Re: [Libmesh-devel] Exact Error Estimator with Fine Grid Solution Patch...

[Derek G. <fri...@gm...>]

On 6/4/07, Roy Stogner <roy...@ic...> wrote:
> I think eventually I'd like to make the coarse-vs-fine grid
> integrations a user-specified option, but for now that looks good; go
> ahead and commit.

The problem is that you can't assume that the coarse grid is a nested
inside the fine... and if it isn't nested then doing the integration
over the fine and accumulating that error back onto the coarse is next
to impossible.  We can surely make it an option at some point in the
future... but I don't think you're ever going to be able to account
for mismatched coarse/fine grids

> No he doesn't, and no you wouldn't.  Vikram's application code is an
> ever-growing miasma of incorrectly factored exectution paths and
> hard-coded unlikely assumptions, whose utility to libMesh lies
> primarily in its value as a warning to others.

LOL!  That is seriously one of the funniest paragraphs I've read in a long time!

I'll just modify ex14 so it has a new option to generate a fine grid
solution...  I'll probably go ahead and check in the estimator first
though... so others can use it (and break it).

Derek