Skip to content

GitLab

Commit f1511cb4 authored by Iker Martín Álvarez's avatar Iker Martín Álvarez
Browse files

Merge branch 'RMA-Distributions' into 'dev' 

RMA functionality and refactor of many of the codes

See merge request martini/malleability_benchmark!4
parents 2f81e29c 6633cd95
Show whitespace changes
Inline Side-by-side
Analysis/CreateIterDataframe.py 0 → 100644
View file @ f1511cb4
import sys
import glob
import numpy as np
import pandas as pd
from enum import Enum
class G_enum(Enum):
TOTAL_RESIZES = 0
TOTAL_GROUPS = 1
TOTAL_STAGES = 2
GRANULARITY = 3
SDR = 4
ADR = 5
DR = 6
RED_METHOD = 7
RED_STRATEGY = 8
SPAWN_METHOD = 9
SPAWN_STRATEGY = 10
GROUPS = 11
FACTOR_S = 12
DIST = 13
STAGE_TYPES = 14
STAGE_TIMES = 15
STAGE_BYTES = 16
ITERS = 17
ASYNCH_ITERS = 18
T_ITER = 19
T_STAGES = 20
T_SPAWN = 21
T_SPAWN_REAL = 22
T_SR = 23
T_AR = 24
T_MALLEABILITY = 25
T_TOTAL = 26
#Malleability specific
NP = 0
NC = 1
#Iteration specific
IS_DYNAMIC = 11
N_PARENTS = 17
#columnsG = ["Total_Resizes", "Total_Groups", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Redistribution_Method", \
# "Redistribution_Strategy", "Spawn_Method", "Spawn_Strategy", "Groups", "FactorS", "Dist", "Stage_Types", "Stage_Times", \
# "Stage_Bytes", "Iters", "Asynch_Iters", "T_iter", "T_stages", "T_spawn", "T_spawn_real", "T_SR", "T_AR", "T_Malleability", "T_total"] #27
columnsL = ["NP", "NC", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Redistribution_Method", \
"Redistribution_Strategy", "Spawn_Method", "Spawn_Strategy", "Is_Dynamic", "FactorS", "Dist", "Stage_Types", "Stage_Times", \
"Stage_Bytes", "N_Parents", "Asynch_Iters", "T_iter", "T_stages"] #20
def copy_iteration(row, dataL_it, group, iteration, is_asynch):
basic_indexes = [G_enum.TOTAL_STAGES.value, G_enum.GRANULARITY.value, \
G_enum.STAGE_TYPES.value, G_enum.STAGE_TIMES.value, G_enum.STAGE_BYTES.value]
basic_asynch = [G_enum.SDR.value, G_enum.ADR.value, G_enum.DR.value]
array_asynch_group = [G_enum.RED_METHOD.value, G_enum.RED_STRATEGY.value, \
G_enum.SPAWN_METHOD.value, G_enum.SPAWN_STRATEGY.value, G_enum.DIST.value]
dataL_it[G_enum.FACTOR_S.value] = row[G_enum.FACTOR_S.value][group]
dataL_it[G_enum.NP.value] = row[G_enum.GROUPS.value][group]
dataL_it[G_enum.ASYNCH_ITERS.value] = is_asynch
dataL_it[G_enum.T_ITER.value] = row[G_enum.T_ITER.value][group][iteration]
dataL_it[G_enum.T_STAGES.value] = list(row[G_enum.T_STAGES.value][group][iteration])
dataL_it[G_enum.IS_DYNAMIC.value] = True if group > 0 else False
for index in basic_indexes:
dataL_it[index] = row[index]
for index in array_asynch_group:
dataL_it[index] = [None, -1]
dataL_it[index][0] = row[index][group]
dataL_it[G_enum.N_PARENTS.value] = -1
if group > 0:
dataL_it[G_enum.N_PARENTS.value] = row[G_enum.GROUPS.value][group-1]
if is_asynch:
dataL_it[G_enum.NC.value] = row[G_enum.GROUPS.value][group+1]
for index in basic_asynch:
dataL_it[index] = row[index]
for index in array_asynch_group:
dataL_it[index][1] = row[index][group+1]
for index in array_asynch_group: # Convert to tuple
dataL_it[index] = tuple(dataL_it[index])
#-----------------------------------------------
def write_iter_dataframe(dataL, name, i, first=False):
dfL = pd.DataFrame(dataL, columns=columnsL)
dfL.to_pickle(name + str(i) + '.pkl')
if first:
print(dfL)
#-----------------------------------------------
def create_iter_dataframe(dfG, name, max_it_L):
it = -1
file_i = 0
first = True
dataL = []
for row_index in range(len(dfG)):
row = dfG.iloc[row_index]
groups = row[G_enum.TOTAL_GROUPS.value]
for group in range(groups):
real_iterations = len(row[G_enum.T_ITER.value][group])
real_asynch = row[G_enum.ASYNCH_ITERS.value][group]
is_asynch = False
for iteration in range(real_iterations-real_asynch):
it += 1
dataL.append( [None] * len(columnsL) )
copy_iteration(row, dataL[it], group, iteration, is_asynch)
is_asynch = True
for iteration in range(real_iterations-real_asynch, real_iterations):
it += 1
dataL.append( [None] * len(columnsL) )
copy_iteration(row, dataL[it], group, iteration, is_asynch)
if it >= max_it_L-1: #Var "it" starts at -1, so one more must be extracted for precise cut
write_iter_dataframe(dataL, name, file_i, first)
dataL = []
file_i += 1
first = False
it = -1
if it != -1:
write_iter_dataframe(dataL, name, file_i)
#-----------------------------------------------
if len(sys.argv) < 2:
print("The files name is missing\nUsage: python3 CreateIterDataframe.py input_file.pkl output_name [max_rows_per_file]")
exit(1)
input_name = sys.argv[1]
if len(sys.argv) > 2:
name = sys.argv[2]
else:
name = "dataL"
print("File names will be: " + name + ".pkl")
if len(sys.argv) > 3:
max_it_L = int(sys.argv[3])
else:
max_it_L = 100000
dfG = pd.read_pickle(input_name)
print(dfG)
create_iter_dataframe(dfG, name, max_it_L)
Analysis/CreateResizeDataframe.py 0 → 100644
View file @ f1511cb4
import sys
import glob
import numpy as np
import pandas as pd
from enum import Enum
class G_enum(Enum):
TOTAL_RESIZES = 0
TOTAL_GROUPS = 1
TOTAL_STAGES = 2
GRANULARITY = 3
SDR = 4
ADR = 5
DR = 6
RED_METHOD = 7
RED_STRATEGY = 8
SPAWN_METHOD = 9
SPAWN_STRATEGY = 10
GROUPS = 11
FACTOR_S = 12
DIST = 13
STAGE_TYPES = 14
STAGE_TIMES = 15
STAGE_BYTES = 16
ITERS = 17
ASYNCH_ITERS = 18
T_ITER = 19
T_STAGES = 20
T_SPAWN = 21
T_SPAWN_REAL = 22
T_SR = 23
T_AR = 24
T_MALLEABILITY = 25
T_TOTAL = 26
#Malleability specific
NP = 0
NC = 1
#Iteration specific
IS_DYNAMIC = 11
N_PARENTS = 17
#columnsG = ["Total_Resizes", "Total_Groups", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Redistribution_Method", \
# "Redistribution_Strategy", "Spawn_Method", "Spawn_Strategy", "Groups", "FactorS", "Dist", "Stage_Types", "Stage_Times", \
# "Stage_Bytes", "Iters", "Asynch_Iters", "T_iter", "T_stages", "T_spawn", "T_spawn_real", "T_SR", "T_AR", "T_Malleability", "T_total"] #27
columnsM = ["NP", "NC", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Redistribution_Method", \
"Redistribution_Strategy", "Spawn_Method", "Spawn_Strategy", "FactorS", "Dist", "Stage_Type", "Stage_Time", \
"Stage_Bytes", "Iters", "Asynch_Iters", "T_iter", "T_stages", "T_spawn", "T_spawn_real", "T_SR", "T_AR", "T_Malleability"] #25
