|System: Four-Way Superscalar||Running: The gcc Compiler (cc1)|
The first ten instructions executing below have direct dependencies: each instruction uses a result produced by its immediate predecessor. Since the first of the ten has reached the ROB head the commit rate is limited to one per cycle while fetch occurs at a faster rate. When the last of the ten commits the processor catches up by commiting other instructions it has completed. The PED plot clearly shows the dependency chain and the ROB plot shows the ROB first filling, then draining. Near the end of the fragment is a misprediction squash.