@Wendell et al.
I watched your recent visit to the AMD campus in Texas. Was there any mention of where they're going with GPU compute? Secondary markets and systems were mentioned. I'm delving into Deep Learning at work and academia, and I'm hoping they come through with something similar to NVidia Digits and NVidia's other similar suites.
At this level, hardware is only half the battle. Robust secondary suites/software is the other. Yes, I can program; no, I do not want to develop a custom framework (test one - sure!). I want to help solve my employer's problems!
Are there any rumblings of running CUDA developed solutions on AMD GPUs (OpenCL?)? Here is something I dug up from 2016 and this 2015 WCCF (I know) article. It looks like the direction might be full migration from CUDA to OpenCL; unfortunately, I do not see an ecosystem that gives me a "warm fuzzy".
If anyone can direct me to information on this topic, I am eager to learn. I'm helping set direction where I work and doing some research. I'd love to give those listening more than one viable (robust, enterprise ready) option.
I hope I'm wrong. Please educate!
Thanks, c0d1st