def copy_resize(row, dataM_it, resize):
basic_indexes = [G_enum.TOTAL_STAGES.value, G_enum.GRANULARITY.value, G_enum.SDR.value, \
G_enum.ADR.value, G_enum.DR.value]
basic_group = [G_enum.STAGE_TYPES.value, G_enum.STAGE_TIMES.value, G_enum.STAGE_BYTES.value]
array_actual_group = [G_enum.FACTOR_S.value, G_enum.ITERS.value, G_enum.ASYNCH_ITERS.value, \
G_enum.T_SPAWN.value, G_enum.T_SPAWN_REAL.value, G_enum.T_SR.value, \
G_enum.T_AR.value, G_enum.T_MALLEABILITY.value, G_enum.T_ITER.value, G_enum.T_STAGES.value]
array_next_group = [G_enum.RED_METHOD.value, G_enum.RED_STRATEGY.value, \
G_enum.SPAWN_METHOD.value, G_enum.SPAWN_STRATEGY.value]
dataM_it[G_enum.NP.value] = row[G_enum.GROUPS.value][resize]
dataM_it[G_enum.NC.value] = row[G_enum.GROUPS.value][resize+1]
dataM_it[G_enum.DIST.value-1] = [None, None]
dataM_it[G_enum.DIST.value-1][0] = row[G_enum.DIST.value][resize]
dataM_it[G_enum.DIST.value-1][1] = row[G_enum.DIST.value][resize+1]
for index in basic_indexes:
dataM_it[index] = row[index]
for index in basic_group:
dataM_it[index-1] = row[index]
for index in array_actual_group:
dataM_it[index-1] = row[index][resize]
for index in array_next_group:
dataM_it[index] = row[index][resize+1]
#-----------------------------------------------
def create_resize_dataframe(dfG, dataM):
it = -1
for row_index in range(len(dfG)):
row = dfG.iloc[row_index]
resizes = row[G_enum.TOTAL_RESIZES.value]
for resize in range(resizes):
it += 1
dataM.append( [None] * len(columnsM) )
copy_resize(row, dataM[it], resize)
#-----------------------------------------------
if len(sys.argv) < 2:
print("The files name is missing\nUsage: python3 CreateResizeDataframe.py input_file.pkl output_name")
exit(1)
input_name = sys.argv[1]
if len(sys.argv) > 2:
name = sys.argv[2]
else:
name = "dataM"
print("File name will be: " + name + ".pkl")
dfG = pd.read_pickle(input_name)
dataM = []
create_resize_dataframe(dfG, dataM)
dfM = pd.DataFrame(dataM, columns=columnsM)
dfM.to_pickle(name + '.pkl')
print(dfG)
print(dfM)
Analysis/MallTimes.py
View file @ f1511cb4
......@@ -12,34 +12,56 @@ class G_enum(Enum):
SDR = 4
ADR = 5
DR = 6
ASYNCH_REDISTRIBUTION_TYPE = 7
SPAWN_METHOD = 8
SPAWN_STRATEGY = 9
GROUPS = 10
FACTOR_S = 11
DIST = 12
STAGE_TYPES = 13
STAGE_TIMES = 14
STAGE_BYTES = 15
ITERS = 16
ASYNCH_ITERS = 17
T_ITER = 18
T_STAGES = 19
T_SPAWN = 20
T_SPAWN_REAL = 21
T_SR = 22
T_AR = 23
T_TOTAL = 24
columnsG = ["Total_Resizes", "Total_Groups", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Asynch_Redistribution_Type", \
"Spawn_Method", "Spawn_Strategy", "Groups", "Factor_S", "Dist", "Stage_Types", "Stage_Times", "Stage_Bytes", \
"Iters", "Asynch_Iters", "T_iter", "T_stages", "T_spawn", "T_spawn_real", "T_SR", "T_AR", "T_total"] #25
RED_METHOD = 7
RED_STRATEGY = 8
SPAWN_METHOD = 9
SPAWN_STRATEGY = 10
GROUPS = 11
FACTOR_S = 12
DIST = 13
STAGE_TYPES = 14
STAGE_TIMES = 15
STAGE_BYTES = 16
ITERS = 17
ASYNCH_ITERS = 18
T_ITER = 19
T_STAGES = 20
T_SPAWN = 21
T_SPAWN_REAL = 22
T_SR = 23
T_AR = 24
T_MALLEABILITY = 25
T_TOTAL = 26
#Malleability specific
NP = 0
NC = 1
#Iteration specific
IS_DYNAMIC = 11
N_PARENTS = 17
columnsG = ["Total_Resizes", "Total_Groups", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Redistribution_Method", \
"Redistribution_Strategy", "Spawn_Method", "Spawn_Strategy", "Groups", "FactorS", "Dist", "Stage_Types", "Stage_Times", \
"Stage_Bytes", "Iters", "Asynch_Iters", "T_iter", "T_stages", "T_spawn", "T_spawn_real", "T_SR", "T_AR", "T_Malleability", "T_total"] #27
#-----------------------------------------------
# Obtains the value of a given index in a splited line
# and returns it as a float values
def get_value(line, index):
return float(line[index].split('=')[1].split(',')[0])
# and returns it as a float values if possible, string otherwise
def get_value(line, index, separator=True):
if separator:
value = line[index].split('=')[1].split(',')[0]
else:
value = line[index]
try:
value = float(value)
if value.is_integer():
value = int(value)
except ValueError:
return value
return value
#-----------------------------------------------
# Obtains the general parameters of an execution and
# stores them for creating a global dataframe
def record_config_line(lineS, dataG_it):
......@@ -48,27 +70,25 @@ def record_config_line(lineS, dataG_it):
offset_line = 2
for i in range(len(ordered_indexes)):
value = get_value(lineS, i+offset_line)
if value.is_integer():
value = int(value)
index = ordered_indexes[i]
dataG_it[index] = value
dataG_it[G_enum.TOTAL_GROUPS.value] = dataG_it[G_enum.TOTAL_RESIZES.value]
dataG_it[G_enum.TOTAL_RESIZES.value] -=1 #FIXME Modificar en App sintetica
dataG_it[G_enum.TOTAL_GROUPS.value] = dataG_it[G_enum.TOTAL_RESIZES.value]+1
#FIXME Modificar cuando ADR ya no sea un porcentaje
dataG_it[G_enum.DR.value] = dataG_it[G_enum.SDR.value] + dataG_it[G_enum.ADR.value]
# Init lists for each column
array_groups = [G_enum.GROUPS.value, G_enum.FACTOR_S.value, G_enum.DIST.value, G_enum.ITERS.value, \
G_enum.ASYNCH_ITERS.value, G_enum.T_ITER.value, G_enum.T_STAGES.value]
array_resizes = [G_enum.ASYNCH_REDISTRIBUTION_TYPE.value, G_enum.SPAWN_METHOD.value, \
G_enum.SPAWN_STRATEGY.value, G_enum.T_SPAWN.value, G_enum.T_SPAWN_REAL.value, \
G_enum.T_SR.value, G_enum.T_AR.value]
G_enum.ASYNCH_ITERS.value, G_enum.T_ITER.value, G_enum.T_STAGES.value, G_enum.RED_METHOD.value, \
G_enum.RED_STRATEGY.value, G_enum.SPAWN_METHOD.value, G_enum.SPAWN_STRATEGY.value,]
array_resizes = [ G_enum.T_SPAWN.value, G_enum.T_SPAWN_REAL.value, G_enum.T_SR.value, G_enum.T_AR.value, G_enum.T_MALLEABILITY.value]
array_stages = [G_enum.STAGE_TYPES.value, \
G_enum.STAGE_TIMES.value, G_enum.STAGE_BYTES.value]
for index in array_groups:
dataG_it[index] = [None]*dataG_it[G_enum.TOTAL_GROUPS.value]
for group in range(dataG_it[G_enum.TOTAL_GROUPS.value]):
dataG_it[G_enum.T_ITER.value][group] = []
for index in array_resizes:
dataG_it[index] = [None]*dataG_it[G_enum.TOTAL_RESIZES.value]
......@@ -76,6 +96,7 @@ def record_config_line(lineS, dataG_it):
for index in array_stages:
dataG_it[index] = [None]*dataG_it[G_enum.TOTAL_STAGES.value]
#-----------------------------------------------
