摘要 |
Disclosed is a parallel computation device for increasing speed of parallel processing by a multi-core processor. A computation unit region selection means (1aa, 1ba, ..., 1na) of each of a plurality of cores (1a, 1b, ..., 1n) sequentially selects an uncomputed computation unit region within a computation region (4). In addition, a computation means (1ab, 1bb, ..., 1nb) of each of the plurality of cores (1a, 1b, ..., 1n) performs computation within a selected computation unit region. Furthermore, the computation means (1ab, 1bb, ..., 1nb) writes computation results (2a, 2b, ...) to a storage device (2) that is accessible from each of the plurality of cores (1a, 1b, ..., 1n). A computation result transmission means (1nc) of a core (1n) acquires computation results within computation result transmission regions (4a, 4b, 4c, 4d) from the storage device (2) and performs transmission processing of the acquired computation results, at different time periods for the plurality of computation result transmission regions (4a, 4b, 4c, 4d). |