Analyzers
Talk to fellow users of Intel Analyzer tools (Intel VTune™ Profiler, Intel Advisor)

Multiple installations of Parallel Studio XE

dkokron
Beginner
828 Views

I am using Parallel Studio 2013 under the non-commercial license.

I installed Parallel Studio 2013 update 3 just the other day and ran into a problem with Vtune crashing.  I would like to reinstall Parallel Studio 2013 update 1 to see if I can track down the cause of this SEGV under Vtune update5.

I am not installing in the default location.  I will be switching between installations during testing using 'modules'.  This is how the admins at work do it so I know it's possible.

The release notes PDF provided with update 1 state on page 4.

"You do not need to uninstall previous versions or updates before installing a newer version –
the new version will coexist with the older versions."

Yet I get the following warning during installation

Step no: 1 of 7 | Previous versions detected
--------------------------------------------------------------------------------
Installation program has detected other version of the product installed which
must be uninstalled to continue installation.
--------------------------------------------------------------------------------
1. Continue with installation [default]

These products will be uninstalled automatically if newer version is selected
for installation:
    Intel(R) Parallel Studio XE 2013 Update 3 for Linux* common files
    Intel(R) VTune(TM) Amplifier XE 2013 Update 5
    Intel(R) Inspector XE 2013 Update 5
    Intel(R) Advisor XE 2013 Update 2

Whom should I believe?

Can you point me to documentation on how to multiple, segregate instalations of Parallel Studio?

Dan

0 Kudos
18 Replies
Steven_L_Intel1
Employee
828 Views

This may be an error in the release notes copied from the Composer XE Release Notes (that would be my fault). As far as I know, only Composer XE supports "side by side installs".

What you might try is downloading just the VTune installer and see if you can install that to a custom location. I am not sure it will let you if there is already an install in a "standard place".  I am going to move this thread to the VTune forum and the support folks there will help you. I will also remove that text from the Parallel Studio XE release notes.

0 Kudos
David_A_Intel1
Employee
828 Views

When it says "version" it is referring to 2011 vs. 2013, not Update 1 vs. Update 5.  Updates cannot coexist.

0 Kudos
Steven_L_Intel1
Employee
828 Views

Right - it does say "versions". Only Composer XE allows updates to exist side-by-side.

0 Kudos
SergeyKostrov
Valued Contributor II
828 Views
>>...I am not installing in the default location. I will be switching between installations during testing using 'modules'. This is how >>the admins at work do it so I know it's possible... Did it have the same problem ( a crash ) for VTune Amplifier ( Update 5 )? Update 5 is the latest Update for VTune Amplifier and it would be nice if you provide more details about the crash.
0 Kudos
dkokron
Beginner
828 Views

The Vtune crashes started happening after I updated to Parallel Studio 2013 update 3.  I did not experience this failure when running Parallel Studio 2013 update 1 on the same box.  The box is running up-to-date Fedora 18 which I know complicates matters since the OS is a moving target.  The Parallel Studio 2013 update 1 version of Vtune (slightly modified to get it to compile) ran okay as recently as kernel version 3.8.3-201.fc18.x86_64.

The tool did generate crash dumps.  If you will remind me where they are stored, I'll post.

Dan

0 Kudos
Bernard
Valued Contributor I
828 Views

@dkokron

Do you have any log file related to the crash?

0 Kudos
dkokron
Beginner
828 Views

Where would they normally be stored on disk?

0 Kudos
dkokron
Beginner
828 Views

I reran the BW experiment to gather another crash dump.  Please see attached.

0 Kudos
SergeyKostrov
Valued Contributor II
828 Views
The log for VTune only confirms that there is a problem ( crash / details are generic ): amplxe-gui *** glibc detected *** amplxe-gui: malloc(): memory corruption: 0x00007fd4ec85c9b0 *** ======= Backtrace: ========= /lib64/libc.so.6[0x30e627dab2] /lib64/libc.so.6(__libc_calloc+0xcd)[0x30e628006d] /usr/lib64/nvidia/libGL.so.1[0x30f2ea918d] ======= Memory map: ======== 00400000-004a1000 r-xp 00000000 fd:01 1579142 /opt/Compliers/Intel/2013.3.163/2013.3.163/vtune_amplifier_xe_2013/bin64/amplxe-gui ...
0 Kudos
Bernard
Valued Contributor I
828 Views

Hi dkokron,

the crash dump file contains almost no information related to the call stack.There is only register context and exception code.Can you resolve with the debugger content of this  address  0x00007fd4ec85c9b0 ? I suppose that the error started when VTune GUI module calls into Nvidia OpenGL library which in turn try to allocate memory with the help of libc.so.6 calloc function .

0 Kudos
SergeyKostrov
Valued Contributor II
828 Views
>>... >>"You do not need to uninstall previous versions or updates before installing a newer version – >>the new version will coexist with the older versions." >>... If this is Not true than this is clearly a bug in the installer. Unfortunately, I didn't have a chance to verify it because I always select the option to install all new Intel software to existing folder and I was able to see that a previous version was completely uninstalled before a new software is installed.
0 Kudos
dkokron
Beginner
828 Views

I am running amplxe-gui under valgrind to see if it can shed any light.  I will report the results when it completes.

0 Kudos
Bernard
Valued Contributor I
828 Views

Thanks dkokron

0 Kudos
dkokron
Beginner
828 Views

amplxe-gui ran much farther under valgrind than without.  Not sure how to proceed.

0 Kudos
Bernard
Valued Contributor I
828 Views

Hi dkokron,

can you explain in more detail what do you mean?

Does amplxe-gui still crashes while running under valgrind?

0 Kudos
dkokron
Beginner
828 Views

The original point of failure that prompted this thread is during the finalization of an experiment.  When I ran amplxe-gui under valgring, it ran past that point.  It was still loading the result, after ten hours, when I wrote that last comment.

Do you have enough information to reproduce the failure inhouse?

Are you willing to provide a debug build of Vtune that I can run on my machine?

How do we proceed?

Dan

0 Kudos
SergeyKostrov
Valued Contributor II
828 Views
>>...Are you willing to provide a debug build of Vtune that I can run on my machine? As far as I know Intel doesn't provide debug builds of Non Open Source software.
0 Kudos
Bernard
Valued Contributor I
828 Views

>>>How do we proceed?>>>

Can you also post full call stack ?

>>>Do you have enough information to reproduce the failure inhouse?>>>

If you are asking me I must tell you that I do not have a Linux version of VTune.

0 Kudos
Reply