<div class="gmail_quote">On Tue, Nov 15, 2011 at 18:03, Matthew Knepley <span dir="ltr"><<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div>What are you on? It is F(x) = 0. This embedding will not work for all problems, but it will work for many, and be easy. If you</div><div>have a more complicated system, by all means us TS, however I am pointing out that there are useful regimes where this</div>
<div>kind of embedding would work.</div></blockquote></div><br><div>Is the underlying transient problem xdot+F(x)=0 or xdot=F(x)?</div><div><br></div><div>What happens if you enforce boundary conditions without eliminating them (like almost every PETSc example)?</div>
<div><br></div><div>Just use the TS interface if time might be involved. Then you can do everything.</div><div><br></div><div>Making assumptions about what the user meant when they were using a lower-level interface doesn't help. And just to get an "enclosing TS", they had to write some setup code. At that point, it's no more difficult to implement the TS callbacks than to implement the SNES callbacks.</div>