Time step cut criteria exceeded (trying with Petsc-dev)

[Das, Rajeev]

UNRESTRICTED / ILLIMITÉE

Hi:

 

I am trying to run a problem (it is an old example the input file is attached) and I am receiving the following error. I know I can change the cuts in timestepper  card, which may help, but I think there is something wrong.

 

Stopping: Time step cut criteria exceeded!.

 

It is arising in the Richards Flow. It could be due to the occurrence of negative term in density (eos_water.F90). I have used a pflotran code  that runs using petsc-dev (not the petsc-maint).

 

Can anyone point to me what I am doing wrong here.

 

Regards,

 

Rajeev Das.

[Peter Lichtner]

I’m curious why you are using so many grid cells:

NXYZ 1 1 5030912

The problem runs fine with nz=100. Probably your grid spacing is too small: dz = 1.3516e-05.

…Peter

> <1DHFex1_5030912.in>

[Rajeev Das]

Hi Peter:

It is bit deliberate. I am performing few scaling tests, and I intend to increase the problem size by reducing the grid spacing. Is there a way I can still retain the same size and not encounter the issue. Yes, you are right in pointing that there will not be a problem if the grid spacing is large. 

Regards,

Rajeev. 

[rajee...@cnl.ca]

Hi All:

It throws the same error with nz=100, my fault I reported wrongly, that it worked with nz=100. I think there is some issue with the code at my end. Any inputs will be really appreciated. I followed a similar post on this forum (posted on Nov 2015, by Emmanuel Leger), and tried a few changes in the deck but no use. My pflotran version is changeset: 9403:cfa0b8daf127 ;tag: tip ; User: Glenn Hammond; date: Tue Jan 19 2016. I must add there are few patches included at my end to make it work with Petsc-dev. 

Thank you.

Regards,

Rajeev

[Peter Lichtner]

You need to send your input file and complete description of error you are getting.

…Peter

[Das, Rajeev]

UNRESTRICTED / ILLIMITÉE

Dear Peter:

 

It is the same file that I attached in this post  1DHFex1_5030912.in the following is the error message that I receive . There is some issue in the residual estimation. The np for the mpirun was set to 1, and my petsc config is  

 

--download-mpich=yes --with-cc=gcc --with-cxx=g++ --with-fc=gfortran --with-shared-libraries=0 --with-debugging=0 --download-parmetis=yes --download-metis=yes --download-hypre=yes --download-fblaslapack=yes  --with-c2html=0 --with-cuda=1 --with-cusp=1 --with-thrust=1 COPTFLAGS=-O3 CXXOPTFLAGS=-O3 FOPTFLAGS=-O3 -march=native -mtune=native

 

Regards,

Rajeev.

 

== RICHARDS FLOW ===============================================================

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  1[  0] t=  0.00000E+00 dt=  5.00000E-07

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  2[  0] t=  0.00000E+00 dt=  2.50000E-07

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  3[  0] t=  0.00000E+00 dt=  1.25000E-07

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  4[  0] t=  0.00000E+00 dt=  6.25000E-08

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  5[  0] t=  0.00000E+00 dt=  3.12500E-08

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  6[  0] t=  0.00000E+00 dt=  1.56250E-08

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  7[  0] t=  0.00000E+00 dt=  7.81250E-09

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  8[  0] t=  0.00000E+00 dt=  3.90625E-09

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut=  9[  0] t=  0.00000E+00 dt=  1.95312E-09

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 10[  0] t=  0.00000E+00 dt=  9.76562E-10

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 11[  0] t=  0.00000E+00 dt=  4.88281E-10

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 12[  0] t=  0.00000E+00 dt=  2.44141E-10

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 13[  0] t=  0.00000E+00 dt=  1.22070E-10

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 14[  0] t=  0.00000E+00 dt=  6.10352E-11

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 15[  0] t=  0.00000E+00 dt=  3.05176E-11

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  -> Cut time step: snes= -3 icut= 16[  0] t=  0.00000E+00 dt=  1.52588E-11

  0 2r: 6.80E-02 2x: 0.00E+00 2u: 0.00E+00 ir: 6.80E-02 iu: 0.00E+00 rsn:   0

  Stopping: Time step cut criteria exceeded!

     icut = 17, max_time_step_cuts= 16

     dt   =  1.5258789E-11, dt_min=  0.0000000E+00

 

FLOW TS BE steps =      0 newton =        0 linear =          0 cuts =      0

FLOW TS BE SNES time = 0.0 seconds

 

TRAN TS BE steps =      0 newton =        0 linear =          0 cuts =      0

TRAN TS BE SNES time = 0.0 seconds

 

Wall Clock Time:  1.9380E-02 [sec]   3.2300E-04 [min]   5.3834E-06 [hr]

[Peter Lichtner]

As it runs fine for me I would assume there is a problem with you installation. Why don’t you try setting up a minimal petsc installation, e.g. I am using on MacOSX 10.11.3:

Using configure Options: --with-cc=clang --with-cxx=clang++ --with-fc=gfortran --download-mpich=1 --download-hdf5=1 --download-metis=1 --download-parmetis=1 --with-debugging=0 --download-sowing-cc=clang --download-sowing-cxx=clang++ --download-c2html-cc=clang --download-c2html-cxx=clang++

with no optimization, cuda etc. and see if that works.

…Peter

[Rajeev Das]

Dear Peter:

Thank you for your response. I am performing scaling tests as mentioned earlier along with a different architecture, so those options are necessary for me. I don't face such an issue with other problems that also involve flow. I will take a look again of where things are going wrong. 

Thank you again.

Regards,

Rajeev.

[Peter Lichtner]

Do you need O3? That could be the problem...

...Peter <iPhone>