The artifact can be executed on any x86 machine with at least 30 GB of main memory and at least 40 GB of disk space. We strongly recommend running the artifact on a workstation with multi-cores and at least 128 GB memory. The artifact needs a Linux environment (preferably Ubuntu) and a compiler that supports the C++14 standard.
Use the following command to download the artifact:
git clone git@github.com:0-EricZhou-0/ece522-mp4.gitInstall the following dependencies:
sudo apt install flex bison tmux python3-pipBuild simulator (the output executable is named sim):
cd src
make clean
make -jEvery configuration file specifies the DNN model and the batch size to be used, as well as other system configuration parameters (such as GPU memory size, SSD Bandwidth, the baseline type, and so on).
To run a single experiment, directly find its corresponding config file and use. For example:
./sim "$relative_path_to_config_file"
# e.g., ./sim configs/example.config