llama.cpp

ggml : extend ggml_can_fuse to work with non-sequential nodes

#16123

Merged