Performance-Portable Autotuning of OpenCL Kernels for Convolutional Layers of Deep Neural Networks | IEEE Conference Publication | IEEE Xplore