Nvidia unveils CUDA Tile to simplify GPU programming for AI developers
Programming model moves from managing thousands of low-level threads to working with high-level ‘tiles of data’
Programming model moves from managing thousands of low-level threads to working with high-level ‘tiles of data’