Compiling with CUDA V7.0
Hi,
I previously had no problems compiling the source code on Ubuntu Linux, but recently updated the CUDA toolkit on my computer. When trying to re-compile after making some minor changes to the source code, I'm getting the following error:
nvcc fatal : Unsupported gpu architecture 'compute_12'
I undid all of my changes to make sure that it wasn't something that I'd messed up, but still get the same error. Do I need to use an older version of CUDA for the code to compile or is there something wrong on my end?
Thanks,
Mike
I previously had no problems compiling the source code on Ubuntu Linux, but recently updated the CUDA toolkit on my computer. When trying to re-compile after making some minor changes to the source code, I'm getting the following error:
nvcc fatal : Unsupported gpu architecture 'compute_12'
I undid all of my changes to make sure that it wasn't something that I'd messed up, but still get the same error. Do I need to use an older version of CUDA for the code to compile or is there something wrong on my end?
Thanks,
Mike
Comments
Cuda V7.0 works very well with ubuntu 14.04. You have to go to the makefile and select your GPU architecture.
Great, thanks. Bit of a newbie question. Appreciate the help!
Mike
as i know , CUDA 7.0 does not support compute_12 anymore. the min version supported is 2.0. the latest vesion supported 1.2 is cuda 6.0
My pleasure
@Henry
I'm not sure but I think that the minium suported is 1.3.