Related papers: GPU-Initiated Networking for NCCL

The Mixture of Experts (MoE) architecture has emerged as a key technique for scaling Large Language Models by activating only a subset of experts per query. Deploying MoE on consumer-grade edge hardware, however, is constrained by limited…

Machine Learning (ML) models execute several parallel computations including Generalized Matrix Multiplication, Convolution, Dropout, etc. These computations are commonly executed on Graphics Processing Units (GPUs), by dividing the…

GPUs are critical for compute-intensive applications, yet emerging workloads such as recommender systems, graph analytics, and data analytics often exceed GPU memory capacity. Existing solutions allow GPUs to use CPU DRAM or SSDs as…

Continuous learning (CL) has emerged as one of the most popular deep learning paradigms deployed in modern cloud GPUs. Specifically, CL has the capability to continuously update the model parameters (through model retraining) and use the…

Mixture-of-experts (MoE) architectures enable trillion-parameter LLMs with sparsely activated experts. Expert parallelism (EP) is a widely adopted MoE training strategy, but it suffers from severe all-to-all communication bottlenecks, which…

Full batch training of Graph Convolutional Network (GCN) models is not feasible on a single GPU for large graphs containing tens of millions of vertices or more. Recent work has shown that, for the graphs used in the machine learning…

Modern LLM serving is increasingly serverless in shape: large model catalogs, long-tail invocations, and multi-tenant demand. Existing GPU serving systems face a tradeoff: dedicated-GPU allocation wastes scarce HBM under sparse traffic,…

Fast-evolving machine learning (ML) workloads have increasing requirements for networking. However, host network transport on RDMA NICs is hard to evolve, causing problems for ML workloads. For example, single-path RDMA traffic is prone to…

This paper presents MoE-Gen, a high-throughput MoE inference system optimized for single-GPU execution. Existing inference systems rely on model-based or continuous batching strategies, originally designed for interactive inference, which…

Modern mobile devices are equipped with high-performance hardware resources such as graphics processing units (GPUs), making the end-side intelligent services more feasible. Even recently, specialized silicons as neural engines are being…

We present a holistic design for GPU-accelerated computation in TrustZone TEE. Without pulling the complex GPU software stack into the TEE, we follow a simple approach: record the CPU/GPU interactions ahead of time, and replay the…

As large language models continue to scale, training demands on compute and system capacity grow rapidly, making single-vendor homogeneous clusters insufficient. This paper presents a technical solution for heterogeneous mixed training in…

There is growing interest in accelerating irregular data-parallel algorithms on GPUs. These algorithms are typically blocking, so they require fair scheduling. But GPU programming models (e.g.\ OpenCL) do not mandate fair scheduling, and…

Effective intra-node GPU communication is essential for optimizing performance in MPI-based HPC applications, especially when leveraging multiple communication paths. In this study, we propose a novel approach that integrates CUDA Graphs…

Modern multi-tenant AI clusters are increasingly communication-bound, driven by high-volume and multi-round GPU-to-GPU collective communication. Consequently, the GPU dispatcher's choice of a physical GPU subset for each tenant largely…

The increasing size of input graphs for graph neural networks (GNNs) highlights the demand for using multi-GPU platforms. However, existing multi-GPU GNN systems optimize the computation and communication individually based on the…

Deep learning training is an expensive process that extensively uses GPUs, but not all model training saturates modern powerful GPUs. Multi-Instance GPU (MIG) is a new technology introduced by NVIDIA that can partition a GPU to better-fit…

In the era of LLMs, dense operations such as GEMM and MHA are critical components. These operations are well-suited for parallel execution using a tilebased approach. While traditional GPU programming often relies on low level interfaces…

Fine-grained, per-micro-batch load balancing is essential for efficient Mixture-of-Experts (MoE) training, yet every prior dynamic scheduling scheme pays for it with extra communication that is hard to hide. Especially on modern…

It is commonly assumed that the end-to-end networking performance of edge offloading is purely dictated by that of the network connectivity between end devices and edge computing facilities, where ongoing innovation in 5G/6G networking can…