b8969
📦 llama-cppView on GitHub →
✨ 1 features🔧 2 symbols
Summary
This release introduces SVE tuned code for a specific matrix multiplication kernel in ggml and includes a minor code style change in repack.cpp.
Migration Steps
- Change arrays to static const in repack.cpp.
✨ New Features
- Added SVE tuned code for the gemm_q8_0_4x8_q8_0() kernel in ggml.