{"slug": "agenticcann-automated-ascend-c-operator-generation-via-knowledge-augmented", "title": "AgenticCANN: Automated Ascend C Operator Generation via Knowledge-Augmented Agentic Evolution", "summary": "Researchers propose AgenticCANN, a knowledge-augmented agentic evolution framework for automated Ascend C operator synthesis on Huawei Ascend 910B NPUs, achieving 90-100% feasibility on elementwise and normalization operators, 56% on fusion operators, and up to 6.65x speedup on 1B Pangu model inference kernels. The framework addresses the platform knowledge deficit by integrating structured domain insights and a stage-adaptive agentic evolution strategy, with knowledge injection improving feasibility from 57% to 86% on elementwise operators.", "body_md": "arXiv:2607.26661v1 Announce Type: new\nAbstract: Ascend C operator optimization is critical for NPU (Neural Processing Unit) inference performance but requires deep hardware expertise.While large language models (LLMs) have shown promise in automated CUDA kernel generation, the fundamentally different programming model of Ascend C introduces unique challenges that remain unexplored. In this paper, we propose AgenticCANN, a knowledge-augmented agentic evolution framework specifically tailored for automated Ascend C operator synthesis in low-corpus NPU environments.To overcome the severe platform knowledge deficit on unfamiliar hardware, AgenticCANN incorporates a knowledge-orchestrated generation system that delivers structured, multi-level domain insights across the development lifecycle to resolve the upstream feasibility bottleneck.Building on this foundation, it features a stage-adaptive agentic evolution strategy that dynamically aligns LLM interaction modes with specific generation and evolution phases, balancing high-exploration candidate discovery with high-convergence performance tuning.Extensive experiments on Huawei Ascend 910B across six operators spanning five pattern categories demonstrate that our method achieves 90 to 100 percent feasibility on elementwise and normalization operators, 56% on fusion operators, and up to 6.65$\\times$ speedup on 1B Pangu model inference kernels. Further analysis reveals that knowledge injection monotonically improves feasibility from 57% to 86% on elementwise operators, demonstrating its general rather than operator-specific benefit.", "url": "https://wpnews.pro/news/agenticcann-automated-ascend-c-operator-generation-via-knowledge-augmented", "canonical_source": "https://arxiv.org/abs/2607.26661", "published_at": "2026-07-31 04:00:00+00:00", "updated_at": "2026-07-31 04:29:40.841084+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "large-language-models", "ai-research", "ai-infrastructure"], "entities": ["AgenticCANN", "Huawei Ascend 910B", "Ascend C", "Pangu", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/agenticcann-automated-ascend-c-operator-generation-via-knowledge-augmented", "markdown": "https://wpnews.pro/news/agenticcann-automated-ascend-c-operator-generation-via-knowledge-augmented.md", "text": "https://wpnews.pro/news/agenticcann-automated-ascend-c-operator-generation-via-knowledge-augmented.txt", "jsonld": "https://wpnews.pro/news/agenticcann-automated-ascend-c-operator-generation-via-knowledge-augmented.jsonld"}}