Credit: Mark Harris - NVIDIA Corporation

Description: A simple kernel which adds two vectors (1D arrays) with the GPU. This is a good example showing off memory allocation and movement use the CUDA C runtime API, while using a very simple kernel function. There are many variations you can use on this example, including using the Thrust library to handle memory allocation and movement instead of cudaMalloc and cudaMemcpy explicitly.

Files: Exercise: exercise.cu - one version of a hands-on lab exercise Answer: answer.cu - the solution to the presented exercise

Compile: Using the NVIDIA C Compiler nvcc -o vectoradd answer.cu

Flow:

1. After having introduced the concepts of cudaMemcpy, cudaMalloc and block and threads, have the students modify the code in exercise.cu to make it compile and do the correct work on the GPU.

References: None