Hi!
I will like to try the beta CUDA support in 5.0. And I have some questions here
1. Are all needed libraries bundled in the release? Or I need to go to CUDA's site to get those dependency? And if I need to, what are those dependency I need to grab and install? 2. I have RH 5 64 bit linux box with single NVIDIA FX card and macpro snow leopard box with dual NVIIDA GT 120. Does CUDA support work for both OS? If they do, any idea what the speed up will be??
Gen ________________________________ This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator
Not to hijack your question, but I have similar questions. I have CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... 32767 CUDA enabled device(s) detected." but, when I run recon-all, e.g.:
recon-all -subjid PC -all -use-gpu
I get the following errors at the mri_em_register_cuda step:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri
mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta
mri_em_register_cuda: Command not found. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux
A quick peak in /usr/freesurfer/bin shows that there is no mri_em_register_cuda binary.
Are cuda binaries supposed to be bundled with the new 5.0 release, or are they created on the fly by FreeSurfer, or are they just plain missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer installed, running on OpenSuse 11.3
recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010
On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen gennan.chen@synarc.comwrote:
Hi!
I will like to try the beta CUDA support in 5.0. And I have some questions here
- Are all needed libraries bundled in the release? Or I need to go to
CUDA’s site to get those dependency? And if I need to, what are those dependency I need to grab and install? 2. I have RH 5 64 bit linux box with single NVIDIA FX card and macpro snow leopard box with dual NVIIDA GT 120. Does CUDA support work for both OS? If they do, any idea what the speed up will be??
Gen
This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator
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.
currently the _cuda binaries are distributed only in the centos4 and centos4_x86_64 distributions. however, those should work in centos5, and i've made a tarball of the _cuda binaries downloadable from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc
look for the file 'fscudabins-linux-centos4_x86_64.tgz' and untar in your freesurfer/bin dir.
! note ! this is a beta release of these binaries! they havent been fully tested across a wide subject set, so don't use the -use-gpu flag in any kind of important group analysis yet.
please let of us know though of any problems you encounter. we need to learn what kind of gpu cards people have out there in order to gauge how far we can optimize our routines.
n.
On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote:
Not to hijack your question, but I have similar questions. I have CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... 32767 CUDA enabled device(s) detected." but, when I run recon-all, e.g.:
recon-all -subjid PC -all -use-gpu
I get the following errors at the mri_em_register_cuda step:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta mri_em_register_cuda: Command not found. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/LinuxA quick peak in /usr/freesurfer/bin shows that there is no mri_em_register_cuda binary.
Are cuda binaries supposed to be bundled with the new 5.0 release, or are they created on the fly by FreeSurfer, or are they just plain missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer installed, running on OpenSuse 11.3
recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010
On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen gennan.chen@synarc.com wrote: Hi!
I will like to try the beta CUDA support in 5.0. And I have some questions here 1. Are all needed libraries bundled in the release? Or I need to go to CUDA’s site to get those dependency? And if I need to, what are those dependency I need to grab and install? 2. I have RH 5 64 bit linux box with single NVIDIA FX card and macpro snow leopard box with dual NVIIDA GT 120. Does CUDA support work for both OS? If they do, any idea what the speed up will be?? Gen ______________________________________________________________ This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator _______________________________________________ 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.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
i forgot to add: make sure you have the latest nvidia driver, which has support for cuda 3.
n.
On Tue, 2010-08-17 at 13:00 -0400, Nick Schmansky wrote:
currently the _cuda binaries are distributed only in the centos4 and centos4_x86_64 distributions. however, those should work in centos5, and i've made a tarball of the _cuda binaries downloadable from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc
look for the file 'fscudabins-linux-centos4_x86_64.tgz' and untar in your freesurfer/bin dir.
! note ! this is a beta release of these binaries! they havent been fully tested across a wide subject set, so don't use the -use-gpu flag in any kind of important group analysis yet.
please let of us know though of any problems you encounter. we need to learn what kind of gpu cards people have out there in order to gauge how far we can optimize our routines.
n.
