摘要 |
<P>PROBLEM TO BE SOLVED: To solve such a problem in a packet communication based method applied to a data processing device performing large scale calculation by using a large capacity of a CPU, that granularity of transfer or overhead is large, and for reducing the granularity of transfer and overhead it requires costs for modifying the CPU and/or an OS. <P>SOLUTION: A data processing device comprises: an initiator IP module; a request transfer circuit; a response transfer circuit; and two or more clusters including a self-cluster numbered register and the like. When there is access to a specific address in response to a request from the initiator IP module, the device is configured to give information necessary for transfer to another cluster and determine whether a transfer destination is its own cluster or the other cluster on the basis of the information to transfer data. For performing access from its own cluster to the other cluster, the device is configured to set a cluster number and an address of the transfer destination before accessing the other cluster. <P>COPYRIGHT: (C)2013,JPO&INPIT |