I made this presentation in the framework of a lecture about the use of GPU for scientific computing in my University.
These slides intends to show some interesting aspects of Thrust : mixing handwritten kernels with generic thrust code, How to harness multi GPU using Thrust, writting generic code for CPU and GPU,...
At the end I present some small usecase in the field of simple linear algebra and variational signal processing.
Code : https://github.com/gnthibault/Cuda_Thrust_Introduction