A first instruction requiring that a data word should be read out from a data memory and be stored in a certain register in a register set, and then a second instruction requiring that two operands, respectively read out from the register and another register in the register set, should be added are pipeline-processed. In a high-speed mode in which an operation clock having a higher frequency is supplied, a data cache intervened between an instruction execution circuit and the data memory is controlled to supply a data word to a WB (write back) stage of the instruction execution circuit within two cycles with respect to an input address associated with the first instruction. In order to execute the second instruction, the data word is supplied from the WB stage to an EX (operation execution) stage of the instruction execution circuit. In a low-speed mode in which an operation clock having a lower frequency is supplied, the data cache is controlled to supply a data word to an MEM (memory access) stage of the instruction execution circuit within one cycle with respect to an input address associated with the first instruction. In order to execute the second instruction, the data word is bypassed from the MEM stage to the EX stage.