J. Bruner, H. Cheong, A. Veidenbaum, and P.-C. Yew, "Chief: a parallel simulation environment for parallel systems," Center for Supercomputing Research and Development, University of Illinois, Urbana, IL, CSRD Report 1050, November 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....time) as its elapsed time is the closest to the lower bound. 5PERFORMANCE STUDY For comparison, we simulated our scheme and earlier techniques using an event driven simulator, called PARSIM, which was developed as a part of the CHIEF simulation environment at CSRD in University of Illinois [4], 5] 5.1 Simulation Model PARSIM simulates a shared memory architecture with M processors and M memory modules interconnected by an Omega network. It can handle the system size up to 256. No cache is considered in this simulation study. Any unsuccessful, blocked request in the network is held ....

....a hot spot and removes undesirable synchronization traffic caused by spinning (in the ADS scheme) or linked list redirection (in the MCS scheme) In addition, the performance gaps between our scheme and the other schemes tend to increase gradually as N grows. Although the simulator used, PARSIM [4], can handle at most 256 processors, we expect the gaps to keep increasing consistently when N goes beyond 256 due to the fact that hot spot contention gets more serious as N becomes large. The effectiveness of our scheme is evident when compared with the lower bound shown by the dashed curve in ....

J.D. Bruner, H. Cheong, A. Veidenbaum, and P.C. Yew, "Chief: A Parallel Simulation Environment for Parallel Systems," Proc. Fifth Int'l Parallel Processing Symp., pp. 568-575, Apr. 1991.

....fashion. Driven by the objectives outlined above, the tool sets should guide the users in developing a correct and efficient parallel program, and tune their programs to achieve the best results. A simulator will enable the user to analyze his her application on a Unix based workstation. See also [4]. It will include: a source level interpreter, a machine assembly level simulator, a performance visualization tools. 7 and some debugging facilities. Our approach will also concentrate on providing structured methods for integrating both domain specific and architecture specific ....

J. Bruner et al. Chief: A parallel simulation environment for parallel systems. Technical report, Center for Supercomputing Research and Development, University of Illinois at Urbana-Champaign, November 1990.

....The network and memory system are modeled at the register transfer level. All algorithms in the experiments are implemented in a hypothetical machine language interpreted by the processor model. The simulators were built using the tools developed as part of CSRD s CHIEF simulation environment [3, 11]. Two performance measures of primary interest presented here are . The improvement of synchronization time . the improvement of network throughput for non synchronization traffic. 18 Packets are composed of some number of 32 bit words. All packets have 64 bits of address information. This ....

Bruner J., Cheong H., Veidenbaum A., and Yew P-C. Chief: A Parallel Simulation Environment for Parallel Systems. CSRD Rep. 1050, Center for Supercomputing Research and Development, Urbana, IL, November 1990.

....prior to building them. Researchers have also used simulation to evaluate new techniques for solving problems, in particular, cache coherency. Much of the development in parallel computer simulation has targeted shared memory systems, which are not of interest to us. Such simulators include Chief [12], MINT [77] SPASM [68] PSIMUL [70] Talisman [3] Tango [24] and CHAPTER 2. LITERATURE SURVEY 9 Threads [53] Unfortunately, far less attention has been given to simulation of distributedmemory (message passing) systems. The known projects that have addressed this problem are the topic of ....

John D. Bruner, Hoichi Cheong, Alexander Veidenbaum, and Pen-Chung Yew. Chief: A parallel simulation environment for parallel systems. In Proceedings of the 5th International Parallel Processing Symposium, pages 568--575, Anaheim, CA, 1991. IEEE Computer Society Press.

No context found.

J. Bruner, H. Cheong, A. Veidenbaum, and P.-C. Yew, "Chief: a parallel simulation environment for parallel systems," Center for Supercomputing Research and Development, University of Illinois, Urbana, IL, CSRD Report 1050, November 1990.

....and analyzing the results of simulations. In Chief, simulated systems are described by connected sets of hierarchically defined components. Each component is defined in CARL (Computer Architecture Research Language) a superset of C designed for the specification of architectural simulations [17]. CARL promotes modularity and reusability, encourages a mix and match philosophy of using existing components in new simulations to explore architectural variations, and enables the cooperative development of large system simulations. In addition, libraries of predefined CARL components ....

....lazy, and direct cancellation) in the global virtual time calculation method, and in the manipulation of the state of logical processes. ParSim is an efficient parallel implementation of the traditional event driven simulation algorithm specifically targeted toward parallel system simulations [17]. In ParSim, the target system is represented by a network of components interconnected via nets. A simulation repeatedly evaluates components with new input values, and subsequently propagates generated output values through nets to the appropriate downstream components. Special attention has ....

