Hi,
configure_cmake.py is a work in progress, but its intentions are
use command line options, avoid punch card era UPPERCASE, have
sensible defaults, ...
cd build
./configure_cmake.py --cuda
I think you can then do this
make pmemd.bla
There is something in the wiki cmake docs about building individual
programs; and guess what ? ! ?
configure_cmake.py -h
will spew that url.
scott
> How do I separately build pmemd.cuda_* without all of its annoying friends
> I don't want to invite to the party?
On Tue, Mar 08, 2022 at 10:23:56AM -0800, Scott Le Grand wrote:
> So I see I have to physicall modify run_cmake to support CUDA. Why not
> parse the command line and override defaults to reproduce the best part
> about the configure script we just had to deprecate because reasons? I
> still cannot build pmemd.cuda_SPFP which is all I want to do here so I can
> integrate the energy drift fix.
>
> On Tue, Mar 8, 2022 at 4:46 AM David A Case <david.case.rutgers.edu> wrote:
>
> > On Mon, Mar 07, 2022, Scott Le Grand wrote:
> >
> > >For example:
> > >gfortran --version
> > >GNU Fortran (Ubuntu 8.4.0-1ubuntu1~16.04.1) 8.4.0
> > >
> > >but dumping the correct variables within cmake I get:
> > >CMAKE_CXX_COMPILER_VERSION=5.4.0
> > >CMAKE_Fortran_COMPILER_VERSION=5.4.0
> > >CMAKE_Fortran_COMPILER_VERSION=5.4.0
> >
> > Did you remove the CMakeCache.txt file that might be left over from a
> > previous run? Of, run the "clean_build" script in the build folder?
> >
> > The lowest version of GNU compilers now accepted is version 6, since (among
> > other things) that is required by the NVIDIA 11.x toolkits. We had
> > discussed getting a early check on this into the cmake build, but I guess
> > that never got accomplished.....
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Mar 08 2022 - 11:00:02 PST
