-
Notifications
You must be signed in to change notification settings - Fork 46
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Faster bound #149
Faster bound #149
Conversation
…ion a bit to move evrything into amr out of geoclaw
travis is showing an error in tests/advection_3d_swirl. I ran this branch and master in that directory and notice that But there must be more differences based on the travis error (which I also get on my laptop):
|
After changing
in
to regenerate the regression data and then do a |
i think they were run with different numbers of cores. When I ran them with the ssame number it agreed completely. But what is the 32000 from? On Oct 15, 2015, at 8:00 PM, Randall J. LeVeque [email protected] wrote:
|
I think even with the same number of cores (> 1) it could be random which gauge prints first at a given time step if they are on grids handled by different cores. At any rate this indicates a problem with our test framework. On my laptop the tests now pass if I set |
As you say the gauge output definitely depends on random ordering of grids and . When I got the same answer with both branches I was running on a single core. Don’t know where the 32000 came from though. Is that from before the filpatch bug was fixed? Marsha On Oct 16, 2015, at 12:01 PM, Randall J. LeVeque [email protected] wrote:
|
The testing framework should compare individual gauges, it does this for the summation test but not the full one (probably because I never tested this with two gauges). I raised an issue clawpack/clawutil#89. |
…x1d from 60 to 32. Also upped initial memsize to 4M
FasterBound updated to incorporate 5.3.1 changes to master
After consulting with @mandli and @mjberger, decided this is ready to merge. Note that geoclaw will be temporarily broken on master until clawpack/geoclaw#173 is merged in. Also need to update clawpack/clawpack to make these work together. |
as the name says, and is sometimes substantially faster for large numbers of grids. HAs to go with matching geoclaw fasterBound branch. Also has lots of other tiny fixes, mostly format statement enlargements, to also go with larger numbers of grids