Subgraph Extraction-based Feedback-guided Iterative Scheduling for HLS
Hanchen Ye, David Z. Pan, Chris Leary, Deming Chen, Xiaoqing Xu
TL;DR
The paper tackles the mismatch between IR-level HLS timing estimates and post-synthesis QoR by introducing ISDC, a feedback-guided iterative SDC scheduling framework. ISDC augments the traditional SDC LP with low-level delays from downstream tools and uses two subgraph extraction strategies (fanout-driven and window-based) to drive a no-human-in-loop feedback loop. Through delay updating and an efficient reformulation, ISDC refines schedules over multiple iterations, achieving a notable 28.5% reduction in register usage on 17 XLS-based benchmarks, at the cost of increased scheduling time. The work demonstrates the value of cross-layer feedback in HLS and provides an open-source implementation to enable broader adoption and further exploration with co-optimization of HLS and logic synthesis.
Abstract
This paper proposes ISDC, a novel feedback-guided iterative system of difference constraints (SDC) scheduling algorithm for high-level synthesis (HLS). ISDC leverages subgraph extraction-based low-level feedback from downstream tools like logic synthesizers to iteratively refine HLS scheduling. Technical innovations include: (1) An enhanced SDC formulation that effectively integrates low-level feedback into the linear-programming (LP) problem; (2) A fanout and window-based subgraph extraction mechanism driving the feedback cycle; (3) A no-human-in-loop ISDC flow compatible with a wide range of downstream tools and process design kits (PDKs). Evaluation shows that ISDC reduces register usage by 28.5% against an industrial-strength open-source HLS tool.
