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

Merge remote-tracking branch 'nccl/master' into 2.8.3

Browse Source 
[ROCm/rccl commit: c985358e11]
This commit is contained in:
Wenkai Du
2021-02-15 19:29:45 +00:00
parent ea2e665834 fc7bdb38a5
commit ab71643c99
23 changed files with 82 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/src/bootstrap.cc
+3 -3
View File
@@ -78,7 +78,7 @@ static ncclResult_t bootstrapNetRecv(int fd, void* data, int size) {
int recvSize;
NCCLCHECK(socketRecv(fd, &recvSize, sizeof(int)));
if (recvSize > size) {
WARN("Message truncated : received %d bytes instead of %d\n", recvSize, size);
WARN("Message truncated : received %d bytes instead of %d", recvSize, size);
return ncclInternalError;
}
NCCLCHECK(socketRecv(fd, data, std::min(recvSize, size)));
@@ -274,7 +274,7 @@ static ncclResult_t remoteAlloc(void** ptr, int fd) {
void* ncclRemoteMemAllocationService(void* args) {
struct remAllocState* state = (struct remAllocState *) args;
if (hipSetDevice(state->cudaDev) != hipSuccess) {
WARN("[Rem Allocator] Failed to set CUDA device %d\n", state->cudaDev);
WARN("[Rem Allocator] Failed to set CUDA device %d", state->cudaDev);
}
// Prepare poll descriptor
@@ -523,7 +523,7 @@ ncclResult_t bootstrapRecv(void* commState, int peer, void* data, int size) {
ncclResult_t bootstrapClose(void* commState) {
struct extState* state = (struct extState*)commState;
if (state->unexpectedConnections != NULL) {
WARN("Unexpected connections are not empty.\n");
WARN("Unexpected connections are not empty");
return ncclInternalError;
}
close(state->extListenFd);