Three-layer optimizations for fast GMM computations on GPU-like parallel processors | IEEE Conference Publication | IEEE Xplore