The Anatomy of a Triton Attention Kernel

• URL: http://arxiv.org/abs/2511.11581v1
• Date: Tue, 07 Oct 2025 13:34:51 GMT
• Title: The Anatomy of a Triton Attention Kernel
• Authors: Burkhard Ringlein, Jan van Lunteren, Radu Stoica, Thomas Parnell,
• Abstract summary: A long-standing goal in both industry and academia is to develop an LLM inference platform that is portable across hardware architectures.<n>We develop a state-of-the-art paged attention kernel, that builds exclusively on the domain-specific just-in-time compiled language Triton.<n>We describe our high-level approach, the key algorithmic and system-level improvements, the parameter auto-tuning required to unlock efficiency, and the integrations into a popular inference server.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: A long-standing goal in both industry and academia is to develop an LLM inference platform that is portable across hardware architectures, eliminates the need for low-level hand-tuning, and still delivers best-in-class efficiency. In this work, we demonstrate that portable, efficient cross-platform LLM inference is indeed possible and share our experience. We develop a state-of-the-art paged attention kernel, the core performance-critical component of many LLM deployments, that builds exclusively on the domain-specific just-in-time compiled language Triton to achieve state-of-the-art performance on both NVIDIA and AMD GPUs. We describe our high-level approach, the key algorithmic and system-level improvements, the parameter auto-tuning required to unlock efficiency, and the integrations into a popular inference server that are necessary to bring the performance of a generic Triton attention kernel from 19.7% of the state-of-the-art to 105.9%. Our results highlight how open-source domain-specific languages can be leveraged to unlock model portability across different GPU vendors.

Related papers

• RooflineBench: A Benchmarking Framework for On-Device LLMs via Roofline Analysis [53.90240071275054]
  The transition toward localized intelligence through Small Language Models (SLMs) has intensified the need for rigorous performance characterization on resource-constrained edge hardware.<n>We propose a systematic framework that unifies architectural primitives and hardware constraints through the lens of operational intensity (OI)<n>By defining an inference-potential region, we introduce the Relative Inference Potential as a novel metric to compare efficiency differences between Large Language Models (LLMs) on the same hardware substrate.
  arXiv  Detail & Related papers  (2026-02-12T03:02:22Z)
• Eliminating Multi-GPU Performance Taxes: A Systems Approach to Efficient Distributed LLMs [61.953548065938385]
  We introduce the ''Three Taxes'' (Bulk Synchronous, Inter- Kernel Data Locality, and Kernel Launch Overhead) as an analytical framework.<n>We propose moving beyond the rigid BSP model to address key inefficiencies in distributed GPU execution.<n>We observe a 10-20% speedup in end-to-end latency over BSP-based approaches.
  arXiv  Detail & Related papers  (2025-11-04T01:15:44Z)
• STARK: Strategic Team of Agents for Refining Kernels [23.717055490630596]
  We introduce an agentic framework for GPU kernel optimization that explores the design space through multi-agent collaboration.<n>This framework mimics the workflow of expert engineers, enabling LLMs to reason about hardware trade-offs, incorporate profiling feedback, and refine kernels iteratively.<n>We evaluate our approach on KernelBench, a benchmark for LLM-based kernel optimization, and demonstrate substantial improvements over baseline agents.
  arXiv  Detail & Related papers  (2025-10-19T20:41:46Z)
• MultiKernelBench: A Multi-Platform Benchmark for Kernel Generation [17.461533973039064]
  MultiKernelBench is a benchmark for the generation of deep learning kernels using large language models (LLMs)<n>It spans 285 tasks across 14 well-defined kernel categories and supports three major hardware platforms.<n>We show significant variation in task difficulty, poor generalization to platforms with less training exposure, and the effectiveness of targeted prompting strategies.
  arXiv  Detail & Related papers  (2025-07-20T00:58:33Z)
• GPU Performance Portability needs Autotuning [0.0]
  LLMs grow in complexity, achieving state-of-the-art performance requires tight co-design across algorithms, software, and hardware.<n>We make the case for combining just-in-time (JIT) compilation with comprehensive kernel parameter autotuning.<n>Our results highlight autotuning as a promising path to unlocking model portability across GPU vendors.
  arXiv  Detail & Related papers  (2025-04-30T12:57:21Z)
• NVLM: Open Frontier-Class Multimodal LLMs [64.00053046838225]
  We introduce NVLM 1.0, a family of frontier-class multimodal large language models (LLMs) that achieve state-of-the-art results on vision-language tasks. We propose a novel architecture that enhances both training efficiency and multimodal reasoning capabilities. We develop production-grade multimodality for the NVLM-1.0 models, enabling them to excel in vision-language tasks.
  arXiv  Detail & Related papers  (2024-09-17T17:59:06Z)
• FusionAI: Decentralized Training and Deploying LLMs with Massive Consumer-Level GPUs [57.12856172329322]
  We envision a decentralized system unlocking the potential vast untapped consumer-level GPU. This system faces critical challenges, including limited CPU and GPU memory, low network bandwidth, the variability of peer and device heterogeneity.
  arXiv  Detail & Related papers  (2023-09-03T13:27:56Z)
• QIGen: Generating Efficient Kernels for Quantized Inference on Large Language Models [22.055655390093722]
  We present an automatic code generation approach for supporting quantized generative inference on LLMs such as LLaMA or OPT on off-the-shelf CPUs. Results on CPU-based inference for LLaMA models show that our approach can lead to high performance and high accuracy, comparing favorably to the best existing open-source solution.
  arXiv  Detail & Related papers  (2023-07-07T17:46:08Z)
• LAMM: Language-Assisted Multi-Modal Instruction-Tuning Dataset, Framework, and Benchmark [81.42376626294812]
  We present Language-Assisted Multi-Modal instruction tuning dataset, framework, and benchmark. Our aim is to establish LAMM as a growing ecosystem for training and evaluating MLLMs. We present a comprehensive dataset and benchmark, which cover a wide range of vision tasks for 2D and 3D vision.
  arXiv  Detail & Related papers  (2023-06-11T14:01:17Z)
• Harnessing Deep Learning and HPC Kernels via High-Level Loop and Tensor Abstractions on CPU Architectures [67.47328776279204]
  This work introduces a framework to develop efficient, portable Deep Learning and High Performance Computing kernels. We decompose the kernel development in two steps: 1) Expressing the computational core using Processing Primitives (TPPs) and 2) Expressing the logical loops around TPPs in a high-level, declarative fashion. We demonstrate the efficacy of our approach using standalone kernels and end-to-end workloads that outperform state-of-the-art implementations on diverse CPU platforms.
  arXiv  Detail & Related papers  (2023-04-25T05:04:44Z)
• EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction [67.11722682878722]
  This work presents EfficientViT, a new family of high-resolution vision models with novel multi-scale linear attention. Our multi-scale linear attention achieves the global receptive field and multi-scale learning. EfficientViT delivers remarkable performance gains over previous state-of-the-art models.
  arXiv  Detail & Related papers  (2022-05-29T20:07:23Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.