Baike.dev
All toolsAI codingTrendingOpen sourceNewsSubmit
Log in
Back to tool/Back to issues
#6981·ncnn

feature plan

Author: nihuiCreated Sep 9, 2026Updated Sep 15, 2026
Labelsenhancement
  • block quantization tooling
  • vulkan int8 block quantization
  • cpu flash attention and persistent kv-cache for arm etc.
  • opaque/linear kvcache layout
  • arm sve infrastructure https://github.com/Tencent/ncnn/pull/6955
  • qcom vulkan cm opt
  • vulkan cm maintenance1 opt

Source: Tencent/ncnn

View original on GitHubView discussion on GitHub