On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote:
Not to hijack your question, but I have similar questions. I have CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... 32767 CUDA enabled device(s) detected." but, when I run recon-all, e.g.:
recon-all -subjid PC -all -use-gpu
I get the following errors at the mri_em_register_cuda step:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta mri_em_register_cuda: Command not found. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/LinuxA quick peak in /usr/freesurfer/bin shows that there is no mri_em_register_cuda binary.
Are cuda binaries supposed to be bundled with the new 5.0 release, or are they created on the fly by FreeSurfer, or are they just plain missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer installed, running on OpenSuse 11.3
recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010
On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen gennan.chen@synarc.com wrote: Hi!
I will like to try the beta CUDA support in 5.0. And I have some questions here 1. Are all needed libraries bundled in the release? Or I need to go to CUDA’s site to get those dependency? And if I need to, what are those dependency I need to grab and install? 2. I have RH 5 64 bit linux box with single NVIDIA FX card and macpro snow leopard box with dual NVIIDA GT 120. Does CUDA support work for both OS? If they do, any idea what the speed up will be?? Gen ______________________________________________________________ This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator _______________________________________________ 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.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Thx. Nick. Will give it a shot. BTW, is there also a plan for using OPENMP, MKL or ATLAS to speed it up a bit? A lot of off the shelf rack mounted servers cannot put those cards in.
Gen
On 8/17/10 10:03 AM, "Nick Schmansky" nicks@nmr.mgh.harvard.edu wrote:
i forgot to add: make sure you have the latest nvidia driver, which has support for cuda 3.
n.
On Tue, 2010-08-17 at 13:00 -0400, Nick Schmansky wrote:
currently the _cuda binaries are distributed only in the centos4 and centos4_x86_64 distributions. however, those should work in centos5, and i've made a tarball of the _cuda binaries downloadable from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc
look for the file 'fscudabins-linux-centos4_x86_64.tgz' and untar in your freesurfer/bin dir.
! note ! this is a beta release of these binaries! they havent been fully tested across a wide subject set, so don't use the -use-gpu flag in any kind of important group analysis yet.
please let of us know though of any problems you encounter. we need to learn what kind of gpu cards people have out there in order to gauge how far we can optimize our routines.
n.
On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote:
Not to hijack your question, but I have similar questions. I have CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... 32767 CUDA enabled device(s) detected." but, when I run recon-all, e.g.:
recon-all -subjid PC -all -use-gpu
I get the following errors at the mri_em_register_cuda step:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta mri_em_register_cuda: Command not found. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/LinuxA quick peak in /usr/freesurfer/bin shows that there is no mri_em_register_cuda binary.
Are cuda binaries supposed to be bundled with the new 5.0 release, or are they created on the fly by FreeSurfer, or are they just plain missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer installed, running on OpenSuse 11.3
recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010
On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen gennan.chen@synarc.com wrote: Hi!
I will like to try the beta CUDA support in 5.0. And I have some questions here 1. Are all needed libraries bundled in the release? Or I need to go to CUDA's site to get those dependency? And if I need to, what are those dependency I need to grab and install? 2. I have RH 5 64 bit linux box with single NVIDIA FX card and macpro snow leopard box with dual NVIIDA GT 120. Does CUDA support work for both OS? If they do, any idea what the speed up will be?? Gen ______________________________________________________________ This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator _______________________________________________ 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.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
________________________________ This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator
Nick-
I Dl'ed the tarball, unpacked, and re-ran. I received the following error message:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri
mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta
mri_em_register_cuda: error while loading shared libraries: libcudart.so.3: wrong ELF class: ELFCLASS32 Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux
recon-all -s PC exited with ERRORS at Tue Aug 17 12:13:42 CDT 2010
Thoughts?
On Tue, Aug 17, 2010 at 12:03 PM, Nick Schmansky nicks@nmr.mgh.harvard.eduwrote:
i forgot to add: make sure you have the latest nvidia driver, which has support for cuda 3.
n.
