llama.cpp b11633 修复 SET_ROWS 多线程缓冲区越界

AI 摘要

llama.cpp 发布 b11633 版本，修复了 ggml-cpu 中 SET_ROWS 操作的转换缓冲区越界问题：该操作原计划以 n_tasks=1 执行，但每个线程都会运行并各自使用 F16 到 F32 转换缓冲区的一行，导致线程号大于 0 的线程在图包含其他多线程算子时写出工作缓冲区之外。修复方式是将 set_rows 移入 n_task = n_threads 分组。该版本同时提供 macOS、Linux、Windows、Android 等多平台预编译包。

正文节选

<details open> ggml-cpu : size set_rows conversion buffer for all threads (#30454) * ggml-cpu : size set_rows conversion buffer for all threads SET_ROWS is planned with n_tasks = 1, but every thread runs it and uses its own row of the F16 -> F32 conversion buffer, so threads > 0 wrote past the work buffer when the graph had another multi-threaded op. * move set_rows to n_task = n_threads group </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/l

发布时间：2026-10-15 11:37

抓取时间：2026-10-15 12:17

来源机构：ggml-org