#!/usr/bin/env python3
# -*- coding: utf-8 -*-
################################################################################
#
# solid_dmft - A versatile python wrapper to perform DFT+DMFT calculations
# utilizing the TRIQS software library
#
# Copyright (C) 2018-2020, ETH Zurich
# Copyright (C) 2021, The Simons Foundation
# authors: A. Hampel, M. Merkel, and S. Beck
#
# solid_dmft is free software: you can redistribute it and/or modify it under the
# terms of the GNU General Public License as published by the Free Software
# Foundation, either version 3 of the License, or (at your option) any later
# version.
#
# solid_dmft is distributed in the hope that it will be useful, but WITHOUT ANY
# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
# PARTICULAR PURPOSE. See the GNU General Public License for more details.
# You should have received a copy of the GNU General Public License along with
# solid_dmft (in the file COPYING.txt in this directory). If not, see
# <http://www.gnu.org/licenses/>.
#
################################################################################
"""
Contains the handling of the QE process. It can start QE, reactivate it,
check if the lock file is there and finally kill QE. Needed for CSC calculations.
"""
import os
import socket
from collections import defaultdict
import shlex
import time
import triqs.utility.mpi as mpi
[docs]def create_hostfile(number_cores, cluster_name):
"""
Writes a host file for the mpirun. This tells mpi which nodes to ssh into
and start VASP on. The format of the hist file depends on the type of MPI
that is used.
Parameters
----------
number_cores: int, the number of cores that vasp runs on
cluster_name: string, the name of the server
Returns
-------
string: name of the hostfile if not run locally and if called by master node
"""
if cluster_name == 'default':
return None
hostnames = mpi.world.gather(socket.gethostname(), root=0)
if mpi.is_master_node():
# create hostfile based on first number_cores ranks
hosts = defaultdict(int)
for hostname in hostnames[:number_cores]:
hosts[hostname] += 1
mask_hostfile = {'openmpi': '{} slots={}', # OpenMPI format
'openmpi-intra': '{} slots={}', # OpenMPI format
'mpich': '{}:{}', # MPICH format
}[cluster_name]
hostfile = 'dft.hostfile'
with open(hostfile, 'w') as file:
file.write('\n'.join(mask_hostfile.format(*i) for i in hosts.items()))
return hostfile
return None
[docs]def find_path_to_mpi_command(env_vars, mpi_exe):
"""
Finds the complete path for the mpi executable by scanning the directories
of $PATH.
Parameters
----------
env_vars: dict of string, environment variables containing PATH
mpi_exe: string, mpi command
Returns
-------
string: absolute path to mpi command
"""
for path_directory in env_vars.get('PATH').split(os.pathsep):
if path_directory:
potential_path = os.path.join(path_directory, mpi_exe)
if os.access(potential_path, os.F_OK | os.X_OK):
return potential_path
return None
[docs]def get_mpi_arguments(mpi_profile, mpi_exe, number_cores, dft_exe, hostfile):
"""
Depending on the settings of the cluster and the type of MPI used,
the arguments to the mpi call have to be different. The most technical part
of the vasp handler.
Parameters
----------
cluster_name: string, name of the cluster so that settings can be tailored to it
mpi_exe: string, mpi command
number_cores: int, the number of cores that vasp runs on
dft_exe: string, the command to start the DFT code
hostfile: string, name of the hostfile
Returns
-------
list of string: arguments to start mpi with
"""
if mpi_profile == 'default':
return [mpi_exe, '-np', str(number_cores)] + shlex.split(dft_exe)
# For the second node, mpirun starts DFT by using ssh
# Therefore we need to handover the env variables with -x
if mpi_profile == 'openmpi':
return [mpi_exe, '-hostfile', hostfile, '-np', str(number_cores),
'-mca', 'mtl', '^psm,psm2,ofi',
'-mca', 'btl', '^vader,openib,usnix',
'-x', 'LD_LIBRARY_PATH',
'-x', 'PATH', '-x', 'OMP_NUM_THREADS'] + shlex.split(dft_exe)
# Run mpi with intra-node communication among ranks (on a single node).
if mpi_profile == 'openmpi-intra':
return [mpi_exe, '-np', str(number_cores),
'--mca', 'pml', 'ob1', '--mca', 'btl', 'self,vader',
'-x', 'LD_LIBRARY_PATH',
'-x', 'PATH', '-x', 'OMP_NUM_THREADS'] + shlex.split(dft_exe)
if mpi_profile == 'mpich':
return [mpi_exe, '-launcher', 'ssh', '-hostfile', hostfile,
'-np', str(number_cores), '-envlist', 'PATH'] + shlex.split(dft_exe)
return None
[docs]def poll_barrier(comm, poll_interval=0.1):
"""
Use asynchronous synchronization, otherwise mpi.barrier uses up all the CPU time during
the run of subprocess.
Parameters
----------
comm: MPI communicator
poll_interval: float, time step for pinging the status of the sleeping ranks
"""
req = comm.Ibarrier()
while not req.Test():
time.sleep(poll_interval)