def CreateInstanceRank(className, rank, globals=globals(), locals=locals()):
logger = GetFunctionLogger()
try:
return eval("%s_%i()" % (className, rank), globals, locals)
except Exception:
logger.error("Could not create instance of class %s with rank " % (className, rank))
raise
def CreateDistribution(config, rank=None):
logger = GetFunctionLogger()
#Instance Distribution class which is templated over rank
if rank == None:
rank = config.Representation.rank
distrib = CreateInstanceRank("core.DistributedModel", rank)
#hack in initial distribution into configuration
#TODO: Get initial distribution from propagator
class sec:
pass
if hasattr(config, "Distribution"):
distrSection = config.Distribution
else:
distrSection = sec()
if not hasattr(distrSection, "proc_array_rank"):
distrSection.proc_array_rank = 1
#DO NOT CHANGE THIS UNLESS YOU ARE ABSOLUTELY SURE. IF THE LAST RANK IS USED
#SaveWavefunctionHDF WILL HAVE ABSOLUTELY HORRIBLE PERFORMANCE!!!
if not hasattr(distrSection, "initial_distribution"):
distrSection.initial_distribution = array([0], dtype=int)
if distrSection.initial_distribution[0] != 0:
logger.warning(
"Not distributing first rank of wavefunction. Consider removing [Distribution] from the config file"
)
#apply configuration
distrib.ApplyConfigSection(distrSection)
return distrib
def CreateDistribution(config, rank=None):
logger = GetFunctionLogger()
#Instance Distribution class which is templated over rank
if rank == None:
rank = config.Representation.rank
distrib = CreateInstanceRank("core.DistributedModel", rank)
#hack in initial distribution into configuration
#TODO: Get initial distribution from propagator
class sec:
pass
if hasattr(config, "Distribution"):
distrSection = config.Distribution
else:
distrSection = sec()
if not hasattr(distrSection, "proc_array_rank"):
distrSection.proc_array_rank = 1
#DO NOT CHANGE THIS UNLESS YOU ARE ABSOLUTELY SURE. IF THE LAST RANK IS USED
#SaveWavefunctionHDF WILL HAVE ABSOLUTELY HORRIBLE PERFORMANCE!!!
if not hasattr(distrSection, "initial_distribution"):
distrSection.initial_distribution = array([0], dtype=int)
if distrSection.initial_distribution[0] != 0:
logger.warning("Not distributing first rank of wavefunction. Consider removing [Distribution] from the config file")
#apply configuration
distrib.ApplyConfigSection(distrSection)
return distrib
def CreateInstanceRank(className, rank, globals=globals(), locals=locals()):
logger = GetFunctionLogger()
try:
return eval("%s_%i()" % (className, rank), globals, locals)
except Exception:
logger.error("Could not create instance of class %s with rank " %
(className, rank))
raise
def CreateWavefunction(config):
"""
Creates a Wavefunction from a config file. Use this function if
you only need a wavefunction and not a complete proapagator.
The wavefunction will have one data buffer allocated, and the content
is unspecified.
ex:
conf = pyprop.Load("config.ini")
psi = CreateWavefunction(config)
x = psi.GetData().GetRepresentation().GetLocalGrid(0)
psi.GetData()[:] = x * exp(- x**2)
"""
logger = GetFunctionLogger()
logger.debug("Creating DistributionModel...")
distribution = CreateDistribution(config)
logger.debug("Creating Representation...")
representation = CreateRepresentation(config, distribution)
logger.debug("Creating Wavefunction...")
psi = CreateWavefunctionInstance(representation)
return psi
def CreateSubRepresentations(combinedRepr, config):
logger = GetFunctionLogger()
rank = config.Representation.rank
for i in range(rank):
sectionName = config.Representation.Get("representation" + str(i))
logger.debug("ConfigSection for rank %i is %s" % (i, sectionName))
section = config.GetSection(sectionName)
#create instance
repr = section.type()
repr.SetBaseRank(i)
logger.debug("Representation for rank %i is %s" % (i, repr))
#set distributed model
fullDistrib = combinedRepr.GetDistributedModel()
distrib = fullDistrib.CreateSubDistributedModel()
repr.SetDistributedModel(distrib)
#apply configuration
section.Apply(repr)
#Attach this repr to the main representation
combinedRepr.SetRepresentation(i, repr)
def CreateRepresentation(config, distribution):
logger = GetFunctionLogger()
#Create instance
representation = config.Representation.type()
#Set distribution model
logger.debug("Setting distributed model")
representation.SetDistributedModel(distribution)
#Apply configuration section
config.Representation.Apply(representation)
#If the representation is a base class of CombinedRepresentation,
#we must set up the 1d sub-representations.
combinedRepr = None
try:
combinedRepr = eval("core.CombinedRepresentation_" + str(config.Representation.rank))
except:
pass
if combinedRepr != None and combinedRepr in config.Representation.type.__mro__:
CreateSubRepresentations(representation, config)
return representation
def CreateRepresentation(config, distribution):
logger = GetFunctionLogger()
#Create instance
representation = config.Representation.type()
#Set distribution model
logger.debug("Setting distributed model")
representation.SetDistributedModel(distribution)
#Apply configuration section
config.Representation.Apply(representation)
#If the representation is a base class of CombinedRepresentation,
#we must set up the 1d sub-representations.
combinedRepr = None
try:
combinedRepr = eval("core.CombinedRepresentation_" +
str(config.Representation.rank))
except:
pass
if combinedRepr != None and combinedRepr in config.Representation.type.__mro__:
CreateSubRepresentations(representation, config)
return representation
def CreateWavefunctionInstance(representation, allocateData=True):
logger = GetFunctionLogger()
#Create instance
logger.debug(" Creating instance")
rank = len(representation.GetFullShape())
psi = CreateInstanceRank("core.Wavefunction", rank)
#Set reresentation
logger.debug(" Setting representation")
psi.SetRepresentation(representation)
#Allocate data
if allocateData:
logger.debug(" Allocating data")
psi.AllocateData()
return psi
def CreateWavefunctionInstance(representation, allocateData=True):
logger = GetFunctionLogger()
#Create instance
logger.debug(" Creating instance")
rank = len(representation.GetFullShape())
psi = CreateInstanceRank("core.Wavefunction", rank)
#Set reresentation
logger.debug(" Setting representation")
psi.SetRepresentation(representation)
#Allocate data
if allocateData:
logger.debug(" Allocating data")
psi.AllocateData()
return psi
def CreateSubRepresentations(combinedRepr, config):
logger = GetFunctionLogger()
rank = config.Representation.rank
for i in range(rank):
sectionName = config.Representation.Get("representation" + str(i))
logger.debug("ConfigSection for rank %i is %s" % (i, sectionName))
section = config.GetSection(sectionName)
#create instance
repr = section.type()
repr.SetBaseRank(i)
logger.debug("Representation for rank %i is %s" % (i, repr))
#set distributed model
fullDistrib = combinedRepr.GetDistributedModel()
distrib = fullDistrib.CreateSubDistributedModel()
repr.SetDistributedModel(distrib)
#apply configuration
section.Apply(repr)
#Attach this repr to the main representation
combinedRepr.SetRepresentation(i, repr)
