Light-weight-ML-DL-Branch-Predictor

Download the CBP-2016 Simulation Infrastructure from here

Unpack the kit:

$ tar -xzvf cbp2016.tar.gz 
$ cd cbp2016 

There should be five directories: sim, scripts, traces, bin, and results.

You will need to install the BOOST library and add a link to the library to the Makefile in the sim directory. See comments in the Makefile.

You will need to download the evaluation traces from here

Extract the traces to cbp2016/traces/ folder

You will have to place our uploaded predictor.h file in sim directory.

We have made all the changes to predictor.h file only. The sim directory contains the simulator:

$ cd sim 
$ make clean; make 

The scripts directory contains scripts that can help you run your predictor for all the traces. Check out the doit.sh file in the scripts directory.

$ cd ../scripts 
$ ./doit.sh

The outputs (MPkI) for various traces will be displayed on the screen along with mean MPkI as well as they are generated in detail in results folder.

We have uploaded our 2.25kB predictor.h file. You can use it to run the predictor with the above mentioned commands and obtain the results.

Normally, it will take around 15-20 mins to obtain the results.

You can as well get the original predictor.h file from the downloaded infrastructure.

The location is /submissions/DanielJimenez1/cbp8KB/predictor.h