Dynamic Sparse Attention on Mobile SoCs

• URL: http://arxiv.org/abs/2508.16703v1
• Date: Fri, 22 Aug 2025 07:41:35 GMT
• Title: Dynamic Sparse Attention on Mobile SoCs
• Authors: Wangsong Yin, Daliang Xu, Mengwei Xu, Gang Huang, Xuanzhe Liu,
• Abstract summary: This paper presents shadowAttn, a system-algorithm codesigned sparse attention module with minimal reliance on CPU/GPU.<n>The key idea is to hide the overhead of estimating the important tokens with a NPU-based pilot compute.
• Score: 11.250584640139998
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: On-device running Large Language Models (LLMs) is nowadays a critical enabler towards preserving user privacy. We observe that the attention operator falls back from the special-purpose NPU to the general-purpose CPU/GPU because of quantization sensitivity in state-of-the-art frameworks. This fallback results in a degraded user experience and increased complexity in system scheduling. To this end, this paper presents shadowAttn, a system-algorithm codesigned sparse attention module with minimal reliance on CPU/GPU by only sparsely calculating the attention on a tiny portion of tokens. The key idea is to hide the overhead of estimating the important tokens with a NPU-based pilot compute. Further, shadowAttn proposes insightful techniques such as NPU compute graph bucketing, head-wise NPU-CPU/GPU pipeline and per-head fine-grained sparsity ratio to achieve high accuracy and efficiency. shadowAttn delivers the best performance with highly limited CPU/GPU resource; it requires much less CPU/GPU resource to deliver on-par performance of SoTA frameworks.

