#!/usr/bin/env python
#########################################################
#
#  Main file for parallel mesh testing.
#
#  This is a modification of the run_parallel_advection.py
# file.
#
#
# *) The (new) files that have been added to manage the
# grid partitioning are
#    +) pmesh_divide.py: subdivide a pmesh
#    +) build_submesh.py: build the submeshes on the host
# processor.
#    +) build_local.py: build the GA mesh datastructure
# on each processor.
#    +) build_commun.py: handle the communication between
# the host and processors
#
# *) Things still to do:
#    +) Overlap the communication and computation: The
# communication routines in build_commun.py should be
# interdispersed in the build_submesh.py and build_local.py
# files. This will overlap the communication and
# computation and will be far more efficient. This should
# be done after more testing and there more confidence in
# the subpartioning.
#    +) Much more testing especially with large numbers of
# processors.
#  Authors: Linda Stals, Steve Roberts and Matthew Hardy,
# June 2005
#
#
#
#########################################################

import sys
import pypar    # The Python-MPI interface
import time

from os import sep
sys.path.append('..'+sep+'pyvolution')

# Numeric arrays

from Numeric import array, zeros, Float

# pmesh

from shallow_water import Domain
from parallel_shallow_water import Parallel_Domain
from pmesh2domain import pmesh_to_domain_instance

# Reuse previous mesh import

from caching import cache

# Mesh partition routines

from pmesh_divide import pmesh_divide, pmesh_divide_steve
from build_submesh import build_submesh, extract_hostmesh
from build_local   import build_local_mesh
from build_commun  import send_submesh, rec_submesh


###############################
# Read in processor information
###############################

numprocs = pypar.size()
myid = pypar.rank()
processor_name = pypar.Get_processor_name()

############################
# Set the initial conditions
############################

rect = zeros( 4, Float) # Buffer for results

class Set_Stage:
    """Set an initial condition with constant water height, for x<x0
    """

    def __init__(self, x0=0.25, x1=0.5, h=1.0):
        self.x0 = x0
        self.x1 = x1
        self.h  = h

    def __call__(self, x, y):
        return self.h*((x>self.x0)&(x<self.x1))

#######################
# Partition the domain
#######################

if myid == 0:

    # Read in the test files

#    filename = 'test-100.tsh'
    filename = 'merimbula_10785_1.tsh'
    nx = numprocs
    ny = 1
    if nx*ny != numprocs:
        print "WARNING: number of subboxes is not equal to the number of proc"

    # Build the whole domain
    
#    domain_full = pmesh_to_domain_instance(filename, Domain)

    domain_full = cache(pmesh_to_domain_instance,
               (filename, Domain),
              dependencies = [filename])

    rect = array(domain_full.xy_extent, Float)
    
    # Initialise the wave
    
#    domain_full.set_quantity('stage', Set_Stage(200.0,300.0,1.0))
    domain_full.set_quantity('stage', Set_Stage(756000.0,756500.0,2.0))

    # Subdivide the domain
    
    nodes, triangles, boundary, triangles_per_proc, quantities = \
         pmesh_divide_steve(domain_full, nx, ny)

    submesh = build_submesh(nodes, triangles, boundary,\
                            quantities, triangles_per_proc)
    
    # Send the mesh partition to the appropriate processor

    for p in range(1, numprocs):
      send_submesh(submesh, triangles_per_proc, p)

    # Build the local mesh for processor 0

    hostmesh = extract_hostmesh(submesh)
    points, vertices, boundary, quantities, ghost_recv_dict, \
            full_send_dict = build_local_mesh(hostmesh, 0, \
                                              triangles_per_proc[0], \
                                              numprocs)

# Read in the mesh partition that belongs to this
# processor (note that the information is in the
# correct form for the GA data structure

else:
    points, vertices, boundary, quantities, ghost_recv_dict, \
            full_send_dict = rec_submesh(0)


###########################################
# Start the computations on each subpartion
###########################################

##    ########### start profile testing
##    import profile, pstats
    
##    # Create a profiler instance.
##    pobject = profile.Profile()
    
    
##    # Profile the test function and print the statistics.
##    s = 'points, vertices, boundary, quantities, ghost_recv_dict, full_send_dict \
##            = rec_submesh(0)'

##    presult = pobject.runctx(s,globals(),locals())
##    presult.dump_stats("/tmp/profile.dat")
    
##    stats = pstats.Stats("/tmp/profile.dat")
##    print stats.sort_stats('time').print_stats(40)

######## end profile testing
    

# The visualiser needs to know the size of the whole domain

pypar.broadcast(rect,0)


domain = Parallel_Domain(points, vertices, boundary,
                         full_send_dict  = full_send_dict,
                         ghost_recv_dict = ghost_recv_dict)

try:
    domain.initialise_visualiser(rect=rect)
    #domain.visualiser.coloring['stage'] = True
    domain.visualiser.scale_z['stage'] = 0.2
    domain.visualiser.scale_z['elevation'] = 0.05
except:
    print 'No visualiser'


domain.default_order = 1

#Boundaries
from parallel_shallow_water import Transmissive_boundary, Reflective_boundary

T = Transmissive_boundary(domain)
R = Reflective_boundary(domain)
domain.set_boundary( {'outflow': R, 'inflow': R, 'inner':R, 'exterior': R, 'open':R} )


domain.set_quantity('stage', quantities['stage'])
domain.set_quantity('elevation', quantities['elevation'])

#domain.store = True
#domain.filename = 'merimbula-%d' %domain.processor

#---------
# Evolution
t0 = time.time()

print 'Processor %d on %s: No of elements %d'%(domain.processor,processor_name,domain.number_of_elements)
yieldstep = 0.05
finaltime = 500.0

yieldstep = 1
finaltime = 30
#processor_name
for t in domain.evolve(yieldstep = yieldstep, finaltime = finaltime):
    if myid == 0:
        domain.write_time()
        #print 'Processor %d, Integral of stage %d'%\
        #       (domain.processor,domain.quantities['stage'].get_integral())


#print 'P%d: That took %.2f seconds' %(myid, time.time()-t0)
#print 'P%d: Communication time %.2f seconds' %(myid, domain.communication_time)
#print 'P%d: Reduction Communication time %.2f seconds' %(myid, domain.communication_reduce_time)
#print 'P%d: Broadcast time %.2f seconds' %(myid, domain.communication_broadcast_time)



if myid == 0:
    print 'That took %.2f seconds' %(time.time()-t0)
    print 'Communication time %.2f seconds'%domain.communication_time
    print 'Reduction Communication time %.2f seconds'%domain.communication_reduce_time
    print 'Broadcast time %.2f seconds'%domain.communication_broadcast_time