GPU Networking Examples (NCCL + DPDK)

This repo contains runnable examples for:

• NCCL collectives (CUDA + NCCL + MPI)
• DPDK packet processing (EAL + RX/TX loops)

Note: These are Linux-first examples. NCCL examples require NVIDIA GPUs and CUDA. DPDK examples require Linux + hugepages + appropriate NIC access.

Build overview (CMake)

Top-level:

mkdir -p build && cd build
cmake ..
cmake --build . -j

NCCL examples

Requires:

• CUDA toolkit
• NCCL
• MPI (e.g., OpenMPI)

Run (single node):

mpirun -np 2 ./nccl_allreduce_mpi

DPDK examples

Requires:

• DPDK installed (and pkg-config able to find it)
• hugepages configured
• appropriate permissions

Run:

sudo ./dpdk_rx_burst_stats -l 0-1 -n 4

See each example's README for details.