[FLASH-USERS] Are the applications listed in SimulationMain/ only test cases?
Dean Townsley
Dean.M.Townsley at ua.edu
Fri Feb 6 15:39:40 EST 2015
Hi Sheng,
As far as I know it is a mixture of the two. Some, maybe even most, of
the things in SimulationMain are just test problems and some are more
real science cases. Though typically in the latter case, as you have
mentioned, the default parameters in the parameter files distributed
with flash may not be similar to those used in science runs, since the
defaults are often those used in small scale testing rather than in full
production.
So both you and your supervisor are correct :)
To give some examples...
Sedov is largely a test problem
Sod is also largely a test problem
StirTurb I believe is a real stirred turbulence simulation used for
science runs or at least similar to those from Benzi et al.
(2008PhRvL.100w4503B <http://adsabs.harvard.edu/abs/2008PhRvL.100w4503B>).
Cellular is a detonation simulation that appears to be at least quite
similar to the basic setup used by Timmes et al. (2000ApJ...543..938T
<http://adsabs.harvard.edu/abs/2000ApJ...543..938T>). Though I think the
default parameters are a little odd due to the necessarily low
resolution and I'm unsure if the way the "noise" is added is quite what
one might expect.
Flame1StageNoise is a test problem that is, I believe, still missing the
"noise" part. (sorry about that!) But it is similar to the setup used
in Townsley et al. (2007ApJ...668.1118T
<http://adsabs.harvard.edu/abs/2007ApJ...668.1118T>) to measure acoustic
noise generated by the flame propagation technique.
RTFlame is a science setup to study Rayleigh-Taylor unstable flames (for
Type Ia supernovae). Similar to those in Khokhlov (1995ApJ...449..695K
<http://adsabs.harvard.edu/abs/1995ApJ...449..695K>), but post-dating
that work by quite a bit. I believe this has been used as an acceptance
test problem on some big machines -- but, again, not with the parameters
in the distributed parameter files.
I'm sure there are other examples from other areas for both test and
science setups, these are just some of the ones that I am familiar
with. I expect the CCSN setup is at least mostly similar to a real core
collapse supernova setup. I think there are some messages in the
mailing list archive discussing this. Also probably the difference
between a "test" problem and a science problem is somewhat blurry in
some cases. Especially when the "science" is development of a
computational technique or algorithm.
Hopefully that helps,
Dean
On 02/05/2015 05:13 PM, Di Sheng wrote:
> Hi all,
> This may be a stupid question.
> Are the applications found in the SimulationMain/ just test cases (toy
> applications)?
>
> In fact, I am looking for some real production HPC applications for
> evaluating our MPI error detector. My supervisor argued that these
> test cases are just toy applications. Instead, I need to use real
> production applications from real users.
> Could any one answer me?
>
> From my point of view, these applications are real world applications.
> The users just need to tune the input parameters to fit their testing
> demand in most of cases. Am I right?
>
> thanks!!
>
> Best
> Sheng
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20150206/cd48f860/attachment-0001.htm>
More information about the flash-users
mailing list