Hi,
On Mon, Oct 11, 2010 at 01:25:43PM -0400, Ben Roberts wrote:
>
> I have found that most of sander (and also pmemd) uses bars for its units, where pconv (in sander/runmd.f and pmemd/runmd.fpp) is set to 1.6604345d+04 * 4.184. In sander, I've updated this constant to use the values in constants.f; the conversion factor can be expressed in terms of Avogadro's number, the Joule-to-calorie conversion, and some orders of magnitude. So far, I haven't done the same for pmemd or pmemd.amba, because there doesn't seem to be the same dependence on a list of constants. In fact, in gbl_constants, I could only find a very few, such as pi.
>
So what happens to the test and benchmark results after your update ?
The consensus from the other am-dev thread on constants seems to be
that historical consistency is more important than improvements to constants.
> At the moment, then, my sander (specifically, runmd) and my pmemd are out of step. Is it best to update the pmemd source so the same values of constants are used in both?
>
What were the test and benchmark results for pmemd vs sander
before your update ? Although it would be nice for sander and pmemd to
define constants identically, interprogram consistency is the highest priority.
scott
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Oct 11 2010 - 22:00:02 PDT
