Reconfigurable chip advantage compared with GPGPU from the compiler perspective

Abstract

This presentation discusses how FPGA or coarse grained reconfigurable processor is superior to CPU/GPGPU from the view point of C compiler. Initially, we introduce the architectural characteristic of CPU, GPGPU and fine grained and coarse grained reconfigurable process with FSM+Datapath model. Then, we explain what kind of applications can be accelerated with "FPGA and C-based High Level Synthesis Tool" better than GPGPU according to the compiler techniques (freedom of compiler parallelization).