llama.cpp releases b9784 with Hexagon MUL_MAT and MUL_MAT_ID rework

Core summary

One sentence to understand this update

llama.cpp has released version b9784, featuring a rework of MUL_MAT and MUL_MAT_ID on Hexagon, including tiled weight repack and fusion updates.

Impact & opportunity

What this could mean

Developers deploying AI models on Qualcomm Hexagon DSP platforms should examine this update for potential performance gains and memory optimizations.