Nowadays, the increasing core number benefits many workloads, but programming limitations to exploiting full performance still remain. A Data-Flow execution model is capable of taking advantage of the full parallelism offered by multicore systems. In such model, the execution can be decomposed in fine-grain threads named Data-Flow Threads (DF-Threads) so that each of them can execute only when their inputs are available. The execution overhead and power consumption is lowered thanks to the reduction of the data push-pull, as well as the burden of thread management.
|Titolo:||From COTSon to HLS: translating timing into an architecture|
|Citazione:||Giorgi, R., KHALILI MAYBODI, F., & Procaccini, M. (2018). From COTSon to HLS: translating timing into an architecture.|
|Appare nelle tipologie:||4.3 Poster|