New AI Trick Cuts Supercomputer Code Translation Costs by Half
This could save companies millions in programming time and energy bills.
Translating parallel scientific codes between HPC programming frameworks is critical but labor-intensive. A new agent-skill framework separates translation, compilation diagnosis, and execution validation into three autonomous agents coordinated by a verdict-driven orchestrator, routing each failure to a targeted repair skill rather than retrying blindly.
An automated output check drives the self-correction loop, complemented by offline LLM-based verification that exposes false positives escaping compiler and runtime checks. Evaluated on 30 HeCBench kernels translated bidirectionally among CUDA, OpenMP, and SYCL with three open-source LLMs, the framework attained a higher verified correctness rate in all 18 model and direction combinations over a state-of-the-art baseline while consuming up to 52.4% fewer tokens, remaining consistent across models and directions without architectural modification.
- Three AI agents team up to translate supercomputer code more accurately than a single AI.
- The method uses up to 52% fewer tokens, cutting AI costs roughly in half.
- Better code translation could speed up scientific research and lower energy use.
Why It Matters
Cheaper, faster supercomputer code translation means quicker scientific discoveries and lower costs for research labs.