Check nvlink_node instead of xgmi_node in xml.cc (#1407)
It seems like here wants to check xgmi_node instead. If checks node for "nvlink", it will verify the link_info everytime.
If checks node for "xgmi", when get yes answer, it won't need check vsmi topo interface.
[ROCm/rccl commit: f2ee8d9132]
This commit is contained in:
@@ -766,7 +766,11 @@ ncclResult_t ncclTopoGetXmlFromGpu(struct ncclXmlNode* pciNode, uint32_t rocmDev
|
|||||||
NCCLCHECK(xmlGetAttrInt(gpuNode, "arch", &arch.value));
|
NCCLCHECK(xmlGetAttrInt(gpuNode, "arch", &arch.value));
|
||||||
|
|
||||||
struct ncclXmlNode* nvlNode = NULL;
|
struct ncclXmlNode* nvlNode = NULL;
|
||||||
|
#if defined(__HIP_PLATFORM_AMD__) || defined(__HIPCC__)
|
||||||
|
NCCLCHECK(xmlGetSub(gpuNode, "xgmi", &nvlNode));
|
||||||
|
#else
|
||||||
NCCLCHECK(xmlGetSub(gpuNode, "nvlink", &nvlNode));
|
NCCLCHECK(xmlGetSub(gpuNode, "nvlink", &nvlNode));
|
||||||
|
#endif
|
||||||
if (nvlNode == NULL) {
|
if (nvlNode == NULL) {
|
||||||
#if defined(__HIP_PLATFORM_AMD__) || defined(__HIPCC__)
|
#if defined(__HIP_PLATFORM_AMD__) || defined(__HIPCC__)
|
||||||
const char* busId;
|
const char* busId;
|
||||||
|
|||||||
Reference in New Issue
Block a user