Optimizing memory bandwidth in OpenVX graph execution on embedded many-core accelerators | IEEE Conference Publication | IEEE Xplore