<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · TensorRT Global Performance Tuner</title>
  <link>https://derekwang2002.github.io/ai-intelligence/projects/tensorrt-global-performance-tuner/</link>
  <description>AI 模型、Agent、工程与基础设施的重要变化，附一手来源与建议等级。</description>
  <language>zh-CN</language>
  <lastBuildDate>Mon, 02 Nov 2026 16:56:00 GMT</lastBuildDate>
  <atom:link href="https://derekwang2002.github.io/ai-intelligence/feeds/projects/tensorrt-global-performance-tuner.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>NVIDIA 开放 TensorRT 精度约束调优</title>
    <link>https://derekwang2002.github.io/ai-intelligence/events/ev-20261023-03/</link>
    <guid isPermaLink="false">ev-20261023-03-published</guid>
    <pubDate>Sat, 24 Oct 2026 15:59:49 GMT</pubDate>
    <category>新事件</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;NVIDIA 在 TensorRT 11.1 发布 Global Performance Tuner，并开放 trtexec 实现。它把精度设为约束，在不同 build route 中搜索 200 多项编译选项。NVIDIA 报告十个工作负载获得 1.14 至 1.63 倍提升，尚无独立复现。&lt;/p&gt;&lt;p&gt;建议等级：TRIAL&lt;/p&gt;&lt;p&gt;来源：&lt;a href=&quot;https://developer.nvidia.com/blog/scaling-nvidia-tensorrt-performance-optimization-beyond-a-single-build-route-with-the-global-performance-tuner-part-1/&quot;&gt;TensorRT Global Performance Tuner, Part 1&lt;/a&gt; · &lt;a href=&quot;https://developer.nvidia.com/blog/scaling-nvidia-tensorrt-performance-optimization-beyond-a-single-build-route-with-the-global-performance-tuner-part-2/&quot;&gt;TensorRT Global Performance Tuner, Part 2&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
