A memory cell mat is divided into a plurality of entries, and an arithmetic logic unit is arranged corresponding to each entry. Between the entries and the corresponding arithmetic logic units, arithmetic/logic operation is executed in bit-serial and entry-parallel manner. Where parallel operation is not very effective, data is transferred in entry-serial and bit-parallel manner to a group of processors provided at a lower portion of the memory mat. In this manner, a large amount of data can be processed at high speed regardless of the contents of operation or data bit width.