Intel® VTune™ Amplifier XE and Intel® VTune™ Amplifier for Systems Help
In a shared-memory machine, instructions and data are stored in the same memory address space. However, for performance, they are cached separately. Large code working set, branch misprediction, including one caused by excessive use of virtual functions, can induce misses into L1I and so cause instruction starvation that badly influence application performance.
A significant number of CPU cycles is spent waiting for code to arrive into L1I. Review application code for the patterns causing instruction starvation and rearrange the code.