2.20.3-1

Add support for alternating rings, allow for cross-nic rings without
cross-rail communication.
Add support for user buffer registration for network send/recv.
Optimize aggregated operations to better utilize all channels.
Add flattening for BCM PCI gen5 switches.
Add support for inter-node NVLink communication
Add support for port fusion in NET/IB.
Add support for ReduceScatter and AllGather using Collnet.
Update net API to v8.
Fix hang during A2A connection.

src/include/ipcsocket.h

@@ -35,4 +35,7 @@ ncclResult_t ncclIpcSocketGetFd(struct ncclIpcSocket* handle, int* fd);
 ncclResult_t ncclIpcSocketRecvFd(struct ncclIpcSocket *handle, int *fd);
 ncclResult_t ncclIpcSocketSendFd(struct ncclIpcSocket *handle, const int fd, int rank, uint64_t hash);
 
+ncclResult_t ncclIpcSocketSendMsg(ncclIpcSocket *handle, void *hdr, int hdrLen, const int sendFd, int rank, uint64_t hash);
+ncclResult_t ncclIpcSocketRecvMsg(ncclIpcSocket *handle, void *hdr, int hdrLen, int *recvFd);
+
 #endif /* NCCL_IPCSOCKET_H */