Processor and branch prediction method
AssigneeKabushiki Kaisha Toshiba
InventorsShigehiro Asano, Takashi Yoshikawa
Life of this patent
Estimating the term…
Loading patent…
AssigneeKabushiki Kaisha Toshiba
InventorsShigehiro Asano, Takashi Yoshikawa
Estimating the term…
A next address computing section contains a selector and is connected to an instruction cache. The instruction cache maintains a predecode result of a branch instruction or predefined settings for a field in this branch instruction. Based on this information maintained in the instruction cache, the selector determines whether the compiler performed a branch prediction about the branch instruction or could not perform that branch prediction. When the compiler could not perform the branch prediction, the selector selects an output from a conditional branch prediction device (saturation counter section). When the compiler performed the branch prediction, the selector selects a prediction result by the compiler for a prediction in Agree mode. These selection results are used for setting a value of a register holding the next address. Based on this next-address register value, an instruction is fetched from the cache then inserted into a pipeline.
A processor comprising:
A branch prediction method of predicting an address of a second instruction to be fetched into an instruction cache when a first instruction previously fetched into the cache is a branch instruction, the method comprising:
Dependencies are inferred from the claim text.