[CIG-SHORT] The pylith_benchmarks reports errors

[Brad Aagaard]

On 3/15/18 4:52 PM, Matthew Knepley wrote:
> On Thu, Mar 15, 2018 at 6:31 AM, tu xiang <tuxiang2016 at outlook.com
> <mailto:tuxiang2016 at outlook.com>> wrote:
>
>     Dear Matt,
>     The version I am running is Trelis V16.0.3 Linux.X64 .
>
>     According to the examples provided by Pylith, I find that there two
>     kinds of methods for setting vertex spacing with increasing spacing
>     away from fault. I wonder which method will be better for a
>     three-dimensional non-planar fault?
>
>
> I do not know the answer. I would try both and look at the condition of
> the cells produced.
>
>   Thanks,
>
>     Matt
>
>
>     One is that you can export the fault surface into Exodus file and
>     then add the  variable  of cellsize which specifies the distance
>     between vertices of a cell edge into the Exodus file.

We export the entire mesh, not just the fault surface, and add the 
variable cellsize to the Exodus file to specify the distance between 
vertices of a cell edge.

The two methods can give equivalent results, so it is not easy to say 
which is better. Ultimately, the method gets you a sufficient quality 
mesh the fastest is the "best" approach for you. It really depends on 
the complexity of your geometry and how quickly you setup and implement 
either approach.

Regards,
Brad