Some issues regarding super-node communication

[EmmaLee-c]

Hello, I have two questions to ask. First, for example, in a supernode environment, I execute osu_bw -d cuda D D for inter-node communication. However, the data buffer in osu_bw is allocated using cudaMalloc instead of using the cuMem* related interfaces. In this case, can UCX still use NVLink for communication instead of using RC for inter-node communication?
Second, I checked the current UCX version, and it seems that only CUDA supports supernode communication, while ROCM does not support it yet. Is there any plan to support supernode communication with ROCM in the future?
Thank you.