On Fri, Jan 24, 2014 at 7:23 PM, Jane Hung jyh...@mit.edu wrote:
Could you give more details about how to extend the overlaps along the top
and bottom of the mesh? Should it be something like this
mesh=Grid2D(nx=500,ny=500,dx=0.25, dy=0.25, overlap=2) ? How could you
specify the top and bottom
Another issue: I tried this Cahn Hilliard example
http://pastebin.com/9UZJ2h24 with trilinos instead of pysparse, and the
results seem to differ. The phase separation doesn't seem to occur with the
trilinos solver.
On Fri, Jan 24, 2014 at 7:23 PM, Jane Hung jyh...@mit.edu wrote:
Could you give
In examples/cahnHilliard/mesh2D.py, we call eq.solve(...,
solver=LinearLUSolver()). I don't know why the default Trilinos solver
(probably GMRES) doesn't show the same evolution, but it's possibly the
nonlinearities in the 2nd-order DiffusionTerm mean we should be sweeping and
GMRES is more
On Wed, Jan 22, 2014 at 10:37 AM, J Hung jyh...@mit.edu wrote:
I decided to try it on Ubuntu and got this error
python -c from fipy.solvers.scipy import DefaultSolver; \ print
DefaultSolver --no-pysparse
Traceback (most recent call last):
File string, line 1, in module
File
Actually I got it working in Ubuntu, but the parallel implementation
doesn't give expected results, as in different from running in serial. The
parallelization setup seems fine from the command line test on the site. Do
periodic grids work when running in parallel? If not, how could periodic
On Fri, Jan 24, 2014 at 10:44 AM, Jane Hung jyh...@mit.edu wrote:
Actually I got it working in Ubuntu, but the parallel implementation doesn't
give expected results, as in different from running in serial. The
parallelization setup seems fine from the command line test on the site. Do
periodic
Is there a way to use Trilinos with windows? PyTrilinos doesn't seem to be
compatible.
When I tried to install Fipy with Trilinos on OSX, I can't seem to run
anything since even
python -c from fipy import *; fipy.test() gives an AttributeError
and in the diffusion/mesh1D.py file, it gives a
On Tue, Jan 21, 2014 at 11:08 AM, Jane Hung jyh...@mit.edu wrote:
Is there a way to use Trilinos with windows? PyTrilinos doesn't seem to be
compatible.
I've never built Trilinos on Windows, but it is supported according to
the docs. You might want to email the Trilinos mailing list if you are
I'm using PySparse
On Jan 15, 2014, at 10:27 PM, Guyer, Jonathan E. Dr.
jonathan.gu...@nist.gov wrote:
On Jan 14, 2014, at 2:05 PM, Jane Hung jyh...@mit.edu wrote:
The vector formulation is still rather slow (much much slower than with
just one equation). Is that expected?
I'm not sure.
My tests show the coupled variant is actually faster (with Trilinos).
As mentioned in ticket 658, the PySparse LU solver uses a lot of memory for
this problem (it always uses a lot, but it's particularly high here). It's
generally quite fast, but if it needs a substantial fraction of the RAM
On Jan 14, 2014, at 2:05 PM, Jane Hung jyh...@mit.edu wrote:
The vector formulation is still rather slow (much much slower than with just
one equation). Is that expected?
I'm not sure. We'll do some tests.
What solver suite are you using? PySparse or Trilinos?
Update: The mesh2DCoupled example did work, but I had to change the grid
dimensions make it look like the pictures online. However, it's also
incredibly slow. I left it running overnight and it only reached elapsed
time of about 0.6. It being so slow for a simple system may explain why it
can't
Coupled Cahn Hilliard has known problems: http://matforge.org/fipy/ticket/582
I don't know whether the issues described in that ticket are the source of what
you're seeing, but it can't help. For your problem, you may want to base what
you're doing on the vector formulation in that example.
On
The vector formulation is still rather slow (much much slower than with
just one equation). Is that expected?
On Jan 14, 2014 12:10 PM, Guyer, Jonathan E. Dr. jonathan.gu...@nist.gov
wrote:
Coupled Cahn Hilliard has known problems:
http://matforge.org/fipy/ticket/582
I don't know whether the
Yes, restricting the time step works. However, whenever I split up the
equation (like d(phi)/dt = Xi, Xi= laplacian(phi)), it is never able to run.
Also, when I run the Cahn-Hilliard mesh2DCoupled example, the results are
that the concentration becomes more and more homogeneous rather than phase
It looks to be stable up to time steps of about 25. You are using the
exponentially increasing stepper from our Cahn Hilliard examples, which are
unconditionally stable (and we have them top out at 100). Because of the
explicit terms in your splitting, you should keep your time steps below the
I tried to start with a simpler system, and it seems like I get the same
problem if I split up the equations at all.
Anyway, I started with a 1 equation system http://pastebin.com/X5tT1RUB and
would like to see the phase separation, but after time ~2000 (see the video
I'm also getting RuntimeError. To get over this, is there a way to
represent the system a different way or does the system itself too
complicated?
What do you mean by know the answer? I have an idea of what the time
evolution of the variables should look like in the 2D case, but I don't
have an
Now that I've changed it as you mentioned, it says Can't expand MemType 0.
What can I do about this?
http://pastebin.com/buFS1NRu
On Fri, Dec 20, 2013 at 10:24 AM, Daniel Wheeler
daniel.wheel...@gmail.comwrote:
On Fri, Dec 20, 2013 at 9:13 AM, Jane Hung jyh...@mit.edu wrote:
That definitely
On Wed, Dec 18, 2013 at 1:37 PM, Jane Hung jyh...@mit.edu wrote:
OK, I can write the equations with grad instead of faceGrad, but it doesn't
seem like the equations can be solved with what I've tried. I tried to use
eq.solve, which gives a TypeError,
This could be a FiPy issue if it works with
On Dec 19, 2013, at 12:59 PM, Daniel Wheeler daniel.wheel...@gmail.com wrote:
I tried this and the residuals decrease quite a bit. Would you expect
the residuals to approach zero at every time step? This seldom happens
with non-linear equations. The absolute magnitude of the residuals
isn't
On Thu, Dec 19, 2013 at 1:33 PM, Guyer, Jonathan E. Dr.
jonathan.gu...@nist.gov wrote:
Do 4th order diffusion terms even work with coupled? I thought they didn't.
You're right, I forgot. So Jane, you'll need to split the equations as
I suggested. Your current implementation combining fourth
On Mon, Dec 16, 2013 at 4:30 PM, J Hung jyh...@mit.edu wrote:
Hi,
I'm trying to find the best way to write my equation, which involves not
only diffusion terms in terms of two order parameters phi and psi, but also
terms like grad(phi) dot grad(laplacian(psi)). Since all these terms are
being
OK, I can write the equations with grad instead of faceGrad, but it doesn't
seem like the equations can be solved with what I've tried. I tried to use
eq.solve, which gives a TypeError, and to use sweep, which does not
converge in the while loop. I've linked the 2 versions of my code:
sweep
Hi,
I'm trying to find the best way to write my equation, which involves not
only diffusion terms in terms of two order parameters phi and psi, but also
terms like grad(phi) dot grad(laplacian(psi)). Since all these terms are
being added together, I'm trying to use faceGrad to make terms like
25 matches
Mail list logo
