2.12.10-1

Fix bug with CollNet
Fix bug with zero-bytes send/recv operations
Fix NCCL_PARAM implementation to avoid taking a lock on every call
Fix bug when setting NCCL_IB_QPS_PER_CONNECTION to more than one.
Improve error reporting for network errors.


[ROCm/rccl commit: 353e8ba446]
此提交包含在:
Sylvain Jeaugey
2022-03-30 02:25:49 -07:00
父節點 7707479804
當前提交 27130280b2
共有 16 個檔案被更改,包括 203 行新增 和 114 行删除
+1 -1
查看文件
@@ -44,7 +44,7 @@ struct ncclSocket {
enum ncclSocketState state;
};
const char *ncclSocketToString(union ncclSocketAddress *addr, char *buf);
const char *ncclSocketToString(union ncclSocketAddress *addr, char *buf, const int numericHostForm = 1);
ncclResult_t ncclGetSocketAddrFromString(union ncclSocketAddress* ua, const char* ip_port_pair);
int ncclFindInterfaceMatchSubnet(char* ifNames, union ncclSocketAddress* localAddrs, union ncclSocketAddress* remoteAddr, int ifNameMaxSize, int maxIfs);
int ncclFindInterfaces(char* ifNames, union ncclSocketAddress *ifAddrs, int ifNameMaxSize, int maxIfs);