Global History Predictor with Index Selection in SimpleScalar

An implementation of gselect branch predictor for the SimpleScalar simulator.

Features

A cost effective (in terms of footprint) global branch predictor that competes with its counterpart gshare. The benchmark analysis is automated by the run.py script under Debian Linux machines. Please refer to the documents found in docs/ for more information. Some setup is required to run the simulator.

How To

The library modules are defined in the requirements.txt. To get started, setup your environment:

$ python3 -m venv proj_env
$ source ./proj_env/bin/activate
$ pip3 install -r ./requirements.txt

NOTE: If your global installation of python3 does not have the module virtualenv installed. Look up how to install python3 modules for your Debian Linux based machine. Now that you have your environment setup, you can start simulating by:

$ ./run.py

Benchmarks

The benchmarks that closely followed McFarling's paper that was available for the SPEC2000 benchmarks was the following:

• li
• gcc
• tomcatv
• fpppp

NOTE: fpppp took too long to run to show in our demo and tomcatv did not work unfortunately. These two benchmarks were excluded in our anaylsis. If more time was given for the project. Then can look in to including these benchmarks.

Reference

• WRL Technical Note TN-36:Combining Branch Predictors, Scott McFarling, 1993
• SimpleScalar Simulator

License

Please refer to the MIT license attached to this repository. We do not own SimpleScalar simulator. It was used only for the purpose for academic research and to further our understanding.