Those changes are already made. I did some checking in the gpu.cpp code
and there are other things there we can do to touch up the barriers.
On Sat, Mar 24, 2018 at 9:25 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
> Ref 2 - probably better to just skip compiling for that architecture. See
> $AMBERHOME/AmberTools/src/configure2
>
> Just change
>
> nvccflags="$sm20flags $sm30flags $sm50flags $sm52flags $sm53flags
> $sm60flags $sm61flags"
>
> to
>
> nvccflags="$sm30flags $sm50flags $sm52flags $sm53flags $sm60flags
> $sm61flags"
>
> Note CUDA 9.0 and later doesn't support SM20 anymore anyway so the
> GetDeviceProperties won't work since if someone compiles with CUDA 9 or
> later it won't run on SM20 anyway.
>
> Note we should also be sure to enable CUDA 9.1 support in configure2.
>
> I can make these changes if you want . Do I just edit directly in the Git
> master and send you the commit log?
>
> All the best
> Ross
>
> > On Mar 24, 2018, at 14:43, Ruxi Qi <ruxiq.uci.edu> wrote:
> >
> > Hi Dave,
> >
> > For number 2, CUDA API function cudaGetDeviceProperties can be used to
> > query the device compute capability. I'd like to help if needed.
> >
> > Best,
> >
> > Ruxi
> >
> >
> > On 03/24/2018 07:38 AM, David A Case wrote:
> >> Hi everyone:
> >>
> >> I was very gratified with the help that was offered in my previous
> >> request for volunteers, so here goes part 2:
> >>
> >> 1. I'd like someone to test on MacOS, with both clang and gnu
> >> compilers. Ray helped out earlier (thanks!), but I'd like a third
> >> opinion here. I'm not sure I trust *my* Mac to be fully representative
> >> of what is out there.
> >>
> >> 2. We no longer support pre-Kepler (sm20) GPUs (e.g. M2070, C2050).
> >> Does anyone know how to check for that (has to be a run-time, not
> >> compile time), and exit gracefully?
> >>
> >> 3. Does anyone besides Scott have access to Intel 18.0 compilers?
> >> Basic things like the dhfr or water tests seem to be failing there. If
> >> this is reproduced, can someone put a trap in the configure2 script?
> >>
> >> 4. The PGI compilers are favorites for some people, and we need a
> >> second opinion here, vs the srb results on the wiki. If anyone has
> >> access to other versions, that might be worth a shot as well.
> >>
> >> ...thx...dac
> >>
> >>
> >>
> >> _______________________________________________
> >> AMBER-Developers mailing list
> >> AMBER-Developers.ambermd.org
> >> http://lists.ambermd.org/mailman/listinfo/amber-developers
> >
> >
> > _______________________________________________
> > AMBER-Developers mailing list
> > AMBER-Developers.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber-developers
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Sat Mar 24 2018 - 19:00:03 PDT
