Deep Learning Language Modeling Workloads: Where Time Goes on Graphics Processors | IEEE Conference Publication | IEEE Xplore