is_outputting_averaged_fields issue

[rclam]

Recently I was running a program and wanted to set is_outputting_averaged_fields = yes for my simulation. I received an error message and the program was killed. I checked to see if the same thing happened when I ran the original benchmark code oedometer-2d.cfg. When the option was set to no, it ran fine. When I changed it to yes, then I got the same error message.

Checking it with valgrind (shown below), it seems that there is leakage occurring. Could someone check their own fork and see if they have the same problem?

[tan2]

Hi Lam, A quick confirmation, what is the revision id of the code you are using? You can check it with the following command: git log -1 HEAD The most recent revision is: commit 5420830

…

On Wed, Dec 6, 2023 at 3:37 AM R.C. Lam ***@***.***> wrote: Recently I was running a program and wanted to set is_outputting_averaged_fields = yes for my simulation. I received an error message and the program was killed. I checked to see if the same thing happened when I ran the original benchmark code oedometer-2d.cfg. When the option was set to no, it ran fine. When I changed it to yes, then I got the same error message. Screen.Shot.2023-12-05.at.1.28.43.PM.png (view on web) <https://github.com/tan2/DynEarthSol/assets/75513951/e6ffe092-e65c-46cd-a87a-e0f5b4d30c5f> Checking it with valgrind (shown below), it seems that there is leakage occurring. Could someone check their own fork and see if they have the same problem? Screen.Shot.2023-12-05.at.1.30.44.PM.png (view on web) <https://github.com/tan2/DynEarthSol/assets/75513951/c7ea01c9-eb20-42d5-8f8f-2da33de8b229> — Reply to this email directly, view it on GitHub <https://github.com/tan2/DynEarthSol/issues/15>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABWQQRCLWUTAI2WMPGPGULYH5ZYRAVCNFSM6AAAAABAIHEUN6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGAZDMOJYGYZDEMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[rclam]

`commit b380249 (HEAD, EhTan/master)
Merge: ac1a328 b071cc6
Author: Eunseo Choi [email protected]
Date: Mon Oct 23 08:25:07 2023 -0500

Merge pull request #12 from chaseshyu/merge_openacc

Preparation of test case, flag, and consistency for merge openacc`

[chaseshyu]

Hi @rclam,

Sorry, I cannot reproduce the Invalid free() / delete / delete[] / realloc() in commit tan2/DynEarthSol@b380249 with valgrind.
Can you provide the commands you are using to compile and debug? And what the compiler do you use?
For instance, my comiler version is: g++ (GCC) 4.8.5 .
I am using following commands to

Compile

make ndims=2 BOOST_ROOT_DIR=/path/to/boost/62 -j4 opt=0

Debug

valgrind --leak-check=full --show-leak-kinds=all --log-file="valgrind-out.txt" ./dynearthsol2d oedometer-2d.cfg

[chaseshyu]

Hi @rclam,

With
max_steps = 200
is_outputting_averaged_fields = yes
quality_check_step_interval = 40

I have run commit tan2/DynEarthSol@b380249 and commit tan2/DynEarthSol@0d20bd9 (the commit before merge).

b380249

0d20bd9

Debug of commits both do not have invaild free() memory access.

[rclam]

Chase, thanks for looking into this. When I last re-ran the same problem using valgrind, it was stuck and didn't give any outputs (just constantly reiteratively trying to do output 1 maybe); when I ran it without valgrind, it gave the same error and killed itself immediately. It's possible the change is because Nefertiti might have had a compiler update between when I submitted the bug and when I checked the compiler versions. How should I approach the issue again? I was thinking to ask someone else on Nefertiti to try running it so I can see if the error is on my version only and I've somehow messed with something (and need to git pull the original all over) or we can see if it's related to Nefertiti's set-up. What do you think? Is there something else I need to check?