Continuing the theme of branch predictors, we saw how a G-Share predictor operates and then did an exercise to solidify a sense of how that works. We then took a look at a basic hybrid predictor, and talked about the engineering challenges of building an actual branch predictor. Although academic research in predictors may inspire actual designers, they are usually faced with such tight timing requirements that the final design looks much different. 

Reminder that the midterm is next Wednesday. 

Slides are here

Handout is here