Page 1 of 1

StressTensor for LDPM in MPI

PostPosted: Fri Oct 07, 2016 3:46 pm
by alnagm
Hi,

I am trying to plot the stress tensor from the paraview plot options. It works for the OMP version but it always fails on MPI at the plot list initialization. Here is the error i get

Initializing list plL-CONC
[drp06:16306] *** Process received signal ***
[drp06:16306] Signal: Segmentation fault (11)
[drp06:16306] Signal code: Invalid permissions (2)
[drp06:16306] Failing at address: 0x7f01f87bd000
[drp06:16306] [ 0] /lib64/libpthread.so.0[0x3cc2c0f7e0]
[drp06:16306] [ 1] /lib64/libc.so.6[0x3cc2489ce2]
[drp06:16306] [ 2] /lib64/libc.so.6(memmove+0x192)[0x3cc2483b22]

Thanks,
Mohammed

Re: StressTensor for LDPM in MPI

PostPosted: Thu Oct 13, 2016 6:51 pm
by zhouxinwei
The bug is fixed in 2016.10.00.

Re: StressTensor for LDPM in MPI

PostPosted: Wed Oct 26, 2016 2:25 pm
by alnagm
Hi,

The simulation run correctly now but there is a problem in output. As the beam deforms it seems like the tet list on which the stresses are plotted does not move correctly, still the top layer is in place while bottom one is moving (see attached image). Basically, the plot seems to have tets at the undeformed position and also the same tets at the deformed one.

stress_tens.jpg
stress_tens.jpg (27.17 KiB) Viewed 12442 times


This looks very strange when plotting stress distributions on a section at the middle of the beam

Thanks,
Mohammed

Re: StressTensor for LDPM in MPI

PostPosted: Wed Oct 26, 2016 6:57 pm
by zhouxinwei
Do you think it's due to the opterations in Paraview? If not, can you please post the input file or the Paraview file (only the first and last time frames are enough) so we can reproduce the issue?

Re: StressTensor for LDPM in MPI

PostPosted: Fri Oct 28, 2016 3:21 am
by alnagm
The files are huge. I have to send them by email to you.

Re: StressTensor for LDPM in MPI

PostPosted: Fri Nov 04, 2016 12:16 am
by zhouxinwei
The bug (each MPI process dumps all tets instead of its own) is fixed in 2016.11.04.