Hi Nick, Richard, Freesurfanados and GPU fans,
I've downloaded the latest _cuda binaries from the tarball you have posted, but when I try to run the recon-all on the mri_ca_register_cuda it appears: * #@# CA Reg Mon Oct 25 10:33:15 CEST 2010 /home/neuro/proc/OutPBS_GPU/AAG_106/mri
mri_ca_register_cuda -nobigventricles -T transforms/talairach.lta -align-after -mask brainmask.mgz norm.mgz /home/neuro/freesurfer/average/RB_all_2008-03-26.gca transforms/talairach.m3z
Acquiring CUDA device Using default device CUDA Error in file 'devicemanagement.cu' on line 46 : no CUDA-capable device is detected. ERROR: mri_ca_register with non-zero status 1 but continuing despite the error #-------------------------------------- #@# CA Reg Inv Mon Oct 25 10:33:15 CEST 2010 /home/neuro/proc/OutPBS_GPU/AAG_106/mri
mri_ca_register_cuda -invert-and-save transforms/talairach.m3z
Acquiring CUDA device Using default device CUDA Error in file 'devicemanagement.cu' on line 46 : no CUDA-capable device is detected. ERROR: mri_ca_register with non-zero status 0 Linux ww-neuro04 2.6.18-194.17.1.el5 #1 SMP Wed Sep 29 12:50:31 EDT 2010 x86_64 x86_64 x86_64 GNU/Linux
recon-all -s AAG_106 exited with ERRORS at Mon Oct 25 10:33:15 CEST 2010*
I think that error was solved in a previous releases of de _cuda binaries, maybe with the new compilation with cuda 3.1 it has been appeared again...
I'm running cuda on a NVIDIA GEFORCE 9800 GT with cuda toolkit 3.2.9...
Thank you!! /Jordi.
2010/10/25 Nick Schmansky nicks@nmr.mgh.harvard.edu
Freesurfanados and GPU fans,
We (namely Richard) have fixed some of the problems with the existing freesurfer gpu binaries, which were the result of building against cuda 3.0 on our build platforms, which seemed to be incompatible with 3.1. We've built against 3.1, although now i see nvidia has come out with 3.2, but i've tested against that and they seem to work.
So you can get the binary tarball here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/linux-centos4_x86_64/fscudabins-linux-centos4_x86_64.tgz
and here for the 32b linux:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/linux-centos4_x86_64/fscudabins-linux-centos4.tgz
copy the extracted contents of those tarball into your freesurfer/bin dir.
The binaries are:
mri_ca_register_cuda mri_em_register_cuda mris_fix_topology_cuda mris_flatten_cuda mris_inflate_cuda mris_sphere_cuda mri_vol2vol_cuda
No doubt people will encounter problems, so let us know. But hopefully you will see a dramatic reduction in the runtime of the subcortical segmentation stage (mri_ca_register).
Also, to remind people, Richard is targeting our gpu development for the Fermi class of gpu card, which has better memory management which leads to better performance of some algorithms, so if you're buying new hardware, be sure to spec that class of card (see nvidia site).
Nick
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.