• 전기전자학회논문지
• /
• 제2권2호
• /
• pp.247-254
• /
• 1998

This paper presents a performance study of two LU decomposition algorithms on two massively parallel SIMD machines: the 16K processor MasPar MP-1 and the 4K processor MasPar MP-2. The paper presents experimental results and an analysis of the algorithms to explain the results. While the blocked and the nonblocked algorithms for LU decomposition have been studied individually by others, we compare the two algorithms and identify the tradeoffs between them. Our analysis of the blocked algorithm shows how the block size affects the interprocessor communication cost and the memory read/write overhead. The analysis in this paper is useful to determine an optimum block size for the blocked algorithm.

• Journal of international Conference on Electrical Machines and Systems
• /
• 제3권3호
• /
• pp.282-288
• /
• 2014

A large capacity power conversion system constructed by using two or more existing power converters has a lot of flexibility in how the power converters are used. However, at the same time, it has a problem of cross current flows between power converters. The cross current must be suppressed by controlling the system while miniaturizing the combination reactor. This paper focuses on two current control methods of a power conversion system constructed by using two power converters connected in parallel supplying the same power. In order to elucidate the control performance of cross current, each control method which are aimed at controlling cross current and not directly controlling it are examined in simulations.

• 한국경영과학회지
• /
• 제14권2호
• /
• pp.47-47
• /
• 1989

This paper considers the parallel machines scheduling problem characterized as a multi-objective combinatorial problem. As this problem belongs to the NP-complete problem, genetic algorithms are applied instead of the traditional analytical approach. The purpose of this study is to show how the problem can be effectively solved by using genetic algorithms with a permutation approach. First, a permutation representation which can effectively represent the chromosome is introduced for this problem . Next, a schedule builder which employs the combination of scheduling theories and a simple heuristic approach is suggested. Finally, through the computer experiments of genetic algorithm to test problems, we show that the niche formation method does not contribute to getting better solutions and that the PMX crossover operator is the best among the selected four recombination operators at least for our problem in terms of both the performance of the solution and the operational convenience.

• 대한산업공학회지
• /
• 제25권2호
• /
• pp.254-265
• /
• 1999

We consider two stage hybrid flowshop scheduling problem when there are two non-identical parallel machines at the first stage, and only one machine at the second stage. Several well-known sequence-first allocate-second heuristics are considered first. We then propose an allocate-first sequence-second heuristic to find minimum makespan schedule. The effectiveness of the proposed heuristic algorithm in finding a minimum makespan schedule is empirically evaluated by comparing with easily computable lower bound. The proposed heuristic algorithm as well as the existing heuristics are evaluated by simulation in four cases which have different processing time distribution, and it is found that the proposed algorithm is more effective in every case.

• 대한산업공학회지
• /
• 제24권3호
• /
• pp.431-446
• /
• 1998

In this paper, we consider the problem of scheduling n jobs with sequence-dependent processing times on a set of parallel-identical machines. The processing time of each job consists of a pure processing time and a sequence-dependent setup time. The objective is to maximize the total remaining machine available time which can be used for other tasks. For the problem, we first propose a dynamic programming(DP) algorithm for sequencing jobs processed on a single machine. The algorithm is then extended to handle jobs on parallel-identical machines. Finally, we developed an improved version of the algorithm which generates optimal solutions using much smaller amount of memory space and computing time. Computational results are provided to illustrate the performance of the DP algorithms.

• Journal of Information Processing Systems
• /
• 제8권2호
• /
• pp.331-346
• /
• 2012

Task-based programming is becoming the state-of-the-art method of choice for extracting the desired performance from multi-core chips. It expresses a program in terms of lightweight logical tasks rather than heavyweight threads. Intel Threading Building Blocks (TBB) is a task-based parallel programming paradigm for multi-core processors. The performance gain of this paradigm depends to a great extent on the efficiency of its parallel constructs. The parallel overheads incurred by parallel constructs determine the ability for creating large-scale parallel programs, especially in the case of fine-grain parallelism. This paper presents a study of TBB parallelization overheads. For this purpose, a TBB micro-benchmarks suite called TBBench has been developed. We use TBBench to evaluate the parallelization overheads of TBB on different multi-core machines and different compilers. We report in detail in this paper on the relative overheads and analyze the running results.

• 경영과학
• /
• 제24권2호
• /
• pp.33-42
• /
• 2007

In this paper we consider an alternate m machine scheduling problem in which each job having at most two eligible machines should be assigned with the objective of makespan minimization. For this problem. we propose a $O(m2^m)$ time rounding algorithm with performance ratio at most 1.5. For a little general problem where each job can be processed in at most three machines, we prove that a polynomial time algorithm does not exist with performance ratio less than 1.5.

• 대한산업공학회지
• /
• 제30권2호
• /
• pp.107-119
• /
• 2004

A study on the following parallel machine problem is addressed in this research. An order is completed only when a given number of processes (cycle) are repeated. Anew cycle is possible only upon the completion of the previous cycle. Orders are classified into job group according to product feature. For a machine to switch to a different job group from the currently processing one a major setup is required while a minor setup time is inserted in between two jobs of the same job group. The objective of the study is to find a schedule that minimizes total weighted tardiness. An initial solution is obtained by the RATCS(Restricted Apparent Tardiness Cost with Setup) rule, and a Tabu search is applied to improve the solution. Numerical examples are also presented.

• Journal of international Conference on Electrical Machines and Systems
• /
• 제2권2호
• /
• pp.216-224
• /
• 2013

This paper describes dual inverter drive for a fractional-slot concentrated winding permanent magnet synchronous machine (PMSM). PMSMs are widely used in many applications from small servo motors to few megawatts generators thanks to its high efficiency and torque density. Especially, fractional-slot concentrated winding PMSM is very popular in the applications where wide operation range is required because it shows very wide constant power speed ratios. High speed operation, however, requires lots of negative daxis current for reducing back-EMF regardless of output torque. Field weakening current does not contribute to the torque generation in surface mounted PMSM case and causes inverter and copper loss. To reduce the losses from field weakening current, this paper proposes PMSM with split stator and parallel dual inverter drive. Proposed parallel dual inverter drive reduces back-EMF and enables efficient drive at high speed and light load situation. Control strategy of proposed dual inverter system is established through loss analysis and simulation. Proposed concept is verified with practical experiment.

• 대한안전경영과학회지
• /
• 제25권4호
• /
• pp.131-140
• /
• 2023

This paper proposes an algorithm for the Unrelated Parallel Machine Scheduling Problem(UPMSP) without setup times, aiming to minimize total tardiness. As an NP-hard problem, the UPMSP is hard to get an optimal solution. Consequently, practical scenarios are solved by relying on operator's experiences or simple heuristic approaches. The proposed algorithm has adapted two methods: a policy network method, based on Transformer to compute the correlation between individual jobs and machines, and another method to train the network with a reinforcement learning algorithm based on the REINFORCE with Baseline algorithm. The proposed algorithm was evaluated on randomly generated problems and the results were compared with those obtained using CPLEX, as well as three scheduling algorithms. This paper confirms that the proposed algorithm outperforms the comparison algorithms, as evidenced by the test results.