On Tue, 2010-08-17 at 13:00 -0400, Nick Schmansky wrote:
currently the _cuda binaries are distributed only in the centos4 and centos4_x86_64 distributions. however, those should work in centos5, and i've made a tarball of the _cuda binaries downloadable from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc
look for the file 'fscudabins-linux-centos4_x86_64.tgz' and untar in your freesurfer/bin dir.
! note ! this is a beta release of these binaries! they havent been fully tested across a wide subject set, so don't use the -use-gpu flag in any kind of important group analysis yet.
please let of us know though of any problems you encounter. we need to learn what kind of gpu cards people have out there in order to gauge how far we can optimize our routines.
n.
On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote:
Not to hijack your question, but I have similar questions. I have CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... 32767 CUDA enabled device(s) detected." but, when I run recon-all, e.g.:
recon-all -subjid PC -all -use-gpu
I get the following errors at the mri_em_register_cuda step:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta mri_em_register_cuda: Command not found. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/LinuxA quick peak in /usr/freesurfer/bin shows that there is no mri_em_register_cuda binary.
Are cuda binaries supposed to be bundled with the new 5.0 release, or are they created on the fly by FreeSurfer, or are they just plain missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer installed, running on OpenSuse 11.3
recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010
On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen gennan.chen@synarc.com wrote: Hi!
I will like to try the beta CUDA support in 5.0. And I have some questions here 1. Are all needed libraries bundled in the release? Or I need to go to CUDA’s site to get those dependency? And if I need to, what are those dependency I need to grab and install? 2. I have RH 5 64 bit linux box with single NVIDIA FX card and macpro snow leopard box with dual NVIIDA GT 120. Does CUDA support work for both OS? If they do, any idea what the speed up will be?? Gen ______________________________________________________________ This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. If you are the intended recipient, please be advised that the content of this message is subject to access, review and disclosure by the sender's Email System Administrator _______________________________________________ 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.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
make sure your LD_LIBRARY_PATH is pointing at:
/usr/local/cuda/lib64
which has the 64b libcudart, and not:
/usr/local/cuda/lib
which has the 32b.
n.
On Tue, 2010-08-17 at 12:17 -0500, joel bruss wrote:
Nick-
I Dl'ed the tarball, unpacked, and re-ran. I received the following error message:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta mri_em_register_cuda: error while loading shared libraries: libcudart.so.3: wrong ELF class: ELFCLASS32 Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux recon-all -s PC exited with ERRORS at Tue Aug 17 12:13:42 CDT 2010Thoughts?
On Tue, Aug 17, 2010 at 12:03 PM, Nick Schmansky nicks@nmr.mgh.harvard.edu wrote: i forgot to add: make sure you have the latest nvidia driver, which has support for cuda 3.
