Parallel Simulation

This work concerns a general technique to enrich parallel version of stochastic simulators for biological systems with tools for online statistical analysis of the results. In particular, within the FastFlow parallel programming framework, we describe the methodology and the implementation of a parallel Monte Carlo simulation infrastructure extended with user-defined on-line data filtering and mining functions. The simulator and the on-line analysis were validated on large multi-core platforms and representative proof-of-concept biological systems.

Parallel simulations executing on a closely coupled network of processors require synchronization of logical processes across all processors. This synchronization overhead requires coordination of host processors and additional message traffic in the host network. We introduce specialized hardware to offload all parallel simulation synchronization overhead from host processors and the host network in a closely coupled network. In this hardware design, critical synchronization information is disseminated in the form of reductions (binary, associative operations) performed on state vectors of values provided by the logical processes. Our hardware description is based on the completed design of a four-node prototype, designed and built at the University of Virginia.

It has been established elsewhere that hardware to support parallel discrete event simulations (PDES) is desirable. We describe the steps leading to the implementation of a hardware-based framework to support PDES. We begin with an exploration of the criteria necessary to make such a framework both practical and useful, concluding that maintenance of sequential consistency is sufficient, while "observable" sequential consistency is more desirable but difficult to attain. We derive a functional design based on these criteria, and from that derive a prototype design. Also, we establish the utility of our design, showing that computation of critical global values, such as global virtual time, can be done in times two orders of magnitude or better than typical event times in discrete event simulations.

Distributed discrete event simulation techniques aim at an acceleration of the execution of a self-contained simulation model by the spatial decomposition of that model and the concurrent simulation of the submodels by so called logical processes (LPs), each executing on a dedicated node of a (closed) multiprocessor system. The dedication of parallel simulators to specific platforms and their adaptation to the respective hardware and software intrinsics has widely preventend an industrial and/or commercial success of such high performance simulation methods. In this work we propose the building of logical prdcess simulators that use the World Wide Web as an execution platform. We have developed and implemented a Java-based simulation engine along the conservative Chandy/Misra/Bryant (CMB) synchronization protocol, allowing for a platform independent, scalable, and performance efficient distributed simulation model execution. Our simulation engine is open and general purpose in the sense that it can be reused in various simulation domains by simply "plugging in" different simulation models, and simulation codes that, once written, can be executed anywhere (on the Internet). The potential performance improvement of Web-based distributed simulation is investigated in a sensitivity analysis conducted on a hypothetical simulation model executed in a Javaenabled workstation LAN. Even for very small simulation models, a speedup of about 3.5 could be attained on a 4 processor (heterogeneous) host LAN.

Lookahead is a key issue in distributed discrete event simulation. It becomes explicit in conservative simulation algorithms, where the two major approaches are the asynchronous null-message (CMB) algorithms and the synchronous window algorithms (CTW). In this paper we demonstrate how a hybrid algorithm can maximize the lookahead capabilities of a model by lookahead accumulation. Furthermore, per processor aggregation of the logical processes allows for tuning of the granularity. A qualitative performance analysis shows that in case of no hop-models our algorithm outperforms the traditional conservative algorithms. This is due to reduced synchronization overhead caused by longer independent computation cycles, that are generated by the lookahead accumulation across the shortest lookahead path.

We investigated the benefit of exploiting the symmetries of graphs for partitioning. We represent the model to be simulated by a weighted graph. Graph symmetries are studied in the theory of permutation groups and can be calculated in polynomial time with the nauty algorithm . We designed an algorithm to extract useful symmetries from the automorphism group, which can be used to create partitions derived from the graph's structure. Our approach is focused on composite graphs, for which identical subgraphs reoccur in the graph. If these identical subgraphs can be mapped onto each other by symmetries, the subgraphs are replaced by equivalent multivertices, resulting in a 'natural' aggregation of vertices. This approach is applied to parallel simulation of a detailed IP-switch with a conservative synchronous algorithm. The experimental results show that even for good partitions, global and temporal load imbalances are inevitable.

