Parallel Closest Pair of Points

🏛 University project for the "High Performance Computing" course by Sandro Luigi Fiore

Using the program 💽

Connecting to the cluster

You'll need a VPN to connect to Trento's VPN to access hpc2.unitn.it

1. Open the GlobalProtect VPN client by running gpclient (be sure to follow the setup instructions from the official repo)

2. Enter as portal address vpn.icts.unitn.it

3. The username is your unitn email without the "studenti" subdomain -e.g., mario.rossi@unitn.it-

4. The password is your unitn password

5. Now you should be able to successfully ping hpc2.unitn.it

Input format

A text file where the first row contains the number of points (unsigned int32). Each subsequent row contains two integers x and y (int32) in the format x<space>y, representing the coordinates of a point.

Example of a file with points [(1, 2), (-3, 89), (42, 137)]:

3
1 2
-3 89
42 137

Output format

Found the closest pair in 100 milliseconds.
The closest pair of points is:
(-3, 89) and (42, 137)
Their distance is:
65.79513

Running the program

Method 1: Github CI

In order to run this program on the cluster, just push here on Github on the main branch. The CI will:

• build the code

• copy it in the cluster

• start the job

• notify us on telegram once it is finished

Note that for this job to run successfully, the following secrets have to be present in the github repo:

• BOT_TOKEN: The token of the telegram bot used for notifying that the job completed
• TELEGRAM_CHAT_ID: The ID of the chat where the bot sends the updates
• UNITN_PASSWORD: Unitn email
• UNITN_USERNAME: Unitn password

Method 2: Manual

Getting the code to the cluster

1. Ensure you can ping the cluster (hpc2.unitn.it). Otherwise you may need to connect to the VPN.

2. Open a ssh connection to the cluster. You can do so by running ssh hpc2.unitn.it -l mario.rossi@unitn.it. Then note the location of your home directory on the cluster, given by the output of the pwd command (should be /home/mario.rossi).

3. Copy the code to the cluster by either

   1. Running on your machine scp -r -o "user=mario.rossi@unitn.it" ./path/to/the/code/on/your/machine hpc2.unitn.it:/home/mario.rossi/HPC_course
   2. Running on the cluster -in the target directory- git clone https://github.com/civts/parallel-closest-pair

4. Compile the code by running (on the cluster) module load mpich-3.2, then mpicc -o parallel_closest_points main.c -lm

Running the program

Start a run of the program (job) by running on the cluster `qsub

Dev environment setup 👨‍💻

Option 1: With Nix ❄️

With Nix and direnv you can get exactly our setup by running this script:

git clone https://github.com/civts/parallel-closest-pair #Downloads this repository
cd parallel-closest-pair
direnv allow #Fetches all the dependencies (and VSCodium)
codium . #Launches VSCodium

Then open the file you want to debug and press F5 to launch it.

Option 2: Manual 🔨

1. Install the following packets:

   • gcc (we used version 12.2.0)
   • glibc (we used version 2.35-163)
   • clang-format (we used version 11.1.0)

2. If you are using VSCode/VSCodium as editor, we recommend you have the following extensions installed:

   • vadimcn.vscode-lldb (we used version 1.7.4)
   • xaver.clang-format (we used version 1.9.0)
   • ms-vscode.cpptools (we used version 1.11.0)
   • franneck94.c-cpp-runner (we used version 3.0.0)

3. Clone this repo, open it in the editor

4. Open the file you want to debug and press F5 to launch it.

Troubleshooting

VS Code

If it does not find gtest/gtest.h, run ./scripts/test.sh, ensure that the test/build is shown in the file explorer on the side (otherwise modify files.exclude in settings.json to not exclude it). Then restart VS Code.