Re: [Libmesh-devel] Matrix free solving

[John P. <jwp...@gm...>]

On Mon, Nov 17, 2008 at 10:21 AM, John Peterson <jwp...@gm...> wrote:
> On Mon, Nov 17, 2008 at 9:49 AM, Tim Kroeger
> <tim...@ce...> wrote:
>> Dear libMesh team,
>>
>> It is done.  Well, at least I think so at the moment.  The shell matrix
>> concept is running well with my application and gets equal results on 1 and
>> on 4 CPUs.
>>
>> Please find attached a patch that contains the current version. Compared
>> with the previous version, a small number of bugs had to be fixed.
>>
>> The patch also contains an example, called ex20, that demonstates how to use
>> the concept.  I didn't spend too much of work in setting up the example, but
>> it should be okay.  In particular it shows the pitfalls.
>>
>> I would suggest that you first (after making sure that you like everything
>> in my patch) commit it to the repository and than later restructure it
>> according to your own ideas.
>>
>> Of course, I am almost sure that I will find more bugs in my code later on,
>> but I will then simply send you more patches.  (-:
>
> I'll work on getting this checked in now.

The patch is all checked in.  The code also worked with a slightly
older version of PETSc, 2.3.1, that I tested it with, so the ifdefs
check against that version instead.

-- 
John