Re: amber-developers: ptraj precision

From: David A. Case <case.scripps.edu>
Date: Mon, 5 Feb 2007 18:02:15 -0800

On Mon, Feb 05, 2007, Robert Duke wrote:

> So are you really doing a sscanf(buf, "%f", &value) where value is a float,
> rather than a sscanf(buf, "%lf", &value) where value is a double?

This looks like it is "it": at least, as I read the readAmberTrajectory()
routine, it is using sscanf( "%f") to read into floats (xx,yy,zz), which are
later copied to arrays x[], y[] and z[] which are doubles. So you have the
worst of both worlds: using 8 bytes of memory for each coordinate in the
array, but only storing a single precision variable there.

Maybe I am missing something here, but this should really be fixed...

...dac
Received on Wed Feb 07 2007 - 06:07:27 PST
Custom Search