Using Runahead Execution to Hide Memory Latency in High Level Synthesis | IEEE Conference Publication | IEEE Xplore