https://hgpu.org/?p=5743
Exploring The Latency and Bandwidth Tolerance of CUDA Applications