You’ll have to keep us apprised of your CUDA programming effort since I’ve only discovered last year the great value of GPU programming as an adjunct to mainline CPU coding. BTW I’ve been following FPGA programming as well as an adjunct to a main processor. Both Microsoft (Azure) and Amazon (AWS) have added to their Cloud environments auxiliary tools such as GPU and FPGA resources that can be dynamically instantiated. I’m wondering how either of these may impact processing of television signals in the future by offloading such work from a main processor.
I basically use TVheadend to record programs (as another DVR resource next to my Tablo) and MiniDLNA to stream the recorded programs (behaves very well with Roku and Kodi). So my Roku sees both the Tablo and RPi as TV sources.
TVheadend uses the 24 hour guide transmitted by the stations but has the option of using XMLTV to supply a two week guide accumulated from a variety of free and legal sources (something I intend to implement later).
BTW I’ve found the Tablo app for Kodi to be more responsive with better picture quality than the Roku. The Kodi box is an Amlogic S912 with hardware video acceleration and the Tablo image appears just a little more fluid and sharper here.