[gmx-users] precision of mixed precision

[Mark Abraham]

On Mon, Sep 15, 2014 at 8:52 PM, Johnny Lu <johnny.lu128 at gmail.com> wrote:

> Hi.
>
> Does gromacs do the following?
>
> 1. I can only write down an integer with 5 digits.
> 2. I want to record 43.41 accurately
> 3. So, I wrote 4341, and remember the actual number should be 100 times
> smaller.
> 4. By writing a number with integer precision, the double preciion number
> 43.41 was recorded accurately.
>

No, that's fixed-point precision (because 100 does not change). GROMACS
uses floating-point precision.


> If my molecule is sufficiently small, can mixed precision gromaces provide
> essentially an equally good precision as double precision gromacs? (for 50
> ns NVE simulation ).
>

Can you define "equally good"? If so, go and measure and do whatever the
measurement suggests. If not, then you should decide whether your
simulation quality is limited by the amount of sampling you can afford, or
the quality lost through the use of mixed precision. The size of your
entire system is indeed a factor here (but nobody has studied that in
GROMACS, yet).

Mark


> Thanks again.
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>