feat(topology): add GPU entries in topology discovery

[staryxchen]

Summary

This PR adds backward compatibility for GPU memory naming by including both "cuda:xx" and "gpu:xx" entries during topology discovery.

Background

In earlier versions of mooncake, the GPU0 memory was referred to as "gpu:0" rather than "cuda:0," so some users would refer to the memory as "gpu:0" when registering it. To maintain compatibility, this PR adds both "cuda:xx" and "gpu:xx" entries during topology discovery.

Changes

• topology.cpp: Modified discoverCudaTopology() function to add "gpu:xx" entries alongside existing "cuda:xx" entries
• transfer_engine.cpp: Output topology information for diagnosing anomalies

Impact

• Maintains backward compatibility with existing code that uses "gpu:xx" naming convention
• Allows users to register memory using either "cuda:xx" or "gpu:xx" format
• No breaking changes to existing functionality

[stmatengss]

Just wonder what the difference is between cuda:x and gpu:x, it seems the same thing in torch.

[staryxchen]

Yes, there is no difference for the Torch API. However, now, in Mooncake, we select the device to use based on the segment buffer name, which is a string comparison. Therefore, if the user registers the memory and declares it as “gpu:”, Mooncake will fall back to using the device corresponding to kWildcardLocation because it cannot find a matching string.

[doujiang24]

nope, I do think people should use cuda: instead of gpu:, see document here: https://github.com/kvcache-ai/Mooncake/blob/main/doc/zh/transfer-engine.md#transferengineregisterlocalmemory

[alogfans]

Changing the name may lead compatibility problem after merging it.

[alogfans]

In other words, you can check whether there is any hard-coded "cuda:" in elsewhere.

[staryxchen]

In other words, you can check whether there is any hard-coded "cuda:" in elsewhere.

mooncake internal (location and topology) uniformly uses cuda:xx.

[stmatengss]

I will review it later. @starychen