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

Merge pull request #685 from ROCmSoftwarePlatform/2.16.5

Browse Source 
Sync up to NCCL 2.16.5

[ROCm/rccl commit: d601c4909c]
This commit is contained in:
Wenkai Du
2023-02-22 10:29:02 -08:00
committed by GitHub
parent 80ce7507c7 0c7a94c462
commit 8aaf6f8e0d
11 changed files with 61 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/src/misc/param.cc
+1 -1
View File
@@ -72,7 +72,7 @@ void ncclLoadParam(char const* env, int64_t deftVal, int64_t uninitialized, int6
value = deftVal;
INFO(NCCL_ALL,"Invalid value %s for %s, using default %lld.", str, env, (long long)deftVal);
} else {
INFO(NCCL_ALL,"%s set by environment to %lld.", env, (long long)value);
INFO(NCCL_ENV,"%s set by environment to %lld.", env, (long long)value);
}
}
__atomic_store_n(cache, value, __ATOMIC_RELAXED);
projects/rccl/src/misc/shmutils.cc
+1 -1
View File
@@ -54,7 +54,7 @@ ncclResult_t ncclShmOpen(char* shmPath, size_t shmSize, void** shmPtr, void** de
const size_t realShmSize = shmSize + refSize;
*handle = *shmPtr = NULL; /* assume shmPtr and handle always set correctly by users. */
EQCHECKGOTO(tmphandle = (struct shmHandleInternal*)malloc(sizeof(struct shmHandleInternal)), NULL, ret, fail);
EQCHECKGOTO(tmphandle = (struct shmHandleInternal*)calloc(1, sizeof(struct shmHandleInternal)), NULL, ret, fail);
if (create) {
/* refcount > 0 means the caller tries to allocate a shared memory. This shared memory segment will have
* refcount references; when the peer attaches, it should pass -1 to reduce one reference count. When it
projects/rccl/src/misc/socket.cc
+10 -3
View File
@@ -498,11 +498,18 @@ static ncclResult_t socketPollConnect(struct ncclSocket* sock) {
memset(&pfd, 0, sizeof(struct pollfd));
pfd.fd = sock->fd;
pfd.events = POLLOUT;
SYSCHECK(ret = poll(&pfd, 1, timeout), "poll");
if (ret == 0) return ncclSuccess;
ret = poll(&pfd, 1, timeout);
if (ret == 0 || (ret < 0 && errno == EINTR)) {
return ncclSuccess;
} else if (ret < 0) {
WARN("socketPollConnect poll() failed with error %s", strerror(errno));
return ncclRemoteError;
} else {
EQCHECK(ret == 1 && (pfd.revents & POLLOUT), 0);
}
/* check socket status */
EQCHECK(ret == 1 && (pfd.revents & POLLOUT), 0);
SYSCHECK(getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void*)&ret, &rlen), "getsockopt");
if (ret == 0) {