This NVIDIA skill helps you add a first native, single-GPU cuDF path to a query engine, or finish its first adapter, in libcudf C++ or cuDF Java. It covers buffer ownership, worker readiness and comparing results against the CPU path, and it lists any proof-of-concept inputs that are missing. It does not cover dataframe analysis, production integration or multi-GPU execution.
Use cases
GPU query acceleration
cuDF
Database engine development
CUDA development
Alternatives
doca-dmaNVIDIA skill for DOCA DMA programming: memcpy tasks, buffer sizing, and permissions.nvidia
tao-run-on-kubernetesNVIDIA TAO skill for running TAO container jobs on Kubernetes with GPU scheduling.nvidia
cdf-auto-labelingHelps set up and debug NVIDIA CDF Auto-Labeling pipelines.nvidia
jetson-video-setupNVIDIA skill for installing, repairing, and verifying native video codec tooling on Jetson.nvidia
nemo-mbridge-multi-node-slurmNVIDIA skill for turning single-node scripts into multi-node Slurm jobs and debugging them.nvidia
nemo-relay-get-startedNVIDIA skill for newcomers trying NeMo Relay with the simplest supported quick start.nvidia
