b8226
📦 llama-cppView on GitHub →
✨ 5 features🐛 1 fixes🔧 3 symbols
Summary
This release focuses heavily on backend improvements for the hexagon architecture, specifically adding and refining the f32 ssm_conv operation and its HVX kernel implementation. It also includes various pre-compiled binaries for broad platform support.
✨ New Features
- Added f32 ssm_conv operation support for hexagon backend.
- Improved the functionality and performance of the ssm-conv HVX kernel on hexagon.
- Added DMA support to the ssm-conv HVX kernel.
- Implemented dynamic scratchpad computation for ssm-conv on hexagon.
- Addressed local context issues and fixed spad indexing within the hex-ssm-conv implementation.
🐛 Bug Fixes
- Fixed various issues in hex-ssm-conv, including spad indexing.