https://hgpu.org/?p=2814
Automated Dynamic Analysis of CUDA Programs