<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>GPUDirect on PlumePHP</title><link>https://plumephp.com/tags/gpudirect/</link><description>Recent content in GPUDirect on PlumePHP</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Mon, 28 Sep 2026 15:00:00 +0800</lastBuildDate><atom:link href="https://plumephp.com/tags/gpudirect/index.xml" rel="self" type="application/rss+xml"/><item><title>NCCL 集合通信：AllReduce、Ring 算法与多卡推理的通信优化</title><link>https://plumephp.com/ai-collective-communication-nccl/</link><pubDate>Mon, 28 Sep 2026 15:00:00 +0800</pubDate><guid>https://plumephp.com/ai-collective-communication-nccl/</guid><description>&lt;p&gt;张量并行把 70B 模型拆到 8 张卡，但「拆」带来的通信开销可能把提速全吃掉——&lt;strong&gt;每层 forward 都要跨卡同步&lt;/strong&gt;。NCCL（NVIDIA Collective Communications Library）就是这场跨卡通信的引擎：AllReduce 归约梯度、AllGather 广播权重、Ring 算法摊平带宽、NVLink/RDMA 打通互连。本文把 NCCL 讲透：有哪些原语、算法怎么收敛、拓扑怎么感知、多卡推理怎么优化通信，以及「通信到底花多少钱、值不值」。&lt;/p&gt;</description></item></channel></rss>