Fovea: Physical-Implication-Aware Wafer-Scale DSE with Decision-Domain-Guided Cross-Fidelity Refinement

2026-08-04

AI Summary

This work addresses the challenge of architectural design in wafer-scale systems, where tightly coupled physical constraints invalidate conventional unconstrained modeling approaches and render high-fidelity evaluation prohibitively expensive while low-fidelity ranking proves unreliable. To overcome these limitations, we propose a physics-aware design space construction method coupled with decision-region-guided cross-fidelity refinement. By integrating constraint-aware modeling, selective high-fidelity evaluation, and sampling-driven reference calibration, our approach efficiently identifies optimal architectures while preserving critical dimensional trade-offs. Empirical evaluation across ten design spaces and seven large-model training workloads demonstrates that our method recovers the global optimum with 100% accuracy using only 10% of the calibration samples required by baseline strategies, achieving an average speedup of 4.13× and up to 7.80× in optimization time.

Abstract

Modern pre-silicon design-space exploration (DSE) follows a coarse-to-fine workflow: low-cost evaluators screen candidate spaces, while detailed evaluation is reserved for a shortlist. Wafer-scale systems strain both stages. Architecture choices induce coupled physical implications for reticle compliance, wafer tiling, die area, D2D capability, boundary access, and placement, so the design space cannot be treated as an unconstrained Cartesian product. Meanwhile, detailed evaluation is too expensive to cover the resulting space, whereas analytical-to-reference ranking inversions make a fixed shortlist unreliable. We present Fovea, a reusable methodology for workload-specific wafer architecture selection rather than a fixed wafer template. Fovea first performs physical-implication-aware design-space formulation to construct a distinct modeled-feasible space while preserving cross-dimensional trade-offs and applying only evaluator-preserving local reductions. It then performs Decision-Domain-guided cross-fidelity refinement. Paired in-domain calibration estimates workload- and space-specific analytical-to-reference disagreement, which parameterizes reference-consistent performance intervals and induces a Decision Domain for selective designated-reference evaluation. Under a valid domain-wide disagreement bound, this domain contains the designated-reference optimum; the sampling-based implementation is evaluated empirically on exhaustive-reference design spaces. Across ten reference-verifiable design spaces and seven LLM-training workloads, Fovea with 10% paired calibration recovers the exhaustive designated-reference optimum in all 70 evaluated pairs while achieving 4.13x average and 7.80x maximum end-to-end speedup.

Problem

• wafer-scale systems
• design-space exploration
• physical implications
• cross-fidelity refinement
• ranking inversion

Innovation

• wafer-scale DSE
• physical-implication-aware
• cross-fidelity refinement
• decision domain
• design-space exploration