https://hgpu.org/?p=5675
Automatic Translation of CUDA to OpenCL and Comparison of Performance Optimizations on GPUs