Segmentation fault with multi threading on Ubuntu/AMD

[thusithaC]

Hi,
I get a segmentation error when xgb is run for some time. Any advice is really welcome. I tried using xgboost built from the source as we as the version from pip.

Stack trace from gdb:

Thread 44 "python" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffb47a7700 (LWP 10273)]
0x00007fffba7aa98a in xgboost::tree::ColMaker<xgboost::tree::GradStats, xgboost::tree::NoConstraint>::Builder::EnumerateSplitCacheOpt(xgboost::SparseBatch::Entry const*, xgboost::SparseBatch::Entry const*, int, unsigned int, std::vector<xgboost::detail::bst_gpair_internal, std::allocator<xgboost::detail::bst_gpair_internal > > const&, std::vector<xgboost::tree::ColMaker<xgboost::tree::GradStats, xgboost::tree::NoConstraint>::ThreadEntry, std::allocator<xgboost::tree::ColMaker<xgboost::tree::GradStats, xgboost::tree::NoConstraint>::ThreadEntry> >&) [clone .constprop.598] ()
from /home/thusitha/anaconda3/lib/python3.6/site-packages/xgboost-0.6-py3.6.egg/xgboost/libxgboost.so

No symbol table info available.
#1 0x00007fffba7ad104 in xgboost::tree::ColMaker<xgboost::tree::GradStats, xgboost::tree::NoConstraint>::Builder::UpdateSolution(xgboost::ColBatch const&, std::vector<xgboost::detail::bst_gpair_internal, std::allocator<xgboost::detail::bst_gpair_internal > > const&, xgboost::DMatrix const&) [clone ._omp_fn.0] ()
from /home/thusitha/anaconda3/lib/python3.6/site-packages/xgboost-0.6-py3.6.egg/xgboost/libxgboost.so
No symbol table info available.
#2 0x00007fffba425b05 in gomp_thread_start (xdata=) at /opt/conda/conda-bld/compilers_linux-64_1505664199673/work/.build/src/gcc-7.2.0/libgomp/team.c:120
team = 0x55555714ba00
task = 0x55555714c350
data =
pool =
local_fn = 0x7fffba7ac2e0 <xgboost::tree::ColMaker<xgboost::tree::GradStats, xgboost::tree::NoConstraint>::Builder::UpdateSolution(xgboost::ColBatch const&, std::vector<xgboost::detail::bst_gpair_internal, std::allocator<xgboost::detail::bst_gpair_internal > > const&, xgboost::DMatrix const&) [clone ._omp_fn.0]>
local_data = 0x7fffffffa9a0
#3 0x00007ffff7bc16ba in start_thread (arg=0x7fffb47a7700) at pthread_create.c:333
__res =
pd = 0x7fffb47a7700
now =
unwind_buf = {cancel_jmp_buf = {{jmp_buf = {140736221312768, 1828178988090501251, 0, 140737488334207, 140736221313472, 140737488334856, -1828012923730548605, -1828161957722740605},
mask_was_saved = 0}}, priv = {pad = {0x0, 0x0, 0x0, 0x0}, data = {prev = 0x0, cleanup = 0x0, canceltype = 0}}}
not_first_call =
pagesize_m1 =
sp =
freesize =
PRETTY_FUNCTION = "start_thread"
#4 0x00007ffff78f73dd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:109

Environment info

Operating System:
ubuntu 16.04
AMD Ryzen 1700x processor
Compiler:
gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.5)
Package used (python/R/jvm/C++):
python Anaconda 5.0.0.1
xgboost version used: from source

If installing from source, please provide

1. The commit hash (git rev-parse HEAD)
   9a81c74
2. Logs will be helpful (If logs are large, please upload as attachment).

If you are using python package, please provide

1. The python version and distribution
   python Anaconda 5.0.0.1

2. The command to install xgboost if you are not installing from source
   from source

Steps to reproduce

What have you tried?