Lookahead is a key issue in distributed discrete event simulation. It becomes explicit in conservative simulation algorithms, where the two major approaches are the asynchronous null-message (CMB) algorithms and the synchronous window algorithms (CTW). In this paper we demonstrate a CMB- based algorithm that maximises the lookahead capabilities of a model, by lookahead accumulation. Furthermore, per processor aggregation of the logical processes allows for tuning of the granularity. A qualitative performance analysis shows that in case of no hop -models our algorithm outperforms the traditional conservative algorithms CMB & CTW. This is due to reduced synchronisation overhead caused by longer independent computation cycles, generated by the lookahead accumulation accros the shortest lookahead path.

Developed many prototype softwares (jointly with other researchers and students). Examples: LOSS: Scalable Parallel simulator for plasma and beams physics using Semi-Lagrangian method (OpenMP+MPI). SPIN: Parallel simulator for high energy physics using Particle In Cell method (MPI). OBIWAN: Adaptive simulator for plasma and beams physics using Semi-Lagrangian method (OpenMP).

of Met Office, Exeter, were prepared as part of their official duties as employees of the UK Government. They are published with the permission of the Controller of Her Majesty's Stationary Office and the Queen's Printer for Scotland. Satellite data are used to quantify and examine the bias in the outgoing long-wave (LW) radiation over North Africa during May-July simulated by a range of climate models and the Met Office global numerical weather prediction (NWP) model. Simulations from an ensemble-mean of multiple climate models overestimate outgoing clear-sky long-wave radiation (LWc) by more than 20 W m -2 relative to observations from Clouds and the Earth's Radiant Energy System (CERES) for May-July 2000 over parts of the west Sahara, and by 9 W m -2 for the North Africa region (20 • W-30 • E, 10-40 • N). Experiments with the atmosphere-only version of the High-resolution Hadley Centre Global Environment Model (HiGEM), suggest that including mineral dust radiative effects removes this bias. Furthermore, only by reducing surface temperature and emissivity by unrealistic amounts is it possible to explain the magnitude of the bias. Comparing simulations from the Met Office NWP model with satellite observations from Geostationary Earth Radiation Budget (GERB) instruments suggests that the model overestimates the LW by 20-40 W m -2 during North African summer. The bias declines over the period 2003-2008, although this is likely to relate to improvements in the model and inhomogeneity in the satellite time series. The bias in LWc coincides with high aerosol dust loading estimated from the Ozone Monitoring Instrument (OMI), including during the GERBILS field campaign (18-28 June 2007) where model overestimates in LWc greater than 20 W m -2 and OMI-estimated aerosol optical depth (AOD) greater than 0.8 are concurrent around 20 • N, 0-20 • W. A model-minus-GERB LW bias of around 30 W m -2 coincides with high AOD during the period 18-21 June 2007, although differences in cloud cover also impact the model-GERB differences.

The Geostationary Earth Radiation Budget Intercomparison of Longwave and Shortwave radiation (GERBILS) was an observational field experiment over North Africa during June 2007. The campaign involved 10 flights by the FAAM BAe‐146 research aircraft over southwestern parts of the Sahara Desert and coastal stretches of the Atlantic Ocean. Objectives of the GERBILS campaign included characterisation of mineral dust geographic distribution and physical and optical properties, assessment of the impact upon radiation, validation of satellite remote sensing retrievals, and validation of numerical weather prediction model forecasts of aerosol optical depths (AODs) and size distributions. We provide the motivation behind GERBILS and the experimental design and report the progress made in each of the objectives. We show that mineral dust in the region is relatively non‐absorbing (mean single scattering albedo at 550 nm of 0.97) owing to the relatively small fraction of iron oxides present (1–3...

This paper outlines the implementation of a distributed billiard ball simulation in C++ using the MPI message passing interface. The implementation is heavily modelled after the discussions and algorithms found in [Lubachevsky 1991] and [MacKenzie and Tropper 1996]. [Lubachevsky 1991] outlines an efficient but non-distributed simple rigid disc simulator that assumes no friction, spin, et cetera, while [MacKenzie and Tropper 1996] discusses a distributed version using shared-memory. The implementation presented in this paper is slightly more complex in that it is distributed with no shared memory.

