摘要 |
PROBLEM TO BE SOLVED: To solve the problem with a microprocessor with a cache memory that makes branch predictions wherein the mounting of a high-capacity branch target buffer for a processor of a multistaged pipeline structure to efficiently execute branch instructions will result in an increase in a circuit scale. SOLUTION: If the access entry of a cache memory 102 varies for an instruction fetch from a CPU 11, a branch target buffer 104 is accessed in advance and a cache access is made to a predicted branch address, whereby a cache is hit when a branch instruction is executed afterward to make a branch prediction. This makes it possible to read instruction data from the cache memory 102 while the branch instruction is being executed for the instruction fetch to the predicted branch address and to either leave no instruction data on the branch address held in the branch target buffer 104 or reduce the amount of data held, so as to reduce the circuit scale. COPYRIGHT: (C)2005,JPO&NCIPI
|