[MARS-UD-100] CreepUD and openMP

[MARS-UD-100] CreepUD and openMP

Postby rwendner » Tue Jul 19, 2016 4:04 pm

There is a problem in running the current creepUD subroutine in combination with openMP.

The simulation with marsO shows excessive increase in kinetic energy. Mohammed observed the same problem in the last weeks (with the earlier versions of that time).
The simulation with marsD runs relatively stably.

The results of a somewhat larger specimen, also run with marsD show a constant frequency, constant amplitude noise (where does it come from ?).
Attachments
Archive.zip
inputs and screenshot of output with marsD
(38.47 KiB) Downloaded 624 times
rwendner
 
Posts: 154
Joined: Wed Jul 15, 2015 4:35 pm

Re: CreepUD and openMP

Postby zhouxinwei » Tue Jul 19, 2016 4:19 pm

When the debug version and OpenMP don't give the same result, it usually means the code is not written correctly. There are a few things to look for: (1) there is a bug in the code; (2) the code is not OpenMP compatible, meaning the code violates how OpenMP handles memory.

To check (2), run the OpenMP version while setting the number of threads to 1. If this gives the same results as the debug version, (2) is the issue, otherwise, check (1).

Also, I noticed you are still importing FE node list into a diffusion node list. I suggest checking my previous replies and use the approaches described in the development notes.
zhouxinwei
 
Posts: 208
Joined: Thu Jul 09, 2015 7:12 pm

Re: [MARS-UD-100] CreepUD and openMP

Postby rwendner » Tue Jul 19, 2016 5:21 pm

any insights regarding the oscillations in the MarsD run, see jpeg file in previous attachment ?
rwendner
 
Posts: 154
Joined: Wed Jul 15, 2015 4:35 pm

Re: [MARS-UD-100] CreepUD and openMP

Postby zhouxinwei » Tue Jul 19, 2016 5:36 pm

there was no jpeg file in the attachment.
zhouxinwei
 
Posts: 208
Joined: Thu Jul 09, 2015 7:12 pm


Return to OpenMP and MPI

Who is online

Users browsing this forum: No registered users and 1 guest

cron