Author Topic: LINK  (Read 2522 times)

ilinca

  • Jr. Member
  • **
  • Posts: 5
LINK
« on: June 29, 2012, 02:34:22 PM »
We have troubles with the LINK command in Feap 8.3
In previous versions of FEAP that I used, a sequence like

link
11,12,,,0,0,1
13,14,,,0,0,1

would link for instance the x and y DOFs of nodes 11 and 12.

However, the same sequence used in Feap8.3 leads to an error message even though the 8.3 manual seems to indicate that the format for the LINK input is still the same:
LINK
node-m1 node-m2 inc-m1 inc-m2 id1 id2 ... id-ndf
node-n1 node-n2 inc-n1 inc-n2 id1 id2 ... id-ndf
... ! repeat for additional node ranges
! blank termination record

The error message we get is
"No file named link0000 exists"

IS there anything that changed and I missed to account for in this input?

Thank you for your help,

Ilinca

FEAP_Admin

  • Administrator
  • Hero Member
  • *****
  • Posts: 386
Re: LINK
« Reply #1 on: June 29, 2012, 03:29:04 PM »
Ilinca,

  The syntax you describe is correct.  Attached is a very simple file+output that shows its use (tested with version 8.3).  If you still have a problem after looking at this example, please try to make a simple input file with which generates the error and post it.  [The input file has two different examples of link which you can play with, one using generation and one without: the without is a nonsense problem, the with simulates a full tie and results in a homogeneous solution.]


ilinca

  • Jr. Member
  • **
  • Posts: 5
Re: LINK
« Reply #2 on: June 30, 2012, 01:06:31 PM »
I am getting the same error with the file you sent me. Like my file (which I also attach here) it runs fine in Feap 8.1 but gives me the error (see run_link_83.png for a screen image) if I try to run it in Feap8.3

Prof. S. Govindjee

  • Administrator
  • Sr. Member
  • *****
  • Posts: 82
Re: LINK
« Reply #3 on: June 30, 2012, 01:16:30 PM »
Ilinca,
  Two thoughts:
   (1) there is a bug in the 8.3 that you have: you could try downloading a new version and see if that fixes your problem. 
   (2) Are you using a semi-recent version of gcc/gfortran?  The newest GNU Compiler Compliation (gcc, gfortran) has an optimization bug that severely affects FEAP by causing seemingly random memory errors. The work around for this compiler bug is to edit 'include/comblk.h' and change hr(1) and mr(1) to hr(1024) and mr(1024). After making the change, please delete all *.o, *.a, and executable files, then rebuild your program.

Prof. S. Govindjee

  • Administrator
  • Sr. Member
  • *****
  • Posts: 82
Re: LINK
« Reply #4 on: June 30, 2012, 01:19:16 PM »
fyi, I ran you file with v8.3 and it worked fine for me:  Here is the link output from my Olink_is file:

    N o d a l    L i n k    O u t p u t s

     Linked  Pairs     DOF Link Pattern (0=link; 1=no link)
    1-Node   2-Node  1-dof  2-dof  3-dof
       44      167      0      0      1
       45      168      0      0      1
       46      169      0      0      1
       47      170      0      0      1
       48      171      0      0      1
       49      172      0      0      1
       50      173      0      0      1
       51      174      0      0      1
       52      175      0      0      1
       53      176      0      0      1
       11       43      0      0      1
       21       55      0      0      1
       32       54      0      0      1
       42       66      0      0      1
      139      267      0      1      1
      141      273      0      1      1
      142      279      0      1      1
      144      285      0      1      1
      239      291      0      1      1
      241      297      0      1      1
      242      303      0      1      1
      244      309      0      1      1
      244      309      0      1      1
     *WARNING* Nodes     244 and     309 already linked for DOF =   1

ilinca

  • Jr. Member
  • **
  • Posts: 5
Re: LINK
« Reply #5 on: June 30, 2012, 03:35:57 PM »
Sanjay,

Thank you for your patience! I am afraid I still could not sort it out.

I downloaded again from http://www.ce.berkeley.edu/projects/feap/webaccess/ver83/feap/ and compiled it in both 32b and 64b and with the modification you suggested in comblk.h ... I tried several times, with different optimization flags too. Same error every time!
It seems to be an output from opnfil.f :

        elseif(iopl.eq.-2) then
          write(*,2002) mac
          return

2002  format('  No file named ',a,' exists.')

where a seems to take the value link0000 in this case

Additional info: I tried on several computers from the lab. They range from a 3 years old laptop intel core duo running OS X 10.5.8 to a one year old intel core i7 running OS X 10.6.8
I am using gcc/ gfortran 4.4.1 20090623
(I remember trying a newer one at some point but I run into troubles at compiling time with it and I reverted back to the above)

If you have any other idea on what I could try please let me know

Thank you
Ilinca
PS ... maybe you have an even newer version of 8.3?




Prof. S. Govindjee

  • Administrator
  • Sr. Member
  • *****
  • Posts: 82
Re: LINK
« Reply #6 on: June 30, 2012, 04:43:41 PM »
I used the downloadable version so the source version is not the problem. This was on a mac (OS 10.6.8, gcc/gfortran 4.6.0). 

This will not be super easy to chase down but if you work with the Ilink file I sent, simplify it
so that there are no commands like ebou or csur (searching commands) then it will be easier to track.
The thing that is going on is that information is supposed to be being written to files like Rlink.link0000
and then read later.  In your case Rlink.link0000 doesn't exist when it should.  The question is why not!


Prof. S. Govindjee

  • Administrator
  • Sr. Member
  • *****
  • Posts: 82
Re: LINK
« Reply #7 on: June 30, 2012, 04:53:38 PM »
I have managed to re-create the error.  I think I see what it happening.  On the first call to create the temporary file for the link data, it does not set the name correctly.  I'll see if I can make an easy fix.

Prof. S. Govindjee

  • Administrator
  • Sr. Member
  • *****
  • Posts: 82
Re: LINK
« Reply #8 on: June 30, 2012, 05:09:45 PM »
As I suspected, the file was being opened with a different name.  I've attached a revised version of program/pcontr.f that fixes this problem and hopefully does not create others.

ilinca

  • Jr. Member
  • **
  • Posts: 5
Re: LINK
« Reply #9 on: June 30, 2012, 07:19:10 PM »
Thank you Sanjay!

It seems to work fine now

Ilinca