Direct GPU

We are working on a system in which an FPGA transfers data directly to the GPU using the supplied kernel API from nvidia.
The question is how can the FPGA notifies the GPU that the buffer is ready without going through the CPU.
How can the GPU notify the FPGA that the buffer is ready for the next data submission. We found 2 APIs:
Our understanding is that with these APIs the cuda stream execution can be synchronized between the FPGA and the GPU. However, we couldn’t find those functions in the cuda installation and they don’t show in the documentation. Can anyone please clarify the situation?