Jeff,

All build types produce the same output format (sizing depends on the file format), so you can load data produced on a Mac with a Linux machine, and data created on a 64bit machine on a 32bit, and vice-versa (ie data is platform independent).

The 64-bit aspect only pertains to the ability of some freesurfer binaries to manipulate huge data structures (>4GB) as in the case of some high-resolution data.

Nick

On Wed, 2006-03-29 at 16:12 -0600, Jeffrey Spielberg wrote:

Hi, thanks for your help. In reference to the centos_x86_64 build, is the output in 32 or 64 bit? The file sizes for the output for the redhat and centos builds were of similar size so I am assuming both are in 32 bit, but I wanted to confirm this, thanks, Jeff

On 3/28/06, Nick Schmansky nicks@nmr.mgh.harvard.edu wrote: Jeff,

    If the centos_x86_64 build works for you, then definitely use
    that.  The
    rh9 build was recommended only because many older SuSE
    installations did
    not have libc++-std.so.6, and commands would fail because of
    this. 

    The differences in precision between the rh9 and the centos
    builds is
    due to differences in floating point optimization in one of
    the math
    libraries we use.  This is being addressed in the next stable
    release, 
    but is not a concern for your existing installation.

    It is also possible that you will see minor differences in
    results due
    to randomness built-in to some of the algorithms that
    Freesurfer uses.
    These differences should be small though, and should not
    affect overall 
    results.

    Sorry, I don't know the answer to #2.

    Nick


    On Tue, 2006-03-28 at 17:34 -0600, Jeffrey Spielberg wrote:
    > Hi, I have a couple of questions I was hoping the list might
    help me
    > with,
    >
    > 1.  I am using SuSE Linux version 10 (X86-64) and previously
    I had
    > been told that the redhat 9 release was the most similar to
    SuSE 10,
    > but with the release of the CentOS 4 X86_64 release I
    wondered if this 
    > would be best to use.  To that end I ran the same subject
    with both
    > the new CentOS 4 X86_64 release and the new redhat 9 release
    and came
    > up with different results.  For example both the wm.mgz
    files produced 
    > by the two releases have small but noticeable
    differences.  Is this
    > because of rounding error due to the 64 vs 32 bit
    difference, and if
    > so does anyone know which release would be best to use, that
    is, is 
    > one going to give a better estimate, or are both going to
    give
    > slightly different, but equally good estimates?  If it's not
    due to
    > rounding error what could be the cause?
    >
    > 2.  I want to map automatic segmentation's to fsl feat
    directories, 
    > but fsl uses LAS orientation, while freesurfer uses
    RAS.  Does the
    > feat2anat or aseg/aparc2feat commands take this into
    account, or
    > should I swap the dimensions of one or the other?  If I
    should swap 
    > dimensions, would it be best to simply put my structural
    scan into
    > freesurfer in LAS format in the first place, or swap things
    after all
    > the freesurfer analysis has been done?
    >
    > thanks, 
    > Jeff
    >
    >
    > _______________________________________________
    > Freesurfer mailing list
    > Freesurfer@nmr.mgh.harvard.edu
    > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer