llama.cpp

ggml-cuda: auto apply iGPU flag for CUDA/HIP if integrated device

#23007

Merged