[petsc-dev] bjacobi bugs ?

[Barry Smith]

On Jan 25, 2012, at 11:53 AM, Satish Balay wrote:

> looks like this piece of code was last changed at:
> 
> http://petsc.cs.iit.edu/petsc/petsc-dev/rev/83c325962928
> 
> the new code added was 'PCSetUp_BJacobi_Multiproc' - different than 'PCSetUp_BJacobi_Multiblock'

   So likely the multiproc code broke the singeblock/multiblock stuff

   Hong can you check this please?


   Barry
> Satish
> 
> On Wed, 25 Jan 2012, Barry Smith wrote:
> 
>> 
>>   In PCSetUp_BJacobi() it has at the end
>> 
>> 
>>     /* ------
>>     Setup code depends on the number of blocks
>>  */
>>  if (jac->n_local == 0) {
>>    ierr = PCSetUp_BJacobi_Singleblock(pc,mat,pmat);CHKERRQ(ierr);
>>  } else {
>>    ierr = PCSetUp_BJacobi_Multiblock(pc,mat,pmat);CHKERRQ(ierr);
>>  }
>> 
>> does this make sense? No. In the normal case it is not using PCSetUp_BJacobi_Singleblock() at all! Something has gone wrong in the logic of the code when multiblock was added.
>> 
>>    Barry
>> 
>> 
>