Dear freesurfer developers,
Anybody know which CUDAversion is the latest one supported by freesurfer? I just found the thread below. it's written in 2014, but experiences same error.Successful detectcuda command, but core dump during very first step of recon-all. https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg36769.html Now I am trying to remove CUDA 8 and installing CUDA 5.5. Next will be the CUDA5.0If anybody knows what is the 'supported' or 'successful' late cuda version, please let me know.Bests,
On 17 November 2016 at 18:10, June Kang cnsla@korea.ac.kr wrote:
Anybody know which CUDAversion is the latest one supported by freesurfer?
I'm trying to provide some support for the CUDA implementations, but I don't have that much time to devote to it.
Now I am trying to remove CUDA 8 and installing CUDA 5.5. Next will be the CUDA5.0
If anybody knows what is the 'supported' or 'successful' late cuda version, please let me know.
I know I had it running on CUDA 7.5 (with K1200), and I'm pretty sure I've run it since CUDA 8.0 came out. Rather than trying to downgrade your CUDA installation, it might be more worthwhile to compile from scratch on the box you want to use. That way, you can be sure that library versions are all matching. It also occurs to me that the nvcc flags might be so old in the default compile as to give trouble with your GPU.
Richard
On 17 November 2016 at 20:23, R Edgar freesurfer.rge@gmail.com wrote:
On 17 November 2016 at 18:10, June Kang cnsla@korea.ac.kr wrote:
Anybody know which CUDAversion is the latest one supported by freesurfer?
I'm trying to provide some support for the CUDA implementations, but I don't have that much time to devote to it.
Now I am trying to remove CUDA 8 and installing CUDA 5.5. Next will be the CUDA5.0
If anybody knows what is the 'supported' or 'successful' late cuda version, please let me know.
I know I had it running on CUDA 7.5 (with K1200), and I'm pretty sure I've run it since CUDA 8.0 came out. Rather than trying to downgrade your CUDA installation, it might be more worthwhile to compile from scratch on the box you want to use. That way, you can be sure that library versions are all matching. It also occurs to me that the nvcc flags might be so old in the default compile as to give trouble with your GPU.
OK, on my box: [rge21@cudastation gpu_utils]$ ./cudadetect Detecting CUDA... There is 1 device supporting CUDA:
Device 0: "Quadro K1200" CUDA Driver Version: 8.0 CUDA Runtime Version: 8.0 CUDA Capability Major revision number: 5 CUDA Capability Minor revision number: 0
Then I can run mri_ca_register: ../mri_ca_register_cuda -nobigventricles -T talairach.lta -align-after -mask brainmask.mgz norm.mgz ../../distribution/average/RB_all_2016-05-10.vc700.gca talairach.m3z nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2016 NVIDIA Corporation Built on Sun_Sep__4_22:14:01_CDT_2016 Cuda compilation tools, release 8.0, V8.0.44
Driver : 8.0 Runtime : 8.0
Acquiring CUDA device Using default device CUDA device: Quadro K1200 not handling expanded ventricles... using previously computed transform talairach.lta renormalizing sequences with structure alignment, equivalent to: -renormalize -regularize_mean 0.500 -regularize 0.500 using MR volume brainmask.mgz to mask input volume...
== Number of threads available to ../mri_ca_register_cuda for OpenMP = 4 == reading 1 input volumes... logging results to talairach.log reading input volume 'norm.mgz'... INFO: MRImask() using MRImaskDifferentGeometry() INFO: MRImask() using MRImaskDifferentGeometry() INFO: MRImask() using MRImaskDifferentGeometry() INFO: MRImask() using MRImaskDifferentGeometry() INFO: MRImask() using MRImaskDifferentGeometry() reading GCA '../../distribution/average/RB_all_2016-05-10.vc700.gca'... label assignment complete, 0 changed (0.00%) det(m_affine) = 1.19 (predicted orig area = 6.7) freeing gibbs priors...done. average std[0] = 5.0 AllocateHost: Warning - not thread safe! RecvAll: Did not reset gca in dst GCAMcopyNodePositions: On GPU RecvAll: Did not reset gca in dst etc.
Richard
freesurfer@nmr.mgh.harvard.edu
-
June Kang -
R Edgar