Optimizing Attention on GPUs by Exploiting GPU Architectural NUMA Effects

• URL: http://arxiv.org/abs/2511.02132v1
• Date: Mon, 03 Nov 2025 23:48:39 GMT
• Title: Optimizing Attention on GPUs by Exploiting GPU Architectural NUMA Effects
• Authors: Mansi Choudhary, Karthik Sangaiah, Sonali Singh, Muhammad Osama, Lisa Wu Wills, Ganesh Dasika,
• Abstract summary: We show how memory latency and bandwidth vary sharply across compute regions, undermining traditional GPU kernel scheduling strategies.<n>We present Swizzled Head-first Mapping, a spatially-aware scheduling strategy that aligns attention heads with GPU NUMA domains.<n>On AMD's MI300X architecture, our method achieves up to 50% higher performance over state-of-the-art attention algorithms.
• Score: 3.124720734513875
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The rise of disaggregated AI GPUs has exposed a critical bottleneck in large-scale attention workloads: non-uniform memory access (NUMA). As multi-chiplet designs become the norm for scaling compute capabilities, memory latency and bandwidth vary sharply across compute regions, undermining the performance of traditional GPU kernel scheduling strategies that assume uniform memory access. We identify how these NUMA effects distort locality in multi-head attention (MHA) and present Swizzled Head-first Mapping, a spatially-aware scheduling strategy that aligns attention heads with GPU NUMA domains to exploit intra-chiplet cache reuse. On AMD's MI300X architecture, our method achieves up to 50% higher performance over state-of-the-art attention algorithms using conventional scheduling techniques and sustains consistently high L2 cache hit rates of 80-97%. These results demonstrate that NUMA-aware scheduling is now fundamental to achieving full efficiency on next-generation disaggregated GPUs, offering a path forward for scalable AI training and inference.

