无法注册GPU算子 | Unable to register the GPU operator

[Ryanwjx]

我试图运行示例代码rknn_custom_gpu_op_demo，但是卡在rknn_init。

uint32_t flag = 0;
flag          = flag | RKNN_FLAG_EXECUTE_FALLBACK_PRIOR_DEVICE_GPU;
int ret = rknn_init(&ctx, model_path, 0, flag, NULL);  //卡住！！！

如果不设置FLAG，而是通过先加载模型，再注册GPU算子，则卡在注册算子：

size_t model_size;
void* model_data = load_file(model_path, &model_size);
int ret = rknn_init(&ctx, model_data, model_size, 0, NULL);
free(model_data);
auto gpu_user_op= get_rknn_custom_op();
ret = rknn_register_custom_ops(ctx, gpu_user_op, 1);   //卡住！！！

---

I tried to run the sample coderknn_custom_gpu_op_demo, but stuck in rknn_init.

uint32_t flag = 0;
flag          = flag | RKNN_FLAG_EXECUTE_FALLBACK_PRIOR_DEVICE_GPU;
int ret = rknn_init(&ctx, model_path, 0, flag, NULL);   // Stuck!!

If you do not set FLAG, but load the model first and then register the GPU operator, the registration operator is stuck:

size_t model_size;
void* model_data = load_file(model_path, &model_size);
int ret = rknn_init(&ctx, model_data, model_size, 0, NULL);
free(model_data);
auto gpu_user_op= get_rknn_custom_op();
ret = rknn_register_custom_ops(ctx, gpu_user_op, 1);   // Stuck!!