lindahl at csb.stanford.edu
Thu Oct 16 02:55:09 CEST 2003
I forgot one thing about netcdf - we discussed using it about a year
ago, but it has one big drawback: it is another big library.
Although it isn't completely rational, I've noticed that many users
prefer to read trajectories "coordinate by coordinate" directly into
their own code instead of using the routines supplied with Gromacs.
This is particularly common for Fortran programmers - I guess the whole
library notion is less common there.
With the common formats they can just read binary numbers (in single or
double - they have to keep track of that themselves outside Gromacs),
possibly with swapping endian.
I'm not ruling out netcdf, but then we might need to create something
like a *single file* in both C and Fortran that contain all code
necessary for reading/writing files.
Anyway, it's just a thought - when I suggest to use the Gromacs
routines people are usually persistent in wanting to know the file
format to do their own IO :-)
More information about the gromacs.org_gmx-developers