----------------------------------------
> Date: Fri, 19 Aug 2016 16:35:07 -0500
> From: royst...@ices.utexas.edu
> To: rwood...@utexas.edu
> CC: libmesh-users@lists.sourceforge.net
> Subject: Re: [Libmesh-users] Where to start learning Libmesh?
>
>
> On Wed, 17 Aug 2016, Ryan Woodall wrote:
>
>> I'm a biomedical engineering PhD student at UT, and I've got a very
>> large FEM domain (~4GB) I'm trying to implement FEM on. I have been
>> pointed to libmesh, and am already using TACC resources.
>
>> Is there any kind of manual or guide to using all the different
>> methods available in libmesh? I have read through and run many of
>> the examples,
>
> I'm afraid you just answered your own question, and I do apologize
> that there isn't a better answer (except for the Doxygen
> documentation, but that's only a half-decent reference and it doesn't
> qualify as a manual or guide).
>
> Most of our users with relatively straightforward problems to solve
> usually do so by picking a libMesh-based framework (i.e. MOOSE or
> GRINS, though I also have a simple little framework that I give out
> for use on simple physics) and writing a new kernel to add their
> physics to it.
>
> Most of our users who need more fine-grained control (or who started
> using the library before MOOSE was open sourced and GRINS matured)
> start with one or another of our examples and modify it.
>
>> but I still can't get my own problem up and running.
>
> Your best source for help here is probably the libmesh-users mailing
> list. I try to be as helpful as possible, but if I'm stuck in
> meetings (as I was Wednesday) or out of town (as I am until tomorrow)
> then I'm pretty slow on the draw, and there are other users and
> developers who are good at taking up the slack.
>
>> If someone were brand new, perhaps so new they didn't know the right
>> questions to ask, where would you first send them to learn this
>> library?
>
> If you're doing things properly, the first step in any simulation is a
> "design document", describing exactly what you're trying to model and
> what the mathematical formulation of the model is.
>
> If you don't have something like that already, you should - aside from
> design philosophy, it'll be the core of your thesis and papers, and
> better written sooner than later.
>
> If you *do* have something like that already, then I'd say chop out
> the bare essentials (PDEs, boundary conditions, etc) and post a link
> to libmesh-users asking how best to get started.
Personally so far I've found gdb to be really useful and would imagine
a good debugger is a good way for many people to learn. (And btw, is
TypeVector size() deprecated? I spent a lot of time adding spurious denormals
together until I realized I got the wrong loop limit since I did not understand
how dphi is organized ).
Getting some simple initial results was easy but it was amazing how hard it
was to get all the loop limits right once I tried to make a better test. If you
already understand FEM it
may be easier for you however.
It is also nice to have quick ways to examine output.
I think originally someone recommended "visit" but I've had very good luck with
"R"
and scripts for trying to understand results. I know how to script R but not
visit.
Curious too what people use for research documents. I wrote some of my own
stuff for building up bibtex files for easy addition into latex docs ( bash and
vim scripts
mostly ). I guess there are better bibliography managers but so far happy with
what I made.
Thanks.
>
>> I really appreciate the help!
>
> Sorry about the delay. I'll Cc: this to libmesh-users so others there
> get a heads up and so you can see the email address; IIRC you'll need
> to subscribe at
> https://sourceforge.net/p/libmesh/mailman/libmesh-users/ before the
> mailman software will allow you to post yourself.
> ---
> Roy
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Libmesh-users mailing list
> Libmesh-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
------------------------------------------------------------------------------
_______________________________________________
Libmesh-users mailing list
Libmesh-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-users
