Slide 16 of 30
Notes:
First, the lookahead PC used by Tango does not advance by one instruction every cycle, but instead it advances from one branch point to the next branch point. This is done by extending the branch target buffer with two additional fields. These fields are BTB indices that point to the branches that follow the current one on the taken and not-taken paths. Each cycle the lookahead PC is updated by following either one of these pointers depending on the prediction information of the branch.
