llama.cpp b11238：音频编码器改用 ggml_pad_ext 左填充
原标题：b11238
AI 摘要
llama.cpp 发布 b11238 版本，主要改动是让 Parakeet、LFM2-Audio、Granite Speech 和 Gemma 4 音频编码器改用 ggml_pad_ext 在左侧填充，替代此前先右填充再 roll 的做法；DFlash2 也借此将零填充块与前序 token 合并为单个节点。改动后 Gemma 4 音频嵌入保持逐位一致，同时跳过 DFlash2 中只读取填充的 tap。该版本同步提供 macOS、Linux、Windows、Android 等多平台预编译包。
正文节选
<details open> models: pad on the left with ggml_pad_ext (#29567) * models: pad on the left with ggml_pad_ext The Parakeet, LFM2-Audio, Granite Speech and Gemma 4 audio encoders build a left padding as a right pad followed by a roll, and DFlash2 concatenates a zero filled block in front of the previous tokens. ggml_pad_ext does both in one node now that every backend supports a left padding. The Gemma 4 audio embeddings are bit identical. * models: skip the DFlash2 taps that only read paddin