n. On Tue, 2010-08-17 at 13:00 -0400, Nick Schmansky wrote: > currently the _cuda binaries are distributed only in the centos4 and > centos4_x86_64 distributions. however, those should work in centos5, > and i've made a tarball of the _cuda binaries downloadable from here: > > ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc > > look for the file 'fscudabins-linux-centos4_x86_64.tgz' and untar in > your freesurfer/bin dir. > > ! note ! this is a beta release of these binaries! they havent been > fully tested across a wide subject set, so don't use the -use-gpu flag > in any kind of important group analysis yet. > > please let of us know though of any problems you encounter. we need to > learn what kind of gpu cards people have out there in order to gauge how > far we can optimize our routines. > > n. > > On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote: > > Not to hijack your question, but I have similar questions. I have > > CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... > > 32767 CUDA enabled device(s) detected." but, when I run recon-all, > > e.g.: > > > > recon-all -subjid PC -all -use-gpu > > > > > > I get the following errors at the mri_em_register_cuda step: > > > > /lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri > > > > mri_em_register_cuda -skull > > nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca > > transforms/talairach_with_skull.lta > > > > mri_em_register_cuda: Command not found. > > Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 > > 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux > > > > A quick peak in /usr/freesurfer/bin shows that there is no > > mri_em_register_cuda binary. > > > > Are cuda binaries supposed to be bundled with the new 5.0 release, or > > are they created on the fly by FreeSurfer, or are they just plain > > missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer > > installed, running on OpenSuse 11.3 > > > > > > recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010 > > > > > > On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen <gennan.chen@synarc.com> > > wrote: > > Hi! > > > > I will like to try the beta CUDA support in 5.0. And I have > > some questions here > > > > 1. Are all needed libraries bundled in the release? Or I > > need to go to CUDA’s site to get those dependency? > > And if I need to, what are those dependency I need to > > grab and install? > > 2. I have RH 5 64 bit linux box with single NVIDIA FX > > card and macpro snow leopard box with dual NVIIDA GT > > 120. Does CUDA support work for both OS? If they do, > > any idea what the speed up will be?? > > > > Gen > > > > ______________________________________________________________ > > This email message is for the sole use of the intended > > recipient(s) and may contain confidential and privileged > > information. Any unauthorized review, use, disclosure or > > distribution is prohibited. If you are not the intended > > recipient, please contact the sender by reply email and > > destroy all copies of the original message. If you are the > > intended recipient, please be advised that the content of this > > message is subject to access, review and disclosure by the > > sender's Email System Administrator > > > > > > _______________________________________________ > > 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. > > > > > > _______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > >
Sorry to be a PITA but I'm still having issues. I pointed at cuda-64 and now I get:
mri_em_register_cuda -skull nu.mgz
/usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta
Acquiring CUDA device Using default device CUDA Error in file 'devicemanagement.cu' on line 46 : CUDA driver version is insufficient for CUDA runtime version. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux
recon-all -s PC exited with ERRORS at Tue Aug 17 13:51:07 CDT 2010
I have the latest CUDA toolkit, 3.1, as well as the latest Nvidia driver, nvidia-gfxG02-kmp-desktop. Any further thoughts?
On Tue, Aug 17, 2010 at 12:51 PM, Nick Schmansky nicks@nmr.mgh.harvard.eduwrote:
make sure your LD_LIBRARY_PATH is pointing at:
/usr/local/cuda/lib64
which has the 64b libcudart, and not:
/usr/local/cuda/lib
which has the 32b.
n.
On Tue, 2010-08-17 at 12:17 -0500, joel bruss wrote:
Nick-
I Dl'ed the tarball, unpacked, and re-ran. I received the following error message:
/lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta mri_em_register_cuda: error while loading shared libraries: libcudart.so.3: wrong ELF class: ELFCLASS32 Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux recon-all -s PC exited with ERRORS at Tue Aug 17 12:13:42 CDT 2010Thoughts?
On Tue, Aug 17, 2010 at 12:03 PM, Nick Schmansky nicks@nmr.mgh.harvard.edu wrote: i forgot to add: make sure you have the latest nvidia driver, which has support for cuda 3.