Related papers

• Optimizing Attention on GPUs by Exploiting GPU Architectural NUMA Effects [3.124720734513875]
  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.
  arXiv  Detail & Related papers  (2025-11-03T23:48:39Z)
• TensorBLEU: Vectorized GPU-based BLEU Score Implementation for Per-Sentence In-Training Evaluation [0.0]
  In this paper, we introduce a novel implementation of the BLEU metric designed from the ground up for this specific use case.<n>Our approach is fully vectorized for GPU-accelerated, per-sentence computation within PyTorch.<n>By creating a compact, batch-specific dictionary of n-grams using texttttorch.unique, our method avoids the prohibitive memory costs of traditional hashing-based vectorization.
  arXiv  Detail & Related papers  (2025-10-07T01:02:46Z)
• ZKProphet: Understanding Performance of Zero-Knowledge Proofs on GPUs [0.7544506709471767]
  Zero-Knowledge Proofs (ZKP) are protocols which construct cryptographic proofs to demonstrate knowledge of a secret input in a computation without revealing any information about the secret.<n>ZKPs enable novel applications in private and verifiable computing such as anonymized cryptocurrencies and blockchain scaling.<n>We find that a systematic characterization of execution bottlenecks in ZKPs, as well as their scalability on modern GPU architectures, is missing in the literature.
  arXiv  Detail & Related papers  (2025-09-17T22:52:19Z)
• HGCA: Hybrid GPU-CPU Attention for Long Context LLM Inference [8.826966369389893]
  We present HGCA, a hybrid CPU- GPU attention mechanism for large language models.<n>We show that HGCA achieves superior scalability, supports longer sequences and larger batch sizes, and outperforms existing sparse attention baselines in both performance and accuracy.<n> Experiments across diverse models and workloads show that HGCA achieves superior scalability, supports longer sequences and larger batch sizes, and outperforms existing sparse attention baselines in both performance and accuracy.
  arXiv  Detail & Related papers  (2025-07-03T20:20:33Z)
• NGPU-LM: GPU-Accelerated N-Gram Language Model for Context-Biasing in Greedy ASR Decoding [54.88765757043535]
  This work rethinks data structures for statistical n-gram language models to enable fast and parallel operations for GPU-optimized inference.<n>Our approach, named NGPU-LM, introduces customizable greedy decoding for all major ASR model types with less than 7% computational overhead.<n>The proposed approach can eliminate more than 50% of the accuracy gap between greedy and beam search for out-of-domain scenarios while avoiding significant slowdown caused by beam search.
  arXiv  Detail & Related papers  (2025-05-28T20:43:10Z)
• FlashBias: Fast Computation of Attention with Bias [70.44379606190569]
  Attention with bias has been widely deployed in vision, language, protein-folding and other advanced scientific models.<n>It disrupts the tightly fused memory-compute pipeline that underlies the speed of accelerators like FlashAttention.<n>This paper presents FlashBias based on the low-rank compressed sensing theory, which can provide fast-exact computation for many widely used attention biases.
  arXiv  Detail & Related papers  (2025-05-17T15:12:50Z)
• KVPR: Efficient LLM Inference with I/O-Aware KV Cache Partial Recomputation [7.204881999658682]
  Key-Value cache is used to store intermediate activations for large language models.<n>The memory required for the KV cache grows rapidly, often exceeding the capacity of GPU memory.<n>Existing methods attempt to address these issues by overlapping GPU computation with I/O or employing CPU-GPU heterogeneous execution.<n>We introduce KVPR, an efficient I/O-aware LLM inference method where the CPU first transfers a partial set of activations.<n> KVPR achieves up to 35.8% lower latency and 46.2% higher throughput during decoding compared to state-of-the-art approaches.
  arXiv  Detail & Related papers  (2024-11-26T04:03:14Z)
• DeepGEMM: Accelerated Ultra Low-Precision Inference on CPU Architectures using Lookup Tables [49.965024476651706]
  DeepGEMM is a lookup table based approach for the execution of ultra low-precision convolutional neural networks on SIMD hardware. Our implementation outperforms corresponding 8-bit integer kernels by up to 1.74x on x86 platforms.
  arXiv  Detail & Related papers  (2023-04-18T15:13:10Z)
• Fluid Batching: Exit-Aware Preemptive Serving of Early-Exit Neural Networks on Edge NPUs [74.83613252825754]
  "smart ecosystems" are being formed where sensing happens concurrently rather than standalone. This is shifting the on-device inference paradigm towards deploying neural processing units (NPUs) at the edge. We propose a novel early-exit scheduling that allows preemption at run time to account for the dynamicity introduced by the arrival and exiting processes.
  arXiv  Detail & Related papers  (2022-09-27T15:04:01Z)
• An Adaptive Device-Edge Co-Inference Framework Based on Soft Actor-Critic [72.35307086274912]
  High-dimension parameter model and large-scale mathematical calculation restrict execution efficiency, especially for Internet of Things (IoT) devices. We propose a new Deep Reinforcement Learning (DRL)-Soft Actor Critic for discrete (SAC-d), which generates the emphexit point, emphexit point, and emphcompressing bits by soft policy iterations. Based on the latency and accuracy aware reward design, such an computation can well adapt to the complex environment like dynamic wireless channel and arbitrary processing, and is capable of supporting the 5G URL
  arXiv  Detail & Related papers  (2022-01-09T09:31:50Z)
• Providing Meaningful Data Summarizations Using Examplar-based Clustering in Industry 4.0 [67.80123919697971]
  We show, that our GPU implementation provides speedups of up to 72x using single-precision and up to 452x using half-precision compared to conventional CPU algorithms. We apply our algorithm to real-world data from injection molding manufacturing processes and discuss how found summaries help with steering this specific process to cut costs and reduce the manufacturing of bad parts.
  arXiv  Detail & Related papers  (2021-05-25T15:55:14Z)
• SpAtten: Efficient Sparse Attention Architecture with Cascade Token and Head Pruning [10.981433334942476]
  We present SpAtten, an efficient algorithm-architecture co-design that leverages token sparsity, head sparsity, and quantization opportunities to reduce the attention computation and memory access. Experiments on 30 benchmarks show that, on average, SpAtten reduces DRAM access by 10.0x with no accuracy loss, and achieves 1.6x, 3.0x, 162x, 347x speedup, and 1,4x, 3.2x, 1193x, 4059x energy savings.
  arXiv  Detail & Related papers  (2020-12-17T18:59:07Z)
• 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.