On Tue, Jun 25, 2013 at 10:55 AM, Clive Wong hycwong@gmail.com wrote:

I'm new to CUDA. I tried to use CUDA in my environment, it gives the message below:

Testing for CUDA device: nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2012 NVIDIA Corporation Built on Fri_Sep_21_17:28:58_PDT_2012 Cuda compilation tools, release 5.0, V0.2.1221

Driver : 5.0 Runtime : 5.0

Acquiring CUDA device Using default device

Must have compute capability 2 for GCAMORPH_ON_GPU CUDA device: Quadro NVS 420

I've checked that the Quadro NVS400 is of compute capability 1.1. and the script devicemanagement.c has blocked devices below compute capability 2.

Is there any workaround to get it work? Or shall I compile the freesurfer from source code? If yes, how could I know how to setup the configuration?

As I recall, it was a technical limitation - I believe that compute capability < 2 limits arguments to GPU kernels to being less than 256 bytes, and a lot of things to do with the GCAMORPHs cleared that by a substantial margin. It should be possible to compile a version of Freesurfer for earlier devices - you just won't get as much acceleration.

Richard