Search papers, labs, and topics across Lattice.
This paper introduces HyperCut, a novel hierarchical partitioning-and-mapping framework that optimizes inter-layer scheduling in deep neural networks (DNNs) by leveraging directed hypergraph abstractions. By allowing early filtering of inter-layer schedules and tightly bounding intra-layer scheduling costs, HyperCut significantly reduces the design space complexity from O(9.899^N) to O(N). The method achieves a remarkable 2.0x performance improvement and an 80.47% reduction in exploration time compared to the existing state-of-the-art scheduler, SET, across ten evaluated cases.
HyperCut slashes design space exploration time by over 80% while doubling performance, revolutionizing inter-layer scheduling for deep neural networks.
As deep neural networks (DNNs) continue to scale, inter-layer scheduling, which orchestrates the spatial allocation of compute resources and the temporal execution order across layers, has become a decisive factor in sustaining high utilization and energy efficiency on tiled accelerators. However, existing inter-layer schedulers defer cost feedback until a complete fine-grained intra-layer scheduling has been resolved. The resulting decoupled flow repeatedly explores sub-optimal or even infeasible inter-layer schedules, and the absence of early pruning during the inter-layer phase remains a critical bottleneck for design-space exploration (DSE) in DNN compilers. Our key observation is that the cost of an intra-layer scheduling can be tightly upper-bounded once the inter-layer cut fixes the sub-mesh shape, which lets us cost every inter-layer candidate without solving the intra-layer problem. Hence, we propose a hierarchical partitioning-and-mapping framework, HyperCut, that enables early filtering of inter-layer schedules based on hypergraph partitioning. Based on the directed hypergraph (DHG) abstraction of DNN, we introduce a unified representation, State, that jointly encodes the DHG partition, tile mesh allocation and tensor batch splitting. Thereby, partitioning and mapping are coupled into a union optimization object. For a DNN with N layers, the resulting theoretical design space is bounded by O(N), compared with O(9.899^N) for the state-of-the-art open-source scheduler SET. Across 10 evaluated cases, HyperCut achieves 2.0x performance improvement and 80.47% exploration time reduction over the SET baseline, measured by geometric mean.