Your article is protected by copyright and all rights are held exclusively by King Fahd University of Petroleum and Minerals. This e-offprint is for personal use only and shall not be self-archived in electronic repositories. If you wish to self-archive your work, please use the accepted author's version for posting to your own website or your institution's repository. You may further deposit the accepted author's version on a funder's repository at a funder's request, provided it is not made publicly available until 12 months after publication.

The SystemC TLM-2.0 standard is widely used in modern electronic system level design for better interoperability and higher simulation speed. However, TLM-2.0 has been identified as an obstacle for parallel SystemC simulation due to the disappearance of channels. Without a containment construct, simulation threads are permitted to directly access data of other modules and that makes it difficult to synchronize such accesses as required by the SystemC execution semantics. In this paper, we propose a compile time approach to statically analyze potential conflicts among threads in SystemC TLM-2.0 loosely-and approximately-timed models. We introduce a new Socket Call Path technique which provides the compiler with socket binding information for precise static analysis. We also propose an algorithm to analyze entangled variable pairs. Experimental results show that our approach is able to support automatically safe parallel simulation of SystemC models with TLM-2.0 Blocking Transport Interface, Direct Memory Interface and Non-blocking Transport Interface, resulting in impressive simulation speeds.

We report the long-time nonlinear dynamical evolution of ultracold atomic gases in the P-band of an optical lattice. A Bose-Einstein condensate (BEC) is fast and efficiently loaded into the Pband at zero quasi-momentum with a non-adiabatic shortcut method. For the first one and half milliseconds, these momentum states undergo oscillations due to coherent superposition of different bands, which are followed by oscillations up to 60ms of a much longer period. Our analysis shows the dephasing from the nonlinear interaction is very conducive to the long-period oscillations induced by the variable force due to the harmonic confinement.

Abstract: We consider massively parallel discrete event simulations where the communication topology among the processing elements is a complex graph. In the case of regular topologies we review recent results on virtual time horizon management. First we analyze the computational scalability of the conservative massively parallel update scheme for discrete event simulations by using the analogy with a well-known surface growth model, then we show that a simple modification of the regular PE communication topology to a ...

Mechanisms for managing message buffers in Time Warp parallel simulations executing on cache-coherent shared-memory multiprocessors are studied. Two simple buffer management strategies called the sender pool and receiver pool mechanisms are examined with respect to their efficiency, and in particular, their interaction with multiprocessor cache-coherence protocols. Measurements of implementations on a Kendall Square Research KSR-2 machine using both synthetic workloads and benchmark applications demonstrate that sender pools offer significant performance advantages over receiver pools. However, it is also observed that both schemes, especially the sender pool mechanism, are prone to severe performance degradations due to poor locality of reference in large simulations using substantial amounts of message buffer memory. A third strategy called the partitioned buffer pool approach is proposed that exploits the advantages of sender pools, but exhibits much better locality. Measurements of this approach indicate that the partitioned pool mechanism yields substantially better performance than both the sender and receiver pool schemes for large-scale, small-granularity parallel simulation applications. The central conclusions from this study are: (1) buffer management strategies play an important role in determining the overall efficiency of multiprocessor-based parallel simulators, and (2) the partitioned buffer pool organization offers significantly better performance than the sender and receiver pool schemes. These studies demonstrate that poor performance may result if proper attention is not paid to realizing an efficient buffer management mechanism.

Many of today's military services and applications run on geographically distributed sites. Before these services and applications can be deployed in an actual network, they need to be tested and evaluated under realistic scenarios with many unpredictable factors. Existing experimental tools cannot meet the requirements for scale, accuracy and timeliness. A network emulation framework called ROSENET is proposed that can meet these requirements by using a remote parallel simulation server to model a wide area network and a local network emulator to provide timely QoS predictions for testing real world applications. This paper discusses the challenges faced in applying ROSENET to defense applications through two case studies. In the first case study we apply synthetic traffic workloads over DARPA's NMS network topology for a large scale simulation and define a metric called remote emulation delay to evaluate and quantify ROSENET's performance. In the second case study we illustrate the procedures using ROSENET to evaluate a contemporary real-time distributed VoIP application, Skype, and present experimental results.

