Advanced-Computer-Architectures

Course material for Advanced Computer Architectures course.

Getting started with the repo

Clone the repository

git clone https://github.com/PARCO-LAB/Advanced-Computer-Architectures.git

Follow the lessons and solve the exercises by keeping your solutions local.

Fork the repository and save the changes

If you want to save your solutions remotely, fork the main repository:

Clone the repository forked in your GitHub profile and solve the exercises. When you’re ready to submit your changes, stage and commit your changes as the following:

git add .
git commit -m "<comment to the changes>"
git push

In this way, your exercise implementations are saved in your personal GitHub profile.

Usage

In the exercise directory, create the build folder, launch the cmake command and compile the sources:

cd <lesson-folder>/<exercise-name>

# Only the first time
mkdir build
cd build
cmake ..

# Every time you have to recompile
make
./<executable_name>

Benchmarking

Add the following line to the main function in the source file:

std::cout << TM_host.duration() << ";" << TM_device.duration() << ";" << TM_host.duration() / TM_device.duration() << std::endl;

Then launch the benchmarking script in the build folder:

../../../utils/benchmark.sh <executable_name>

If you are using a Jetson board, make sure to set the power profile to max, typing:

sudo nvpmodel -m<n> # (The number may change, depending on the device, for TX2 <n>=0)
sudo jetson_clocks

Hardware details (only for NVIDIA Jetson)

To have a complete overview of the Jetson hardware status, type:

sudo jtop

If it isn't already installed, type:

sudo -H pip3 install -U jetson-stats
sudo systemctl restart jetson_stats.service