Proposal for expanding FPGA offloading targets for environment adaptive software

Abstract

We have proposed environment-adaptive software that automatically converts code written by programmers for regular CPUs according to the deployment environment, enabling high-performance operation. This paper focuses on automatic offloading to FPGAs according to the computation type of processing, such as image processing and signal processing. The existing application to be offloaded is semantically analyzed using pattern matching and an abstract syntax tree to determine whether the computation type has an OpenCL that can be replaced. If OpenCL is found, the application is replaced with that OpenCL and performance improvements are confirmed. The effectiveness of the proposed method for automatic offloading is confirmed by measuring processing time using an actual Intel Stratix FPGA.

IEICE Technical Report, NS2025-187, Dec. 2025, Copyright(C)2025 IEICE