yfxiee / perceptron-learning-cache-reuse-predictor

A cache reuse and bypass predictor for LLC using the perceptron learning algorithm. Based on Teran et al., MICRO 2016

This is a C++ based implementation of two cache reuse predictors:

1. Perceptron-learning based Reuse Prediction

2. Sampling-dead Block Prediction

Master branch contains the Perceptron learning predictor and branch sdbp contains the SDBP code. SDBP code is still WIP.

The traces to run the program can be found at: http://faculty.cse.tamu.edu/djimenez/614/traces.tar PS: This is not an enterprise server. Please be judicious while downloading. Also, the files can be taken down anytime.

To run a single trace, run:

./run_single.sh <location-to-trace-file>

To run all traces, run:

./run_traces.sh <location-to-traces-directory>

To generate a bar-graph comparing the geometric mean speed-up w.r.t LRU, run:

python calc_gmean.py

Similarly, run the following for Arithmetic Mean of MPKI values per trace:

python calc_amean.py

The files to check out are: replacement_state.*