clawpack/public

Kyle Mandli

@mandli

@ANDESD There's a bunch of stuff that could be going wrong although the most likely problem is that your parameters are not being saved out replacing the old ones. Are you running make .data after you change the parameters?

David Ketcheson

@ketch

@gadam01 what is the error you are getting? Can you post your code on Github?

David Ketcheson

@ketch

@gadam01 here is an example of how to use your own Riemann solver: ketch/pyclaw@b1e3561

chqiao

@chqiao

Hello, everyone! I noticed that there is a 'run_tests.py' file in plane_wave folder of multi-layer example of geoclaw, when I run this file, it prompt there no module named 'batch', then I searched it in Synaptic manager of Ubuntu, but no appropriate items found. So can anyone give me some suggestion? Any suggestion will be appreciate.

Kyle Mandli

@mandli

Heh, sorry about that. The batch module is an add-on you can get from my github account. I use it to run parameter sweeps or testing easily with Clawpack setups.

chqiao

@chqiao

Hi, mandli. Thank you so much for your help, I will take a try.

Kyle Mandli

@mandli

We eventually would like to incorporate this functionality into Clawpack but we wanted to make it more broadly applicable and replace some standard Python modules.

chqiao

@chqiao

Batch is a utility module and the two-layer SWE is so attractive. I need more time to acquainted with it.

chqiao

@chqiao

Hi, mandli, in the multi-layer example of geoclaw i.e. the plane_wave folder, the 'setrun.py' use a function named transform_p2c(x, y, x0, y0, theta). Just as the name implied the function do a rotation transform, but I don't know exactly the 'theta' is the angle between which? Angle between line which connecting (x0, y0) (x,y) and the x-axis? Wish your help!

Kyle Mandli

@mandli

If I remember right theta refers to the angle between the x-axis and a vector pointing in the direction of travel of the plane-wave. This means theta = 0.0 will produce a plane-wave that would be uniform in y and vary in x only. theta = pi / 2 would do the opposite.

chqiao

@chqiao

Thanks, this helpful!

tyler-abbot

@tyler-abbot

Just reading through the documentation and curioud about reflecting boundaries. Is this implied by option '3', or 'wall' ? eg if I wanted partial q/partial x = 0 on a boundary?

tyler-abbot

@tyler-abbot

And on that note, I am having trouble using a wall in the simple advection example. When I simply change the bc to extrap, it works fine, but when I change it to wallI get the following traceback:

  File "/home/tmabbot/anaconda/lib/python2.7/site-packages/clawpack/pyclaw/solver.py", line 474, in _bc_lower
    array[self.reflect_index[idim],i,...] = -array[self.reflect_index[idim],2*self.num_ghost-1-i,...] # Negate normal velocity
IndexError: index 1 is out of bounds for axis 0 with size 1

Still trying to get the feel for everything Clawpack is doing, so any guidance would be appreciated.

David Ketcheson

@ketch

@tyler-abbot What does a "wall" boundary condition mean for advection? I think the code is right to complain there.

Wall does mean "reflecting", but it assumes that reflecting boundary conditions can be implemented in a very particular way; see http://www.clawpack.org/bc.html?highlight=wall

tyler-abbot

@tyler-abbot

@ketch No idea, just thought I'd try and see if it worked. I'm trying to understand how the functions work and thought I would try different boundary conditions. I guess I just don't understand what the wall boundary is.

tyler-abbot

@tyler-abbot

So if I'm reading solver.py correctly, the wall is a reflecting boundary, fixing the gradient to be zero along the normal vector to the boundary?

If so, my question then is why this function throws an error. I understand why mathematically, but am curious what the mechanics of the program that cause this error.

Kyle Mandli

@mandli

The technical reason above is that the default reflected index is 1 (the second element of the array) where as the simple advection examples only have one field. By convention problems with a wall boundary condition often have some scalar quantity (such as density) as the first element and a momenta or velocity in the second hence the default.

tyler-abbot

@tyler-abbot

Thanks @mandli. I see now that since the boundary only has one element, this is out of range, but can't find where reflect_index is defined. Probably unimportant, but just curious.

Kyle Mandli

@mandli

I had to remind myself of this as well. The defaults are set in the subclasses's __init__ function. For the Lax-Wendroff (classic) solvers these are in src/pyclaw/classic/solver.py.

tyler-abbot

@tyler-abbot

@mandli Thanks! I was struggling to find that. One other question: is there a built in function to plot characteristics?

