b9458
📦 llama-cppView on GitHub →
✨ 1 features🐛 1 fixes🔧 1 symbols
Summary
This release refactors Vulkan pipeline compilation to improve concurrency by reducing the time the device mutex is held. It also cleans up internal code by removing the 'needed' variable.
✨ New Features
- Optimized Vulkan pipeline compilation by decoupling the device mutex usage; the lock is now only held during pipeline traversal and lazy initialization, not during the compilation phase itself.
🐛 Bug Fixes
- Removed the unused 'needed' variable/concept.