More with Less: Efficient Parallel Scientific Code Translation via an Agent-Skill Workflow
Executive Abstract Synthesis
Translating parallel scientific codes between HPC programming frameworks is a critical but labor-intensive task. Existing LLM-based approaches rely on fixed sequential pipelines with undifferentiated repair strategies and correctness criteria that do not match scientific computing practice. We present an agent-skill framework that 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 an offline LLM-based verification that exposes false positives escaping compiler and runtime checks. We evaluate the framework on 30 HeCBench kernels translated bidirectionally among CUDA, OpenMP, and SYCL with three open-source LLMs. Our framework attains the 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.
Key Empirical Findings & Benchmarks
- Peer-indexed open-access research brief synthesized by the ARGUS Scientific Intelligence Desk.
- Demonstrates empirical methodology and reproducible benchmarks across computational and physical models.
- Evaluated for cross-sectoral impact on technology infrastructure, computational hardware, and policy.
Strategic & Hardware Implementation Impact
Provides researchers, engineers, and enterprise decision-makers with early visibility into frontier scientific developments.