# Obtains the parameters of a stage line
# and stores it in the dataframe
# Is needed to indicate in which stage is
......@@ -86,144 +107,194 @@ def record_stage_line(lineS, dataG_it, stage):
offset_lines = 2
for i in range(len(array_stages)):
value = get_value(lineS, i+offset_lines)
if value.is_integer():
value = int(value)
index = array_stage[i]
index = array_stages[i]
dataG_it[index][stage] = value
#-----------------------------------------------
# Obtains the parameters of a resize line
# and stores them in the dataframe
# Is needed to indicate to which group refers
# the resize line
def record_resize_line(lineS, dataG_it, group):
array_stages = [G_enum.ITERS.value, G_enum.GROUPS.value, G_enum.FACTOR_S.value, G_enum.DIST.value, \
G_enum.ASYNCH_REDISTRIBUTION_TYPE.value, G_enum.SPAWN_METHOD.value, G_enum.SPAWN_STRATEGY.value]
def record_group_line(lineS, dataG_it, group):
array_groups = [G_enum.ITERS.value, G_enum.GROUPS.value, G_enum.FACTOR_S.value, G_enum.DIST.value, \
G_enum.RED_METHOD.value, G_enum.RED_STRATEGY.value, G_enum.SPAWN_METHOD.value, G_enum.SPAWN_STRATEGY.value]
offset_lines = 2
for i in range(len(array_stages)):
for i in range(len(array_groups)):
value = get_value(lineS, i+offset_lines)
if value.is_integer():
value = int(value)
index = array_stage[i]
index = array_groups[i]
dataG_it[index][group] = value
#-----------------------------------------------
def record_time_line(lineS, dataG_it):
T_names = ["T_spawn:", "T_spawn_real:", "T_SR:", "T_AR:", "T_total:"]
T_values = [G_enum.T_SPAWN.value, G_enum.T_SPAWN_REAL.value, G_enum.T_SR.value, G_enum.T_AR.value, G_enum.T_TOTAL.value]
T_names = ["T_spawn:", "T_spawn_real:", "T_SR:", "T_AR:", "T_Malleability:", "T_total:"]
T_values = [G_enum.T_SPAWN.value, G_enum.T_SPAWN_REAL.value, G_enum.T_SR.value, G_enum.T_AR.value, G_enum.T_MALLEABILITY.value, G_enum.T_TOTAL.value]
if not (lineS[0] in T_names): # Execute only if line represents a Time
return
index = T_names.index(linesS[0])
index = T_names.index(lineS[0])
index = T_values[index]
offset_lines = 1
for i in range(len(dataG_it[index])):
value = get_value(lineS, i+offset_lines)
dataG_it[index][i] = value
len_index = 1
if dataG_it[index] != None:
len_index = len(dataG_it[index])
for i in range(len_index):
dataG_it[index][i] = get_value(lineS, i+offset_lines, False)
else:
dataG_it[index] = get_value(lineS, offset_lines, False)
#-----------------------------------------------
def read_global_file(f, dataG, it):
resizes = 0
timer = 0
previousNP = 0
def record_multiple_times_line(lineS, dataG_it, group):
T_names = ["T_iter:", "T_stage"]
T_values = [G_enum.T_ITER.value, G_enum.T_STAGES.value]
if not (lineS[0] in T_names): # Execute only if line represents a Time
return
index = T_names.index(lineS[0])
index = T_values[index]
offset_lines = 1
if index == G_enum.T_STAGES.value:
offset_lines += 1
total_iters = len(lineS)-offset_lines
stage = int(lineS[1].split(":")[0])
if stage == 0:
dataG_it[index][group] = [None] * total_iters
for i in range(total_iters):
dataG_it[index][group][i] = [None] * dataG_it[G_enum.TOTAL_STAGES.value]
for i in range(total_iters):
dataG_it[index][group][i][stage] = get_value(lineS, i+offset_lines, False)
else:
total_iters = len(lineS)-offset_lines
for i in range(total_iters):
dataG_it[index][group].append(get_value(lineS, i+offset_lines, False))
#-----------------------------------------------
def read_local_file(f, dataG, it, runs_in_file):
offset = 0
real_it = 0
group = 0
for line in f:
lineS = line.split()
if len(lineS) > 0:
if lineS[0] == "Config": # CONFIG LINE
it += 1
dataG.append([None]*(25+1))
#dataG[it][-1] = None Indicates if local data has been recorded(1) or not(None)
record_config(lineS, dataG[it])
resize = 0
stage = 0
elif lineS[0] == "Stage":
record_stage_line(lineS, dataG[it], stage)
stage+=1
elif lineS[0] == "Resize":
record_resize_line(lineS, dataG[it], resize)
resize+=1
elif lineS[0] == "T_total:":
value = get_value(lineS, 1)
dataG[it][G_enum.T_TOTAL.value] = value
if lineS[0] == "Group": # GROUP number
offset += 1
real_it = it - (runs_in_file-offset)
group = int(lineS[1].split(":")[0])
elif lineS[0] == "Async_Iters:":
offset_line = 1
dataG[real_it][G_enum.ASYNCH_ITERS.value][group] = get_value(lineS, offset_line, False)
else:
record_time_line(lineS, dataG[it])
return it
record_multiple_times_line(lineS, dataG[real_it], group)
#-----------------------------------------------
def read_local_file(f, dataG, it):
resizes = 0
timer = 0
previousNP = 0
def read_global_file(f, dataG, it):
runs_in_file=0
for line in f:
lineS = line.split()
if len(lineS) > 0:
if lineS[0] == "Config": # CONFIG LINE
it += 1
record_config(lineS, dataG[it], dataM[it])
resize = 0
runs_in_file += 1
group = 0
stage = 0
dataG.append([None]*len(columnsG))
record_config_line(lineS, dataG[it])
elif lineS[0] == "Stage":
record_stage_line(lineS, dataG[it], stage)
stage+=1
elif lineS[0] == "Resize":
record_resize_line(lineS, dataG[it], resize)
resize+=1
elif lineS[0] == "T_total:":
value = get_value(lineS, 1)
dataG[it][G_enum.T_TOTAL.value] = value
elif lineS[0] == "Group":
record_group_line(lineS, dataG[it], group)
group+=1
else:
record_time_line(lineS, dataG[it])
return it
return it,runs_in_file
#-----------------------------------------------
#columnsG = ["Total_Resizes", "Total_Groups", "Total_Stages", "Granularity", "SDR", "ADR", "DR", "Asynch_Redistribution_Type", \\
# "Spawn_Method", "Spawn_Strategy", "Groups", "Dist", "Stage_Types", "Stage_Times", "Stage_Bytes", \\
# "Iters", "Asynch_Iters", "T_iter", "T_stages", "T_spawn", "T_spawn_real", "T_SR", "T_AR", "T_total"] #24
#-----------------------------------------------
def convert_to_tuples(dfG):
array_list_items = [G_enum.GROUPS.value, G_enum.FACTOR_S.value, G_enum.DIST.value, G_enum.ITERS.value, \
G_enum.ASYNCH_ITERS.value, G_enum.RED_METHOD.value, G_enum.RED_STRATEGY.value, G_enum.SPAWN_METHOD.value, \
G_enum.SPAWN_STRATEGY.value, G_enum.T_SPAWN.value, G_enum.T_SPAWN_REAL.value, G_enum.T_SR.value, \
G_enum.T_AR.value, G_enum.STAGE_TYPES.value, G_enum.STAGE_TIMES.value, G_enum.STAGE_BYTES.value]
#TODO Falta T_malleability?
