Teaching An Old Dog New Tricks: Porting Legacy Code to Heterogeneous Compute Architectures With Automated Code Translation

2025-02-07

AI Summary

To address the high cost and intrusive nature of porting legacy scientific simulation codes—such as Fortran 2003 multigrid solvers—to heterogeneous architectures like GPUs, this paper proposes a zero-intrusion automatic porting method. Our approach leverages compile-time source-code parsing integrated with the Loopy framework to construct an end-to-end Fortran-to-GPU code generation pipeline, producing portable C++/CUDA kernels. We further introduce a lightweight translation layer and a custom MPI-GPU co-runtime to enable multi-node scheduling. Crucially, our method requires no modifications to the original Fortran source code, achieving seamless heterogeneity adaptation. Experimental evaluation demonstrates 2–3× speedup on a single GPU node and up to 6× strong scaling across multiple nodes. This work significantly lowers the barrier and engineering overhead for migrating scientific computing applications to heterogeneous systems.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Hardware-aware ML
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Legacy codes are in ubiquitous use in scientific simulations; they are well-tested and there is significant time investment in their use. However, one challenge is the adoption of new, sometimes incompatible computing paradigms, such as GPU hardware. In this paper, we explore using automated code translation to enable execution of legacy multigrid solver code on GPUs without significant time investment and while avoiding intrusive changes to the codebase. We developed a thin, reusable translation layer that parses Fortran 2003 at compile time, interfacing with the existing library Loopy to transpile to C++/GPU code, which is then managed by a custom MPI runtime system that we created. With this low-effort approach, we are able to achieve a payoff of an approximately 2-3x speedup over a full CPU socket, and 6x in multi-node settings.

Problem

• Automated translation of legacy code
• Enabling GPU execution of Fortran
• Achieving speedup with minimal code changes

Innovation

• Automated code translation for GPUs
• Fortran 2003 to C++/GPU conversion
• Custom MPI runtime system management