OmpSs and FPGAs

Speaker: Carlos Alvarez (BSC)
Date: Thursday, 3 December 2015, 15:30-17:30
Session: FPGAs
Talk type: Project talk (30 min)

Abstract: In this talk two projects that deal with OmpSs and FPGAs are going to be explained. The first project aims at using OmpSs to simplify the programming of FPGAs as program accelerators. In this sense FPGAs are used as heterogeneous accelerators in the same way as GPUs using Mercurium compiler and Nanos++ runtime to take care of all the small details that are cumbersome to programmers. The second project uses the FPGA directly as an accelerator for Nanos++ runtime and in fact can be used to accelerate any dataflow task-based programming model. It aims to demonstrate the usefulness of hardware support to alleviate the overheads of runtime dependence management.