[AMBER] To All - Consider Switching to NetCDF

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Thu, 11 Nov 2010 08:58:30 -0500

Hi everyone,

Given the recent series of emails on the list regarding trajectory
corruption I would like to take a moment to urge people to switch to
the NetCDF trajectory format in Amber. I feel like this is
particularly relevant for people who are doing relatively long
simulations and/or studying very large systems. When standard text
trajectory file sizes start getting into the GB range the chances of a
text trajectory becoming corrupted are much greater (all it takes is a
few bytes written as NULL somewhere around byte 103240244 and poof,
there goes everything after that point). Even if that doesn't happen,
if your large system starts translating off into the distance and you
get ***'s in your trajectory, the result is the same.

Unless there is a specific reason for using the old trajectory format
(like you do all your visualization in a Windows OS or you enjoy
opening the file in a text editor so you can see millions of numbers
or something) I urge everyone to switch to the NetCDF trajectory
format (ioutfm=1 in SANDER and PMEMD). The NetCDF format is higher
precision, more compact, faster to process, and far less error prone
than the text format (corruption of one frame doesn't mess up
subsequent frames).

Just my personal opinion.


AMBER mailing list
Received on Thu Nov 11 2010 - 06:00:04 PST
Custom Search