| 摘要 |
PROBLEM TO BE SOLVED: To reduce the arithmetic cost of a parallel programming language, specially, the total number (work) of operations which is executed by a program and the longest sequence length (depth) of dependency from both the viewpoints of time and space. SOLUTION: The sequential order of tasks is determined for processing, first. Then, priority is assigned to possible tasks according to the sequential order. Tasks more than all parallel processing elements which are usable are selected out of all the possible tasks having the top priority. The selected tasks are divided into groups as many as the usable parallel processing elements. The parallel processing elements execute the above tasks. |
