https://hgpu.org/?p=3130
CUDA-based AES parallelization with fine-tuned GPU memory utilization