Function create_backend

smallthinker/tools/rpc/rpc-server.cpp:216–255

}

static ggml_backend_t create_backend(const rpc_server_params & params) {
    ggml_backend_t backend = nullptr;

    if (!params.device.empty()) {
        ggml_backend_dev_t dev = ggml_backend_dev_by_name(params.device.c_str());
        if (dev) {
            backend = ggml_backend_dev_init(dev, nullptr);
            if (!backend) {
                fprintf(stderr, "Failed to create backend for device %s\n", params.device.c_str());
                return nullptr;
            }
        }
    }

    // try to initialize a GPU backend first
    if (!backend) {
        backend = ggml_backend_init_by_type(GGML_BACKEND_DEVICE_TYPE_GPU, nullptr);
    }

    // if there aren't GPU backends fallback to CPU backend
    if (!backend) {
        backend = ggml_backend_init_by_type(GGML_BACKEND_DEVICE_TYPE_CPU, nullptr);
    }

    if (backend) {
        fprintf(stderr, "%s: using %s backend\n", __func__, ggml_backend_name(backend));

        // set the number of threads
        ggml_backend_dev_t dev = ggml_backend_get_device(backend);
        ggml_backend_reg_t reg = dev ? ggml_backend_dev_backend_reg(dev) : nullptr;
        if (reg) {
            auto ggml_backend_set_n_threads_fn = (ggml_backend_set_n_threads_t) ggml_backend_reg_get_proc_address(reg, "ggml_backend_set_n_threads");
            if (ggml_backend_set_n_threads_fn) {
                ggml_backend_set_n_threads_fn(backend, params.n_threads);
            }
        }
    }

    return backend;
}

static void get_backend_memory(ggml_backend_t backend, size_t * free_mem, size_t * total_mem) {
    ggml_backend_dev_t dev = ggml_backend_get_device(backend);