[petsc-dev] PETSC application

[Barry Smith]

http://www.mcs.anl.gov/petsc/petsc-2/snapshots/petsc-dev/docs/manualpages/Mat/MatCreateShell.html#MatCreateShell

   This allows you to create a Mat that can be passed to  
KSPSetOperators() that provides matrix-vector products (and possibly  
other operations) but does not explicitly store the nonzero entries of  
the matrix.  This is exactly what you need to use for the A.

    Barry

On Mar 7, 2010, at 12:36 PM, David sheehan wrote:

> Can you give me any more details about it? thanks.
>
> David
>
> On Sun, Mar 7, 2010 at 12:22 PM, Matthew Knepley <knepley at gmail.com>  
> wrote:
> On Sun, Mar 7, 2010 at 12:20 PM, David sheehan <david.sheehanjr at gmail.com 
> > wrote:
> >I mean Ax =b doesn't have A in explicit form, but in expression A=B 
> +C1+C2+...+Cn,
> >where B is main part of A in an explicit form, C1, C2,... and Cn  
> are small portions
> >comparing with B and they don't have an explicit form. Thus I want   
> to  make a
> >preconditioner from B, then apply the preconditioner to Ax=b.  
> Please tell me how I
> >can do by PETSC sentences for PETSC solver, thanks.
>
> >   call KSPSetOperators(ksp,A,B,......)
>
> However, A doesn't have an explicit matrix form, how to call
> call KSPSetOperators(ksp,A,B,......)
>
> Use MATSHELL for A.
>
>    Matt
>
> ?  thanks.
>
> Based on what I thought, once I got preconditioner of B, I am  
> wondering if I can
> define a special matrix-vector multiplication for A=B+C1+C2+...+Cn  
> from a PETSC
> sentence, such that it can do following
> 1) x1=B*x
> 2) x2=x1+ C1*x, where C1*x is a sequence of matrix-vector  
> multiplications
> 3) x3= x2+C2*x,......
> .....
> 4) xn+1=xn+Cn*x,where Cn*x is a squence of matrix-vector  
> multiplications
>
> Then select a KSP iterative method such as GMRES to use the above  
> matrix-vector
> multiplication and preconditioner that I made from B.
>
> Could you please tell me if it is possible? if so (I hope), how can  
> I do it ? thanks.
>
> David
>
> On Sun, Mar 7, 2010 at 11:12 AM, Barry Smith <bsmith at mcs.anl.gov>  
> wrote:
>
> On Mar 7, 2010, at 10:57 AM, David sheehan wrote:
>
>> >First, configure with Hypre .
>> You mean, configure for the installation of PETSC with hypre? thanks.
>    yes, add --download-hypre
>
>>
>> >Then -pc_type hypre -pc_hypre_type boomeramg
>> You mean, without adding any information into code include, just  
>> compile and link
>> the code, to get an excuteable file, e.g. "a.out", then
>> a.out -pc_type hypre -pc_hypre_type boomeramg
>> is it right? thanks.
>
>    Yes,
>
>>
>> >I cannot understand exactly what you mean. However, if you want to  
>> create a PC
>> > based upon a >matrix which is not the system matrix A, you can  
>> specify it in
>> > KSPSetOperators().
>>
>> I mean Ax =b doesn't have A in explicit form, but in expression A=B 
>> +C1+C2+...+Cn,
>> where B is main part of A in an explicity form, C1, C2,... and Cn  
>> are small portions
>> comparing with B and they don't have an explicit form. Thus I want   
>> to  make a
>> preconditioner from B, then apply the preconditioner to Ax=b.  
>> Please tell me how I
>> can do by PETSC sentences for PETSC solver, thanks.
>
>    call KSPSetOperators(ksp,A,B,......)
>
>    Barry
>
>
>>
>>
>> David
>>
>>
>> On Sun, Mar 7, 2010 at 12:32 AM, Matthew Knepley  
>> <knepley at gmail.com> wrote:
>> On Sat, Mar 6, 2010 at 11:49 PM, David sheehan <david.sheehanjr at gmail.com 
>> > wrote:
>> Hi,
>> I need your helps about PETSC as follows:
>> 1) how can I call boomerAMG of HYPRE through PETSC as a solver and  
>> a preconditioner?
>>
>> First, configure with Hypre . Then -pc_type hypre -pc_hypre_type  
>> boomeramg
>>
>> 2) how can I make a precondioner  from B matrix,  but apply the  
>> preconditioner to Ax=b in PETSC?
>> Thanks in advance.
>>
>> I cannot understand exactly what you mean. However, if you want to  
>> create a PC based upon a matrix
>> which is not the system matrix A, you can specify it in  
>> KSPSetOperators().
>>
>>   Matt
>>
>> David
>> -- 
>> 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
>>
>
>
>
>
>
> -- 
> 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
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20100307/d6649adb/attachment.html>