PhotonTestAMR

[david collins]

Hi, all--

I've found a strange phenomenon in PhotonTestAMR.  If one adds

float * test[100];

to Grid.h, the answer changes on the test at about the 5% level.  I've mentioned this to John Wise, who has done some exploration of this.  I'm looking into it with valgrind, but haven't had much success yet.  

I don't think this is related to the other stochastic processes in Enzo. The result is bitwise across many different changesets that don't change the contents of Grid.h.  

Any input is appreciated.  I'll report back if/when I find out more.

I found this when trying to make sure I didn't mess anything up with the MHD-CT PR. While the bug it probably not due to that PR, I don't like causing changes of this order if it's not an improvement.

Thanks,

d.

[Nathan Goldbaum]

I see the same thing in enzo-3.0, probably because of the changes we've had to make in Grid.h to support active particles.  The answer test results for my run are pasted here: http://paste.yt-project.org/show/3371/

[John Wise]

Hi everyone,

I fixed this bug this morning, and now adding this test array of pointers to the grid class doesn't change the answer. An array wasn't being updated and was being accessed out of its bounds.

Nathan tells me that it does change the answer to PhotonTestAMR. The gold standard will need to be updated.

I'm going to issue a pull request now.

Cheers,
John

> --
> You received this message because you are subscribed to the Google Groups "enzo-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to enzo-dev+u...@googlegroups.com.
> To post to this group, send email to enzo...@googlegroups.com.
> Visit this group at http://groups.google.com/group/enzo-dev?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

--
John Wise
Assistant Professor of Physics
Center for Relativistic Astrophysics, Georgia Tech
http://cosmo.gatech.edu