Analyzers
Talk to fellow users of Intel Analyzer tools (Intel VTune™ Profiler, Intel Advisor)
4975 Discussions

Intel Advisor crashes when trying to display captured survey data

LM11
Beginner
2,370 Views

Hello,

 

Having issues with Intel Advisor 2022.3.0 on a Linux RHEL 7.9 system.  When I attempted to display data that was collected from the command line survey the GUI crashes with a segfault error and the GUI closes out.

 

The application is a C++ code baseline that contains a number of function calls and only runs on one core (not parallel).

 

The scenarios I have tested are below:

1) A small sized (~1500 Walltime seconds for the run) survey capture and displaying of data works just fine 

2) A medium sized (~3500 Walltime seconds for the run) survey capture works, but displaying analysis data fails with segfault

3) A medium sized (~3500 Walltime seconds for the run) survey capture and displaying of data works with the sampling interval increased to 500ms 

4) A large sized (~4800 Walltime seconds for the run) survey capture works, but displaying analysis data fails with a segfault and unable to make it work with the maximum of 1000ms sampling interval 

 

The advisor-gui does not appear to be running out memory and is only using ~4-5GB prior to the crash. The system has 64GB available memory. The ulimit for stack size is also set to unlimited.

 

It appears that there is some sort of limitation on the amount of data that can displayed. Are there any environment settings or configuration that should be applied when running large data analyses? Is Advisor able to display this much analysis data? 

0 Kudos
13 Replies
Ruslan_M_Intel
Employee
2,340 Views

Hello,

 

Could you please try to generate advisor report? (See this

 

E.g.: 

advisor --report=survey --format=text --report-output=./survey.txt --project-dir=./advi_results

 

I assume it could fail here too. Anyway it is important to know where exactly it fails - in GUI or somewhere else.

0 Kudos
JaideepK_Intel
Moderator
2,336 Views

Hi,


Thank you for posting in intel communities.


could you please try the above suggestion provided by Ruslan.


If issue still persist, please provide the below information.

  • Could you please upgrade your Advisor to the latest version, i.e., 2023.0?
  • Since RHEL 7.9 was deprecated, could you please try on a supported OS?
  • Please provide a sample reproducer (a replica of your sample) so that we can reproduce your issue from our end.


Regards,

Jaideep


0 Kudos
LM11
Beginner
2,318 Views

I was able to run the following command with my paths:

 

advisor --report=survey --format=text --report-output=./survey.txt --project-dir=./advi_results

 

The text file was created successfully I believe, although it was not very readable given the tabbing and vertical spacing/paging. Is that to be expected?

 

--- Additional Notes on the Crash ---

Prior to the GUI crash, Advisor is able to start preparing the data and finalize the frequently used data on the Survey and Roofline tab. It will show a glimpse of the results, times, percentages, etc..., then crash and close out. It seems like it may be having issues after the report data is finalized and displayed.

 

I also used gdb to run the advisor-gui to see if I could get any details without full debug information.  I found that the segfault is potentially occurring in the "/opt/intel/oneapi/advisor/2022.3.0/lib64/libadvixe_discmodels_recomendations_2.479.so"

 

 

We did see this same issue on RHEL7.9 with the older 2021.4 release of Advisor.  We made the upgraded to 2022.3 in the hopes of fixing this issue, but it still persisted.

 

Upgrading the OS right now may be a bit of a challenge, but I will see if I can get a sample case that reproduces the same issue. If so, the sample case could be ported into a test environment with RHEL 8 and 2023.0 and tested.

 

0 Kudos
Ruslan_M_Intel
Employee
2,087 Views

Could you please share exact commands you used? If possible please also share the app you used and/or gathered profiles you have. Otherwise please describe what the app does (e.g. heavy compute ops, a lot of sleep commands, single/multithread etc.) and I can try to create something similar for testing purposes

0 Kudos
LM11
Beginner
2,049 Views

Text Report  Command:

$ advisor --report=survey --format=text --report-output=./survey.txt --project-dir=/home/<userID>/intel/advixe/projects/AdvTest/

 

GDB Data Capture Steps/Commands:

$ gdb advisor-gui

Select "Show Result"

The Summary page is displayed showing the data including the "Top Time Consuming Loops" which has the current collected stats

Select "Survey & Roofline" Tab

The status bar is almost 100% complete and spinning.  It also shows "Preparing frequently used data"

Then the window closes out and display the segmentation fault on the terminal window

 

Application Information:

The code baseline that is causing the crash is a C++ application that is doing an event based simulation that is not extremely heavy in compute. The code fully utilizes object oriented coding aspects with numerous method calls. The entire baseline is around 2.5M lines of code. The application is also single threaded and runs on one core.

 

We have also started the process for attempting to run this on RHEL8 with the 2022.3 and the latest Advisor version. 

 

0 Kudos
Andrey_K
Employee
1,817 Views

Hi,

 

You may also try to collect your results from CLI and then open it in GUI, if it makes sense.

 

To generate command line use the button on workflow, at the right of the start button. Or "advisor --help collect".

0 Kudos
JaideepK_Intel
Moderator
1,918 Views

Hi,


Are you able to run your application on Advisor2023.0 successfully on RHEL8 without any issues?

Could you please share a sample reproducer so that we can try to reproduce it from our end.


Thanks,

Jaideep


0 Kudos
LM11
Beginner
1,883 Views

We are in the process of setting up a RHEL 8 system to test the configuration. Hopefully we will have some results to share shortly.

0 Kudos
JaideepK_Intel
Moderator
1,769 Views

Hi,


Good day to you.

Have you tried installing the latest version of advisor on RHEL8? Is it working fine?


Regards,

Jaideep




0 Kudos
LM11
Beginner
1,719 Views

We have had a bit of a set back on the RHEL 8 instance, but this effort is still underway. Hoping to have something running in the next week.

0 Kudos
JaideepK_Intel
Moderator
1,702 Views

Hi,

 

Good day to you.

In the meantime, could you please optimize your code? Can you make it run on multiple threads because it is currently running on a single thread? 

 

Thanks,

Jaideep


0 Kudos
JaideepK_Intel
Moderator
1,625 Views

Hi,


Good day to you.

Have you tried installing the latest version of advisor on RHEL8? Is it working fine?


Regards,

Jaideep


0 Kudos
JaideepK_Intel
Moderator
1,480 Views

Hi,


Good day to you.


We have not heard back from you. This thread will no longer be monitored by Intel. If you need further assistance, please post a new question.


Regards,

Jaideep.



0 Kudos
Reply