Recently, a considerable amount of effort in the U.S. Department of Defense has been devoted to defining the High Level Architecture (HLA) for distributed simulations. This paper describes the time management component of the HLA that defines the means by which individual simulations (called federates) advance through time. Time management includes synchronization mechanisms to ensure event ordering when this is needed. The principal challenge of the time management structure is to support interoperability among federates using different local time management mechanisms such as that used in DIS, conservative and optimistic mechanisms developed in the parallel simulation community, and real-time hardware-in-theloop simulations.

We show how random Boolean networks, a well-known model of the interaction among genes within a living cell, can be generalized by relaxing the Boolean approximation, therefore de®ning a class of continuous genetic networks, and by explicitly describing gene±chemical interactions. It is also shown that population growth can be modelled by describing in detail the interactions among a limited number of genes, while the eects of the remaining genes are represented by a collective variable. The model can describe speci®c metabolic systems, as shown by the case of a microorganism able to degrade aromatic compounds, and it can also be used to study the generic properties of genetic networks. For this purpose, theoretical and simulation results are given in the case where all the genes are arranged on a 2-D regular square topology, with connections among neighbouring sites, which represents a cellularautomata like model. Parallel simulation on a MIMD machine shows a good speed-up, thus con®rming that these models are amenable to ecient parallel implementation.

We study the second sound dipole mode in a partially Bose-Einstein condensed gas. This mode is excited by spatially separating and releasing the center-of-mass of the Bose-Einstein condensate (BEC) with respect to the thermal cloud, after which the equilibration is observed. The oscillation frequency and the damping rate of this mode is studied for different harmonic confinements and temperatures. The measured damping rates close to the collisionless regime are found to be in good agreement with Landau damping. For increasing hydrodynamicity of the cloud we observe an increase of the damping. Comment: Submitted to Phys. Rev. Lett. with 5 figures

It is risky to implement a large-scale wireless sensor system without predicating its behavior beforehand through simulation, an indispensable step toward the final system. Since in wireless simulation, over 90% simulation time is spent on the characterization of the MAC layer behaviors, abstraction at the MAC layer becomes the key to the scalability of wireless sensor network simulations. In this work, we employ an abstract simulation mode, established by online data collection, to describe the MAC behaviors efficiently. Observing that the abstract simulation mode becomes inaccurate when network traffic changes significantly, we adopt a hybrid approach. We switch between the abstract simulation mode and full simulation mode in response to the traffic changes. Statistics are collected online under full simulation mode to update the abstract simulation mode to be used. In evaluation, we demonstrate that our approach is 9 times faster than the original simulation with a loss of accuracy less than 10% in the end-to-end delay and 1% in the number of messages successfully transmitted across the network.

We present a dynamic verification tool MCC for Multicore Communication API applications -a new API for communication among cores. MCC systematically explores all relevant interleavings of an MCAPI application using a tailormade dynamic partial order reduction algorithm (DPOR). Our contributions are (i) a way to model the non-overtaking message matching relation underlying MCAPI calls with a high level algorithm to effect DPOR for MCAPI that controls the lower level details so that the intended executions happen at runtime; and (ii) a list of default safety properties that can be utilized in the process of verification. To our knowledge, this is the first push button model checker for MCAPI application writers that, at present, deals with an interesting subset of MCAPI calls. Our result is the demonstration that we can indeed develop a dynamic model checker for MCAPI that can directly control the non-deterministic behavior at runtime that is inherent in any implementation of the library without additional API modifications or additions..

The aim of this thesis was to develop a model of a receiver which could be quickly used to analyze radiated interference levels from data captured at the output of the antenna equipment used to measure radiated energy. Active circuits were mainly used in developing this model for the ease with which the design and simulations could be carried out in OrCAD.The guiding document for the thesis work has been CISPR 16-1-1 (International Special Committee on Radio Interference part 16-1-1) which specifies the characteristics and performance of equipment for the measurement of radiated interference.The testing of this receiver model was carried out as far as possible based on the test setupsrecommended in CISPR, and all results have been compared with the standards laid down for the model.Using the results, it will be shown that a CISPR EMI receiver can be modeled as a simple EMI receiver consisting of filtering, mixing and detecting circuits built according to specifications.

