Use Of Parallel Resources Research Articles

A novel batch-selection strategy for parallel global optimization

ABSTRACT Bayesian-based parallel global optimization methods have become increasingly popular for solving black-box functions using parallel resources. When the function exhibits multimodality, it is difficult to identify the global optimum. This article proposes a novel batch selection strategy for parallel global optimization called Sampling–Optimizing–Selecting (SOS) to address this problem. SOS, utilizing minimum energy design, gradient optimization and hypersphere clustering, can adaptively explore multiple local optimal regions of the objective function simultaneously and make full use of parallel resources to obtain the global optimum efficiently. Additionally, the SOS algorithm maintains the sparsity between sample points to a certain extent, which enhances the accuracy of the surrogate model and thus improves the robustness of the optimization effect. Several numerical simulations and optimization experiments involving array decoy jamming are presented to demonstrate the superiority of the proposed strategy.

Implementation and optimization of hybrid parallel strategy in HNET

For the self-developed three-dimensional whole-core High-fidelity NEutron Transport calculation program HNET, although the numerical acceleration algorithms can improve the computation performance of high-fidelity neutron transport in terms of algorithms and models, it still faces some critical issues such as long computing time and enormous memory requirement. The rapid development of high-performance clusters provides a foundation for the application of massively parallel computing. Most current MOC programs are based on a single-type variable to achieve efficient parallelism, and spatial domain decomposition methods are the most common parallel schemes. However, its parallelism is limited and cannot fully utilize the current state-of-the-art computer resources. To solve this problem, hybrid parallel strategies are implemented in HNET to further expand the parallel degree, improve the speed of computation, and reduce the memory requirement. A hybrid MPI/OpenMP method based on domain decomposition and characteristic rays is proposed for the method of characteristics (MOC). For domain decomposition, the simulation domain is divided into spatial subdomains, with each subdomain handled by different processes. On this basis, the characteristic ray parallelism is implemented taking advantage of the inherent parallelism of the characteristic rays. Meanwhile, the optimization of the hybrid parallel strategy further improves the computation speed by eliminating atomic operations and using private pointer arrays and other techniques. In addition, in the framework of generalized equivalence theory (GET) based two-level coarse mesh finite difference method (CMFD), there is also a certain time consumption in solving CMFD linear system. Hence, for CMFD, a hybrid MPI/OpenMP method based on domain decomposition and secondary domain decomposition can be used. Using the secondary domain decomposition method, each subdomain is divided into sub-subdomains, with each sub-subdomain handled by threads, which allows CMFD to utilize the resources of characteristic ray parallelism in MOC, and also increases the speed of CMFD computation. Numerical results show that for both steady-state and transient calculations, the hybrid MPI/OpenMP in HNET can further expand the parallelism and accelerate the computation. It can take full use of parallel resources and achieve large-scale parallelism.

Two-Step Predict and Correct Non-Intrusive Parametric Model Order Reduction for Changing Well Locations Using a Machine Learning Framework

The objective of this paper is to develop a two-step predict and correct non-intrusive Parametric Model Order Reduction (PMOR) methodology for the problem of changing well locations in an oil field that can eventually be used for well placement optimization to gain significant computational savings. In this work, we propose a two-step PMOR procedure, where, in the first step, a Proper Orthogonal Decomposition (POD)-based strategy that is non-intrusive to the simulator source code is introduced, as opposed to the convention of using POD as a simulator intrusive procedure. The non-intrusiveness of the proposed technique stems from formulating a novel Machine Learning (ML)-based framework used with POD. The features of the ML model (Random Forest was used here) are designed such that they take into consideration the temporal evolution of the state solutions and thereby avoid simulator access for the time dependency of the solutions. The proposed PMOR method is global, since a single reduced-order model can be used for all the well locations of interest in the reservoir. We address the major challenge of the explicit representation of the well location change as a parameter by introducing geometry-based features and flow diagnostics-inspired physics-based features. In the second step, an error correction model based on reduced model solutions is formulated to correct for discrepancies in the state solutions at well grid blocks expected from POD basis for new well locations. The error correction model proposed uses Artificial Neural Networks (ANNs) that consider the physics-based reduced model solutions as features, and is proved to reduce the error in QoI (Quantities of Interest), such as oil production rates and water cut, significantly. This workflow is applied to a simple homogeneous reservoir and a heterogeneous channelized reservoir using a section of SPE10 model that showed promising results in terms of model accuracy. Speed-ups of about 50×–100× were observed for different cases considered when running the test scenarios. The proposed workflow for Reduced-Order Modeling is “non-intrusive” and hence can increase its applicability to any simulator used. Additionally, the method is formulated such that all the simulation time steps are independent and hence can make use of parallel resources very efficiently and also avoid stability issues that can result from error accumulation over time steps.

