After compiling GCC6, you need to compile LAPACK with GCC6 and add it to library path. You will likely get some gomp related errors. I got around this by manually copying files like libgomp.spec and libgomp.so into library paths.
In the git cloned directory, do:
CXX=icpc CXXL=icpc LDSHARED="icc -shared" CC=icc python setup.py build_ext -i --compiler=intelem -c icc -vvv
and add to the Python path.
On systems like Niagara, an Intel version of fftw is the default module. This has a bug which makes it impossible to FFT multi-dimensional arrays.
If you are going to compile FFTW yourself, then you have to use gcc, and this is what it will typically look like:
``
./configure CC=gcc CXX=g++ CFLAGS=-fPIC –enable-openmp –enable-mpi –enable-sse –enable-sse2 –enable-avx –enable-avx2 –enable-float –enable-shared –prefix=/home/r/rbond/msyriac/.local/lib make -j24 CFLAGS=-fPIC make install make clean
./configure CC=gcc CXX=g++ CFLAGS=-fPIC –enable-openmp –enable-mpi –enable-sse2 –enable-avx –enable-avx2 –enable-shared –prefix=/home/r/rbond/msyriac/.local make -j24 CFLAGS=-fPIC make install make clean ./configure CC=gcc CXX=g++ CFLAGS=-fPIC –enable-openmp –enable-mpi –enable-long-double –enable-shared –prefix=/home/r/rbond/msyriac/.local make -j24 CFLAGS=-fPIC make install make clean ``
and pyfftw with
pip install pyfftw --user