This paper presents the first step toward a computational steering environment based on CORBA. This environment, called EPSN 1 , allows the control, the data exploration and the data modification for numerical simulations involving an iterative process. In order to be as generic as possible, we introduce an abstract model of steerable simulations. This abstraction allows us to build steering clients independently of a given simulation. This model is described with an XML syntax and is used in the simulation by some source code annotations. EPSN takes advantage of the CORBA technology to design a communication infrastructure with portability, interoperability and network transparency. In addition, the in-progress parallel CORBA objects will give us a very attractive framework for extending the steering to parallel and distributed simulations.

The on-line visualization and the computational steering of parallel simulations come up against a serious coherence problem. Indeed, data distributed over parallel processes must be accessed carefully to ensure they are presented to the visualization system in a meaningful way. In this paper, we present a solution to the coherence problem for structured parallel simulations. We introduce a hierarchical task model that allows to better grasp the complexity of simulations, too often considered as "single-loop" applications. Thanks to this representation, we can schedule in parallel the request treatments on the simulation processes and satisfy the temporal coherence.

The on-line visualization and the computational steering of parallel simulations come up against a serious coherence problem. Indeed, data distributed over parallel processes must be accessed carefully to ensure they are presented to the visualization system in a meaningful way. In this paper, we present a solution to the coherence problem for structured parallel simulations. We introduce a hierarchical task model that allows to better grasp the complexity of simulations, too often considered as "single-loop" applications. Thanks to this representation, we can schedule in parallel the request treatments on the simulation processes and satisfy the temporal coherence.

We propose a bottom-up computer engineering approach to the Grand Challenge of understanding the Architecture of Brain and Mind as a viable complement to top-down modelling and alternative approaches informed by the skills and philosophies of other disciplines. Our approach starts from the observation that brains are built from spiking neurons and then progresses by looking for a systematic way to deploy spiking neurons as components from which useful information processing functions can be constructed, at all stages being informed (but not constrained) by the neural structures and microarchitectures observed by neuroscientists as playing a role in biological systems. In order to explore the behaviours of large-scale complex systems of spiking neuron components we require high-performance computing equipment, and we propose the construction of a machine specifically for this task -a massively parallel computer designed to be a universal spiking neural network simulation engine.

Ce travail concerne la résolution numérique de l'équation de Vlasov en utilisant une grille de l'espace des phases. Contrairement aux méthodes Particle In Cell (PIC) qui sont connues pour être bruitée, nous proposons une méthode basée sur la méthode semi-Lagrangienne pour discrétiser l'équation de Vlasov en deux dimensions de l'espace des phases. Ce type de méthode étant très coûteuse numériquement, on propose d'effectuer les simulations sur des machines parallèles. Pour cela, on présente une méthode de décomposition de domaine, chaque sous-domaine étant dédié à un processeur. Des conditions de type Hermite aux bords permettent alors d'obtenir une bonne approximation de la solution globale. Plusieurs résultats numériques montrent la précision et la bonne scalabilité de la méthode jusqu'à 64 processeurs.

Ce travail concerne la résolution numérique de l'équation de Vlasov en utilisant une grille de l'espace des phases. Contrairement aux méthodes Particle In Cell (PIC) qui sont connues pour être bruitée, nous proposons une méthode basée sur la méthode semi-Lagrangienne pour discrétiser l'équation de Vlasov en deux dimensions de l'espace des phases. Ce type de méthode étant très coûteuse numériquement, on propose d'effectuer les simulations sur des machines parallèles. Pour cela, on présente une méthode de décomposition de domaine, chaque sous-domaine étant dédié à un processeur. Des conditions de type Hermite aux bords permettent alors d'obtenir une bonne approximation de la solution globale. Plusieurs résultats numériques montrent la précision et la bonne scalabilité de la méthode jusqu'à 64 processeurs.

