Build Jar artifact of cuDF

Build the docker image

Prerequisite

1. Docker should be installed.
2. nvidia-docker should be installed.

Build the docker image

In the root path of cuDF repo, run below command to build the docker image.

docker build -f java/ci/Dockerfile.rocky --build-arg CUDA_VERSION=12.9.1 -t cudf-build:12.9.1-devel-rocky8 .

The following CUDA versions are supported w/ CUDA Enhanced Compatibility:

• CUDA 12.2+

Change the --build-arg CUDA_VERSION to what you need. You can replace the tag "cudf-build:12.9.1-devel-rocky8" with another name you like.

Start the docker then build

Start the docker

Run below command to start a docker container with GPU.

nvidia-docker run -it cudf-build:12.9.1-devel-rocky8 bash

Download the cuDF source code

You can download the cuDF repo in the docker container or you can mount it into the container. Here I choose to download again in the container.

git clone --recursive https://github.com/rapidsai/cudf.git -b main

Build cuDF jar with devtoolset

cd cudf
export WORKSPACE=`pwd`
source java/ci/env.sh
${sclCMD} "java/ci/build-in-docker.sh"

The output

You can find the cuDF jar in java/target/ like cudf-26.06.0-SNAPSHOT-cuda12.jar.