Hi,
I want to shrink the computational domain at recovery from checkpoint. For example, I want to change the domain from initial domain size:
coordbase::xmin = 0.0 coordbase::xmax = 50.00 coordbase::ymin = 0.0 coordbase::ymax = 50.0 coordbase::zmin = 0.0 coordbase::zmax = 50.00
to
coordbase::xmin = 0.0 coordbase::xmax = 30.00 coordbase::ymin = 0.0 coordbase::ymax = 30.0 coordbase::zmin = 0.0 coordbase::zmax = 30.00
Is it possible? Is there already such a mechanism?
Thanks.
Kentaro TAKAMI
On 26 Mar 2012, at 09:43, Kentaro Takami wrote:
Hi,
I want to shrink the computational domain at recovery from checkpoint. For example, I want to change the domain from initial domain size:
coordbase::xmin = 0.0 coordbase::xmax = 50.00 coordbase::ymin = 0.0 coordbase::ymax = 50.0 coordbase::zmin = 0.0 coordbase::zmax = 50.00
to
coordbase::xmin = 0.0 coordbase::xmax = 30.00 coordbase::ymin = 0.0 coordbase::ymax = 30.0 coordbase::zmin = 0.0 coordbase::zmax = 30.00
Is it possible? Is there already such a mechanism?
As far as I know this is not possible using checkpoint/recovery. You might be able to use the "file reader" mechanism to make CarpetIOHDF5 read data from the checkpoint files as new "initial" data, and in that case, I think the grid does not have to be identical, though the points have to line up (there is no interpolation). I'm not sure how to use this method -- maybe someone else can help with that.
Hi, Ian,
As far as I know this is not possible using checkpoint/recovery. You might be able to use the "file reader" mechanism to make CarpetIOHDF5 read data from the checkpoint files as new "initial" data, and in that case, I think the grid does not have to be identical, though the points have to line up (there is no interpolation). I'm not sure how to use this method -- maybe someone else can help with that.
Thank you for your suggestion.
Finally I could do this by CarpetIOHDF5 reader. Exactly speaking, we can use not checkpoint files but HDF5 3D data files with only a iteration slice, because reading checkpoint files derive unexpected behavior.
Kentaro TAKAMI
users@lists.einsteintoolkit.org