array_multiple_list_items = [G_enum.T_ITER.value, G_enum.T_STAGES.value]
for item in array_list_items:
name = columnsG[item]
values = dfG[name].copy()
for index in range(len(values)):
values[index] = tuple(values[index])
dfG[name] = values
for item in array_multiple_list_items:
name = columnsG[item]
values = dfG[name].copy()
for i in range(len(values)):
for j in range(len(values[i])):
if(type(values[i][j][0]) == list):
for r in range(len(values[i][j])):
values[i][j][r] = tuple(values[i][j][r])
values[i][j] = tuple(values[i][j])
values[i] = tuple(values[i])
dfG[name] = values
#-----------------------------------------------
if len(sys.argv) < 2:
print("The files name is missing\nUsage: python3 iterTimes.py resultsName directory csvOutName")
print("The files name is missing\nUsage: python3 MallTimes.py commonName directory OutName")
exit(1)
common_name = sys.argv[1]
if len(sys.argv) >= 3:
BaseDir = sys.argv[2]
print("Searching in directory: "+ BaseDir)
else:
BaseDir = sys.argv[2]
BaseDir = "./"
if len(sys.argv) >= 4:
print("Csv name will be: " + sys.argv[3] + "G.csv & " + sys.argv[3] + "M.csv")
name = sys.argv[3]
else:
name = "data"
print("File name will be: " + name + "G.pkl")
insideDir = "Run"
lista = glob.glob("./" + BaseDir + insideDir + "*/" + sys.argv[1]+ "*Global.o*")
lista += (glob.glob("./" + BaseDir + sys.argv[1]+ "*Global.o*")) # Se utiliza cuando solo hay un nivel de directorios
lista = glob.glob(BaseDir + insideDir + "*/" + common_name + "*_Global.out")
lista += (glob.glob(BaseDir + common_name + "*_Global.out")) # Se utiliza cuando solo hay un nivel de directorios
print("Number of files found: "+ str(len(lista)));
it = -1
dataG = []
dataM = []
columnsG = ["N", "%Async", "Groups", "NP", "NS", "Dist", "Matrix", "CommTam", "Cst", "Css", "Time", "Iters", "TE"] #13
columnsM = ["N", "%Async", "NP", "NS", "Dist", "Matrix", "CommTam", "Cst", "Css", "Time", "Iters", "TC", "TH", "TS", "TA"] #15
for elem in lista:
f = open(elem, "r")
it = read_file(f, dataG, dataM, it)
id_run = elem.split("_Global.out")[0].split(common_name)[1]
path_to_run = elem.split(common_name)[0]
lista_local = glob.glob(path_to_run + common_name + id_run + "_G*NP*.out")
it,runs_in_file = read_global_file(f, dataG, it)
f.close()
for elem_local in lista_local:
f_local = open(elem_local, "r")
read_local_file(f_local, dataG, it, runs_in_file)
f_local.close()
#print(data)
dfG = pd.DataFrame(dataG, columns=columnsG)
dfG.to_csv(name + 'G.csv')
convert_to_tuples(dfG)
print(dfG)
dfG.to_pickle(name + 'G.pkl')
dfM = pd.DataFrame(dataM, columns=columnsM)
#dfM = pd.DataFrame(dataM, columns=columnsM)
#Poner en TC el valor real y en TH el necesario para la app
cond = dfM.TH != 0
dfM.loc[cond, ['TC', 'TH']] = dfB.loc[cond, ['TH', 'TC']].values
dfM.to_csv(name + 'M.csv')
#cond = dfM.TH != 0
#dfM.loc[cond, ['TC', 'TH']] = dfM.loc[cond, ['TH', 'TC']].values
#dfM.to_csv(name + 'M.csv')
Analysis/analyser.ipynb deleted 100644 → 0
View file @ 2f81e29c
This source diff could not be displayed because it is too large. You can view the blob instead.
