This paper proposes an object oriented model scheduling for parallel computing in media MultiProcessors System on Chip(MPSoC).Firstly,the Coarse Grain Data Flow Graph(CGDFG) parallelprogramming model is used in this ...
详细信息
This paper proposes an object oriented model scheduling for parallel computing in media MultiProcessors System on Chip(MPSoC).Firstly,the Coarse Grain Data Flow Graph(CGDFG) parallelprogramming model is used in this ***,this approach has the feature of unified abstraction for software objects implementing in processor and hardware objects implementing in ASICs,easy for mapping CGDFG programming on *** approach cuts down the kernel overhead and reduces the code size *** principle of the oriented object model,the method of scheduling,and how to map a parallelprogramming through CGDFG to the MPSoC are analyzed in this *** approach also compares the code size and execution cycles with conventional control flow scheduling,and presents respective management overhead for one application in me-dia-SoC.
We report on performance tests of pair interaction lattice gas automata in two and three dimensions coded in FORTRAN and C. The programs have been run on ALLIANT/FX-80, ALLIANT/FX-2800, CONVEX C2, CRAY-YMP, NEC/SX3, a...
详细信息
We report on performance tests of pair interaction lattice gas automata in two and three dimensions coded in FORTRAN and C. The programs have been run on ALLIANT/FX-80, ALLIANT/FX-2800, CONVEX C2, CRAY-YMP, NEC/SX3, and SUN/IPC. The maximum update rates are 200 million site updates per second on the NEC/SX3 (FORTRAN), 117 (2D version) and 29 (3D version) on the CRAY-YMP (C). As a byproduct we give results for the performance of integer arithmetic and bit operations. Usually the C-programs were somewhat faster than the FORTRAN-programs except on the NEC/SX3 where the C-compiler was not able to vectorize the main loops.
暂无评论