Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

remove unreached code of AllreduceRobust::CheckAndRecover (#108)

Browse Source
This commit is contained in:
Xu Xiao 2019-09-19 11:06:59 +08:00 committed by Jiaming Yuan
parent d4ce6807c7
commit e92641887b
1 changed files with 5 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
src/allreduce_robust.cc
View File

@ -611,28 +611,12 @@ bool AllreduceRobust::CheckAndRecover(ReturnType err_type) {
if (err_type == kSuccess) return true;
utils::Assert(err_link != NULL, "must know the error source");
recover_counter += 1;
{
// simple way, shutdown all links
for (size_t i = 0; i < all_links.size(); ++i) {
if (!all_links[i].sock.BadSocket()) all_links[i].sock.Close();
}
ReConnectLinks("recover");
return false;
}
// this was old way
// TryResetLinks still causes possible errors, so not use this one
while (err_type != kSuccess) {
switch (err_type.value) {
case kGetExcept: err_type = TryResetLinks(); break;
case kSockError: {
TryResetLinks();
ReConnectLinks();
err_type = kSuccess;
break;
}
default: utils::Assert(false, "RecoverLinks: cannot reach here");
}
// simple way, shutdown all links
for (size_t i = 0; i < all_links.size(); ++i) {
if (!all_links[i].sock.BadSocket()) all_links[i].sock.Close();
}
ReConnectLinks("recover");
return false;
}
/*!