https://hgpu.org/?p=17094
Comparing Programmer Productivity in OpenACC and CUDA: an Empirical Investigation