Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SuperCode: Sustainability PER AI-driven CO-DEsign

P. Chris Broekema, Rob V. van Nieuwpoort

TL;DR

The paper tackles the environmental impact of data-intensive science by proposing SuperCode, an AI-driven co-design framework that uses LLMs to generate energy-efficient code for emerging hardware. It defines sustainability as a KPI and introduces $M_S = \frac{TVO}{TCO}$ to balance scientific value against environmental cost. Two radio astronomy use-cases (terrestrial LOFAR/SKA and space-based OLFAR/ALO/DSL) will be used to validate the approach, with energy efficiency as a primary objective. The methodology combines AI-assisted translation and porting across architectures, LoRA fine-tuning, retrieval-augmented generation, and reinforcement learning driven by measured energy, plus human-in-the-loop to ensure correctness. If successful, the project could accelerate sustainable data-intensive science and be applicable to climate research, remote sensing, and earth observation, with open dissemination.

Abstract

Currently, data-intensive scientific applications require vast amounts of compute resources to deliver world-leading science. The climate emergency has made it clear that unlimited use of resources (e.g., energy) for scientific discovery is no longer acceptable. Future computing hardware promises to be much more energy efficient, but without better optimized software this cannot reach its full potential. In this vision paper, we propose a generic AI-driven co-design methodology, using specialized Large Language Models (like ChatGPT), to effectively generate efficient code for emerging computing hardware. We describe how we will validate our methodology with two radio astronomy applications, with sustainability as the key performance indicator. This paper is a modified version of our accepted SuperCode project proposal. We present it here in this form to introduce the vision behind this project and to disseminate the work in the spirit of Open Science and transparency. An additional aim is to collect feedback, invite potential collaboration partners and use-cases to join the project.

SuperCode: Sustainability PER AI-driven CO-DEsign

TL;DR

The paper tackles the environmental impact of data-intensive science by proposing SuperCode, an AI-driven co-design framework that uses LLMs to generate energy-efficient code for emerging hardware. It defines sustainability as a KPI and introduces to balance scientific value against environmental cost. Two radio astronomy use-cases (terrestrial LOFAR/SKA and space-based OLFAR/ALO/DSL) will be used to validate the approach, with energy efficiency as a primary objective. The methodology combines AI-assisted translation and porting across architectures, LoRA fine-tuning, retrieval-augmented generation, and reinforcement learning driven by measured energy, plus human-in-the-loop to ensure correctness. If successful, the project could accelerate sustainable data-intensive science and be applicable to climate research, remote sensing, and earth observation, with open dissemination.

Abstract

Currently, data-intensive scientific applications require vast amounts of compute resources to deliver world-leading science. The climate emergency has made it clear that unlimited use of resources (e.g., energy) for scientific discovery is no longer acceptable. Future computing hardware promises to be much more energy efficient, but without better optimized software this cannot reach its full potential. In this vision paper, we propose a generic AI-driven co-design methodology, using specialized Large Language Models (like ChatGPT), to effectively generate efficient code for emerging computing hardware. We describe how we will validate our methodology with two radio astronomy applications, with sustainability as the key performance indicator. This paper is a modified version of our accepted SuperCode project proposal. We present it here in this form to introduce the vision behind this project and to disseminate the work in the spirit of Open Science and transparency. An additional aim is to collect feedback, invite potential collaboration partners and use-cases to join the project.

Paper Structure

This paper contains 16 sections, 1 equation, 4 figures.

Table of Contents

  1. abstract
  2. Introduction
  3. Problem statement
  4. Illustrative Example
  5. Vision and methodology
  6. Sustainability
  7. Sustainability as Key Performance Indicator
  8. AI-driven co-design
  9. Emerging Technologies
  10. Use-cases
  11. Use-case 1: Terrestrial large-scale distributed radio telescopes
  12. Use-case 2: Space-based radio telescopes
  13. Scientific and Societal Impact
  14. Related work
  15. Conclusions
  16. ...and 1 more sections

Figures (4)

  • Figure 1: Adding science to the hardware/software co-design loop and leveraging AI to facilitate this process
  • Figure 2: A high-level representation of our AI-driven co-design vision
  • Figure 3: The core of the LOFAR telescope.
  • Figure 4: The OLFAR roadmap, picture courtesy bentum_roadmap_2020.