HLSFactory

HLSFactory是由佐治亚理工学院等机构联合开发的一个综合框架，旨在促进高质量HLS设计数据集的策划和生成。该数据集包含257条设计，涵盖了从简单到复杂的设计，使用AMD/Xilinx和Intel工具流进行合成。HLSFactory通过三个主要阶段（设计空间扩展、设计合成和数据聚合）确保数据点的广泛覆盖，并支持多供应商工具的互操作性。用户可以通过提供的API轻松贡献自己的HLS设计或合成结果。该框架不仅适用于ML训练，还可用于各种需要大量多样HLS运行的任务，如回归测试和学术HLS工具的基准测试。

HLSFactory is a comprehensive framework jointly developed by the Georgia Institute of Technology and other research institutions, dedicated to facilitating the curation and generation of high-quality HLS design datasets. This dataset includes 257 designs ranging from simple to complex, which are synthesized using both AMD/Xilinx and Intel toolflows. HLSFactory ensures comprehensive coverage of data points through three core stages: design space exploration, design synthesis, and data aggregation, and supports interoperability across multi-vendor tools. Users can easily contribute their own HLS designs or synthesis results via the provided API. This framework is applicable not only to ML training, but also to various tasks requiring a large number of diverse HLS runs, such as regression testing and benchmarking of academic HLS tools.