CUDA H.264 decoding RTMP streams

Hi, I’m looking for a solution for hardware decoding H.264 RTMP. It looks like this is possible with the FFMPEG libs, however I’d like to find a solution using just the NVidia video decoder fed by an RTMP stream (the missing part). I need to decode over 100 640x360 streams in real-time, ideally in one computer using ~4 GPUs if possible. After decoding, the frames will be composited for display on multiple monitors- ideally decoding to textures which can then be quickly drawn to the frame buffers for display (nothing coming back across the bus).

Thanks for any tips!

John