Community
cancel
Showing results for 
Search instead for 
Did you mean: 
pjfloyd
Beginner
93 Views

new user problems

Hi

I'm trying to use vtune 9.0.0 on a RHEL3 machine
cpu family : 15
model : 3
model name : Intel Xeon CPU 3.00GHz
physical id : 0
siblings : 2
runqueue : 0
stepping : 4
(accoring to /proc/cpuinfo)

My working dir is in /tmp. App under test compiled with GCC 4.0.2 (built inhouse), almost all plain C.

When sampling, there is a HUGE system load (40-50)
- lots of sampling of gkrellmd and kernel processes, however the app under test is only getting ~0% CPU, and is starved of CPU to such an extent that after 1 hour of wall clock time, it still hasn't even finished its startup code. The run that I'm instrumenting takes around 20 minutes outside of instrumentation.

When sampling finishes, a little popup window shows. If I've changed virtual windows on my desktop machine (a Sun Ultra 45 with Solaris 10), then the vtune eclipse GUI crashes.

I haven't managed to use callgraph. Our app spawns a license server process, which seems to be dying with

/view/PFView/vobs/adms/aol-dbg/com/det/aol/lib/mgls_asynch: error while loading shared libraries: /opt/intel/vtune/rdc/analyzer/bin/libLauncherPreload32e.so: cannot open shared object file: No such file or directory

Any ideas how to get meaningful results with sampling and any results with callgraph?

Paul
0 Kudos
2 Replies
Corey_A_Intel
Employee
93 Views

Quoting - pjfloyd
Hi

I'm trying to use vtune 9.0.0 on a RHEL3 machine
cpu family : 15
model : 3
model name : Intel Xeon CPU 3.00GHz
physical id : 0
siblings : 2
runqueue : 0
stepping : 4
(accoring to /proc/cpuinfo)

My working dir is in /tmp. App under test compiled with GCC 4.0.2 (built inhouse), almost all plain C.

When sampling, there is a HUGE system load (40-50)
- lots of sampling of gkrellmd and kernel processes, however the app under test is only getting ~0% CPU, and is starved of CPU to such an extent that after 1 hour of wall clock time, it still hasn't even finished its startup code. The run that I'm instrumenting takes around 20 minutes outside of instrumentation.

When sampling finishes, a little popup window shows. If I've changed virtual windows on my desktop machine (a Sun Ultra 45 with Solaris 10), then the vtune eclipse GUI crashes.

I haven't managed to use callgraph. Our app spawns a license server process, which seems to be dying with

/view/PFView/vobs/adms/aol-dbg/com/det/aol/lib/mgls_asynch: error while loading shared libraries: /opt/intel/vtune/rdc/analyzer/bin/libLauncherPreload32e.so: cannot open shared object file: No such file or directory

Any ideas how to get meaningful results with sampling and any results with callgraph?

Paul

Hi,

Please try VTune 9.1 for Linux Update 2 at: https://registrationcenter.intel.com/

Regards,
Corey
geoffrey-burling
Employee
93 Views

Quoting - pjfloyd
Hi

I'm trying to use vtune 9.0.0 on a RHEL3 machine
cpu family : 15
model : 3
model name : Intel Xeon CPU 3.00GHz
physical id : 0
siblings : 2
runqueue : 0
stepping : 4
(accoring to /proc/cpuinfo)

My working dir is in /tmp. App under test compiled with GCC 4.0.2 (built inhouse), almost all plain C.

When sampling, there is a HUGE system load (40-50)
- lots of sampling of gkrellmd and kernel processes, however the app under test is only getting ~0% CPU, and is starved of CPU to such an extent that after 1 hour of wall clock time, it still hasn't even finished its startup code. The run that I'm instrumenting takes around 20 minutes outside of instrumentation.

When sampling finishes, a little popup window shows. If I've changed virtual windows on my desktop machine (a Sun Ultra 45 with Solaris 10), then the vtune eclipse GUI crashes.

I haven't managed to use callgraph. Our app spawns a license server process, which seems to be dying with

/view/PFView/vobs/adms/aol-dbg/com/det/aol/lib/mgls_asynch: error while loading shared libraries: /opt/intel/vtune/rdc/analyzer/bin/libLauncherPreload32e.so: cannot open shared object file: No such file or directory

Any ideas how to get meaningful results with sampling and any results with callgraph?

Paul

One detail which Corey may have overlooked (or simply ignored -- upgrading to the latest release is always one of the best first steps in resolving issues) is that it sounds as if you are sending theXWindow output toyour Sun Ultra. If this is true, *and* it persistsafter upgrading tothe latest version, pleasereport back here. (This is a useful feature of X which people who are not familiar with that GUI environment often are unaware exists.)

Geoff
Reply