llama.cpp b12064：Hexagon 新增分块 HVX 转置拷贝

AI 摘要

llama.cpp 发布 b12064 版本，为 Hexagon NPU 后端新增分块 HVX 转置路径，用于处理带置换的 CONT/CPY 拷贝。此前这类拷贝会退化为逐元素 DMA 描述符，速度仅约 45 MB/s；新内核通过 2D DMA、vgather 转置和双缓冲提升性能，并补充了 f32/f16 多种尺寸的测试。该改动主要优化视频 VAE 中通道优先与通道末位之间的置换拷贝。

正文节选

<details open> hexagon: tiled hvx permute (#31446) * hexagon: add tiled HVX transpose path for permuted CONT/CPY Permuted copies where dst dim 0 is strided in src, but some other src dim is contiguous, used to fall back to HTP_COPY_KERNEL_RESHAPE. That issues one DMA descriptor per element. The channel-first <-> channel-last permutes around per-channel norms (e.g. [W,H,T,C] <-> [C,W,H,T] in video VAEs) ht this path, and these copies ran at about 45 MB/s. Add HTP_COPY_KERNEL_TRANSPOSE for sam

发布时间：2026-11-03 13:42

抓取时间：2026-11-03 14:42

来源机构：ggml-org