On latency in GPU throughput microarchitectures | IEEE Conference Publication | IEEE Xplore