https://hgpu.org/?p=7538
Effects of Compiler Optimizations in OpenMP to CUDA Translation