Speculation and replication in temperature accelerated dynamics

Abstract

Mth: Codesigned Hardware/Software Support for Fine Grain Threads

Multi-core processors are ubiquitous in all market segments from embedded to high performance computing, but only few applications can efficiently utilize them. Existing parallel frameworks aim to support thread-level parallelism in applications, but the imposed overhead prevents their usage for small problem instances. This work presents Micro-threads ( Mth ) a hardware-software proposal focused on a shared thread management model enabling the use of parallel resources in applications that have small chunks of parallel code or small problem inputs by a combination of software and hardware: delegation of the resource control to the application, an improved mechanism to store and fill processor's context, and an efficient synchronization system. Four sample applications are used to test our proposal: HSL filter (trivially parallel), FFT Radix2 (recursive algorithm), LU decomposition (barrier every cycle) and Dantzig algorithm (graph based, matrix manipulation). The results encourage the use of Mth and could smooth the use of multiple cores for applications that currently can not take advantage of the proliferation of the available parallel resources in each chip.

Proactive task scheduling and stealing in master-slave based load balancing for parallel contingency analysis

With increasing emphasis on analyzing N−k contingencies, use of parallel resources has become imperative. Parallelization imposes the requirement of load-balancing for achieving high resource usage efficiency. Conventional static allocation based scheduling techniques fail to achieve load balancing. To address this limitation, master-slave scheduling (MSS) has been used; however, in MSS, after task completion, slave processors wait for the next task to arrive leading to idle-wait. In the case of contention at master, the idle-wait could become significant and degrade the performance of the MSS algorithm.We present a technique to combine the advantage of proactive task scheduling and stealing with the simplicity of MSS. We refer to it as PTMSS. In PTMSS, master proactively queues an extra task at the slave processor, such that on completion of a task, the next task is immediately started. Further, when master runs out of the tasks, it steals a queued task from one slave and allocates to another slave which has completed its tasks. Simulation experiments have been conducted on a large power system with 13,029 buses and thousands of contingencies have been analyzed. The results show that PTMSS performs better than conventional MSS and also offers significant computational gains over serial execution.

Parallel Dantzig–Wolfe decomposition of petroleum production allocation problems

This article discusses the optimization of a petroleum production allocation problem through a parallel Dantzig–Wolfe algorithm. Petroleum production allocation problems are problems in which the determination of optimal production rates, lift gas rates and well connections are the central decisions. The motivation for modelling and solving such optimization problems stems from the value that lies in an increased production rate and the current lack of integrated software that considers petroleum production systems as a whole. Through our computational study, which is based on realistic production data from the Troll West field, we show the increase in computational efficiency that a parallel Dantzig–Wolfe algorithm offers. In addition, we show that previously implemented standard parallel algorithms lead to an inefficient use of parallel resources. A more advanced parallel algorithm is therefore developed to improve efficiency, making it possible to scale the algorithm by adding more CPUs and thus approach a reasonable solution time for realistic-sized problems.

Estimation of service reliability and performance in grid environment

Service time for the tasks with a predefined deadline is an important parameter in the evaluation of a grid resource manager (GRMS). This parameter is affected directly by the failures of connection links and computational resources. In this paper, a new approach to estimate the reliability of a service and mathematical expectation of service time considering failure existence is proposed. In this approach unlike the current approaches which only consider permanent failures, both permanent and transient failures are considered. Also, in the previous works in order to increase the reliability and consequently to decrease the service time, the use of parallel resources was suggested. Nevertheless, due to the cost constraint the use of parallel resources is not practical. In this paper, the use of passive parallel systems to increase the reliability of grid services is suggested and the reliability and service time in the presence of passive parallel replica resources are analyzed. A passive parallel system has lower impact on reducing the service time, but it is more economical and consequently more applicable.

Optimistic parallel simulation of reliable multicast protocols

Parallel discrete-event simulation offers the promise of harnessing the computational power of multiple processors in order to reduce the time needed for simulation-based performance studies. In this paper, we investigate the use of optimistic parallel simulation techniques in simulating reliable multicast communication network protocols. Through empirical studies (using the TeD simulation programming language, the Georgia Tech time warp simulator, and a 12-processor SGI Challenge), we find that these parallelized simulations can run noticeably faster than a uniprocessor simulation and, in a number of cases, can make effective use of parallel resources. These results are somewhat surprising because reliable multicast protocols require considerable communication (and hence synchronization) among different network entities.

Session C3: AI architecture and applications I

The continued advances in processing capabilities through VLSI and the increasing interest for symbolic processing and declarative languages have resulted in a tremendous increase in the sophistication of AI processing hardware. To fully utilize this advanced capability, advances in the use of parallel resources must also be made. Today, it can be described by a research effort towards new machine architectures for more intelligence, more efficiency and more fault tolerance. All papers, submitted to the session “AI Architectures and Applications I, II” of Euromicro 87 deal more or less with these subjects. The programming language Prolog is being increasingly used in the field of artificial intelligence and knowledge engineering and has been chosen as the main programming language in the japanese fifth computer generation project. Therefore, most of the papers in both sessions deal with modelling, architecture and implementation aspects of Prolog machines and expert systems, because there is a low efficiency of conventional machines in execution of these programs.