Analysis/bt_scheme.py 0 → 100644
View file @ f1511cb4
'''
Created on Oct 24, 2016
@author: David Llorens (dllorens@uji.es)
(c) Universitat Jaume I 2016
@license: GPL2
'''
from abc import ABCMeta, abstractmethod
infinity = float("infinity")
## Esquema para BT básico --------------------------------------------------------------------------
class PartialSolution(metaclass=ABCMeta):
@abstractmethod
def is_solution(self)-> "bool":
pass
@abstractmethod
def get_solution(self) -> "solution":
pass
@abstractmethod
def successors(self) -> "IEnumerable<PartialSolution>":
pass
class BacktrackingSolver(metaclass=ABCMeta):
@staticmethod
def solve(initial_ps : "PartialSolution") -> "IEnumerable<Solution>":
def bt(ps):
if ps.is_solution():
yield ps.get_solution()
else:
for new_ps in ps.successors():
yield from bt(new_ps)
yield from bt(initial_ps)
class BacktrackingSolverOld(metaclass=ABCMeta):
def solve(self, initial_ps : "PartialSolution") -> "IEnumerable<Solution>":
def bt(ps):
if ps.is_solution():
return [ps.get_solution()]
else:
solutions = []
for new_ps in ps.successors():
solutions.extend(bt(new_ps))
return solutions
return bt(initial_ps)
## Esquema para BT con control de visitados --------------------------------------------------------
class PartialSolutionWithVisitedControl(PartialSolution):
@abstractmethod
def state(self)-> "state":
# the returned object must be of an inmutable type
pass
class BacktrackingVCSolver(metaclass=ABCMeta):
@staticmethod
def solve(initial_ps : "PartialSolutionWithVisitedControl") -> "IEnumerable<Solution>":
def bt(ps):
seen.add(ps.state())
if ps.is_solution():
yield ps.get_solution()
else:
for new_ps in ps.successors():
state = new_ps.state()
if state not in seen:
yield from bt(new_ps)
seen = set()
yield from bt(initial_ps)
## Esquema para BT para optimización ----------------------------------------------------------------
class PartialSolutionWithOptimization(PartialSolutionWithVisitedControl):
@abstractmethod
def f(self)-> "int or double":
# result of applying the objective function to the partial solution
pass
class BacktrackingOptSolver(metaclass=ABCMeta):
@staticmethod
def solve(initial_ps : "PartialSolutionWithOptimization") -> "IEnumerable<Solution>":
def bt(ps):
nonlocal best_solution_found_score
ps_score = ps.f()
best_seen[ps.state()] = ps_score
if ps.is_solution() and ps_score < best_solution_found_score: #sólo muestra una solución si mejora la última mostrada
best_solution_found_score = ps_score
yield ps.get_solution()
else:
for new_ps in ps.successors():
state = new_ps.state()
if state not in best_seen or new_ps.f() < best_seen[state]:
yield from bt(new_ps)
best_seen = {}
best_solution_found_score = infinity
yield from bt(initial_ps)
Analysis/iterTimes.py deleted 100644 → 0
View file @ 2f81e29c
import sys
import glob
import numpy as numpy
import pandas as pd
#-----------------------------------------------
def read_file(f, dataA, dataB, itA, itB):
compute_tam = 0
comm_tam = 0
sdr = 0
adr = 0
dist = 0
css = 0
cst = 0
time = 0
recording = False
it_line = 0
aux_itA = 0
aux_itB = 0
iters = 0
np = 0
np_par = 0
ns = 0
array = []
columnas = ['Titer','Ttype','Top']
#print(f)
for line in f:
lineS = line.split()
if len(lineS) > 1:
if recording and lineS[0].split(':')[0] in columnas: #Record data
aux_itA = 0
lineS.pop(0)
if it_line==0:
for observation in lineS:
dataA.append([None]*15)
dataA[itA+aux_itA][0] = sdr
dataA[itA+aux_itA][1] = adr
dataA[itA+aux_itA][2] = np
dataA[itA+aux_itA][3] = np_par
dataA[itA+aux_itA][4] = ns
dataA[itA+aux_itA][5] = dist
dataA[itA+aux_itA][6] = compute_tam
dataA[itA+aux_itA][7] = comm_tam
dataA[itA+aux_itA][8] = cst
dataA[itA+aux_itA][9] = css
dataA[itA+aux_itA][10] = time
dataA[itA+aux_itA][11] = iters
dataA[itA+aux_itA][12] = float(observation)
array.append(float(observation))
aux_itA+=1
elif it_line==1:
deleted = 0
for observation in lineS:
dataA[itA+aux_itA][13] = float(observation)
if float(observation) == 0:
array.pop(aux_itA - deleted)
deleted+=1
aux_itA+=1
else:
for observation in lineS:
dataA[itA+aux_itA][14] = float(observation)
aux_itA+=1
it_line += 1
if(it_line % 3 == 0): # Comprobar si se ha terminado de mirar esta ejecucion
recording = False
it_line = 0
itA = itA + aux_itA
if ns != 0: # Solo obtener datos de grupos con hijos
dataB.append([None]*14)
dataB[itB][0] = sdr
dataB[itB][1] = adr
dataB[itB][2] = np
dataB[itB][3] = np_par
dataB[itB][4] = ns
dataB[itB][5] = dist
dataB[itB][6] = compute_tam
dataB[itB][7] = comm_tam
dataB[itB][8] = cst
dataB[itB][9] = css
dataB[itB][10] = time
dataB[itB][11] = iters
dataB[itB][12] = tuple(array)
dataB[itB][13] = numpy.sum(array)
itB+=1
array = []
if lineS[0] == "Config:":
compute_tam = int(lineS[1].split('=')[1].split(',')[0])
comm_tam = int(lineS[2].split('=')[1].split(',')[0])
sdr = int(lineS[3].split('=')[1].split(',')[0])
adr = int(lineS[4].split('=')[1].split(',')[0])
css = int(lineS[6].split('=')[1].split(',')[0])
cst = int(lineS[7].split('=')[1].split(',')[0])
time = float(lineS[8].split('=')[1])
elif lineS[0] == "Config":
recording = True
iters = int(lineS[2].split('=')[1].split(',')[0])
dist = int(lineS[4].split('=')[1].split(',')[0])
np = int(lineS[5].split('=')[1].split(',')[0])
np_par = int(lineS[6].split('=')[1].split(',')[0])
ns = int(float(lineS[7].split('=')[1]))
return itA,itB
#-----------------------------------------------
#Config: matrix=1000, sdr=1000000000, adr=0, aib=0 time=2.000000
#Config Group: iters=100, factor=1.000000, phy=2, procs=2, parents=0, sons=4
#Ttype: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
if len(sys.argv) < 2:
print("The files name is missing\nUsage: python3 iterTimes.py resultsName directory csvOutName")
exit(1)
if len(sys.argv) >= 3:
BaseDir = sys.argv[2]
print("Searching in directory: "+ BaseDir)
else: #FIXME
BaseDir = sys.argv[2]
if len(sys.argv) >= 4:
print("Csv name will be: " + sys.argv[3] + ".csv and "+ sys.argv[3] + "_Total.csv")
name = sys.argv[3]
else:
name = "data"
insideDir = "Run"
lista = glob.glob("./" + BaseDir + insideDir + "*/" + sys.argv[1]+ "*ID*.o*")
print("Number of files found: "+ str(len(lista)));
itA = itB = 0
dataA = []
dataB = [] #0 #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14
columnsA = ["N", "%Async", "NP", "N_par", "NS", "Dist", "Compute_tam", "Comm_tam", "Cst", "Css","Time", "Iters", "Ti", "Tt", "To"] #15
columnsB = ["N", "%Async", "NP", "N_par", "NS", "Dist", "Compute_tam", "Comm_tam", "Cst", "Css","Time", "Iters", "Ti", "Sum"] #14
for elem in lista:
f = open(elem, "r")
itA,itB = read_file(f, dataA, dataB, itA, itB)
f.close()
#print(data)
dfA = pd.DataFrame(dataA, columns=columnsA)
dfB = pd.DataFrame(dataB, columns=columnsB)
dfA['N'] += dfA['%Async']
dfA['%Async'] = (dfA['%Async'] / dfA['N']) * 100
dfA.to_csv(name + '.csv')
dfB['N'] += dfB['%Async']
dfB['%Async'] = (dfB['%Async'] / dfB['N']) * 100
dfB.to_csv(name + '_Total.csv')
Analysis/joinDf.py
View file @ f1511cb4
......@@ -3,22 +3,19 @@ import glob
import numpy as numpy
import pandas as pd
if len(sys.argv) < 3:
print("The files name is missing\nUsage: python3 joinDf.py resultsName1.csv resultsName2.csv csvOutName")
print("The files name is missing\nUsage: python3 joinDf.py resultsName1.pkl resultsName2.pkl OutName")
exit(1)
if len(sys.argv) >= 4:
print("Csv name will be: " + sys.argv[3] + ".csv")
name = sys.argv[3]
else:
name = "dataJOINED"
df1 = pd.read_csv( sys.argv[1] )
df2 = pd.read_csv( sys.argv[2] )
print("File name will be: " + name + ".pkl")
df1 = pd.read_pickle( sys.argv[1] )
df2 = pd.read_pickle( sys.argv[2] )
frames = [df1, df2]
df3 = pd.concat(frames)
df3 = df3.drop(columns=df3.columns[0])
df3.to_csv(name + '.csv')
df3.to_pickle(name + '.pkl')
Analysis/new_analyser.ipynb 0 → 100644
View file @ f1511cb4
This source diff could not be displayed because it is too large. You can view the blob instead.
