ggml-impl.h: Fix build issues on AArch64 with CUDA version 12#12872

zhengjun-xing wants to merge 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

zhengjun-xing:fix-build-aarch64-cuda12zhengjun-xing/llama.cpp:fix-build-aarch64-cuda12Copy head branch name to clipboard