Add fgout grid capability and support for output_format binary32 #533
Conversation
When checking times include relative buffer of 1d-13 to avoid rounding error issues in which grids at some levels get updated but not other levels at the same time. This gave some spurious looking results. Note: many debug write statements still included to help understand flow, and will be useful for further cleaning up and refactoring, then should be removed.
fg%output_arrival_time and fg%output_surface_max no longer supported. Use fgmax grids instead for these capabilities
Now based on v5.8.1 and renamed fgout for new development. * fixedgrid_output: fix comment in f90 remove debug write statements move fgrid_out to tick.f and remove deprecated max options Fix some things with fixedgrid output to make more robust
…able names Making fgout more similar to fgmax. Still need to split up module into separate subroutines and improve input/output formats.
only supports point_style==2, i.e. uniform 2d grid of points so far.
…ting Requires fgout branch of clawutil, pyclaw, visclaw
no need to define iaddqeta function as was done in valout.f
add new fgout%output_times array
grid is now cell centered, not points, to conform to output, interpolate by pw const in space by default (method=0), still need to fix up pw bilinear (method=1) near shore. fix 2x2 arrays passed in to interpolate, may have been wrong order still has problems at some fgout times coincident with frame times
Made FGOUT_ttol larger (1e-13 to 1e-6) and changed to use this uniformly in several routines so that fgout is properly updated at all levels just before an output frame time. This tolerance may need more thought.
…exception Many old setrun files instantiated but didn't use this capability, in which case only a warning is issued. But if len(self.fixedgrids)>0 i.e. the user is trying to use it, an exception is raised.
Define properties of FGoutFrame that allow access to x,y, etc. from associated FGoutGrid rather than using additional storage for every frame read in.
… when making big animation
rjleveque
changed the title
|
I still plan to add some more examples, but I think this is working well enough to merge after others review. |
|
@rjleveque I was attempting to test this and am running into problems with the |
* master: add support for binary32 gauge output revert code that prints gauges in critical section update claw_git_status for tests/dtopo1 redo tests/dtopo1 so binary gauge data produced is compared with archived ascii data for portability added binary gauge output to tests/dtopo1 support binary gauge output in gauges_module.f90 declare variable that should be integer change a comment about gauge critical section put print_gauges in critical block and use same output unit number for all
rjleveque
changed the title
|
This PR has been updated to also include support for specifying In the process, I also added support for writing the standard output frames (all AMR grids) using In An additional change is that the These changes to frame output are mixed in with the PR for new fgout capabilities because the same tools are used to read them in. For fgout grids the files produced are called To test this PR, you also need:
@mandli, the tests in |
|
The multilayer test is failing as are the examples. I think it should be an easy fix as it looks like the Makefile just needs to be adjusted as the error during compilation is that occurs in both the test and the examples. |
|
@mandli: the Makefile was already fixed to include |
…_format to fort.t file Also fixed comments in setrun.py files
|
@mandli, please see if this looks ok now. |
This (together with fgmax grids) replaces the old fixedgrid capability of GeoClaw, which has now been removed as an option.
This PR requires related PRs on visclaw, and pyclaw in order for the visualizations to work in the example provided in
examples/tsunami/chile2010_fgmax-fgout.Some documentation will soon appear in the
devbranch at https://www.clawpack.org/dev/fgout.html.This is still being actively tested and there may be more updates.