HLSDataset
收藏资源简介:
HLSDataset是由德克萨斯大学奥斯汀分校计算机体系结构实验室创建的开源数据集,专为使用高层次综合(HLS)的机器辅助FPGA设计。该数据集包含约9000个Verilog设计样本,来源于Polybench、Machsuite、CHStone和Rosetta等广泛使用的HLS C基准。数据集通过应用多种指令如循环展开、循环流水线和数组分区,确保覆盖优化和实际设计。此外,数据集还包括CSV文件,包含HLS和实现度量,以及原始C源代码、Verilog设计、HLS后报告和实现后报告,便于ML模型的直接使用。HLSDataset旨在通过减少数据集生成的时间消耗,使研究人员能够更专注于ML模型的训练,适用于资源使用、时序和功率估计等多种应用。
HLSDataset is an open-source dataset developed by the Computer Architecture Laboratory at The University of Texas at Austin, specifically tailored for machine-aided FPGA design using high-level synthesis (HLS). It contains approximately 9,000 Verilog design samples sourced from widely adopted HLS C benchmarks including Polybench, Machsuite, CHStone, and Rosetta. A range of HLS optimizations such as loop unrolling, loop pipelining, and array partitioning have been applied to the dataset, ensuring coverage of both optimized and practical real-world design scenarios. Additionally, the dataset includes CSV files storing HLS and implementation metrics, alongside the original C source code, Verilog designs, post-HLS reports, and post-implementation reports, which enable direct utilization for machine learning (ML) models. HLSDataset aims to reduce the time and effort spent on dataset generation, allowing researchers to focus more on ML model training, and supports a wide array of applications including resource utilization estimation, timing analysis, and power consumption estimation.

- 1HLSDataset: Open-Source Dataset for ML-Assisted FPGA Design using High Level Synthesis德克萨斯大学奥斯汀分校计算机体系结构实验室 · 2023年



