Fix XGMI detection (#699)

* Fix XGMI detection

* Increase stack size

* Temporarily disable signal hangler in CI

[Process: 17281] Inside handler function signal: Segmentation fault (11)
BFD: DWARF error: section .debug_info is larger than its filesize! (0x93ef57 vs 0x530ea0)
BFD: DWARF error: section .debug_info is larger than its filesize! (0x93ef57 vs 0x530ea0)
Cette révision appartient à :
Wenkai Du a créé et GitHub a révisé 2023-03-08 14:08:07 -08:00
1 Parent 79a2031951
révision 22b81fbaae
3 fichiers modifiés
+3 -3

Aucun fichier ne correspond à votre recherche

+1 -1
Voir le fichier
@@ -1187,7 +1187,7 @@ ncclResult_t ncclTopoGetLinkType(struct ncclTopoSystem* system, int cudaDev1, in
struct ncclTopoNode* remNode = link->remNode;
if (remNode->gpu.dev == cudaDev2) {
*isXGMI = (link->type == LINK_NVL);
return ncclSuccess;
if (*isXGMI) return ncclSuccess;
}
}
}
+1 -1
Voir le fichier
@@ -1431,7 +1431,7 @@ fail:
#ifdef USE_INDIRECT_FUNCTION_CALL
NCCL_PARAM(SetStackSize, "SET_STACK_SIZE", 1);
RCCL_PARAM(StackSizeOverride, "STACK_SIZE_OVERRIDE", 256);
RCCL_PARAM(StackSizeOverride, "STACK_SIZE_OVERRIDE", 512);
#else
NCCL_PARAM(SetStackSize, "SET_STACK_SIZE", 0);
RCCL_PARAM(StackSizeOverride, "STACK_SIZE_OVERRIDE", 0);