Codes/IOcodes/read_ini.c
View file @ f1511cb4
......@@ -42,9 +42,9 @@ static int handler(void* user, const char* section, const char* name,
} else if (MATCH("general", "Granularity")) {
pconfig->granularity = atoi(value);
} else if (MATCH("general", "SDR")) { // TODO Refactor a nombre manual
pconfig->sdr = atoi(value);
pconfig->sdr = strtoul(value, NULL, 10);
} else if (MATCH("general", "ADR")) { // TODO Refactor a nombre manual
pconfig->adr = atoi(value);
pconfig->adr = strtoul(value, NULL, 10);
} else if (MATCH("general", "Rigid")) {
pconfig->rigid_times = atoi(value);
......@@ -72,8 +72,10 @@ static int handler(void* user, const char* section, const char* name,
aux_value = MALL_DIST_SPREAD;
}
pconfig->groups[pconfig->actual_group].phy_dist = aux_value;
} else if (MATCH(resize_name, "Asynch_Redistribution_Type") && LAST(pconfig->actual_group, pconfig->n_groups)) {
pconfig->groups[pconfig->actual_group].at = atoi(value);
} else if (MATCH(resize_name, "Redistribution_Method") && LAST(pconfig->actual_group, pconfig->n_groups)) {
pconfig->groups[pconfig->actual_group].rm = atoi(value);
} else if (MATCH(resize_name, "Redistribution_Strategy") && LAST(pconfig->actual_group, pconfig->n_groups)) {
pconfig->groups[pconfig->actual_group].rs = atoi(value);
} else if (MATCH(resize_name, "Spawn_Method") && LAST(pconfig->actual_group, pconfig->n_groups)) {
pconfig->groups[pconfig->actual_group].sm = atoi(value);
} else if (MATCH(resize_name, "Spawn_Strategy") && LAST(pconfig->actual_group, pconfig->n_groups)) {
......
Codes/IOcodes/results.c
View file @ f1511cb4
......@@ -41,14 +41,14 @@ void comm_results(results_data *results, int root, size_t resizes, MPI_Comm inte
* En concreto son tres escalares y dos vectores de tamaño "resizes"
*/
void def_results_type(results_data *results, int resizes, MPI_Datatype *results_type) {
int i, counts = 6;
int blocklengths[] = {1, 1, 1, 1, 1, 1};
int i, counts = 7;
int blocklengths[] = {1, 1, 1, 1, 1, 1, 1};
MPI_Aint displs[counts], dir;
MPI_Datatype types[counts];
// Rellenar vector types
types[0] = types[1] = types[2] = types[3] = types[4] = types[5] = MPI_DOUBLE;
blocklengths[2] = blocklengths[3] = blocklengths[4] = blocklengths[5] = resizes;
types[0] = types[1] = types[2] = types[3] = types[4] = types[5] = types[6] = MPI_DOUBLE;
blocklengths[2] = blocklengths[3] = blocklengths[4] = blocklengths[5] = blocklengths[6] = resizes;
// Rellenar vector displs
MPI_Get_address(results, &dir);
......@@ -59,6 +59,7 @@ void def_results_type(results_data *results, int resizes, MPI_Datatype *results_
MPI_Get_address(results->async_time, &displs[3]);
MPI_Get_address(results->spawn_real_time, &displs[4]);
MPI_Get_address(results->spawn_time, &displs[5]);
MPI_Get_address(results->malleability_time, &displs[6]);
for(i=0;i<counts;i++) displs[i] -= dir;
......@@ -87,6 +88,7 @@ void set_results_post_reconfig(results_data *results, int grp, int sdr, int adr)
} else {
results->async_time[grp-1] = 0;
}
results->malleability_time[grp-1] = results->malleability_end - results->malleability_time[grp-1];
}
/*
......@@ -100,6 +102,7 @@ void set_results_post_reconfig(results_data *results, int grp, int sdr, int adr)
*/
void reset_results_index(results_data *results) {
results->iter_index = 0;
results->iters_async = 0;
}
//=============================================================== FIXME BORRAR?
......@@ -162,17 +165,18 @@ void compute_results_stages(results_data *results, int myId, int numP, int root,
int i;
if(myId == root) {
for(i=0; i<stages; i++) {
MPI_Reduce(MPI_IN_PLACE, results->stage_times[i], results->iter_index, MPI_DOUBLE, MPI_SUM, root, comm);
for(size_t j=0; j<results->iter_index; j++) {
MPI_Reduce(MPI_IN_PLACE, results->stage_times[i], results->iter_index, MPI_DOUBLE, MPI_MAX, root, comm);
/* for(size_t j=0; j<results->iter_index; j++) {
results->stage_times[i][j] = results->stage_times[i][j] / numP;
}
}*/
}
}
else {
for(i=0; i<stages; i++) {
MPI_Reduce(results->stage_times[i], NULL, results->iter_index, MPI_DOUBLE, MPI_SUM, root, comm);
MPI_Reduce(results->stage_times[i], NULL, results->iter_index, MPI_DOUBLE, MPI_MAX, root, comm);
}
}
//MPI_Barrier(comm); //FIXME Esto debería de borrarse
}
//======================================================||
......@@ -189,12 +193,12 @@ void compute_results_stages(results_data *results, int myId, int numP, int root,
void print_iter_results(results_data results) {
size_t i;
printf("Async_Iters: %ld\n", results.iters_async);
printf("T_iter: ");
for(i=0; i< results.iter_index; i++) {
printf("%lf ", results.iters_time[i]);
}
printf("\nAsync_Iters: %ld\n", results.iters_async);
printf("\n");
}
/*
......@@ -240,6 +244,11 @@ void print_global_results(results_data results, size_t resizes) {
printf("%lf ", results.async_time[i]);
}
printf("\nT_Malleability: ");
for(i=0; i < resizes; i++) {
printf("%lf ", results.malleability_time[i]);
}
printf("\nT_total: %lf\n", results.exec_time);
}
......@@ -262,6 +271,7 @@ void init_results_data(results_data *results, size_t resizes, size_t stages, siz
results->spawn_real_time = calloc(resizes, sizeof(double));
results->sync_time = calloc(resizes, sizeof(double));
results->async_time = calloc(resizes, sizeof(double));
results->malleability_time = calloc(resizes, sizeof(double));
results->wasted_time = 0;
results->iters_size = iters_size + RESULTS_EXTRA_SIZE;
......@@ -280,20 +290,24 @@ void realloc_results_iters(results_data *results, size_t stages, size_t needed)
int error = 0;
double *time_aux;
size_t i;
if(results->iters_size >= needed) return;
time_aux = (double *) realloc(results->iters_time, needed * sizeof(double));
if(time_aux == NULL) error = 1;
for(i=0; i<stages; i++) { //TODO Comprobar que no da error el realloc
results->stage_times[i] = (double *) realloc(results->stage_times[i], needed * sizeof(double));
if(results->stage_times[i] == NULL) error = 1;
}
if(time_aux == NULL) error = 1;
if(error) {
fprintf(stderr, "Fatal error - No se ha podido realojar la memoria de resultados\n");
MPI_Abort(MPI_COMM_WORLD, 1);
}
results->iters_time = time_aux;
results->iters_size = needed;
}
/*
......@@ -318,6 +332,10 @@ void free_results_data(results_data *results, size_t stages) {
free(results->async_time);
results->async_time = NULL;
}
if(results->malleability_time != NULL) {
free(results->malleability_time);
results->malleability_time = NULL;
}
if(results->iters_time != NULL) {
free(results->iters_time);
......
