llama.cpp Release b11716 Adds Hexagon DSP Support for Q1_0 and Q2_0

The llama.cpp project released version b11716, introducing native support for Q1_0 and Q2_0 quantization formats on the Qualcomm Hexagon DSP backend. This update includes optimized vector dot products, non-square tiling adjustments, and layout refinements specifically tailored for mobile hardware acceleration. This release enables extreme low-bit model compression to run directly on mobile and edge devices utilizing Qualcomm Snapdragon hardware. By improving execution paths on Hexagon digital signal processors, it helps run large language models more efficiently under tight power and memory constraints. The update enables Q1_0 weights in the 2D HMX matrix multiplication path and adds the Q2_0 format with generalized tiling and reduced unpacking overhead. Developers can deploy these low-bit configurations using updated Linux and Android Snapdragon packages.