Highly-efficient parallel convolution acceleration by using multiple GPUs | IEEE Conference Publication | IEEE Xplore