Codes/IOcodes/results.h
View file @ f1511cb4
......@@ -14,8 +14,9 @@ typedef struct {
// Spawn, Thread, Sync, Async and Exec time
double spawn_start, *spawn_time, *spawn_real_time;
double sync_start, sync_end, *sync_time;
double async_start, async_end, *async_time;
double sync_end, *sync_time;
double async_end, *async_time;
double malleability_end, *malleability_time;
double exec_start, exec_time;
double wasted_time; // Time spent recalculating iter stages
} results_data;
......
Codes/Main/Main.c
View file @ f1511cb4
......@@ -12,6 +12,8 @@
#include "../malleability/malleabilityManager.h"
#include "../malleability/malleabilityStates.h"
#define DR_MAX_SIZE 1000000000
int work();
double iterate(int async_comm);
double iterate_relaxed(double *time, double *times_stages);
......@@ -37,6 +39,7 @@ int main(int argc, char *argv[]) {
int numP, myId, res;
int req;
int im_child;
size_t i;
int num_cpus, num_nodes;
char *nodelist = NULL;
......@@ -54,6 +57,8 @@ int main(int argc, char *argv[]) {
if(req != MPI_THREAD_MULTIPLE) {
printf("No se ha obtenido la configuración de hilos necesaria\nSolicitada %d -- Devuelta %d\n", req, MPI_THREAD_MULTIPLE);
fflush(stdout);
MPI_Abort(MPI_COMM_WORLD, -50);
}
init_group_struct(argv, argc, myId, numP);
......@@ -66,11 +71,27 @@ int main(int argc, char *argv[]) {
set_benchmark_configuration(config_file);
set_benchmark_results(results);
if(config_file->n_groups > 1) {
set_malleability_configuration(config_file->groups[group->grp+1].sm, config_file->groups[group->grp+1].ss,
config_file->groups[group->grp+1].phy_dist, config_file->groups[group->grp+1].rm, config_file->groups[group->grp+1].rs);
set_children_number(config_file->groups[group->grp+1].procs); // TODO TO BE DEPRECATED
malleability_add_data(&(group->grp), 1, MAL_INT, 1, 1);
malleability_add_data(&run_id, 1, MAL_INT, 1, 1);
malleability_add_data(&(group->iter_start), 1, MAL_INT, 1, 1);
if(config_file->sdr) {
for(i=0; i<group->sync_data_groups; i++) {
malleability_add_data(group->sync_array[i], group->sync_qty[i], MAL_CHAR, 0, 1);
}
}
if(config_file->adr) {
for(i=0; i<group->async_data_groups; i++) {
malleability_add_data(group->async_array[i], group->async_qty[i], MAL_CHAR, 0, 0);
}
}
}
MPI_Barrier(comm);
results->exec_start = MPI_Wtime();
} else { //Init hijos
......@@ -82,6 +103,7 @@ int main(int argc, char *argv[]) {
// TODO Refactor - Que sea una unica funcion
// Obtiene las variables que van a utilizar los hijos
void *value = NULL;
size_t entries;
malleability_get_data(&value, 0, 1, 1);
group->grp = *((int *)value);
......@@ -91,7 +113,25 @@ int main(int argc, char *argv[]) {
malleability_get_data(&value, 2, 1, 1);
group->iter_start = *((int *)value);
if(config_file->sdr) {
malleability_get_entries(&entries, 0, 1);
group->sync_array = (char **) malloc(entries * sizeof(char *));
for(i=0; i<entries; i++) {
malleability_get_data(&value, i, 0, 1);
group->sync_array[i] = (char *)value;
}
}
if(config_file->adr) {
malleability_get_entries(&entries, 0, 0);
group->async_array = (char **) malloc(entries * sizeof(char *));
for(i=0; i<entries; i++) {
malleability_get_data(&value, i, 0, 0);
group->async_array[i] = (char *)value;
}
}
group->grp = group->grp + 1;
realloc_results_iters(results, config_file->n_stages, config_file->groups[group->grp].iters);
}
//
......@@ -105,14 +145,15 @@ int main(int argc, char *argv[]) {
MPI_Comm_rank(comm, &(group->myId));
group->grp = group->grp + 1;
set_benchmark_grp(group->grp);
if(group->grp != 0) {
obtain_op_times(1); //Obtener los nuevos valores de tiempo para el computo
obtain_op_times(0); //Obtener los nuevos valores de tiempo para el computo
set_results_post_reconfig(results, group->grp, config_file->sdr, config_file->adr);
}
if(config_file->n_groups != group->grp + 1) { //TODO Llevar a otra funcion
set_malleability_configuration(config_file->groups[group->grp+1].sm, config_file->groups[group->grp+1].ss,
config_file->groups[group->grp+1].phy_dist, config_file->groups[group->grp+1].at, -1);
config_file->groups[group->grp+1].phy_dist, config_file->groups[group->grp+1].rm, config_file->groups[group->grp+1].rs);
set_children_number(config_file->groups[group->grp+1].procs); // TODO TO BE DEPRECATED
if(group->grp != 0) {
......@@ -122,11 +163,11 @@ int main(int argc, char *argv[]) {
res = work();
if(res == MALL_ZOMBIE) break;
if(res==1) { // Se ha llegado al final de la aplicacion
MPI_Barrier(comm); // TODO Posible error al utilizar SHRINK
MPI_Barrier(comm);
results->exec_time = MPI_Wtime() - results->exec_start - results->wasted_time;
}
print_local_results();
reset_results_index(results);
} while(config_file->n_groups > group->grp + 1 && config_file->groups[group->grp+1].sm == MALL_SPAWN_MERGE);
......@@ -180,8 +221,8 @@ int work() {
state = malleability_checkpoint();
iter = 0;
while(state == MALL_DIST_PENDING || state == MALL_SPAWN_PENDING || state == MALL_SPAWN_SINGLE_PENDING || state == MALL_SPAWN_ADAPT_POSTPONE) {
if(iter < config_file->groups[group->grp+1].iters) {
while(state == MALL_DIST_PENDING || state == MALL_SPAWN_PENDING || state == MALL_SPAWN_SINGLE_PENDING || state == MALL_SPAWN_ADAPT_POSTPONE || state == MALL_SPAWN_ADAPT_PENDING) {
if(group->grp+1 < config_file->n_groups && iter < config_file->groups[group->grp+1].iters) {
iterate(state);
iter++;
group->iter_start = iter;
......@@ -227,6 +268,7 @@ double iterate(int async_comm) {
results->iters_async += 1;
}
// TODO Pasar el resto de este código a results.c
if(results->iter_index == results->iters_size) { // Aumentar tamaño de ambos vectores de resultados
realloc_results_iters(results, config_file->n_stages, results->iters_size + 100);
}
......@@ -235,6 +277,7 @@ double iterate(int async_comm) {
results->stage_times[i][results->iter_index] = times_stages_aux[i];
}
results->iter_index = results->iter_index + 1;
// TODO Pasar hasta aqui
free(times_stages_aux);
......@@ -395,6 +438,8 @@ void init_group_struct(char *argv[], int argc, int myId, int numP) {
* se comunican con los padres para inicializar sus datos.
