Does CUDA or OpenCL acceleration have any meaningful advantage over the other when it comes to exporting video using current GPUs?