<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Thu, Nov 10, 2016 at 9:38 AM, Jed Brown <span dir="ltr"><<a href="mailto:jed@jedbrown.org" target="_blank">jed@jedbrown.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>> writes:<br>
<br>
> On Thu, Nov 10, 2016 at 1:31 AM, Julian Andrej <<a href="mailto:juan@tf.uni-kiel.de">juan@tf.uni-kiel.de</a>> wrote:<br>
><br>
>> I'm getting the correct solution now, thanks! There are still a few<br>
>> open questions.<br>
>><br>
>> 1. The mass term is necessary _and_ using the u_tShift value if i<br>
>> provide the jacobian. After reading the manual countless times, i<br>
>> still don't get what the u_tShift value tells me in this context. Are<br>
>> there any resources which you could point me to?<br>
>><br>
><br>
> The manual is always a lagging resource because we are trying things<br>
> out.<br>
<br>
</span>This has been explained in the manual similarly to your email for<br>
several years. If anyone has a suggestion for how to make it better,<br>
we'd like to hear. The typical syndrome is that once you learn it, the<br>
description suddenly makes sense and you wonder why you were ever<br>
confused. It takes feedback from people just learning the material to<br>
make the explanation more clear.<br>
<span class=""><br>
> I learned about this by reading examples. The idea is the<br>
> following. We have an implicit definition of our timestepping method<br>
><br>
> F(u, grad u, u_t, x, t) = 0<br>
<br>
</span>It doesn't make sense to write grad u as a separate argument here.<br></blockquote><div><br></div><div>Sorry, that is just how I think of it, not the actual interface.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Also, is `x` meant to be coordinates or some other statically prescribed<br></blockquote><div><br></div><div>Coordinates, as distinct from u. Again this is my fault.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
data? It can't be actively changing if you expect a generic TS to be<br>
convergent. (It's not an argument to the function.) So really, you<br>
have:<br>
<br>
F(u, u_t, t) = 0<br>
<span class=""><br>
> which is not unlike a Lagrangian description of a mechanical system. The<br>
> Jacobian can be considered<br>
> formally to have two parts,<br>
><br>
> dF/du and dF/du_t<br>
><br>
> just as in the Lagrangian setting. The u_tShift variable is the multiplier<br>
> for dF/du_t so that you actually<br>
> form<br>
><br>
> dF/du + u_tShift dF/du_t<br>
<br>
</span>We're taking the total derivative of F with respect to u where u_t has<br>
been _defined_ as an affine function of u, i.e., shift*u+affine.<br>
u_tShift comes from the chain rule. When computing the total<br>
derivative, we have<br>
<br>
dF/du + dF/du_t du_t/du. </blockquote><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<span class=""><br>
>> 2. Is DMProjectFunction necessary _before_ TSSolve? This acts like an<br>
>> initial condition if i'm not mistaken.<br>
>><br>
><br>
> Yes, this is the initial condition.<br>
><br>
><br>
>> 3. A more in depth question i guess. Where is the "mass action"<br>
>> formed? As i could see by stepping through the debugger, there is just<br>
>> a formed action for the residual equation. It seems way over my<br>
>> understanding how this formulation works out. I'd also appreciate<br>
>> resources on that if thats possible.<br>
>><br>
><br>
> Jed is the only one who understands this completely.<br>
<br>
</span>That's a stretch -- several other people have developed sophisticated TS<br>
implementations.</blockquote><div><br></div><div>Matt does not understand this completely.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class=""><br>
> However, I guess by "mass action" you mean the dF/du_t term. In the<br>
> explicit methods, you just have u_t + ..., so<br>
><br>
> dF/du_t = M<br>
><br>
> for finite element methods. So you are putting that there in<br>
> FormIJacobian(). In the simplest<br>
> case of backwards Euler then u_tShift would be 1/dt.<br>
<br>
</span>Specifically, for implicit methods, and many semi-implicit methods, we<br>
don't need M separate.<br>
</blockquote></div><br>Yes.</div><div class="gmail_extra"><br></div><div class="gmail_extra"> Matt<br clear="all"><div><br></div>-- <br><div class="gmail_signature" data-smartmail="gmail_signature">What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>-- Norbert Wiener</div>
</div></div>