NVIDIA and US National Labs will open source the results. When I was finishing up my time at ORNL, my colleagues were working to add code to push our matrix operations out to CUDA cores in the big ...