Installation pitfalls, MPI issues, compiler quirks, and runtime troubleshooting.
tddft=yes?Enables real-time TDDFT support via ce-tddft. Real-time TDDFT also supports adding or replacing external potentials during propagation.
tddft=yes python -m pip install ./QEpy
No. Use the traditional ./configure && make workflow for Quantum ESPRESSO.
Error in routine allocate_fft or fft_type_setImport qepy before mpi4py. Add import qepy as the first line of your script.
export LD_PRELOAD=/opt/intel/oneapi/mkl/latest/lib/libmkl_rt.so
export LD_PRELOAD=/opt/intel/oneapi/mpi/latest/lib/libmpifort.so
wf_collect. When in doubt, read with a single processor.qepy==6.5.0.Upgrade f90wrap to the latest release:
python -m pip install --upgrade f90wrap
QEpy and mpi4py must be built with the same Fortran/MPI compiler. Reinstall mpi4py:
python -m pip install mpi4py --no-cache-dir
BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSESWithout mpi4py, QEpy can still run in parallel (launched by MPI), but you cannot control the communicator from Python. Each new job requires restarting the script or Jupyter kernel. With mpi4py, pass comm to Driver or use comm.py2f() in lower-level calls. See tutorials.
If you see lto1: fatal error, disable link-time optimization:
FFLAGS='-fPIC -fno-lto' CFLAGS='-fPIC -fno-lto' ...
Add -fallow-argument-mismatch to FFLAGS.
zdotc segfaultAppend -ff2c to QE CFLAGS. See the QE wiki.
__malloc__ attribute errors: Intel 2021.x may need GCC 4.8–9.2 compatibility headers.mca_patcher_overwriteUpdate OpenMPI or patch shared libraries with patchelf (see legacy docs for OpenMPI 2.1.1 workaround).
Illegal Instruction: 4: add -mmacosx-version-min=10.14 to FFLAGS.clang: error: no input files: set CPP=gcc -E in make.inc.-headerpad_max_install_names to LDFLAGS.mbd_c_api.F90 with ifxDisable libmbd C API before building QE:
export LIBMBD_C_API=0
-D__OLDXML): deprecated since QE 6.4. Last QEpy version with support: qepy==6.5.0.Still stuck? Open an issue on GitHub or contact the developers.