Hi FSL

I'm processing HCP data (subject 50022 in this example) with FSL version 5.0.9 on a Debian 3.2.60 with bedpostx.  I'm running into issues that seem to be slice-dependent.  I've set FSLPARALLEL=true and FSLREMOTECALL=ssh.  For some slices, notably the first 10 or so slices (depends on the subject, but always within the first 10), bedpostx fails to process those slices.

Here is my call to bedpostx:

bedpostx 500222/Diffusion_5.0.9_Test_2/ -n 3 -model 3 -b 3000 -g --rician

Here are the log and error files:

log file for log0002

Log directory is: 

500222/Diffusion_5.0.9_Test_2.bedpostX/diff_slices/data_slice_0002

Rician noise model requested. Non-linear parameter initialization will be performed, overriding other initialization options!

1/80

log file for bedpostx.e*.2 file

bedpostx_single_slice.sh: line 106: 28893 Killed                  ${FSLDIR}/bin/xfibres --data=$subjdir/data_slic

e_$slicezp --mask=$subjdir/nodif_brain_mask_slice_$slicezp --seed=$RSEED -b $subjdir/bvals -r $subjdir/bvecs --forcedir --logdir=$subjdir.bedpostX/diff_slices/dat

a_slice_$slicezp $opts > $subjdir.bedpostX/logs/log$slicezp

terminate called after throwing an instance of 'std::bad_alloc'

  what():  std::bad_alloc

log0002 does not produce any output beyond the first "1/80" and bedpostx.e*2 shows what appears to be a memory error message.  However, multiple slices produces similar errors, while others beyond the first 10 or so complete successfully.  I should note that I thought it might be a random seed-specific error and have attempted multiple random seeds -- each seed produces the same error for these slices that initially produced the error.  The call to bedpostx is what the HCP recommends for processing their data.  I'll note that I cannot reproduce these errors with model=1, but model=2 and model=3 both produce these errors.

Do you have any insight into this?

Thanks

Kristian