shared memory promises to be an active topic of research in the coming years. Alternative scalable approaches to implementing shared memory proposed in [61, 69, 77] use hardware mechanisms that allocate directory pointers dynamically. These schemes do not require ....

Matthew J. Frank and Mary K. Vernon. A Hybrid Shared Memory/Message Passing Parallel Machine. In Proceedings of the 22nd International Conference on Parallel Processing. IEEE, August 1993.

No context found.

M. I. Frank and M. K. Vernon. A hybrid shared memory / message passing parallel machine. In Proceedings of the 1993 International Conference on Parallel Processing, pages I:232--237, August 1993.

No context found.

M. I. Frank and M. K. Vernon. A hybrid shared memory / message passing parallel machine. In Proceedings of the 1993 International Conference on Parallel Processing, pages I:232-- 237, August 1993.

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