CYBER WATCH / EQUINOXE INTELLIGENCE

← All articles

b12103: opencl – improve q2_0 GEMV for Adreno (#31577)

llama.cpp Releases · 2026-11-04 16:01:22

This update optimizes q2_0 decoding and small 32‑bit floating‑point matrix multiplications on Adreno 8xx GPUs via OpenCL enhancements.

Read original source ↗