Zero right-hand-side vector

[EDMUNDOmex]

Hi

I compilated feap 83 on opensuse, there weren't errors and I run an example in the Output file appears this warning

*WARNING* Zero right-hand-side vector

and I don't find the error. It implies that the stiffness matrix is zero.

Could you help me?

Regards

Edmundo

[K.Li]

Can you post your input file?

[Prof. S. Govindjee]

It implies that there are no loads on your mesh.

[EDMUNDOmex]

Thanks for your attention

It is my script, in fact is the patch test

I think the problem is the compiler!!!

Originally I was using gfortran and gcc, after that, I used mpif90 and mpigcc and I've got the same results


Warnings like that

Warning: Possible change of value in conversion from REAL( to INTEGER(4) at (1)

vblke.f:232:0: warning: iteration 1 invokes undefined behavior [-Waggressive-loop-optimizations]

Warning: Unused dummy argument 'v12' at (1)



I have run the same input file on FEAP74, and it works very well

So, I hope you can help me.

Thanks

[Prof. S. Govindjee]

I tried your file.  It works just fine.  Converges properly.  Results in a linear displacement field.

[EDMUNDOmex]

Hi Professor

I tried the same file in Windows and it works fine as you say, but when I tried on Linux the same file doesn´t work. I am working on opensuse.

Best regards

[Prof. R.L. Taylor]

What version of feap is causing the problem?  Also what version of the compilers are you using?

Have you checked the output file to make sure that all the data for the mesh appears properly?  If any part is not processed you do not have a correctly defined input.

We know that many compilers now complain about changing precision or type in an assignment statement; thus statements like

   i = d

where "i" is integer*4 and "d" is a real*8 variable will be tagged as you note.  In the latest release we have tried to fix these using a form like
   i = nint(d)

This is probably not the cause of your zero RHS -- are you using the correct include files with your compile (usually now the include and the include/integer8 ones).

Please check what works properly, not just that a zero right hand side occurs.

[EDMUNDOmex]

Hi Prof. Taylor

I am working with Feap82 on Opensuse.

I used integer8 and I changed the value of ipr to 2, first, I used gfortran and gcc

I installed Feap, and there weren't errors, So I tried with the patch test (Previously posted) and FEAP doesn't work, something was wrong,

after that, I changed the compiler to mpif90 and mpigcc, the results were the same (Oejem82.txt), where you can see that all the data for the mesh appears properly.

Looking for a solution, I found a post in INSTALLATION FEAP on the forum, in which recommended change the size of the arrays mr( ) and hr( ) in include/comblk.h to be of length
1024. 

I did it, and Feap works fine.

I tried with other examples and compared them with the same compiled on Windows, and the results are the same.

Thanks for your attention

Best regards.

[FEAP_Admin]

Version 8.2 is very old.  You should consider upgrading to the current release.