摘要 |
PROBLEM TO BE SOLVED: To provide an improved implementation technique of a processing pipeline capable of minimizing disadvantages which lose a processing capability caused by a conventional processing architecture. SOLUTION: A processing method for commands in multiprocessor system comprises a step of transferring one or more of data blocks between a shared memory 106 and a plurality of parallel processors 102 including a local memory 104, a step of executing one or more of programs in the local memory 104 of one or more of processors 102 so that one or more of programs can be coded without depending on data caching in the processor 102, and a step of reserving three or less of commands from the local memory 104 prepared for command buffer of the processor 102 to a buffer. The command buffer of each processor 102 can be adjusted to practically and effectively execute a command when one or more of programs are coded without depending on the data caching in the processor 102. COPYRIGHT: (C)2006,JPO&NCIPI
|