Automating Change Propagation in Software Applications for Heterogeneous Computing
Simone Bauco, Guglielmo De Angelis, and Alessandro Pellegrini
Published in: 23rd IEEE International Conference on Software Architecture, ICSA 2026 - Companion
Abstract:
Developing applications for heterogeneous systems is challenging because maintaining semantic consistency between CPU and GPU code bases is difficult. We address this by proposing a software architecture grounded on Model-Driven Engineering that isolates logical models from physical implementations. Our approach transforms a high-level Domain-Specific Language into an intermediate representation based on the Actor Model, enabling the automatic generation of both C and CUDA code from a single source. We evaluate the framework using Parallel Discrete Event Simulation benchmarks. Results confirm that the architecture successfully propagates functional changes to all targets, ensuring the synthesised code is highly performant and functionally equivalent across platforms.
BibTeX Entry:
author = {Bauco, Simone and De Angelis, Guglielmo and Pellegrini, Alessandro},
title = {Automating Change Propagation in Software Applications for Heterogeneous Computing},
booktitle = {23rd {IEEE} International Conference on Software Architecture, {ICSA} 2026 - Companion},
publisher = {IEEE},
year = {2026},
month = jun
}