glmark2 produces Segmentation fault in GPU PC about glmark2

[afrantzis]

Hi! Are you able to run glmark2 with gdb on the remote server? Perhaps that would give some kind of clue (e.g., a backtrace) what's going on.

But, taking a step back, I am trying to understand what you are trying to achieve with X11 forwarding. Do you want to benchmark the GPU of the remote machine? If that's the case, then doing ssh -X from your local machine to your remote machine and running glmark2 in that session won't help, since you will be actually using (or at least trying to use) your local machine's display/gpu hardware, not the remote GPU.

The only reasonable way to benchmark the remote GPU is to run glmark2 natively in the remote machine (i.e., without X11 forwarding).

[uddmorningsun]

Thanks your reply really! @afrantzis About glmark2 running method in X11 forwarding, I see.
Below followings are related backtrace info ：

$ glmark2
Segmentation fault

$ gdb /usr/local/bin/glmark2
Reading symbols from /usr/local/bin/glmark2...done.
(gdb) l
133	    benchmark_collection.populate_from_options();
134	
135	    MainLoopValidation loop(canvas, benchmark_collection.benchmarks());
136	
137	    while (loop.step());
138	}
139	
140	int
141	main(int argc, char *argv[])
142	{
(gdb) r
Starting program: /usr/local/bin/glmark2 

Program received signal SIGSEGV, Segmentation fault.
0x0000000000000000 in ?? ()
Missing separate debuginfos, use: debuginfo-install glibc-2.17-196.el7.x86_64 libX11-1.6.5-1.el7.x86_64 libXau-1.0.8-2.1.el7.x86_64 libXext-1.3.3-3.el7.x86_64 libgcc-4.8.5-16.el7.x86_64 libjpeg-turbo-1.2.90-5.el7.x86_64 libpng12-1.2.50-10.el7.x86_64 libstdc++-4.8.5-16.el7.x86_64 libxcb-1.12-1.el7.x86_64 zlib-1.2.7-17.el7.x86_64
(gdb) bt
#0  0x0000000000000000 in ?? ()
#1  0x00007ffff7983221 in init () from /lib64/libdl.so.2
#2  0x00007ffff7983667 in _dlerror_run () from /lib64/libdl.so.2
#3  0x00007ffff7983128 in dlsym () from /lib64/libdl.so.2
#4  0x00007ffff5c9912f in ?? () from /usr/lib64/nvidia/libGLdispatch.so.0
#5  0x00007ffff5c998b9 in ?? () from /usr/lib64/nvidia/libGLdispatch.so.0
#6  0x00007ffff7dea4bf in _dl_init_internal () from /lib64/ld-linux-x86-64.so.2
#7  0x00007ffff7ddc1aa in _dl_start_user () from /lib64/ld-linux-x86-64.so.2
#8  0x0000000000000001 in ?? ()
#9  0x00007fffffffe626 in ?? ()
#10 0x0000000000000000 in ?? ()

I find libGLdispatch.so.o, so I try:

$  cd /etc/ld.so.conf.d/
$  ls
cuda-8-0.conf  dyninst-x86_64.conf  kernel-3.10.0-693.el7.x86_64.conf  mariadb-x86_64.conf  nvidia-lib64.conf
$ ldconfig -p | grep nvidia | grep libGL   
... ...
$ mv nvidia-lib64.conf   /tmp
$ ldconfig

And try again , it work . Thanks! But I remove file nvidia-lib64.conf, whether it will have an effect or not? Or glmark2 don't care GPU related file, it only care GPU hardware?

[uddmorningsun]

About remote server, I find a better way: startx /usr/local/bin/glmark2 > glmark2.txt

refer to: https://ubuntuforums.org/showthread.php?t=2131425&page=15&p=13642458#post13642458

[afrantzis]

But I remove file nvidia-lib64.conf, whether it will have an effect or not? Or glmark2 don't care GPU related file, it only care GPU hardware?

By removing the nvidia-lib64.conf file you are probably falling back to the Mesa drivers. When running with X11 forwarding this allows the forwarding to work, but as I mentioned above, it doesn't matter since you are going to use the drivers of the local pc anyway, so this doesn't do what you want.

However, this has an effect when running natively in the remote server, since it means that you are not going to use the nvidia drivers to run glmark2, so you won't get the results you want.

About remote server, I find a better way: startx /usr/local/bin/glmark2 > glmark2.txt

Seems like the way to go. Just make sure that you restore nvidia-lib64.conf. You should see NVIDIA in the GL_VERSION field of the glmark2 output.

[uddmorningsun]

Thanks ! I got it, I will close it.