AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Improve robustness of PCI detection

Browse Source 
Fallback to default values when class/speed is unknown.


[ROCm/rccl commit: 23a9fbb788]
This commit is contained in:
Sylvain Jeaugey
2020-04-16 14:27:50 -07:00
parent acd868dfcb
commit 5df2502deb
2 changed files with 5 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/src/graph/topo.cc
+2 -4
View File
@@ -303,8 +303,8 @@ ncclResult_t ncclTopoAddGpu(struct ncclXmlNode* xmlGpu, struct ncclTopoSystem* s
return ncclSuccess; return ncclSuccess;
} }
struct kvDict kvDictPciClass[] = { { "0x060400", PCI }, { "0x068000", NVS }, { "0x068001", CPU }, { "0x03", GPU }, { "0x02", NIC }, { NULL, 0 } }; struct kvDict kvDictPciClass[] = { { "0x060400", PCI }, { "0x068000", NVS }, { "0x068001", CPU }, { "0x03", GPU }, { "0x02", NIC }, { NULL, PCI /* Default fallback value */ } };
struct kvDict kvDictPciGen[] = { { "2.5 GT/s", 15 }, { "5 GT/s", 30 }, { "8 GT/s", 60 }, { "16 GT/s", 120 }, { NULL, 0 } }; // x100 Mbps per lane struct kvDict kvDictPciGen[] = { { "2.5 GT/s", 15 }, { "5 GT/s", 30 }, { "8 GT/s", 60 }, { "16 GT/s", 120 }, { NULL, 60 /* Default fallback */ } }; // x100 Mbps per lane
ncclResult_t ncclTopoAddPci(struct ncclXmlNode* xmlPci, struct ncclTopoSystem* system, struct ncclTopoNode* parent) { ncclResult_t ncclTopoAddPci(struct ncclXmlNode* xmlPci, struct ncclTopoSystem* system, struct ncclTopoNode* parent) {
const char* str; const char* str;
@@ -356,8 +356,6 @@ ncclResult_t ncclTopoAddPci(struct ncclXmlNode* xmlPci, struct ncclTopoSystem* s
// Manage cases where speed was not indicated in /sys // Manage cases where speed was not indicated in /sys
if (width == 0) width = 16; if (width == 0) width = 16;
if (strlen(str) == 0 || strcasecmp(str, "Unknown speed") == 0) str = "8 GT/s";
NCCLCHECK(kvConvertToInt(str, &speed, kvDictPciGen)); // Values in 100Mbps, per lane (we want GB/s in the end) NCCLCHECK(kvConvertToInt(str, &speed, kvDictPciGen)); // Values in 100Mbps, per lane (we want GB/s in the end)
NCCLCHECK(ncclTopoConnectNodes(node, parent, LINK_PCI, width*speed/80.0)); NCCLCHECK(ncclTopoConnectNodes(node, parent, LINK_PCI, width*speed/80.0));
projects/rccl/src/graph/xml.h
+3 -2
View File
@@ -218,8 +218,9 @@ static ncclResult_t kvConvertToInt(const char* str, int* value, struct kvDict* d
} }
d++; d++;
} }
WARN("KV Convert to int : could not find value of '%s' in dictionary", str); INFO(NCCL_GRAPH, "KV Convert to int : could not find value of '%s' in dictionary, falling back to %d", str, d->value);
return ncclInternalError; *value = d->value;
return ncclSuccess;
} }
static ncclResult_t kvConvertToStr(int value, const char** str, struct kvDict* dict) { static ncclResult_t kvConvertToStr(int value, const char** str, struct kvDict* dict) {
struct kvDict* d = dict; struct kvDict* d = dict;