A large code working set size or a high degree of
branch misprediction can induce instruction delivery stalls at the front-end,
such as misses in the L1I. Such stalls are called Instruction Starvation. This
metric is the ratio of cycles generated when no instruction was issued by the
front-end to all cycles.
A significant number of CPU cycles is spent waiting for code to be
delivered due to L1I misses or other problems. Look for ways to reduce the code
working set, branch misprediction, and the use of virtual functions.