Conference Paper 1991 IEEE : Institute of Electrical and Electronics Engineers
Parallel processing scheme for a Fortran program on a multiprocessor system OSCAR
マルチプロセッサシステムOSCAR上でのFortranプログラムの並列処理スキーム
Hiroki Honda, Akiyoshi Mogi, Akio Ogura, Hironori Kasahara, Seinosuke Narita IEEE Pacific Rim Conference on Communications, Computers and Signal Processing Conference Proceedings
【抄録】A parallel processing scheme is proposed for a Fortran program on a multiprocessor system named OSCAR (Optimally SCheduled Advanced multiprocessoR). The scheme combines parallel processing of fine grain tasks (each of which consists of an assignment statement), macro-data-flow computation (which uses parallelism among coarse grain tasks), and the traditional loop concurrentization. A parallelizing compiler that accomplishes the proposed scheme has been implemented on OSCAR. OSCAR has been designed to effectively support task-scheduling, which has been an obstacle to the realization of efficient parallel processing. A performance evaluation of the scheme on OSCAR is also described.
【抄録】This paper describes a parallel processing scheme for the calculation of optical flow and the determination of camera motion parameters. The optical flow calculation includes the solution of linear equations with sparse coefficient matrix. Camera motion parameters are determined by the Jacobi method in case of translational camera-movement and by Gauss elimination in rotational camera-movement. These calculations are decomposed into tasks, which are assigned on an arbitrary number of processors by the optimal multiprocessor scheduling algorithms developed by authors. The performance evaluation of the parallel processing scheme for a multiprocessor system named OSCAR shows the scheme allows us real-time optical flow calculation and determination of the camera motion parameters.
PARALLEL PROCESSING FOR REAL-TIME CONTROL AND SIMULATION OF DCCS.(First author)
DCCの実時間制御とシミュレーションのための並列処理技術
Hironori Kasahara, S. Narita
【抄録】This paper concerns itself with the resource allocation and task scheduling problems for real-time control and simulation of distributed computer control systems. The problems are formulated and solved at two levels. The Interblock Problem deals with the allocation of computing resources to a set of interrelated blocks (program modules). A solution algorithm is presented to minimize the total processing time, with due consideration given to interprocessor communication times. The Intrablock Problem essentially is an optimum resource allocation and task scheduling problem in a parallel processing environment. Heuristic algorithms are combined with a rigorous solution algorithm to reduce the overall solution time. The proposed solution algorithms are applied to a few real-time control and simulation problems to demonstrate their usefulness in the design, implementation, and operation stages of DCCS.