b9255
📦 llama-cppView on GitHub →
✨ 2 features🐛 1 fixes🔧 2 symbols
Summary
This release focuses on reworking the HMX quantized matmul implementation on Hexagon, including updates to dequant logic and removal of non-pipelined versions. It also includes minor platform-specific updates and bug fixes.
✨ New Features
- Updated Snapdragon scripts to bump default ubatch-size to 1K.
- Combined HMX and power and clock settings into a single set_power call on Hexagon.
🐛 Bug Fixes
- Fixed an editconf error on Hexagon.