gpu support of llama.cpp for Vivante, through opencl

[Mallakbasheer]

I am working on deploying a small language model on the edge itself. Our device is equipped with Vivante GC7000UL gpu and npu. GPU can be accessed using opencl, and Npu using openvx. Gpu access through Vulkan is not supported for compute pipeline, got to know this from the manufacturer. Is there anyway, I can run llama.cpp either on gpu or npu?
Thanks

[jeffbolznv]

Gpu access through Vulkan is not supported for compute pipeline

This is an odd way to phrase it, there's no way to have a Vulkan implementation without compute pipeline support.