الفهرس 
conclusionsOnly 14 pages are availabe for public view
 |  | from | 142 |  |  |
| | | from | 142 | | |
Abstract
This thesis gives an overview of the object allocation approaches in a multiprocessor network concentrating on the heuristic approach. The target of this thesis is to introduce a contribution in the field of heuristic allocation.
The thesis proposes three different general-purpose heuristic algorithms. The three algorithms have the capability of dealing with any precedence graph and any processor network confoguration. The proposed algorithms cluster and allocate simultaneously while considering processor load balance.
The first two algorithms have a common goal which is minimizing the communication cost regardless of execution time. The third algoritms aims at minimizing the parallel execution time of the problem graph. It has a look-ahead feature which gives better performance. It is known that the third algorithm is more suitable for heterpgeneous networks.
The proposed algorithms have been applied to different size problem graphs and allocated to various multiprocessor architectures. Their performance was evaluted using simulation based on shortest execution time. Load balance was also heuristic algorithms regarding their performance.
All simulations are implemented by AweSim and Visual Slam software on a pentium 11 computer.