<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>AI-Accelerator on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/ai-accelerator/</link><description>Recent content in AI-Accelerator on 兜忆轩 | Cedric's Compiler Notes</description><image><title>兜忆轩 | Cedric's Compiler Notes</title><url>https://douyixuan.github.io/images/papermod-cover.png</url><link>https://douyixuan.github.io/images/papermod-cover.png</link></image><generator>Hugo -- 0.146.0</generator><language>en-us</language><lastBuildDate>Wed, 21 Oct 2026 12:04:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/ai-accelerator/index.xml" rel="self" type="application/rss+xml"/><item><title>036 · 跨抽象层更快，是选对表示还是扩大了搜索？</title><link>https://douyixuan.github.io/posts/ai-silicon-papers/2026-10-21/</link><pubDate>Wed, 21 Oct 2026 12:04:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-silicon-papers/2026-10-21/</guid><description>TileCas 让 kernel Agent 在高层 DSL 与低层原生代码之间切换，并保留可回访的候选分支。它用随机路由和同检查点干预说明：收益不只来自多一种表示，还来自何时切换、从哪个历史候选继续。</description></item></channel></rss>