Exploiting Large Ineffectual Instruction Sequences

A processor executes the full dynamic instruction stream in order to compute the final output of a program, yet we observe equivalent, smaller instruction streams that produce the same correct output. Based on this observation, we attempt to identify large, dynamically-contiguous regions of instructions that are ineffectual as a whole: they either contain no writes, writes that are never referenced, or writes that do not modify the value of a location. The architectural implication is that instruction fetch/execution can quickly bypass predicted-ineffectual regions, while another thread of control verifies that the implied branch predictions in the region are correct and that the region is truly ineffectual.