Section: Highlights of the Year
Highlights of the Year
We designed a dataflow transformation which always recovers all the FIFO in our dataflow model (DPN) after a loop tiling , , a program transformation widely used in automatic parallelization. This is an important enabling transformation which reinforces DPN as an intermediate representation in the CASH HLS project.
We obtained new results on the comparison between different forms of synchronisation on futures, bringing a better understanding on the impact dataflow synchronisation and future typing on program synchronisation.