[petsc-dev] [petsc-users] SAMG support

[Patrick Sanan]

Awesome - I wasn't aware of the shared library option, but that indeed
makes things even more attractive.

On Wed, Apr 27, 2016 at 1:47 PM, Matthew Knepley <knepley at gmail.com> wrote:
> On Wed, Apr 27, 2016 at 5:51 AM, Lisandro Dalcin <dalcinl at gmail.com> wrote:
>>
>> On 27 April 2016 at 11:53, Patrick Sanan <patrick.sanan at gmail.com> wrote:
>> >
>>
>> > includes), and have them add a single registration line after
>> > PetscInitialize which makes the new solver usable just as the built-in
>> > ones are (via the options database and command line arguments).
>>
>> Actually, if you build your plugin as a shared library, then you can just
>>
>> -dll_append pcsupercool.so
>>
>> and get it registered, no need to modify code.
>
>
> Note that this is how I organized our class project:
>
>   https://github.com/jag20/CAAM520SRM
>
> and as you can see from the build system, its really easy.
>
>   Thanks,
>
>     Matt
>
>>
>>
>> >
>> > I don't know of any packages that actually do this - are there some?
>> >
>>
>> Don't know. But I would argue that many packages currently in PETSc
>> could be reworked as plugins very easily.
>>
>> Oh! Hold on. We have some support for writing "plugins" in Python, and
>> we dynamically load the Python shared library to bootstrap the Python
>> runtime and then petsc4py. And IMHO, this is quite a bit more
>> complicated that plain C plugins.
>>
>> >
>> > Is this a hassle for reasons I haven't thought through fully yet?
>> >
>>
>> The only hassle would be to updating it after new PETSc releases. And
>> BTW, you could keep your code compatible with older PETSc releases
>> with proper processor guards.
>>
>> A minor issue with dynamically loadable plugins is that using the
>> command line is the only way to configure the package is through the
>> command line. But that's a fair price to pay for not having to ever
>> modify (nor recompile) your code.
>>
>> > The specific interest I have in this is that a student of Olaf's has
>> > been working with an interface to (non-MKL) PARDISO which I think
>> > could be useful, and I'm also interested in there being an interface
>> > to ILUPack.
>> >
>>
>> Go for it. I can help with comments and review.
>>
>> PS: I'm not 100% sure about the status of dynamic lib loading in
>> PETSc, but I can help with any required refactoring of that machinery
>> to better support plugins.
>>
>> --
>> Lisandro Dalcin
>> ============
>> Research Scientist
>> Computer, Electrical and Mathematical Sciences & Engineering (CEMSE)
>> Extreme Computing Research Center (ECRC)
>> King Abdullah University of Science and Technology (KAUST)
>> http://ecrc.kaust.edu.sa/
>>
>> 4700 King Abdullah University of Science and Technology
>> al-Khawarizmi Bldg (Bldg 1), Office # 0109
>> Thuwal 23955-6900, Kingdom of Saudi Arabia
>> http://www.kaust.edu.sa
>>
>> Office Phone: +966 12 808-0459
>
>
>
>
> --
> 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