[petsc-users] Obtain the preconditioned matrix

[我]

Hi Matt,
I have tested my code with a really small matrix(10*10) just now and the preconditioned matrix is obtained. But for my case, the output is still equal to zero. I'm thinking if it is because this coefficient matrix is too ill. In my case, due to the Dirichlet boundary condition, I multiply a really huge number(e.g. 1e+10) at the matrix's given position. So this matrix is combined with some really huge numbers. I don't know if it will influence the preconditioned matrix output.
The viewer is PETSC_VIEWER_STDOUT_WORLD.
Thanks again!
daye







At 2018-03-24 19:26:34, "Matthew Knepley" <knepley at gmail.com> wrote:

On Sat, Mar 24, 2018 at 8:15 AM, 我 <dayedut123 at 163.com> wrote:

Hi Matt,
Sorry to distribute you again. I still have trouble with KSPComputeExplicitOperator. I reduced the matrix size to around 1000*1000. The time-cost of this function is acceptable, but when I output the preconditioned matrix, all the values are equal to zero! Here is part of my codes, do I operate error with this function?

PC pc;
KSP ksp;
Mat A,PA;
Vec b,x;
ierr=KSPCreate(PETSC_COMM_WORLD,&ksp);CHKERRQ(ierr);
ierr=KSPSetOperators(ksp,A,A);CHKERRQ(ierr);
ierr=KSPGetPC(ksp,&pc);CHKERRQ(ierr);
ierr=PCSetType(pc,PCHYPRE);CHKERRQ(ierr);
ierr=KSPSetType(ksp,KSPGMRES);CHKERRQ(ierr);
ierr=KSPSetFromOptions(ksp);CHKERRQ(ierr);
ierr=KSPSolve(ksp,b,x);CHKERRQ(ierr);
ierr=KSPComputeExplicitOperator(ksp,&PA);



There is no need to assemble the output matrix.
 
ierr=MatAssemblyBegin(PA,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
ierr=MatAssemblyEnd(PA,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);



What kind of viewer are you using? The code above should work. I would step through
in the debugger to see why one column is all zeros, or try it in an example like
SNES ex5 to make sure its not a problem in your code.


  Thanks,


    Matt
 
ierr=MatView(PA,viewer);CHKERRQ(ierr);

Thank you very much!
Daye








At 2018-03-23 07:41:26, "Matthew Knepley" <knepley at gmail.com> wrote:

On Thu, Mar 22, 2018 at 8:29 PM, 我 <dayedut123 at 163.com> wrote:

Hi all,
I want to analyze the preconditioned matrix. But the KSPComputeExplicitOperator costs too much time to obtain the matrix. My origin matrix is about 3000*3000 sparse one. I noticed this function is applicable for the relative small system. What's the matrix-size limitation for this function? And for my case, any suggestions to solve it with less time?



This is just a really expensive operation. There is no way around it.


   Matt
 
Thank you very much!
daye






 






--

What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
-- Norbert Wiener


https://www.cse.buffalo.edu/~knepley/





 






--

What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
-- Norbert Wiener


https://www.cse.buffalo.edu/~knepley/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20180324/9c14d98a/attachment-0001.html>