https://hgpu.org/?p=18344
Analyzing Memory Accesses for Performance and Correctness of Parallel Programs