J. Bruner, H. Cheong, A. Veidenbaum, and P.-C. Yew, "Chief: A Parallel Simulation Environment for Parallel Systems," in Proceedings of the 5th International Parallel Processing Symposium, pp. 568--575, April 1991.

....of available parallelism with the least possible overhead. One way to simulate synchronous systems efficiently, is to parallelize the traditional synchronous event driven algorithm [15] Parsim is one example of such a parallel implementation. Parsim is a parallel, synchronous event driven method [4]. In Parsim the studied system is modeled as a set of components each of which represents part of the physical system. A component has zero or more inputs, from where it receives new values, and zero or more outputs where it sends new values. A component may also have internal state which provides ....

J. Bruner, H. Cheong, A. Veidenbaum, and P.-C. Yew. Chief: A Parallel Simulation Environment for Parallel Systems. Proceedings of the 5th International Parallel Processing Symposium, pages 568--575, April, 1991.

....simulation (EDS) EDS in EPG sim utilizes the ETG techniques described in Section 3.3 to form execution driven event generators. These event generators are coupled with parallel discrete event simulators to implement EDS. The overall structure of an EPG sim EDS is shown in Figure 3. CHIEF [BCVY90, Brun90] parallel discrete event simulators are used to construct system models and to perform simulation. The CHIEF system is portable, allowing simulations to be executed on a variety of uniprocessor and parallel host machines. A CHIEF simulator consists of a simulation model and a kernel library that ....

Bruner, J., Cheong, H., Veidenbaum, A., and Yew, P.-C., "CHIEF: A Parallel Simulation Environment for Parallel Systems", Center for Supercomputing Research and Development, University of Illinois at Urbana-Champaign, CSRD Report No. 1050, November 1990.

....the different approaches is the method used to determine which events are safe to process. Interesting synchronous parallel simulation methods include the use of directional distances [7] the three phase algorithm [1] the bounded lag algorithm [6] Breathing Time Buckets [9] and ParSim [3]. A more detailed discussion on synchronous parallel simulation methods can be found in [5] Synchronous parallel simulation methods introduce two kinds of overheads into a simulation. First, they use multiple global synchronization operations in order to coordinate the processors executing the ....

J. Bruner, H. Cheong, A. Veidenbaum, and P.-C. Yew. Chief: A Parallel Simulation Environment for Parallel Systems. Proceedings of the 5th IPPS, pages 568--575, April, 1991.

....simulation (EDS) EDS in EPG sim utilizes the ETG techniques described in Section 3.3 to form execution driven event generators. These event generators are coupled with parallel discrete event simulators to implement EDS. The overall structure of an EPG sim EDS is shown in Figure 1. CHIEF [16] parallel discrete event simulators are used to construct system models and to perform simulation. The CHIEF system is portable, allowing simulations to be executed on a variety of uniprocessor and parallel host machines. A CHIEF simulator consists of a simulation model and a kernel library that ....

Bruner, J., Cheong, H., Veidenbaum, A., and Yew, P.-C., "CHIEF: A Parallel Simulation Environment for Parallel Systems", Center for Supercomputing Research and Development, University of Illinois at Urbana-Champaign, CSRD Report No. 1050, November 1990.

.... of which were proposed by Chandy and Sherman, and Breathing Time Buckets a hybrid method proposed by Steinman [37, 38] However, the most obvious and commonly used parallel synchronous simulation algorithm is the straightforward parallel implementation of the traditional event driven algorithm [9, 35, 36]. Under this algorithm, simulation evolves in steps which are divided into two phases. During the first phase, we concurrently evaluate all events occurring at the earliest simulation time. Then, during the second phase, we simulate in parallel all the components whose inputs have changed by an ....

....in parallel all the components whose inputs have changed by an event evaluation. Barrier synchronizations between the two phases guarantee the correctness of the simulation. An efficient implementation of this parallel centralized time algorithm is Parsim, a simulation method developed at CSRD [8, 9]. Because of its efficiency, we use Parsim as the basis of our study of the proposed parallel simulation method. 2.2 Parallel Architectural Simulations As computer systems become increasingly more complex, studying their architectures and exploring design alternatives through simulation becomes ....

J. Bruner, H. Cheong, A. Veidenbaum, and P.-C. Yew. Chief: A Parallel Simulation Environment for Parallel Systems. In Proceedings of the 5th InternationalParallel ProcessingSymposium, pages 568--575, 1991.

No context found.

J. Bruner, H. Cheong, A. Veidenbaum, and P.-C Yew. Chief: A parallel simulation environment for parallel systems. In Proceedings of the 5 th IPPS, pages 568--575, April 1991.

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