Search papers, labs, and topics across Lattice.
This paper introduces Hawk, a training-free framework designed to generate high-performance kernels for Neural Processing Units (NPUs) by addressing the challenges posed by hardware constraints and memory hierarchies. By integrating three innovative modules鈥擱un-Time Knowledge Synthesis, Bottleneck-Aware Knowledge Retrieval, and Effect-Driven Knowledge Distillation鈥擧awk significantly enhances kernel generation accuracy from 49.4% to 80.0% and achieves up to a 2.2x speedup in execution over existing methods. These improvements highlight the importance of hardware-aware knowledge in automating NPU kernel development, a critical bottleneck in the industry.
Hawk boosts NPU kernel generation accuracy by over 30% while doubling execution speed, revolutionizing how we approach hardware-specific programming.
Developing high-performance kernels for Neural Processing Units (NPUs) is a critical industry bottleneck, requiring developers to manually navigate implicit hardware constraints and strict memory hierarchies. While large language models offer immense automation potential, they fail catastrophically on NPUs due to a fundamental lack of hardware-specific priors. Naively transplanting code snippets from similar NPU kernels may pass the compiler, but it consistently triggers runtime crashes and performance degradation by blindly violating underlying hardware constraints. To overcome this, we introduce Hawk, a training-free framework that harnesses hardware-aware knowledge through three core modules: (1) Run-Time Knowledge Synthesis Module, which employs a Triple-Part Executable Knowledge Representation to inherently couple the error context with executable semantics; (2) Bottleneck-Aware Knowledge Retrieval Module, which implements a 2D-Retrieval paradigm to project queries into orthogonal syntactic and hardware-aligned semantic spaces; and (3) Effect-Driven Knowledge Distillation Module, which leverages LLM-driven semantic arbitration to continuously distill the knowledge by pruning errors and consolidating redundancies based on the empirical execution feedback. Extensive evaluations on real-world NPU workloads demonstrate that Hawk elevates generation accuracy from 49.4% to 80.0%, while achieving up to a 2.2x execution speedup over state-of-the-art baselines.
