Fine-grained Computation-Communication Overlap via Tile-level Signaling and Scheduling for Mixture-of-Experts

• URL: http://arxiv.org/abs/2607.19539v1
• Date: Tue, 21 Jul 2026 19:40:06 GMT
• Title: Fine-grained Computation-Communication Overlap via Tile-level Signaling and Scheduling for Mixture-of-Experts
• Abstract summary: Mixture-of-Experts (MoE) architectures increase model capacity without proportionally increasing computation cost.<n>We present a fine-grained approach that overlaps expert compute with the second all-to-all via tile-level signaling and scheduling.<n>Our approach achieves up to 2.64x end-to-end speedup and 2.74x MoE-layer speedup on a 4-A100 GPU platform.
• Score: 0.9940728137241214
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Mixture-of-Experts (MoE) architectures increase model capacity without proportionally increasing computation cost and have become a key building block for scaling large language models (LLMs) to trillion-parameter regimes. Efficient deployment of these MoE models relies on distributed execution across multiple GPUs, where each MoE layer involves two all-to-all communications: dispatching tokens to expert ranks and returning the expert outputs to their source ranks. Conventional MoE implementations launch this return all-to-all after expert compute completes, exposing communication latency on the critical path and reducing GPU utilization. We present a fine-grained approach that overlaps expert compute with the second all-to-all via tile-level signaling and scheduling. Our producer-consumer co-design combines: (1) a persistent per-rank computation kernel (producer) that covers all local experts on the rank to eliminate repeated kernel launch overhead and prioritizes remote-critical tiles, and (2) a persistent communication kernel (consumer) on a small dedicated partition of streaming multiprocessors (SMs) that issues segment-granular transfers as tiles become ready. Our co-design avoids intrusive changes to the underlying computation operators or communication primitives, making it practical for improving distributed MoE execution efficiency on multi-GPU systems. On a 4-A100 GPU platform, evaluated on three MoE models against four state-of-the-art MoE systems, our approach achieves up to 2.64x end-to-end speedup and 2.74x MoE-layer speedup. Compared with a conventional non-overlap baseline, our approach consistently improves both operator- and MoE-layer-level performance across varying GEMM shapes, router modes, and a broad range of producer/consumer SM partitions, while preserving correctness.

Related papers

• Accelerating Unified Multimodal Models with Core-Expansion Routing and Unified Computation Scheduling [61.28599393707943]
  Unified multimodal models jointly support understanding and generation, but incur substantial redundant computation across tokens, layers, and generation timesteps.<n>We propose a task-shared core scorer and progress-conditioned generation expansions, optimized through generation decomposition and cross-task core alignment.<n> Experiments on two representative UMM architectures demonstrate consistent quality--efficiency improvements across both tasks, retaining 98.03% of dense understanding performance with a 1.93times$ end-to-end inference speedup.
  arXiv  Detail & Related papers  (2026-08-29T14:27:55Z)
• DisagMoE: Computation-Communication overlapped MoE Training via Disaggregated AF-Pipe Parallelism [26.73062475753107]
  We present DisagMoE, a training system that jointly optimize model placement and scheduling for maximal efficiency.<n>DisagMoE improves training efficiency across multiple MoE models with up to 1.8x speedup on 16-node 8xH800 clusters.
  arXiv  Detail & Related papers  (2026-05-10T05:57:05Z)
• Piper: Efficient Large-Scale MoE Training via Resource Modeling and Pipelined Hybrid Parallelism [1.0130502379786543]
  Training MoE models on HPC platforms is hindered by large memory footprints, frequent large-scale communication across heterogeneous networks, and severe workload imbalance.<n>We develop a mathematical model that quantifies memory, compute, and communication requirements for MoE configurations under various parallelization schemes.<n>We introduce Piper, a framework that leverages resource modeling to identify efficient training strategies for MoE models on target HPC platforms.
  arXiv  Detail & Related papers  (2026-05-06T15:47:14Z)
• 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)
• MOBIUS: Big-to-Mobile Universal Instance Segmentation via Multi-modal Bottleneck Fusion and Calibrated Decoder Pruning [91.90342432541138]
  Scaling up model size and training data has advanced foundation models for instance-level perception.<n>High computational cost limits adoption on resource-constrained platforms.<n>We introduce a new benchmark for efficient segmentation on both high-performance computing platforms and mobile devices.
  arXiv  Detail & Related papers  (2025-10-16T18:00:00Z)
• Comet: Fine-grained Computation-communication Overlapping for Mixture-of-Experts [8.80408909878008]
  Mixture-of-experts (MoE) has been extensively employed to scale large language models to trillion-plus parameters.<n>Existing methods suggest the communication in a MoE layer to be pipelined with the computation for overlapping.<n>We present COMET, an optimized MoE system with fine-grained communication-computation overlapping.
  arXiv  Detail & Related papers  (2025-02-27T06:36:45Z)
• Read-ME: Refactorizing LLMs as Router-Decoupled Mixture of Experts with System Co-Design [59.00758127310582]
  We propose a novel framework Read-ME that transforms pre-trained dense LLMs into smaller MoE models. Our approach employs activation sparsity to extract experts. Read-ME outperforms other popular open-source dense models of similar scales.
  arXiv  Detail & Related papers  (2024-10-24T19:48:51Z)
• FusionLLM: A Decentralized LLM Training System on Geo-distributed GPUs with Adaptive Compression [55.992528247880685]
  Decentralized training faces significant challenges regarding system design and efficiency. We present FusionLLM, a decentralized training system designed and implemented for training large deep neural networks (DNNs) We show that our system and method can achieve 1.45 - 9.39x speedup compared to baseline methods while ensuring convergence.
  arXiv  Detail & Related papers  (2024-10-16T16:13:19Z)
• EPS-MoE: Expert Pipeline Scheduler for Cost-Efficient MoE Inference [49.94169109038806]
  This paper introduces EPS-MoE, a novel expert pipeline scheduler for MoE that surpasses the existing parallelism schemes.<n>Our results demonstrate at most 52.4% improvement in prefill throughput compared to existing parallel inference methods.
  arXiv  Detail & Related papers  (2024-10-16T05:17:49Z)
• Shortcut-connected Expert Parallelism for Accelerating Mixture-of-Experts [4.629608387540524]
  ScMoE is a novel shortcut-connected MoE architecture integrated with an overlapping parallelization strategy.<n>Compared to the prevalent top-2 MoE baseline, ScMoE achieves speedups of 1.49 times in training and 1.82 times in inference.
  arXiv  Detail & Related papers  (2024-04-07T17:17:23Z)

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.