Refactor `valout` routine

clawpack / amrclaw

AMR version of Clawpack

http://www.clawpack.org

BSD 3-Clause "New" or "Revised" License

26 stars 46 forks source link

Refactor `valout` routine #202

Closed mandli closed 6 years ago

mandli commented 7 years ago

This refactor is the basis of an effort to add other output formats, in particular either HDF5, SILO, and/or NetCDF.

mandli commented 6 years ago

This PR now incorporates the recording of timing as outlined in clawpack/geoclaw#280. A couple of things:

Does the output look ok?
It was mentioned that we may want to also move the timing from fort.amr to it's own file as well.

Once this is figured out we can also add this to GeoClaw.

rjleveque commented 6 years ago

Yes, I think what now appears at the end of fort.amr with overall timing info could go to a new file timing.txt.

I like the timing.csv format, but it seems like a column is missing for total wall time?

I will update my plot_timing_stats.py from https://github.com/clawpack/geoclaw/pull/280 to read in this file.

The original point of this PR was to support HDF5, is this part fully working?

mandli commented 6 years ago

[x] I will modify this PR to pipe the output into a file called timing.txt.
I thought total wall time was being recorded, will look.
It was but I think all of this works independently of that right now and I think it’s better merge than to wait for that to happen.

rjleveque commented 6 years ago

We also need a column in timings.csv for the simulation time (the value printed to fort.t*) at each output time, since the plots show e.g. CPU time vs. simulation time.

mandli commented 6 years ago

Does the line call cpu_time(t_CPU_overall) not get the simulation time?

mandli commented 6 years ago

I modified the end of amr*.f90 to output to a separate file.

rjleveque commented 6 years ago

Don't merge until https://github.com/mandli/amrclaw/pull/2 is merged in to this.

rjleveque commented 6 years ago

I am going to merge this, but at some point more work is needed to make restarting timers work in 1d and 3d, see #219.

See also the discussion in https://github.com/mandli/amrclaw/pull/2