gcc versions later than 6 are not supported

[mdhimes]

Hello,

When trying to build the software via make, I get the following error: In file included from /usr/include/host_config.h:50:0, from /usr/include/cuda_runtime.h:78, from :0: /usr/include/crt/host_config.h:121:2: error: #error -- unsupported GNU version! gcc versions later than 6 are not supported!

Any plans to update the software to be compatible with gcc versions > 6?

[sbastrakov]

Hello @mdhimes . This limitation might come not from our build scripts, but from your CUDA version, perhaps it is < 9.1.85 ? At least, according to our experience gcc > 6 is not compatible with earlier CUDA versions.

[mdhimes]

Thanks for the prompt response. I am using a Titan Xp, so it's certainly possible that I do have CUDA version < 9.1.85. I will check this once I am able to do so. I am currently running memtest on the machine to rule out issues with the RAM, and will check the GPU RAM after.

[ax3l]

Thanks for your report and thank you for using cuda_memtest.

Yes, so this is not a problem with cuda_memtest but with your local setup. You can either proceed by:

• Using and older host compiler, e.g. via nvcc -ccbin <path/to/compiler> something.cu or export CXX=$(which g++-5); export CC=$(which gcc-5) before CMake, or by
• upgrading your CUDA drivers and Toolkit to CUDA 9.2 or newer, which as linked above works well with newer GCC releases and therefore this project.

CUDA/NVCC host-compiler restrictions are set by Nvidia and we do not have influence on that. Basically, it's usually the compiler versions that were available on a CUDA release, so Nvidia could adopt for and test against those.