Parallel computers today are designed with larger number of processors than ever before, connected by large scale Interconnection Networks (INs). Communication is the key to achieving high performance on such machines, making the study of Interconnection Networks more important. Parallel simulations of Interconnection Networks present a unique problem characterized by fine-grained computation and a strong dependence among events. The absence of large lookaheads makes it unsuitable to use a conservative simulation. Using an optimistic Parallel Discrete Event Simulation (PDES) allows us to extract reasonable parallelism from this simulation. In this paper we present BigNetSim, an Interconnection Network simulator. We analyze its performance and present techniques related to enhancing performance and scaling it to a large number of processors on different artificial traffic patterns and real application logs. In spite of the overheads of a parallel optimistic simulation, we have achieved a breakeven point with sequential simulation at 4 processors and demonstrate perfect scaling to 128 processors. 1

Networks on Chip (NoC) come along with increased complexity from the implementation and management perspective. This leads to higher energy consumption and programming complexity of NoC architectures. This work introduces communication aware programming to address communication resource management and efficient programming of NoC architectures. A programming interface is introduced to express communication requirements at the language level. These requirements are evaluated by an operating system component, which configures the communication hardware accordingly. The proposed concept enables an intuitive use of NoC features like end-to-end connections and Direct Memory Access (DMA). The presented results show that communication aware programming can improve performance and energy consumption.

Different numerical optimization strategies were used to find an optimized parameter setting for the sheet metal forming process. A parameterization of a time-dependent blank-holder force was used to control the deep-drawing simulation. Besides the already wellestablished gradient and direct search algorithms and the response surface method the novel Kriging approach was used as an optimization strategy. Results for two analytical and two sheet metal forming test problems reveal that the new Kriging approach leads to a fast and stable convergence of the optimization process. Parallel simulation is perfectly supported by this method.

For parallel simulation of VLSI circuits on transistor level a sophisticated partitioning of the circuits into subcircuits is crucial. Each net connecting the subcircuits causes additional communication and computation effort. As the slave processors simulating the subcircuits advance synchronously in time, the computation effort for each subcircuit should be approximately the same. In this paper a new approach for partitioning VLSI circuits on transistor level yielding a low number of interconnects between the subcircuits and balanced subcircuit sizes is presented. Simulation of industrial circuits using this partitioning is up to 41% faster than with other known partitioning approaches for parallel analog simulation.

In this paper, an Adaptive Code NonUniform Modulation with Orthogonal Frequency Division Multiplexing (OFDM) System was evaluated in the underwater acoustic communication channel. This system was adopted from the Subcarrier arrangement of the OFDM standard of IEEE 802.11a, and improved with some modifications. There are three schemes in the proposed adaptive system. The first is the Hamming code with Binary phase-shift keying (BPSK) modulated, which is mapped for 12-bit subcarrier data. The second is the Bose-Chaudhuri-Hocquenghem (BCH) code with Quadrature phase-shift keying (QPSK) modulation which is for the other 12-bit subcarrier data. And the third is the Reed Solomon code with 16-Quadrature Amplitude Modulation (16- QAM) modulation that is mapped for 24-bit subcarrier data. The system modeling was evaluated in the shallow water acoustic environments. Bit error rate (BER) and signal-to-noise ratio (SNR) were analyzed to evaluate the proposed system performance. The evaluation r...

Simultaneous events are the events scheduled to occur at the same simulation time. This paper proposes a new event ordering mechanism for handling simultaneous events of DEVS models in distributed simulation. The DEVS formalism provides a formal framework for specifying discrete event models in a modular, hierarchical form. Thus, the formalism can ease the model verification and validation problems of distributed simulation. Also, the formalism separates models from underlying simulation algorithms. Hence, DEVS models can be simulated in both sequential and distributed environments without any modification. One important issue for such framework is to obtain the same results in both simulation environments. However, in distributed simulation of DEVS models, the processing order of simultaneous events may affect the simulation results. Thus, some ordering mechanism of events is required for well-defined simulation results. The proposed mechanism orders simultaneous events correctly with respect to their causal relationships in distributed DEVS simulation. Also, the mechanism guarantees the same ordering of simultaneous events in both sequential and distributed simulation environments. @ 1997 Elsevier Science B.V.

