On Jan 20, 2012, at 10:50 AM, Jed Brown wrote:

> On Fri, Jan 20, 2012 at 10:43, Barry Smith <bsmith at mcs.anl.gov> wrote:
>   The whole reason for Jed putting in the flag is to clean up the bad no 
> preallocation in PETSc. 
> 
> This also won't work because it sets the option on the wrong matrix.
>  
> 
>    Please do not do this kind of stuff.
> 
>    This needs to be fixed by putting the correct preallocation in. And why is 
> converting to SeqAIJ using a basic converter? Isn't there a converter to  
> SeqAIJ that does preallocation?
> 
> There is no MatConvert_SeqDense_SeqAIJ

   But what about a MatConvert_whatever_SeqAIJ() that uses MatGetRow() to 
determine counts for preallocation. We don't have such a beast?????


   Barry



Reply via email to