[Gate-users] Trouble installing GATE with CUDA enabled

Pedro Luis Esquinas plesqui at gmail.com
Fri Apr 13 19:37:25 CEST 2018


Hi Jacob and David,

Thanks to your suggestion:

'Second: I am not at a GPU expert, but when I was testing the develop
> branch in preparation for the 8.1 release, I ran in to the same
> problem, using the same CUDA compiler (the current release, 9.1). I
> did some googling and found reports about similar/identical errors,
> and it looks like this is due to an internal CUDA compiler bug: nvcc
> crashes with a segfault when trying to compile our code. Compilers
> really shouldn't crash like that, no matter how
> bad/evil/weird/outdated our code might be. For me the error went away
> if I add "-G" to the list of compiler options (edit the CMakeLists.txt
> file, insert "-G;" in the option string in the 'SET(CUDA_NVCC_FLAGS
> ...' line). However, that is a pretty ugly blunt axe solution. It
> selects "debug" mode, which makes it *much* slower, which is very
> probably very unwelcome. Maybe this compiler bug will be fixed in the
> next CUDA release? You could maybe also try downgrading to an older
> CUDA version, maybe that makes the crash go away without speed
> penalty. I did not yet find time to try this myself, unfortunately.'

I was able to install Gate 8.0 with GPU support! (Cuda 9.0)

Thank you!

Pedro
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/pipermail/gate-users/attachments/20180413/9cfe3f74/attachment.html>


More information about the Gate-users mailing list