[gmx-users] [gmx-developers] GTX 680 non detected on OS X

Mark Abraham mark.j.abraham at gmail.com
Tue May 31 11:31:35 CEST 2016


Hi,

The point of the patch that I linked was to specifically disable static
linking in what sounds like your case. What linking resulted from it? Do
you even have shared libraries installed? What does NVIDIA do/recommend on
Mac?

Mark

On Tue, 31 May 2016 11:25 Uliano Guerrini <uliano.guerrini at unimi.it> wrote:

>
> >
> > The output you pasted into that email clearly shows that libcudart is
> > *statically* linked, so it's normal that it does not show up in the otool
> > output (you can list the symbols in your binary with some OS X tool
> similar
> > to nm if you want to verify).
> >
> > I have *no Idea* of the mechanism that is able to load libcudart without
> >> having it listed by otool -L
> >>
> >
> > The mechanism is called binary loader (if you linked statically) ;)
>
> actually it seems that libcudart *is* statically linked to libgromacs
>
> configuring with:
>
> uliano at MacPro:~/local/src/gromacs_build$ cmake ../gromacs-5.1.2
> -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DGMX_GPU=on
> -DCMAKE_PREFIX_PATH="/opt/local;/Developer/NVIDIA/CUDA-7.5/lib"
> -DCUDA_TOOLKIT_ROOT_DIR=/Developer/NVIDIA/CUDA-7.5
> -DCMAKE_INSTALL_PREFIX=/opt/gromacs  -DCUDA_HOST_COMPILER=/usr/bin/clang++
>
> I get:
>
> uliano at MacPro:~$ nm  /opt/gromacs/lib/libgromacs.1.dylib | grep cudart |
> head
> 000000000090cd70 t __ZN6cudart10configData11addArgumentEPKvmm
> 000000000090cd30 t __ZN6cudart10configData5resetE4dim3S1_mP11CUstream_st
> 000000000090cc90 t __ZN6cudart10configDataC1Ev
> 000000000090ccd0 t __ZN6cudart10configDataC2Ev
> 000000000090cd10 t __ZN6cudart10configDataD1Ev
> 000000000090cd20 t __ZN6cudart10configDataD2Ev
> 000000000092c160 t __ZN6cudart10cuosCallocEmm
> 000000000092a3e0 t __ZN6cudart10cuosGetEnvEPKcPcm
> 000000000092c140 t __ZN6cudart10cuosMallocEm
> 000000000092a3d0 t __ZN6cudart10cuosSetEnvEPKcS1_
>
> but still:
>
> NOTE: Error occurred during GPU detection:
>       CUDA driver version is insufficient for CUDA runtime version
>       Can not use GPU acceleration, will fall back to CPU kernels.
>
> so, maybe this is not a linking problem
>
> uliano
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>


More information about the gromacs.org_gmx-users mailing list