always use double precision

[Giovanni Bussi]

Dear all,

I always use GROMACS in double precision. While I was helping Michel
Cuendet in porting PLUMED to the next version of GROMACS, I realized
that for some variables in PLUMED (such as RMSD) it is quite crucial
to employ double precision.

I guess in a future version we will implement in PLUMED completely in
double precision (except from variables passed back and forth from
GROMACS). Until that, I warmly suggest PLUMED users to compile gromacs
with --enable-double, especially if you use "complex" variables such
as RMSD with many atoms.

Giovanni

[Andrea Spitaleri]

Hi Giovanni,
I have been using plumed along with dihedral CV. is this a problem? I
will test the same simulation in double precision and I will let you
know in case of problem.
The problems you are talking about is related to the FES or to what?

thanks in advance

Regards

andrea

2009/11/20 Giovanni Bussi <giovann...@gmail.com>:

> --
>
> You received this message because you are subscribed to the Google Groups "PLUMED users" group.
> To post to this group, send email to plumed...@googlegroups.com.
> To unsubscribe from this group, send email to plumed-users...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/plumed-users?hl=.
>
>
>



--
----------------------------------------------------------------
http://sites.google.com/site/andreaspitaleri/
----------------------------------------------------------------
http://it.linkedin.com/in/andreaspitaleri
----------------------------------------------------------------

[Giovanni Bussi]

Hi Andrea,

I think the problem should be relevant only with "difficult" CVs. I
would guess no problem with dihedrals. Anyway, you can also use the
"-rerun" GROMAS option (or the plumed driver) to recalculate the
COLVARs with a double precision version. If you have numbers which are
very similar it's ok.

(indeed I did this on a run of Michel with RMSD and with single
precision I was seeing some spike in the trajectory)

Maybe RMSD experts can say something about using RMSD in single precision.

Giovanni

[Andrea Spitaleri]

Hi Giovanni,
thanks for your reply. I had the chance to test plumed in double
precision and I confirm that dihedral CV is not affected by this
"problem". The FES generated by double precision binary is identical
to the one in single precision.

However, I would like to know why with the -rerun option the file
HILLS and COLVAR are written every 6 ps rather than 0.12 (as the
original one) since my plumed.dat is such:

PRINT W_STRIDE 600
HILLS HEIGHT 1.0 W_STRIDE 60

thanks in advance

Regards

andrea

2009/11/23 Giovanni Bussi <giovann...@gmail.com>:

[Fabio Pietrucci]

Dear Andrea,

probably this is because your atomic trajectory file was saved every 6ps, so with -rerun you cannot have higher resolution than this.

ciao!

    fabio

For more options, visit this group at http://groups.google.com/group/plumed-users?hl=en.

--
____________________
CECAM Lausanne
homepage: http://sites.google.com/site/fabiopietrucci