# -*- coding: utf-8 -*-
"""Classes for calcs e wfls analysis. hybrid AiiDA and not_AiiDA...hopefully"""
from __future__ import absolute_import
import numpy as np
from scipy.optimize import curve_fit
from matplotlib import pyplot as plt, style
import pandas as pd
import copy
import cmath
try:
from aiida.orm import Dict, Str, load_node, KpointsData
from aiida.plugins import CalculationFactory, DataFactory
from aiida_yambo.utils.common_helpers import *
from aiida_yambo.utils.parallelism_finder import *
except:
pass
################################################################################
'''
PAR_def_mode= "balanced" # [PARALLEL] Default distribution mode ("balanced"/"memory"/"workload")
'''
################################################################################
[docs]def fix_parallelism(resources, failed_calc):
'''
bands, qp, last_qp, runlevels = find_gw_info(failed_calc.inputs)
nscf = find_pw_parent(failed_calc,calc_type=['nscf'])
occupied = gap_mapping_from_nscf(nscf.pk,)['valence']
mesh = nscf.inputs.kpoints.get_kpoints_mesh()[0]
kpoints = gap_mapping_from_nscf(nscf.pk,)['number_of_kpoints']
if 'gw0' or 'HF_and_locXC' in runlevels:
new_parallelism, new_resources = find_parallelism_qp(resources['num_machines'], resources['num_mpiprocs_per_machine'], \
resources['num_cores_per_mpiproc'], bands, \
occupied, qp, kpoints,\
last_qp, namelist = {})
elif 'bse' in runlevels:
pass
'''
pop_list = []
for p in failed_calc.inputs.parameters.get_dict()['variables']:
for k in ['CPU','ROLEs']:
if k in p:
pop_list.append(p)
new_parallelism, new_resources = {'PAR_def_mode': "balanced"}, resources
return new_parallelism, new_resources, pop_list
[docs]def fix_memory(resources, failed_calc, exit_status, max_nodes, iteration):
#bands, qp, last_qp, runlevels = find_gw_info(failed_calc.inputs)
#nscf = find_pw_parent(failed_calc,calc_type=['nscf'])
#occupied = gap_mapping_from_nscf(nscf.pk,)['valence']
#mesh = nscf.inputs.kpoints.get_kpoints_mesh()[0]
#kpoints = gap_mapping_from_nscf(nscf.pk,)['number_of_kpoints']
increase_nodes = False
if max_nodes > resources['num_machines']: # and iteration < 2:
increase_nodes = True
#tot_cores = resources['num_machines']*resources['num_mpiprocs_per_machine']*resources['num_cores_per_mpiproc']
'''if resources['num_mpiprocs_per_machine']==1 or failed_calc.outputs.output_parameters.get_dict()['has_gpu'] or iteration > 1: #there should be a limit
new_nodes = int(2*resources['num_machines'])
#new_nodes += new_nodes%2
if new_nodes <= max_nodes:
resources['num_machines'] = new_nodes
elif max_nodes==0:
resources['num_machines'] = int(1)
else:
resources['num_machines'] = int(max_nodes)'''
if resources['num_mpiprocs_per_machine']>1:
if increase_nodes: resources['num_machines'] = min(max_nodes,int(resources['num_machines']*1.5))
if not failed_calc.outputs.output_parameters.get_dict()['has_gpu']:
resources['num_cores_per_mpiproc'] = int(resources['num_cores_per_mpiproc']*2)
resources['num_mpiprocs_per_machine'] = int(resources['num_mpiprocs_per_machine']/2)
pop_list = []
for p in failed_calc.inputs.parameters.get_dict()['variables']:
for k in ['CPU','ROLEs']:
if k in p:
pop_list.append(p)
#if 'gw0' or 'HF_and_locXC' in runlevels:
# new_parallelism, new_resources = find_parallelism_qp(resources['num_machines'], resources['num_mpiprocs_per_machine']/2, \
#resources['num_cores_per_mpiproc']*2, bands, \
#occupied, qp, kpoints,\
#last_qp, namelist = {})
#elif 'bse' in runlevels:
# pass
new_parallelism, new_resources = {'PAR_def_mode': "balanced"}, resources #balanced is more efficient the memory for now
return new_parallelism, new_resources, pop_list
[docs]def fix_time(options, restart, max_walltime):
options['max_wallclock_seconds'] = \
int(options['max_wallclock_seconds']*1.5*restart)
if options['max_wallclock_seconds'] > max_walltime:
options['max_wallclock_seconds'] = int(max_walltime)
return options
