Example GPU kernel implementations

This repository contains a implementations of some common GPU kernels in CUDA C++.

Installation

Compile the kernels using

mkdir build
make

Then run the kernels, for instance using

./build/scan

This have only been tested on nvcc 12.9 on an L4 (compute capability 8.9).

Profiling

The kernels can be profiled using Nsight Compute or Nsight Systems:

# compute
make ncu

# systems
make nsys