n. On Tue, 2010-08-17 at 13:00 -0400, Nick Schmansky wrote: > currently the _cuda binaries are distributed only in the centos4 and > centos4_x86_64 distributions. however, those should work in centos5, > and i've made a tarball of the _cuda binaries downloadable from here: > > ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc > > look for the file 'fscudabins-linux-centos4_x86_64.tgz' and untar in > your freesurfer/bin dir. > > ! note ! this is a beta release of these binaries! they havent been > fully tested across a wide subject set, so don't use the -use-gpu flag > in any kind of important group analysis yet. > > please let of us know though of any problems you encounter. we need to > learn what kind of gpu cards people have out there in order to gauge how > far we can optimize our routines. > > n. > > On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote: > > Not to hijack your question, but I have similar questions. I have > > CUDA toolkit 3.1 installed, "cudadetect" returns: "Detecting CUDA... > > 32767 CUDA enabled device(s) detected." but, when I run recon-all, > > e.g.: > > > > recon-all -subjid PC -all -use-gpu > > > > > > I get the following errors at the mri_em_register_cuda step: > > > > /lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri > > > > mri_em_register_cuda -skull > > nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca > > transforms/talairach_with_skull.lta > > > > mri_em_register_cuda: Command not found. > > Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 > > 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux > > > > A quick peak in /usr/freesurfer/bin shows that there is no > > mri_em_register_cuda binary. > > > > Are cuda binaries supposed to be bundled with the new 5.0 release, or > > are they created on the fly by FreeSurfer, or are they just plain > > missing? FWIW, I have CentOS 5 x86_64 version of FreeSurfer > > installed, running on OpenSuse 11.3 > > > > > > recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 CDT 2010 > > > > > > On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen <gennan.chen@synarc.com> > > wrote: > > Hi! > > > > I will like to try the beta CUDA support in 5.0. And I have > > some questions here > > > > 1. Are all needed libraries bundled in the release? Or I > > need to go to CUDA’s site to get those dependency? > > And if I need to, what are those dependency I need to > > grab and install? > > 2. I have RH 5 64 bit linux box with single NVIDIA FX > > card and macpro snow leopard box with dual NVIIDA GT > > 120. Does CUDA support work for both OS? If they do, > > any idea what the speed up will be?? > > > > Gen > > > > ______________________________________________________________ > > This email message is for the sole use of the intended > > recipient(s) and may contain confidential and privileged > > information. Any unauthorized review, use, disclosure or > > distribution is prohibited. If you are not the intended > > recipient, please contact the sender by reply email and > > destroy all copies of the original message. If you are the > > intended recipient, please be advised that the content of this > > message is subject to access, review and disclosure by the > > sender's Email System Administrator > > > > > > _______________________________________________ > > 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. > > > > > > _______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > >
no problem, eventually we'll have a help page once we gather all the new issues with this new gpu stuff.
what kind of card do you have? is it one of these:
http://www.nvidia.com/object/cuda_gpus.html
make sure your driver is the one from the site. the driver name 'nvidia-gfxG02-kmp-desktop' doesnt look familiar to me. it should be named something like 'NVIDIA-Linux-x86_64-256.44', which you should be able to get with 'nvidia-installer --update'.
n.
On Tue, 2010-08-17 at 14:18 -0500, joel bruss wrote:
Sorry to be a PITA but I'm still having issues. I pointed at cuda-64 and now I get:
mri_em_register_cuda -skull nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca transforms/talairach_with_skull.lta Acquiring CUDA device Using default device CUDA Error in file 'devicemanagement.cu' on line 46 : CUDA driver version is insufficient for CUDA runtime version. Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux recon-all -s PC exited with ERRORS at Tue Aug 17 13:51:07 CDT 2010I have the latest CUDA toolkit, 3.1, as well as the latest Nvidia driver, nvidia-gfxG02-kmp-desktop. Any further thoughts?
