<span id="mailbox-conversation"><div>awesome, great to hear.</div>
<div><br></div>
<div>Thanks!</div>
<div><br></div>
<div>-Andrew</div></span><div class="mailbox_signature"><br></div>
<br><br><div class="gmail_quote"><p>On Tue, Mar 31, 2015 at 7:16 PM, Emil Constantinescu <span dir="ltr"><<a href="mailto:emconsta@mcs.anl.gov" target="_blank">emconsta@mcs.anl.gov</a>></span> wrote:<br></p><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><p>Hi Andrew - we found a bug in TSComputeRHSFunctionLinear.
<br><br>In your code you need to provide the context to TSSetRHSFunction:
<br>TSSetRHSFunction( tss, NULL, TSComputeRHSFunctionLinear, &c );
<br><br>For now you can check out a branch that has implemented the fix by
<br>running git checkout emil/fix-ts-dae-rebased-v2 and then make everything.
<br><br>The fix is at
<br>https://bitbucket.org/petsc/petsc/commits/557d2b190687a728e86290bdc604ce0d2dc53473
<br><br>This is working for us.
<br><br>We'll push the fix soon in master.
<br><br>Emil
<br><br>--
<br>Emil M. Constantinescu, Ph.D.
<br>Computational Mathematician
<br>Argonne National Laboratory
<br>Mathematics and Computer Science Division
<br>9700 S Cass Avenue,
<br>Argonne, IL 60439, USA
<br><br>Ph: 630-252-0926
<br>http://www.mcs.anl.gov/~emconsta
<br><br>On 3/30/15 9:32 PM, Andrew Spott wrote:
<br>> I’m sorry, I didn’t want to imply you were taking too much time. I was
<br>> just interested in trying to solve the problem myself, especially if it
<br>> was a simpler fix.
<br>>
<br>> I don’t have enough of an understanding of the internals of PETSc
<br>> though, if it depends on the compiler flags.
<br>>
<br>> Good luck!
<br>>
<br>> -Andrew
<br>>
<br>> —
<br>> Andrew
<br>>
<br>>
<br>> On Mon, Mar 30, 2015 at 7:23 PM, Emil Constantinescu
<br>> <emconsta@mcs.anl.gov <mailto:emconsta@mcs.anl.gov>> wrote:
<br>>
<br>> Hi Andrew,
<br>>
<br>> Sorry for the delay; things are going slower with me over the spring
<br>> break but we are working on it.
<br>>
<br>> The problem that you see may not be related to the integrator, which
<br>> runs well (we get expected convergence rates and good results) for
<br>> other
<br>> problems, in particular, for problems with real variables (and no
<br>> c++11). So the high level th->endpoint implementation seems fine.
<br>>
<br>> For some reason the linear solver is misbehaving when using cn. If you
<br>> use option -snes_mf the problem disappears.
<br>>
<br>> Please do not use -ts_theta_adapt until it is fixed.
<br>>
<br>> In conclusion I don't think that looking at theta.c will be productive.
<br>> We are currently looking at what happens at the lower levels.
<br>>
<br>> Thanks,
<br>> Emil
<br>>
<br>> On 3/30/15 4:25 PM, Andrew Spott wrote:
<br>> > Emil: If I want to fix this myself, do you have any idea where I
<br>> should
<br>> > start looking? I assume in the th->endpoint section of
<br>> TSStep_Theta in
<br>> >
<br>> http://www.mcs.anl.gov/petsc/petsc-current/src/ts/impls/implicit/theta/theta.c.html?
<br>>
<br>> >
<br>> > Thanks,
<br>> >
<br>> > -Andrew
<br>> >
<br>> >
<br>> >
<br>>
<br>>
<br></p></blockquote></div><br>