[MINC-users] mritoself xcorr memory problem

[Andrew Janke]

Hi mishkin

This memory issue is known to happen with the current mincmorph + hdf
combination.  It is due to a combination of chunking sizes and file
size.

You will find that it is not due to filetype but the size of the file
itself.  This is of course assuming the problem is what I think it is.

This problem is fixed in mincmorph 2.1 which is due for release either
this week or early next week.

Andrew.

On 2010-06-23, Mishkin Derakhshan <mishkind at gmail.com> wrote:
> Hi,
> I have found that every now and then when I run mritoself with the
> xcorr option it will take up as much memory as it can find and never
> complete.
>
> I've managed to narrow the problem down to a data type problem,
> specifically if my source image is of type long.
> In this case, mritoself, or more precisely volume_stats according to
> top, will just eat up all 8GB of my memory.
>
> This fails:
> mritoself brain.mnc.gz head.mnc.gz out.xfm -clobber -xcorr
>
> where mincinfo brain.mnc.gz gives:
> file: brain.mnc.gz
> image: signed__ long -2147483648 to 2147483647
> image dimensions: zspace yspace xspace
>     dimension name         length         step        start
>     --------------         ------         ----        -----
>     zspace                     60            3          -48
>     yspace                    256     0.976562     -100.586
>     xspace                    256     0.976562     -124.023
>
> Whereas if I do this, then it works:
> mincresample brain.mnc.gz -short short.mnc
> mincinfo short.mnc
> file: short.mnc
> image: signed__ short -32768 to 32767
> image dimensions: zspace yspace xspace
>     dimension name         length         step        start
>     --------------         ------         ----        -----
>     zspace                     60            3          -48
>     yspace                    256     0.976562     -100.586
>     xspace                    256     0.976562     -124.023
>
> mritoself short.mnc head.mnc.gz out.xfm -clobber -xcorr
>
>
> I've managed to find a fix for my problem but perhaps someone with a
> bit more knowledge of what is going on can fix this bug (or feature
> depending on what you wnt to call it).
>
> mishkin
>
> fyi:
>
> opus[~]$ uname -a
> Linux opus 2.6.24-24-generic #1 SMP Tue Aug 18 16:22:17 UTC 2009
> x86_64 GNU/Linux
>
> opus[~]$ mritoself -version
> mritoself 0.99.3 built from:
> Package mni_autoreg 0.99.3, compiled by ladmin at ccr-twodogs
> (x86_64-unknown-linux-gnu) on Wed Feb 11 10:46:24 EST 2009
>
> opus[~]$ mincinfo -version
> program: 2.0.17
> libminc: 2.0.17
> netcdf : "3.6.2" of Feb  8 2008 15:52:05 $
> HDF5   : 1.6.5
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>