ABINIT(1)

✨ ABINIT(1) · User Commands ✨

okay bestie, let's talk about abinit 💖 it's the Ab initio atomic-scale simulation software and it is giving main character energy 💅

---

✨ NAME ✨

abinit - Ab initio atomic-scale simulation software 🔬✨

---

✨ DESCRIPTION ✨

okay so here's every flag abinit takes, and I'm not skipping a single one 🫶

• -v, --version — shows the version number, then bounces 👋
• -b, --build — shows the build parameters, then dips 🛠️
• -d, --dry-run — validates your input file and then exits. no vibes lost 🙈
• -j, --omp-num-threads — sets the number of OpenMP threads 🧵
• --use-xgemm3m[=yesno] — uses the ZGEMM3M routines instead of ZGEMM. Default: no 🌸
• --use-mpi-in-place[=yesno] — turns usage of MPI_IN_PLACE on or off, e.g. in xmpi_sum. Default: no 🫧

⚠️ bestie, listen. Some MPI libs, like intel-mpi, may not implement this feature correctly, so it is advised to test this option with e.g. structural relaxations before running production calculations. Don't skip the test 😭

• --ipi — activates a socket-driven calculation using the i-pi protocol 🔌 For a UNIX socket use --ipi {unixsocket}:UNIX, and for an INET socket use --ipi {host}:{port}. Usage example: abinit run.abi --ipi {unixsocket}:UNIX > run.log

NB: this one requires ionmov 28 and some tuning of input variables. See: https://wiki.fysik.dtu.dk/ase/dev/ase/calculators/socketio/socketio.html 🧠

• --log — enables log files and status files in parallel execution 📝
• --netcdf-classic — uses netcdf classic mode for new files if parallel-IO isn't needed. Default is netcdf4/hdf5 🗂️
• --enforce-fortran-io — uses Fortran-IO instead of MPI-IO when operating on Fortran files. It's useful for reading files when the MPI-IO library isn't efficient 🐢

🚫 DON'T USE this option when the code needs to write large files, e.g. WFK. Seriously, not for that 💀

• -t, --timelimit — sets the timelimit for the run. It accepts Slurm syntax: days-hours, days-hours:minutes, days-hours:minutes:seconds, minutes, minutes:seconds, hours:minutes:seconds. For now, only GS, relaxations and MD runs support this option ⏰
• --mem-per-cpu=<size>[units] — sets memory per cpu using Slurm syntax. Default units are megabytes. You can specify different units with the suffix [K|M|G|T] 💾
• --mem=<size>[units] — sets memory per node using Slurm syntax. Default units are megabytes. It requires ntasks-per-node and is not compatible with -mem-per-cpu 🚫💾
• --ntasks-per-node=INT — sets the number of tasks per node. Used in conjunction with --mem 🧩
• --verbose — enables verbose mode in argparse 🗣️
• -h, --help — shows this help and exits 🆘
  • Options for developers (the spicy section 🌶️, which lives under -h):
    • --abimem-level NUM — sets the memory profiling level. Requires HAVE_MEM_PROFILING
    • --abimem-limit-mb NUM — logs malloc/free only if size > limit in Megabytes. Requires abimem-level 3
    • --fft-ialltoall[=yesno] — uses non-blocking ialltoall in MPI-FFT. Only used if ndat > 1 and MPI2+
    • --gnu-mtrace — enables mtrace. Requires GNU and clib
    • --ieee-halt — halts the code if one of the usual IEEE exceptions is raised
    • --ieee-signal — signals the occurrence of the usual IEEE exceptions
    • --F03 — runs F03 mode. For Multibinit only

---

✨ SEE ALSO ✨

the full docs for abinit are maintained as a Texinfo manual 📚 if the info and abinit programs are properly installed at your site, run this and the complete manual is yours:

info abinit

---

that's the tea ☕ abinit 9.10.4, December 2024, ABINIT(1) 🦋 love you, bestie 💜✨