b8459
📦 llama-cppView on GitHub →
✨ 1 features🔧 1 symbols
Summary
This release optimizes performance for PPC architectures by applying explicit inlining to tinyBLAS accumulator operations, preventing unnecessary stack spills. It also provides a comprehensive set of pre-built binaries for numerous operating systems and hardware configurations.
✨ New Features
- Improved performance on PPC architectures by ensuring MMA accumulator disassembly stays within the kernel's register context via explicit inlining.