Someone should tell Qualcomm then: https://www.qualcomm.com/developer/blog/2024/11/introducing-...

I believe the llama.cpp Vulkan backend is inoperable on Adreno GPUs