https://lilinji.github.io/tags/%E6%80%A7%E8%83%BD%E8%B0%83%E4%BC%98/Recent content in 性能调优 on Ringi's LogHugo -- 0.152.2zh-cnMon, 31 Aug 2026 00:00:00 +0800https://lilinji.github.io/2026/08/%E7%AC%AC11%E8%AE%B2gpu-%E5%88%A9%E7%94%A8%E7%8E%87-99-%E7%9A%84%E5%8D%A1%E4%B8%BA%E4%BB%80%E4%B9%88%E8%AE%AD%E7%BB%83%E8%BF%98%E6%98%AF%E8%B7%91%E4%B8%8D%E5%BF%AB%E4%BB%8E-kineto-%E5%BA%95%E5%B1%82%E6%8E%A2%E9%92%88trace-%E6%97%B6%E9%97%B4%E7%BA%BF%E5%9B%9B%E9%87%8D%E7%97%85%E7%81%B6%E5%88%B0-pytorch-profiler-%E6%80%A7%E8%83%BD%E4%BD%93%E6%A3%80%E5%AE%9E%E6%88%98/Mon, 31 Aug 2026 00:00:00 +0800https://lilinji.github.io/2026/08/%E7%AC%AC11%E8%AE%B2gpu-%E5%88%A9%E7%94%A8%E7%8E%87-99-%E7%9A%84%E5%8D%A1%E4%B8%BA%E4%BB%80%E4%B9%88%E8%AE%AD%E7%BB%83%E8%BF%98%E6%98%AF%E8%B7%91%E4%B8%8D%E5%BF%AB%E4%BB%8E-kineto-%E5%BA%95%E5%B1%82%E6%8E%A2%E9%92%88trace-%E6%97%B6%E9%97%B4%E7%BA%BF%E5%9B%9B%E9%87%8D%E7%97%85%E7%81%B6%E5%88%B0-pytorch-profiler-%E6%80%A7%E8%83%BD%E4%BD%93%E6%A3%80%E5%AE%9E%E6%88%98/建立大厂 AI Infra 性能诊断第一性原理：以生活直觉与极简小算盘拆穿 GPU 利用率假象，从 Kineto/CUPTI 底层硬件时间戳探针、四阶 Schedule 采集周期、Chrome Trace / Perfetto 时间线四大病灶排障，到 TensorBoard 性能看板与 4 大生产级性能体检实战。