[petsc-dev] Problem mixing sieve and TS

[Blaise Bourdin]

> On Wed, Oct 26, 2011 at 12:45, Blaise Bourdin <bourdin at lsu.edu> wrote:
> Any suggestion? FWIW, in this test, both the IJacobian matrix and RHS Vector are constant, but this will not be the case once in the future.
> 
> Something else bugging me is that the Jacobian function does not seem to be ever called (even in ex22 from src/ts/examples/tutorial).
> 
> How are you coming to that conclusion? The default method for ex22 is a method with an implicit part. If you don't set any type at all, then the default method is forward Euler (we should probably change this, maybe to a Rosenbrock-W), which does not evaluate a Jacobian, but that code sets the default type to ARKIMEX which does evaluate the IJacobian.
That is correct. In both cases, I was setting a breakpoint at the wrong location.

>  
> Is that expected? Is there a written description (or link to the literature) for each of the ODE solvers implemented in TS?
> 
> The man page for each method should have a citation to a paper that introduced it or discussed it. Some of the methods in there were designed by Emil and have not been published yet.
Thanks,
I had look at the manual and the man page for ARKIMEX but not the individual pages.

At this point we have overcome the original issue but are still having difficulties. We followed the lines of ex22 to setup our TS, with the most notable difference being that our DM is a DMMesh and not a DMDA. For some reason, the second time the Ifunction is called, the Udot argument is a Vec of NaN. We checked the TS in the debugger and dan't catch anything obviously wrong. We followed the flow of the code in the debugger, and the NaN get generated here:

(gdb) 
713       ierr = VecAXPBYPCZ(ark->Ydot,-ark->shift,ark->shift,0,ark->Z,X);CHKERRQ(ierr); /* Ydot = shift*(X-Z) */
(gdb) where
#0  SNESTSFormFunction_ARKIMEX (snes=0x18b3850, X=0x1951230, F=0x1919800, ts=0x1896910) at /opt/HPC/petsc-dev/src/ts/impls/arkimex/arkimex.c:713
#1  0x00007f707689675b in SNESTSFormFunction (snes=0x18b3850, X=0x1951230, F=0x1919800, ctx=0x1896910) at /opt/HPC/petsc-dev/src/ts/interface/ts.c:2536
#2  0x00007f70767ab7f2 in SNESComputeFunction (snes=0x18b3850, x=0x1951230, y=0x1919800) at /opt/HPC/petsc-dev/src/snes/interface/snes.c:1314
#3  0x00007f70767e7c2a in SNESSolve_LS (snes=0x18b3850) at /opt/HPC/petsc-dev/src/snes/impls/ls/ls.c:609
#4  0x00007f70767bb791 in SNESSolve (snes=0x18b3850, b=0x197ac70, x=0x1951230) at /opt/HPC/petsc-dev/src/snes/interface/snes.c:2850
#5  0x00007f707684a5df in TSStep_ARKIMEX (ts=0x1896910) at /opt/HPC/petsc-dev/src/ts/impls/arkimex/arkimex.c:576
#6  0x00007f7076890411 in TSStep (ts=0x1896910) at /opt/HPC/petsc-dev/src/ts/interface/ts.c:1813
#7  0x00007f7076891860 in TSSolve (ts=0x1896910, x=0x15c16b0, ftime=0x7fff05be8a00) at /opt/HPC/petsc-dev/src/ts/interface/ts.c:1933
#8  0x00007f7076831a08 in tssolve_ (ts=0x7fff05be89f0, x=0x7fff05be8698, ftime=0x7fff05be8a00, __ierr=0x7fff05be7dec) at /opt/HPC/petsc-dev/src/ts/interface/ftn-auto/tsf.c:397
#9  0x000000000040d75c in solvetransient (appctx=...) at m_TransientHeat.F90:493
#10 0x000000000040be2e in transientheat () at TransientHeat.F90:98
#11 0x000000000047d5ea in main ()
#12 0x00007f7072ee8c4d in __libc_start_main (main=<value optimized out>, argc=<value optimized out>, ubp_av=<value optimized out>, init=<value optimized out>, fini=<value optimized out>, 
    rtld_fini=<value optimized out>, stack_end=0x7fff05be8b98) at libc-start.c:226
#13 0x00000000004061f9 in _start ()

Do you have any suggestion of potential causes? At this point, we are running out of ideas.

Blaise
 

-- 
Department of Mathematics and Center for Computation & Technology
Louisiana State University, Baton Rouge, LA 70803, USA
Tel. +1 (225) 578 1612, Fax  +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin







-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111026/ea5df152/attachment.html>