Related papers

• GPU-Accelerated Algorithms for Graph Vector Search: Taxonomy, Empirical Study, and Research Directions [54.570944939061555]
  We present a comprehensive study of GPU-accelerated graph-based vector search algorithms.<n>We establish a detailed taxonomy of GPU optimization strategies and clarify the mapping between algorithmic tasks and hardware execution units.<n>Our findings offer clear guidelines for designing scalable and robust GPU-powered approximate nearest neighbor search systems.
  arXiv  Detail & Related papers  (2026-02-10T16:18:04Z)
• Evolutionary Mapping of Neural Networks to Spatial Accelerators [64.13809409887254]
  We introduce the first evolutionary, hardware-in-the-loop mapping framework for neuromorphic accelerators.<n>We evaluate our approach on Intel Loihi 2, a representative spatial accelerator featuring 152 cores in a 2D mesh.<n>Our method achieves up to 35% reduction in total latency compared to default cores on two sparse multi-layer perceptron networks.
  arXiv  Detail & Related papers  (2026-02-04T16:28:08Z)
• AIRES: Accelerating Out-of-Core GCNs via Algorithm-System Co-Design [6.554916179445241]
  Graph convolutional networks (GCNs) are fundamental in various scientific applications, ranging from biomedical protein-protein interactions (PPI) to large-scale recommendation systems.<n>An essential component for modeling graph structures in GCNs is sparse general matrix-matrix multiplication (SpGEMM)<n>SpGEMMs are often conducted in an out-of-core fashion due to limited GPU memory space in resource-constrained systems.<n>We propose AIRES, a novel algorithm-system co-design solution to accelerate out-of-core SpGEMM computation for GCNs.
  arXiv  Detail & Related papers  (2025-07-02T00:35:43Z)
• Power Law Guided Dynamic Sifting for Efficient Attention [0.8198759882422455]
  We propose SiftAttention, a novel approximate attention method that replaces the top-k$ step with a computationally efficient element-wise filtering operation based on a threshold value.<n>Our evaluation demonstrates that SiftAttention preserves model quality better than existing approximate attention methods while reducing memory usage when loading vectors.
  arXiv  Detail & Related papers  (2025-06-05T17:50:32Z)
• HAPM -- Hardware Aware Pruning Method for CNN hardware accelerators in resource constrained devices [44.99833362998488]
  The present work proposes a generic hardware architecture ready to be implemented on FPGA devices. The inference speed of the design is evaluated over different resource constrained FPGA devices. We demonstrate that our hardware-aware pruning algorithm achieves a remarkable improvement of a 45 % in inference time compared to a network pruned using the standard algorithm.
  arXiv  Detail & Related papers  (2024-08-26T07:27:12Z)
• Multi-GPU RI-HF Energies and Analytic Gradients $-$ Towards High Throughput Ab Initio Molecular Dynamics [0.0]
  This article presents an optimized algorithm and implementation for calculating resolution-of-the-identity Hartree-Fock energies and analytic gradients using multiple Graphics Processing Units (GPUs) The algorithm is especially designed for high throughput emphab initio molecular dynamics simulations of small and medium size molecules (10-100 atoms)
  arXiv  Detail & Related papers  (2024-07-29T00:14:10Z)
• FULL-W2V: Fully Exploiting Data Reuse for W2V on GPU-Accelerated Systems [5.572152653851948]
  FULL-W2V exploits the opportunities for data reuse in the W2V algorithm to reduce access to low memory levels and improve temporal locality. Our prototype implementation achieves 2.97X speedup when ported from Nvidia Pascal P100 to Volta V100 cards, and outperforms the state-of-the-art by 5.72X on V100 cards with the same embedding quality.
  arXiv  Detail & Related papers  (2023-12-12T21:22:07Z)
• Constant Memory Attention Block [74.38724530521277]
  Constant Memory Attention Block (CMAB) is a novel general-purpose attention block that computes its output in constant memory and performs updates in constant computation. We show our proposed methods achieve results competitive with state-of-the-art while being significantly more memory efficient.
  arXiv  Detail & Related papers  (2023-06-21T22:41:58Z)
• Adaptable Butterfly Accelerator for Attention-based NNs via Hardware and Algorithm Co-design [66.39546326221176]
  Attention-based neural networks have become pervasive in many AI tasks. The use of the attention mechanism and feed-forward network (FFN) demands excessive computational and memory resources. This paper proposes a hardware-friendly variant that adopts a unified butterfly sparsity pattern to approximate both the attention mechanism and the FFNs.
  arXiv  Detail & Related papers  (2022-09-20T09:28:26Z)
• The Architectural Implications of Distributed Reinforcement Learning on CPU-GPU Systems [45.479582612113205]
  We show how to improve the performance and power efficiency of RL training on CPU-GPU systems. We quantify the overall hardware utilization on a state-of-the-art distributed RL training framework. We also introduce a new system design metric, CPU/GPU ratio, and show how to find the optimal balance between CPU and GPU resources.
  arXiv  Detail & Related papers  (2020-12-08T04:50:05Z)
• Real-time Semantic Segmentation with Fast Attention [94.88466483540692]
  We propose a novel architecture for semantic segmentation of high-resolution images and videos in real-time. The proposed architecture relies on our fast spatial attention, which is a simple yet efficient modification of the popular self-attention mechanism. We show that results on multiple datasets demonstrate superior performance with better accuracy and speed compared to existing approaches.
  arXiv  Detail & Related papers  (2020-07-07T22:37:16Z)
• Heterogeneous CPU+GPU Stochastic Gradient Descent Algorithms [1.3249453757295084]
  We study training algorithms for deep learning on heterogeneous CPU+GPU architectures. Our two-fold objective -- maximize convergence rate and resource utilization simultaneously -- makes the problem challenging. We show that the implementation of these algorithms achieves both faster convergence and higher resource utilization than on several real datasets.
  arXiv  Detail & Related papers  (2020-04-19T05:21:20Z)
• Faster than FAST: GPU-Accelerated Frontend for High-Speed VIO [46.20949184826173]
  This work focuses on the applicability of efficient low-level, GPU hardware-specific instructions to improve on existing computer vision algorithms. Especially non-maxima suppression and the subsequent feature selection are prominent contributors to the overall image processing latency.
  arXiv  Detail & Related papers  (2020-03-30T14:16:23Z)

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.