aiqbench analysis
This research addresses a critical bottleneck in automated kernel optimization for emerging accelerators like NPUs by moving beyond surface-level profiling metrics. The proposed framework uses LLMs to perform hierarchical, compiler-grounded diagnosis that links runtime symptoms directly to intermediate representation structures before suggesting source rewrites. This approach significantly improves performance on Ascend 950 benchmarks while providing transparent insights into why specific optimizations fail at the backend level.
Compiler-Grounded Hierarchical Diagnosis for LLM-Based Triton Kernel Optimization
arXiv:2607.23089v1 Announce Type: new Abstract: Recent advances in large language models (LLMs) have enabled automated kernel generation and optimization, but most existing approaches rely on surface signals such as compilation feedback and profiling metrics. These signals reveal that a kernel is slow, but not why the backend compiler fails to realize a profitable optimization, especially on emerging accelerators such as NPUs. We therefore formulate kernel optimization as a progressive cross-layer diagnosis problem that links runtime symptoms to IR structure and compiler behavior before rewriting source. Based on this insight, we present our system, a compiler-grounded and hierarchical optimization framework for Triton kernels. the system escalates from lightweight pattern triage and profiling diagnosis to IR attribution and compiler-grounded analysis only when deeper evidence is needed, then proposes evidence-backed source-level rewrites. We implement the system on Triton for Ascend NPUs and evaluate it on 37 successfully converted entries from a standardized NPUKernelBench-derived Ascend 950 benchmark. Across these entries, the system attains a geometric-mean speedup of 4.35$\times$ and a median speedup of 2.73$\times$ from the initial to optimized Triton kernel; 22/37 exceed 2$\times$ and 13/37 exceed 5$\times$. The complete distribution ranges from near-baseline entries to large wins, motivating transparent reporting of the current system's scope and limitations.
Abstract:Recent advances in large language models (LLMs) have enabled automated kernel generation and optimization, but most existing approaches rely on surface signals such as compilation feedback and profiling metrics. These signals reveal that a kernel is slow, but not why the backend compiler fails to realize a profitable optimization, especially on emerging accelerators such as NPUs. We therefore formulate kernel optimization as a progressive cross-layer diagnosis problem that links runtime symptoms to IR structure and compiler behavior before rewriting source. Based on this insight, we present our system, a compiler-grounded and hierarchical optimization framework for Triton kernels. the system escalates from lightweight pattern triage and profiling diagnosis to IR attribution and compiler-grounded analysis only when deeper evidence is needed, then proposes evidence-backed source-level rewrites. We implement the system on Triton for Ascend NPUs and evaluate it on 37 successfully converted entries from a standardized NPUKernelBench-derived Ascend 950 benchmark. Across these entries, the system attains a geometric-mean speedup of 4.35$\times$ and a median speedup of 2.73$\times$ from the initial to optimized Triton kernel; 22/37 exceed 2$\times$ and 13/37 exceed 5$\times$. The complete distribution ranges from near-baseline entries to large wins, motivating transparent reporting of the current system's scope and limitations.
Artificial Intelligence (cs.AI); Programming Languages (cs.PL)
arXiv:2607.23089 [cs.AI]
(or arXiv:2607.23089v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.23089 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 220 words
Related coverage
Translating CUDA Tile Operations from Python to Rust Using Agentic AI
NVIDIA · Sep 16, 2026
Performance Profiling on AMD GPUs – Part 5: Profiling-Driven Kernel Optimization with an AI Code-Assist Tool
AMD · Jul 16, 2026
AMDKernelVault: Large-Scale Datasets and Agentic Training for AMD GPU Kernel Optimization
arXiv · Sep 14, 2026
How Generative Recommenders Are Redefining RecSys at Scale
NVIDIA · Aug 20, 2026