llama.cpp 发布: b11047

来源摘要

cuda : fix CUB argsort corruption caused by in-place keys (#28389) argsort_f32_i32_cuda_cub called the one-shot DeviceRadixSort::SortPairs API with d_keys_in == d_keys_out (temp_keys, temp_keys). CUB's internal double-buffer ping-pong requires distinct key buffers: with aliased buffers the sort partially overwrites its own input mid-pass and emits a corrupted permutation, surfacing as intermittent garbage indices (e.g. backend top_k over a 248k-column vocab on Maxwell/CUDA 12.5/CCCL 2.x, which then triggered out-of-bounds gathers in downstream get_rows). Use a distinct keys-out buffer for all six call sites (plain and segmented, ascending and descending, size-query and execute). --------- Co-authored-by: Claude Opus 4.6 Co-authored-by: Oliver Simons **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b11047/llama-b11047-bin-ubuntu-vulkan-x64.tar.gz) - [Ubun