Kyle Mandli

@mandli

Not that I am aware of unfortunately.

tyler-abbot

@tyler-abbot

Ah! Too bad. Maybe I can come up with something general, but the plotting functions seem pretty dense. Thanks again for your help @mandli and @ketch

sophia-wright-blue

@sophia-wright-blue

Hi, I have a question on the computation time to run Example 9.4 in the book (the layered medium problem). How long does it take to run the example for dx=t=0.05 for total t = 100, for 2400 cells (as described in the example)? Thank you,

David Ketcheson

@ketch

@sophia-wright-blue It's 1D and that's a mere 2000 time steps, so should be quite fast. Depends on your computer, of course.

sophia-wright-blue

@sophia-wright-blue

thanks for replying @ketch ; i followed the method explained in the book to solve the problem using a code that I wrote (python with numpy arrays) and it takes a very long time (hours) to get to t=100 for dx=0.05. Do you have any general recommendations to speed up Python code - Numba, Cython, PyPy?

David Ketcheson

@ketch

@sophia-wright-blue The first thing I would check is how well you have vectorized the Python/numpy code itself. If you post it to Github I'd be happy to take a quick look. For something faster, why not use PyClaw?

sophia-wright-blue

@sophia-wright-blue

Thank you @ketch, I'm gonna try and vectorize the code completely. I haven't used pyclaw yet - I'm new to Numerical Modeling, and am still working with the basic Hyp PDE algorithms, so I'm writing code myself.

David Ketcheson

@ketch

@sophia-wright-blue Okay, good luck. For some ways to accelerate your code (beyond vectorization), here is a quick demo: https://gist.github.com/ketch/ae87a94f4ef0793d5d52

sophia-wright-blue

@sophia-wright-blue

@ketch Thank you for sharing that - its very helpful!

rizosss

@rizosss

hi! i tried to install clawpack in windows 8 but it does not support it. Now i'm trying to download the VM from the link that you provide, http://www.clawpack.org/VM , but appears 404 error, page do not found. Could anyone tell me where I could find the VM specific for Clawpack? Thank you in advance

Kyle Mandli

@mandli

Hi @rizosss, unfortunately we have not updated the VM available for the new version of Clawpack yet. There is an experimental docker image that's available.

http://www.clawpack.org/docker_image.html?highlight=docker

There is some interest in trying to support windows more directly with the new support for a true bash shell on windows 10 (https://blogs.windows.com/buildingapps/2016/03/30/run-bash-on-ubuntu-on-windows/#xjcmhw8ztkLhSFIy.97)

AnmolHaque

@AnmolHaque

hey hi!

I am a new user of geoclaw and also opening output files for analyzing results ( fort.bxxxx or fort.qxxxx )files

how can i open these files from an output folder and work with it?

i m using python to do the output extraction and analysis

new programmer

salikrosing

@salikrosing

I've been trying to use the BoussClaw add-on model to GeoClaw (https://www.sciencedirect.com/science/article/pii/S0378383917300546#s0010). However, it doesn't seem to gel very well with the version of ClawPack that I have installed - it gives a lot of errors for undefined references to modules in AMR.

Could this be an issue with the version I'm using (5.4.1), where BoussClaw might be developed for an older version? Does anyone have experience with this?

salikrosing

@salikrosing

Oh, wait, I just stumbled upon this: http://www.clawpack.org/claw46to50.html

I will give it a try tomorrow

salikrosing

@salikrosing

Having tried the converter tool, it seems that the issue is not with the clawpack version, at least the setrun.py file in the BoussClaw test directory uses the clawpack-5.4.1 syntax.

The error messages I am getting are all to do with the amr module, where I get undefined reference errors for amr_module_MOD _listofgrids, _amr_module_MOD _liststart and amr_module_MOD _bndlist and a few others sprinkled throughout the code.

Did those functions get replaced? And if so, by what?

David Ketcheson

@ketch

This room is officially deprecated now; please use the Clawpack Slack channel instead.

Jose Marie Antonio Minoza

@JomaMinoza

Hi @ketch, is there any invite link for Clawpack Slack channel? Thanks!

gebray21

@gebray21

I am a new user of geoclaw who spent the last few days reading through the documentation for Clawpack v5.6. I am trying to run the example in ~/clawpack/clawpack-v5.6.0/geoclaw/examples/tsunami/chile2010 and any command with make gives the warning "Makefile:65: warning: overriding recipe for target 'all'". Even make .help starts with the same warning. "Make all" gives a fatal error of " Fatal Error: Reading module ‘geoclaw_module’ at line 1 column 1: Unexpected EOF"

I can't get what is wrong with my installation.

David Ketcheson

@ketch

Here’s the slack link:

https://join.slack.com/t/clawpack/shared_invite/enQtNTAwMjQxNTk5NjY5LTJjMzJkMzhjZTQ0NzQ2Y2JkZTg3YWFhN2M4ZjIwMDA3YWNiNDM5NTA2NGNlZDA2YzlmNDkwMzA5ZGFkODgyMzc

Where communities thrive

People

Repo info

Activity