Does OpenCV in C++ take full advantage of Multi-GPUs?

edit

I have code in C++ that uses OpenCV, and want to use some of the GPU/CUDA versions of some OpenCV functions to speedup the code. The function that takes up the most time is fastnlmeansdenoising

If I have multiple GPUs, would OpenCV use them at full capacity for fastnlmeansdenoising? Or am I better off just sticking to a single more powerful GPU?