XXXiong/ChatHLS-HLSTuner
ChatHLS-HLSTuner by XXXiong is a specialized large language model fine-tuned from Qwen/Qwen2.5-Coder-14B-Instruct for High-Level Synthesis (HLS) C/C++ code optimization. It functions as an optimization analysis agent within the ChatHLS framework, focusing on inserting `#pragma` directives to enhance loops and arrays. The model aims to improve performance, power, and area (PPA) metrics by suggesting modifications for lower latency and optimized resource utilization in HLS designs. It is specifically designed for systematic design automation and optimization in HLS.
Loading preview...
ChatHLS-HLSTuner: HLS C/C++ Code Optimization
ChatHLS-HLSTuner, developed by XXXiong, is a specialized large language model fine-tuned from Qwen/Qwen2.5-Coder-14B-Instruct. Its core function is to act as an optimization analysis agent within the broader ChatHLS framework, as detailed in its source paper, "ChatHLS: Towards Systematic Design Automation and Optimization for High-Level Synthesis" (ACL 2026).
Key Capabilities
- HLS Code Optimization: Specifically designed to analyze and optimize High-Level Synthesis (HLS) C/C++ code.
- Pragma Insertion: Recommends the insertion of
#pragmadirectives (e.g., Loop Pipeline, Loop Unroll, Array Partition) to improve code efficiency. - PPA Enhancement: Aims to enhance Performance, Power, and Area (PPA) metrics by suggesting modifications that lead to lower latency and optimized hardware resource utilization.
- Optimization Analysis: Provides detailed reasoning for suggested pragmas, including their expected impact on PPA.
Ideal Use Cases
- Automated HLS Design: Developers working on automated design and optimization flows for HLS.
- Hardware Acceleration: Engineers seeking to optimize C/C++ code for FPGA or ASIC synthesis through HLS.
- Performance Tuning: Users needing systematic suggestions for improving latency and resource usage in HLS projects.