The results obtained by Calder et al. for the proposed architecture without the BTB, are not as good as that obtained for the BTB-based architecture. Calder et al. explain this by reasoning that the proposed architecture is not able to handle the indirect jumps. They substantiate this with the fact that C++ programs and eqntott have the most degradation with the new architecture and that these programs have a numerous indirect jumps. After profiling, better results are obtained for these programs. Calder et al. come up with a solution for this proposing a small BTB in conjunction with the other mechanisms. The BTB only stores indirect jumps and hence is very small which Calder et al. suppose would not increase the overall processor time.