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.


[ROCm/rccl commit: b6475625fb]
Dieser Commit ist enthalten in:
Sylvain Jeaugey
2024-02-05 05:06:02 -08:00
Ursprung 382ae08419
Commit 2ab8a3a750
74 geänderte Dateien mit 4632 neuen und 2165 gelöschten Zeilen
+2 -2
Datei anzeigen
@@ -132,7 +132,7 @@ ncclResult_t ncclIpcSocketRecvMsg(ncclIpcSocket *handle, void *hdr, int hdrLen,
WARN("UDS: Receiving data over socket failed : %d", errno);
return ncclSystemError;
}
if (handle->abortFlag && *handle->abortFlag) return ncclInternalError;
if (handle->abortFlag && __atomic_load_n(handle->abortFlag, __ATOMIC_RELAXED)) return ncclInternalError;
}
if (recvFd != NULL) {
@@ -221,7 +221,7 @@ ncclResult_t ncclIpcSocketSendMsg(ncclIpcSocket *handle, void *hdr, int hdrLen,
WARN("UDS: Sending data over socket %s failed : %s (%d)", temp, strerror(errno), errno);
return ncclSystemError;
}
if (handle->abortFlag && *handle->abortFlag) return ncclInternalError;
if (handle->abortFlag && __atomic_load_n(handle->abortFlag, __ATOMIC_RELAXED)) return ncclInternalError;
}
return ncclSuccess;