Change8

b9265

📦 llama-cppView on GitHub →
1 features🐛 1 fixes🔧 2 symbols

Summary

This release focuses on internal optimizations and fixes for the SSM-CONV backend, including better handling for large prompts and resolving an issue in hex-rope related to cache initialization.

Migration Steps

  1. If you rely on the previous implementation details of ssm-conv, note that gathers have been removed and VTCM handling has been improved.

✨ New Features

  • Improved handling of large prompts in hexagon:ssm-conv.

🐛 Bug Fixes

  • Fixed an issue where rope_cache_init was inlined, causing breakage after rebasing with SSM_CONV changes in hex-rope.

Affected Symbols