On Tue, Aug 17, 2010 at 12:51 PM, Nick Schmansky nicks@nmr.mgh.harvard.edu wrote: make sure your LD_LIBRARY_PATH is pointing at:
/usr/local/cuda/lib64 which has the 64b libcudart, and not: /usr/local/cuda/lib which has the 32b. n. On Tue, 2010-08-17 at 12:17 -0500, joel bruss wrote: > Nick- > > I Dl'ed the tarball, unpacked, and re-ran. I received the following > error message: > > /lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri > > mri_em_register_cuda -skull > nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca > transforms/talairach_with_skull.lta > > mri_em_register_cuda: error while loading shared libraries: > libcudart.so.3: wrong ELF class: ELFCLASS32 > Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT 2010-06-29 > 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux > > recon-all -s PC exited with ERRORS at Tue Aug 17 12:13:42 CDT > 2010 > > Thoughts? > > > On Tue, Aug 17, 2010 at 12:03 PM, Nick Schmansky > <nicks@nmr.mgh.harvard.edu> wrote: > i forgot to add: make sure you have the latest nvidia driver, > which has > support for cuda 3. > > n. > > > On Tue, 2010-08-17 at 13:00 -0400, Nick Schmansky wrote: > > currently the _cuda binaries are distributed only in the > centos4 and > > centos4_x86_64 distributions. however, those should work in > centos5, > > and i've made a tarball of the _cuda binaries downloadable > from here: > > > > ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc > > > > look for the file 'fscudabins-linux-centos4_x86_64.tgz' and > untar in > > your freesurfer/bin dir. > > > > ! note ! this is a beta release of these binaries! they > havent been > > fully tested across a wide subject set, so don't use the > -use-gpu flag > > in any kind of important group analysis yet. > > > > please let of us know though of any problems you encounter. > we need to > > learn what kind of gpu cards people have out there in order > to gauge how > > far we can optimize our routines. > > > > n. > > > > On Tue, 2010-08-17 at 11:35 -0500, joel bruss wrote: > > > Not to hijack your question, but I have similar > questions. I have > > > CUDA toolkit 3.1 installed, "cudadetect" returns: > "Detecting CUDA... > > > 32767 CUDA enabled device(s) detected." but, when I run > recon-all, > > > e.g.: > > > > > > recon-all -subjid PC -all -use-gpu > > > > > > > > > I get the following errors at the mri_em_register_cuda > step: > > > > > > > /lcn2/lesion_warp.dir/MAP3_Nifti/freesurfer_subs.dir/PC/mri > > > > > > mri_em_register_cuda -skull > > > > nu.mgz /usr/freesurfer/average/RB_all_withskull_2008-03-26.gca > > > transforms/talairach_with_skull.lta > > > > > > mri_em_register_cuda: Command not found. > > > Linux linux-lvmb 2.6.34-12-desktop #1 SMP PREEMPT > 2010-06-29 > > > 02:39:08 +0200 x86_64 x86_64 x86_64 GNU/Linux > > > > > > A quick peak in /usr/freesurfer/bin shows that there is no > > > mri_em_register_cuda binary. > > > > > > Are cuda binaries supposed to be bundled with the new 5.0 > release, or > > > are they created on the fly by FreeSurfer, or are they > just plain > > > missing? FWIW, I have CentOS 5 x86_64 version of > FreeSurfer > > > installed, running on OpenSuse 11.3 > > > > > > > > > recon-all -s PC exited with ERRORS at Tue Aug 17 11:14:35 > CDT 2010 > > > > > > > > > On Tue, Aug 17, 2010 at 11:07 AM, Gennan Chen > <gennan.chen@synarc.com> > > > wrote: > > > Hi! > > > > > > I will like to try the beta CUDA support in 5.0. > And I have > > > some questions here > > > > > > 1. Are all needed libraries bundled in the > release? Or I > > > need to go to CUDA’s site to get those > dependency? > > > And if I need to, what are those > dependency I need to > > > grab and install? > > > 2. I have RH 5 64 bit linux box with single > NVIDIA FX > > > card and macpro snow leopard box with dual > NVIIDA GT > > > 120. Does CUDA support work for both OS? > If they do, > > > any idea what the speed up will be?? > > > > > > Gen > > > > > > > ______________________________________________________________ > > > This email message is for the sole use of the > intended > > > recipient(s) and may contain confidential and > privileged > > > information. Any unauthorized review, use, > disclosure or > > > distribution is prohibited. If you are not the > intended > > > recipient, please contact the sender by reply > email and > > > destroy all copies of the original message. If you > are the > > > intended recipient, please be advised that the > content of this > > > message is subject to access, review and > disclosure by the > > > sender's Email System Administrator > > > > > > > > > _______________________________________________ > > > 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. > > > > > > > > > _______________________________________________ > > > Freesurfer mailing list > > > Freesurfer@nmr.mgh.harvard.edu > > > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > _______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > > > >
freesurfer@nmr.mgh.harvard.edu
-
Gennan Chen -
joel bruss -
Nick Schmansky