llama.cpp b11877:Vulkan 为 conv3d 增加更大 tile size
原标题:b11877
AI 摘要
llama.cpp 发布 b11877 版本,主要变更为 Vulkan 后端为 conv3d 增加更大的 tile size,该优化仅适用于 coopmat2 且不需要共享内存存储输出的场景,并仅对较新驱动版本生效。同时新增了一些具有代表性的后端测试,据提交说明这些测试由 codex 参考 sd.cpp 的 Wan VAE 生成。该版本还提供了覆盖 macOS、Linux、Windows、Android 等多平台及 CUDA、ROCm、Vulkan、SYCL、OpenVINO 等后端的预编译包。
正文节选
<details open> vulkan: Add a larger tile size for conv3d (coopmat2 only) (#29920) * vulkan: Add a larger tile size for conv3d (coopmat2 only) This adds a larger tile size, in practice it is only for coopmat2 where we don't need shared memory to store the output (would use more than 64KB). Add some additional representative backend tests. At least, this is what codex came up with from looking at sd.cpp Wan VAE. * only apply large tile size for newer driver versions </details> **Website:**