We discuss an approach for creating a federated network simulation that eases the burdens on the simulator user that typically arise from more traditional methods for defining spaceparallel simulations. Previous approaches have difficulties that arise from the need for global topology knowledge when forwarding simulated packets between the federates. In all but the simplest cases, proper packet forwarding decisions between federates requires routing tables of size O(mn) (m is the number of nodes modeled in a particular simulator instance, and n is the total number of network nodes in the entire topology) in order to determine how packets should be routed between federates. Further, the benefits of the well-known NIx-Vector routing approach cannot be fully achieved without global knowledge of the overall topology. We seek to overcome these difficulties by utilizing a topology partitioning methodology that uses Ghost Nodes. A ghost node is a simulator object in a federate that represents a simulated network node that is spatially assigned to some other federate, and thus that other federate is responsible for maintaining all state associated with the node. However, ghost nodes do retain topology connectivity information with other nodes, allowing all federate in a space-parallel simulation to obtain a global picture of the network topology. We show with experimental results that the memory overhead associated with the ghosts is minimal relative to the overall memory footprint of the simulation.

A distributed simulation is said to be repeatable if successive executions utilizing the same inputs produce exactly the same outputs. Repeatability is a highly desirable property, particularly for analytic simulation models. This paper discusses the question of repeatability in distributed simulations in general, and in the context of the High Level Architecture in particular. Specifically, allowing zero lookahead, a feature not supported in the baseline HLA, has important ramifications with respect to achieving repeatable executions. Extensions to the existing time management services in the HLA are proposed that allow (1) repeatable executions, (2) federate control over the ordering of simultaneous events, and (3) zero lookahead. The extensions proposed in this paper are currently under consideration by DMSO for possible future inclusion in the High Level Architecture time management services.

Multiobjective combinatorial optimization problems in various disciplines remain to be the focus of extensive research due to their inherent hard nature and difficulty of finding near-optimal solutions. Iterative heuristics like Tabu Search (TS) and Simulated Evolution (SimE) have successfully been employed to solve a range of such optimization problems [1]. These heuristics are able to obtain high quality solutions, but for most real-life large size problems they may have huge runtime requirements. Parallelization of these heuristics is one of the adopted practical approach to achieve the solutions within acceptable runtimes. In this paper we address a hard multiobjective optimization problem namely, VLSI cell placement [2] with three possibly conflicting objectives: interconnect wirelength, power dissipation, and timing performance. Two heuristics namely, parallel tabu search (TS) and parallel simulated evolution (SimE) are presented. Fuzzy rules are used to design a multiobjective aggregate cost function. The parallel TS implementation is a based on a synchronous candidate list partitioning model, whereas the parallel SimE implementation is based on random distribution of rows to processors [3, 4]. For comparison purposes, a parallel genetic algorithm (GA) based on the island model [5], and a parallel simulated annealing (SA) based on the asynchronous multiple-Markov chain [6] are also implemented. Results of experiments on ISCAS-85/89 benchmark circuits are presented, with solution quality and speedup used as metrics for the comparative/relative evaluation of the presented heuristics.

The key innovation in this paper is an open-source, high-performance iterative solver for high contrast, strongly anisotropic elliptic partial differential equations implemented within dune-pdelab. The iterative solver exploits a robust, scalable two-level additive Schwarz preconditioner, GenEO (Spillane et al. 2014). The development of this solver has been motivated by the need to overcome the limitations of commercially available modeling tools for solving structural analysis simulations in aerospace composite applications. Our software toolbox dune-composites encapsulates the mathematical complexities of the underlying packages within an efficient C++ framework, providing an application interface to our new high-performance solver. We illustrate its use on a range of industrially motivated examples, which should enable other scientists to build on and extend dune-composites and the GenEO preconditioner for use in their own applications. We demonstrate the scalability of the solver on more than 15,000 cores of the UK national supercomputer Archer, solving an aerospace composite problem with over 200 million degrees of freedom in a few minutes. This scale of computation brings composites problems that would otherwise be unthinkable into the feasible range. To demonstrate the wider applicability of the new solver, we also confirm the robustness and scalability of the solver on SPE10, a challenging benchmark in subsurface flow/reservoir simulation.