Hi,
Using current GIT master also fails with Intel 15.0.3 compilers and
Python 3.4.2 :: Anaconda 2.0.1 (64-bit)
producing somewhat different errors:
...
make[3]: Leaving directory `/tmp/amber/AmberTools/src/pysander'
(cd pymdgx && /usr/local/anaconda/anaconda3/bin/python setup.py install --prefix=/tmp/amber)
Error compiling Cython file:
------------------------------------------------------------
...
int nCPUcluster
# lgrp* CPUcluster
# imat SystemCPUs
# int* MySystemDomain
Coordinates ReadRst(AmberPrmtop *tp, char* source);
^
------------------------------------------------------------
mdgx/mdgx.pxd:129:54: Syntax error in C variable declaration
... many more like above which also smells like a missing mdgx.h
Compiling mdgx/mdgx.pyx because it changed.
Cythonizing mdgx/mdgx.pyx
Traceback (most recent call last):
File "setup.py", line 40, in <module>
compiler_directives=cython_directives,
File "/usr/local/anaconda/anaconda3/lib/python3.4/site-packages/Cython/Build/Dependencies.py", line 785, in cythonize
cythonize_one(*args[1:])
File "/usr/local/anaconda/anaconda3/lib/python3.4/site-packages/Cython/Build/Dependencies.py", line 902, in cythonize_one
raise CompileError(None, pyx_file)
Cython.Compiler.Errors.CompileError: mdgx/mdgx.pyx
make[2]: *** [pymdgx] Error 1
Note that there is an mdgx.h in the logical place:
56 -rw-r--r-- 1 me apes 56652 Dec 7 13:31 /tmp/amber/include/mdgx.h
scott
On Tue, Dec 13, 2016 at 12:59:34AM -0500, Scott Brozell wrote:
> FWIW, which probably isn't much, my routine building has been working:
> make veryclean
> ./configure -noX11 --skip-python intel
> make
> icc version 15.0.3 (gcc version 4.4.7 compatibility)
>
> I'll try a non-skipping python build.
>
> scott
>
> On Mon, Dec 12, 2016 at 11:54:46PM -0500, David Cerutti wrote:
> > I don't have access to the most current Intel compilers, I think. But I
> > have just recompiled the serial version with both GNU 4.9 and an older
> > Intel compiler on different systems without problems. Dan, can you advise
> > me on what steps I should take, or if you don't mind looking into the
> > problem a bit from your end, suggest things I could do or tell you about
> > that might help you solve it?
> >
> > On Mon, Dec 12, 2016 at 4:32 PM, David Cerutti <dscerutti.gmail.com> wrote:
> >
> > > When I try to compile with intel, I am getting stuck at the configure
> > > stage:
> > >
> > > Here's the output:
> > > Checking NetCDF...
> > > Using bundled NetCDF library.
> > > NetCDF must be rebuilt.
> > > Configuring NetCDF C interface (may be time-consuming)...
> > > Compiling the NetCDF C interface (may be time-consuming)...
> > > Configuring NetCDF Fortran interface (may be time-consuming)...
> > > Compiling the NetCDF Fortran interface (may be time-consuming)...
> > > Error: Could not compile with NetCDF Fortran interface.
> > > /opt/software/ClusterStudio/2013sp1/composer_xe_2013_sp1.0.080/bin/intel64/ifort
> > > -fPIC -I/opt/software/ClusterStudio/2013sp1/composer_xe_2013_sp1.0.080/mkl/include/intel64/lp64
> > > -I/opt/software/ClusterStudio/2013sp1/composer_xe_2013_sp1.0.080/mkl/include/fftw
> > > -I/opt/software/ClusterStudio/2013sp1/composer_xe_2013_sp1.0.080/mkl/include
> > > -I/mnt/home/cerutti/amber/include -o testp testp.f90
> > > /mnt/home/cerutti/amber/lib/libnetcdff.a /mnt/home/cerutti/amber/lib/
> > > libnetcdf.a
> > > Compile error follows:
> > > /mnt/home/cerutti/amber/lib/libnetcdff.a(netcdf.o): In function
> > > `netcdf_mp_nf90_inq_libvers_':
> > > netcdf.f90:(.text+0x25): undefined reference to `nf_inq_libvers_'
> > >
> > > There is a similar thread on the Amber listserv, unanswered. I've tried
> > > extending LD_LIBRARY_PATH, but to no avail, and the thing configures and
> > > compiles just fine with GNU. If I did anything in the past six months that
> > > would have created this problem, I might not have seen it due to being
> > > limited to GNU. Does anyone have an idea of what could be the problem
> > > compiling with the NetCDF Fortran interface?
> > >
> > > On Mon, Dec 12, 2016 at 3:08 PM, Hai Nguyen <nhai.qn.gmail.com> wrote:
> > >
> > >> On Mon, Dec 12, 2016 at 2:55 PM, Daniel Roe <daniel.r.roe.gmail.com>
> > >> wrote:
> > >>
> > >> > Using current GIT master (3e7cad1d67407ab152e5faa072984f6750855a86)
> > >> > with Intel 17.0.0 compilers.
> > >> >
> > >> > Failure:
> > >> >
> > >> > ```
> > >> > building 'mdgx.mdgx' extension
> > >> > gcc -pthread -fno-strict-aliasing -g -O2 -DNDEBUG -g -fwrapv -O3 -Wall
> > >> > -Wstrict-prototypes -fPIC
> > >> > -I/uufs/chpc.utah.edu/common/home/u0827715/Amber/GIT/amber-
> > >> > lonepeak/include
> > >> > -I/uufs/chpc.utah.edu/common/home/u0827715/Amber/GIT/amber-
> > >> > lonepeak/miniconda/include/python2.7
> > >> > -c
> > >> > <http://chpc.utah.edu/common/home/u0827715/Amber/GIT/amber-l
> > >> onepeak/include%0A-I/uufs/chpc.utah.edu/common/home/u0827715
> > >> /Amber/GIT/amber-lonepeak/miniconda/include/python2.7%0A-c>
> > >> > mdgx/mdgx.c -o build/temp.linux-x86_64-2.7/mdgx/mdgx.o -O0 -ggdb
> > >> > mdgx/mdgx.c:461:18: fatal error: mdgx.h: No such file or directory
> > >> > #include "mdgx.h"
> > >> > ```
> > >> >
> > >> >
> > >>
> > >> > So mdgx.h is not found,
> > >>
> > >>
> > >> Need Dave to fix mdgx. pymdgx finds mdgx.h in $AMBERHOME/inblude
> > >> https://github.com/Amber-MD/pymdgx/blob/master/setup.py#L14
> > >>
> > >>
> > >>
> > >> > and also it's using gcc instead of icc.
> > >> >
> > >> >
> > >> You catch me. I have never used icc (since I can not test it on travis).
> > >> May be I just exclude pymdgx build if using icc.
> > >>
> > >> Hai
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Dec 13 2016 - 00:00:03 PST
