Deployment_Guide.md

N-ways to GPU programming Deployment Guide

The N-Ways to GPU Programming Bootcamp covers the basics of GPU programming and provides an overview of different methods for porting scientific application to GPUs using NVIDIA® CUDA®, OpenACC, standard languages, OpenMP offloading, and/or CuPy and Numba. Throughout the bootcamp, attendees with learn how to analyze GPU-enabled applications using NVIDIA Nsight™ Systems and participate in hands-on activities to apply these learned skills to real-world problems.

Deploying the materials

Prerequisites

To run this tutorial, you will need a machine with NVIDIA GPU.

• Install the latest Docker or Singularity.

• Install the latest NVIDIA Nsight™ Systems.

• The base containers required for the lab may require users to create an NGC account and generate an API key (https://docs.nvidia.com/ngc/ngc-catalog-user-guide/index.html#registering-activating-ngc-account)

The material is also tested to be working with NVIDIA V100 and T4 GPUs, please contact us if you require assistance in deploying the content.

Tested environment

These materials was tested with both Docker and Singularity on both NVIDIA A100 and H100 GPUs in an x86-64 platform installed with a driver version of 550.90.07.

Deploying with container

These materials can be deployed with either Docker or Singularity container, refer to the respective sections for the instructions.

Docker Container

To build a docker container, specify the dockerfile name using -f flag: sudo docker build -f <dockerfile name> -t <imagename>:<tagnumber> .

For instance:

• To build the docker container, for N-Ways to GPU Programming-Python, follow the below steps:

  1. sudo docker build -f nways_Dockerfile_python -t openhackathons:nways_python .
  2. sudo docker run --rm -it --gpus=all -p 8888:8888 openhackathons:nways_python
  3. To access the labs, run: jupyter-lab --ip 0.0.0.0 --port 8888 --no-browser --allow-root
  4. Now, open the jupyter lab in browser: http://localhost:8888, and start working on the lab by clicking on the _start_nways.ipynb notebook

• To build the docker container, for N-Ways to GPU Programming-C-Fortran, follow the below steps:

  1. sudo docker build -f nways_Dockerfile -t openhackathons:nways_CFortran .
  2. sudo docker run --rm -it --gpus=all -p 8888:8888 openhackathons:nways_CFortran
  3. To access the labs, run: jupyter-lab --ip 0.0.0.0 --port 8888 --no-browser --allow-root
  4. Now, open the jupyter lab in browser: http://localhost:8888, and start working on the lab by clicking on the _start_nways.ipynb notebook

Please note, if you are to run both contents, you would need to change the ports to access them seperately.

Singularity Container

• To build the singularity container, for N-Ways to GPU Programming-Python, follow the below steps:

  1. singularity build --fakeroot nways_python.simg nways_Singularity_python
  2. singularity run nways_python.simg cp -rT /labs ~/labs
  3. singularity run --nv nways_python.simg jupyter-lab --notebook-dir=~/labs
  4. Now, open the jupyter lab in browser: http://localhost:8888, and start working on the lab by clicking on the _start_nways.ipynb notebook

• To build the singularity container, for N-Ways to GPU Programming-C-Fortran, follow the below steps:

  1. singularity build --fakeroot nways_CFortran.simg nways_Singularity
  2. singularity run nways_CFortran.simg cp -rT /labs ~/labs
  3. singularity run --nv nways_CFortran.simg jupyter-lab --notebook-dir=~/labs
  4. Now, open the jupyter lab in browser: http://localhost:8888, and start working on the lab by clicking on the _start_nways.ipynb notebook

Known issues

• Please go through the list of exisiting bugs/issues or file a new issue at Github.