The peLIFOLite interface class (written in C++) is provided below. The details of this algorithm can be found here (see Section 5). The code for set sampling is designed specifically for sixteen set samples per policy in a bank-pair. For a different number of samples, the code should be modified appropriately. From my experience, this is where the major tuning is needed. Too many set samples usually degrade the performance of workloads that are not peLIFOLite-friendly.

The README file details the expected interface with the LLC and shows an example.

README
peLIFOLite.h
peLIFOLite.cc (Note on tuning here)

Updated: 06-MAR-2010