*/
void init_application() {
int i, last_index;
if(group->argc < 2) {
printf("Falta el fichero de configuracion. Uso:\n./programa config.ini id\nEl argumento numerico id es opcional\n");
MPI_Abort(MPI_COMM_WORLD, -1);
......@@ -407,10 +452,29 @@ void init_application() {
results = malloc(sizeof(results_data));
init_results_data(results, config_file->n_resizes, config_file->n_stages, config_file->groups[group->grp].iters);
if(config_file->sdr) {
malloc_comm_array(&(group->sync_array), config_file->sdr , group->myId, group->numP);
group->sync_data_groups = config_file->sdr % DR_MAX_SIZE ? config_file->sdr/DR_MAX_SIZE+1 : config_file->sdr/DR_MAX_SIZE;
group->sync_qty = (int *) malloc(group->sync_data_groups * sizeof(int));
group->sync_array = (char **) malloc(group->sync_data_groups * sizeof(char *));
last_index = group->sync_data_groups-1;
for(i=0; i<last_index; i++) {
group->sync_qty[i] = DR_MAX_SIZE;
malloc_comm_array(&(group->sync_array[i]), group->sync_qty[i], group->myId, group->numP);
}
group->sync_qty[last_index] = config_file->sdr % DR_MAX_SIZE ? config_file->sdr % DR_MAX_SIZE : DR_MAX_SIZE;
malloc_comm_array(&(group->sync_array[last_index]), group->sync_qty[last_index], group->myId, group->numP);
}
if(config_file->adr) {
malloc_comm_array(&(group->async_array), config_file->adr , group->myId, group->numP);
group->async_data_groups = config_file->adr % DR_MAX_SIZE ? config_file->adr/DR_MAX_SIZE+1 : config_file->adr/DR_MAX_SIZE;
group->async_qty = (int *) malloc(group->async_data_groups * sizeof(int));
group->async_array = (char **) malloc(group->async_data_groups * sizeof(char *));
last_index = group->async_data_groups-1;
for(i=0; i<last_index; i++) {
group->async_qty[i] = DR_MAX_SIZE;
malloc_comm_array(&(group->async_array[i]), group->async_qty[i], group->myId, group->numP);
}
group->async_qty[last_index] = config_file->adr % DR_MAX_SIZE ? config_file->adr % DR_MAX_SIZE : DR_MAX_SIZE;
malloc_comm_array(&(group->async_array[last_index]), group->async_qty[last_index], group->myId, group->numP);
}
obtain_op_times(1);
......@@ -440,13 +504,29 @@ void obtain_op_times(int compute) {
* Libera toda la memoria asociada con la aplicacion
*/
void free_application_data() {
if(config_file->sdr) {
size_t i;
if(config_file->sdr && group->sync_array != NULL) {
for(i=0; i<group->sync_data_groups; i++) {
free(group->sync_array[i]);
group->sync_array[i] = NULL;
}
free(group->sync_qty);
group->sync_qty = NULL;
free(group->sync_array);
group->sync_array = NULL;
}
if(config_file->adr) {
if(config_file->adr && group->async_array != NULL) {
for(i=0; i<group->async_data_groups; i++) {
free(group->async_array[i]);
group->async_array[i] = NULL;
}
free(group->async_qty);
group->async_qty = NULL;
free(group->async_array);
group->async_array = NULL;
}
free_malleability();
free_results_data(results, config_file->n_stages);
......
Codes/Main/Main_datatypes.h
View file @ f1511cb4
......@@ -15,13 +15,14 @@ typedef struct {
unsigned int grp;
int iter_start;
int argc;
size_t sync_data_groups, async_data_groups;
int numS; // Cantidad de procesos hijos
MPI_Comm children, parents;
char *compute_comm_array, *compute_comm_recv;
char **argv;
char *sync_array, *async_array;
char **sync_array, **async_array;
int *sync_qty, *async_qty;
} group_data;
......@@ -48,7 +49,7 @@ typedef struct
typedef struct
{
int iters, procs;
int sm, ss, phy_dist, at;
int sm, ss, phy_dist, rm, rs;
float factor;
} group_config_t;
......@@ -57,7 +58,8 @@ typedef struct
size_t n_groups, n_resizes, n_stages; // n_groups==n_resizes+1
size_t actual_group, actual_stage;
int rigid_times;
int granularity, sdr, adr;
int granularity;
size_t sdr, adr;
MPI_Datatype config_type, group_type, iter_stage_type;
iter_stage_t *stages;
......
Codes/Main/configuration.c
View file @ f1511cb4
......@@ -71,7 +71,8 @@ void malloc_config_resizes(configuration *user_config) {
user_config->groups[i].sm = 0;
user_config->groups[i].ss = 1;
user_config->groups[i].phy_dist = 0;
user_config->groups[i].at = 0;
user_config->groups[i].rm = 0;
user_config->groups[i].rs = 1;
user_config->groups[i].factor = 1;
}
def_struct_groups(user_config);
......@@ -135,18 +136,14 @@ void free_config(configuration *user_config) {
}
}
//Liberar tipos derivados
if(user_config->config_type != MPI_DATATYPE_NULL) {
MPI_Type_free(&(user_config->config_type));
user_config->config_type = MPI_DATATYPE_NULL;
}
if(user_config->group_type != MPI_DATATYPE_NULL) {
MPI_Type_free(&(user_config->group_type));
user_config->group_type = MPI_DATATYPE_NULL;
}
if(user_config->iter_stage_type != MPI_DATATYPE_NULL) {
MPI_Type_free(&(user_config->iter_stage_type));
user_config->iter_stage_type = MPI_DATATYPE_NULL;
}
free(user_config->groups);
free(user_config->stages);
......@@ -162,17 +159,17 @@ void free_config(configuration *user_config) {
void print_config(configuration *user_config) {
if(user_config != NULL) {
size_t i;
printf("Config loaded: R=%zu, S=%zu, granularity=%d, SDR=%d, ADR=%d\n",
printf("Config loaded: R=%zu, S=%zu, granularity=%d, SDR=%zu, ADR=%zu\n",
user_config->n_resizes, user_config->n_stages, user_config->granularity, user_config->sdr, user_config->adr);
for(i=0; i<user_config->n_stages; i++) {
printf("Stage %zu: PT=%d, T_stage=%lf, bytes=%d, T_capped=%d\n",
i, user_config->stages[i].pt, user_config->stages[i].t_stage, user_config->stages[i].real_bytes, user_config->stages[i].t_capped);
}
for(i=0; i<user_config->n_groups; i++) {
printf("Group %zu: Iters=%d, Procs=%d, Factors=%f, Dist=%d, AT=%d, SM=%d, SS=%d\n",
printf("Group %zu: Iters=%d, Procs=%d, Factors=%f, Dist=%d, RM=%d, RS=%d, SM=%d, SS=%d\n",
i, user_config->groups[i].iters, user_config->groups[i].procs, user_config->groups[i].factor,
user_config->groups[i].phy_dist, user_config->groups[i].at, user_config->groups[i].sm,
user_config->groups[i].ss);
user_config->groups[i].phy_dist, user_config->groups[i].rm, user_config->groups[i].rs,
user_config->groups[i].sm, user_config->groups[i].ss);
}
}
}
......@@ -194,16 +191,16 @@ void print_config_group(configuration *user_config, size_t grp) {
sons = user_config->groups[grp+1].procs;
}
printf("Config: granularity=%d, SDR=%d, ADR=%d\n",
printf("Config: granularity=%d, SDR=%zu, ADR=%zu\n",
user_config->granularity, user_config->sdr, user_config->adr);
for(i=0; i<user_config->n_stages; i++) {
printf("Stage %zu: PT=%d, T_stage=%lf, bytes=%d, T_capped=%d\n",
i, user_config->stages[i].pt, user_config->stages[i].t_stage, user_config->stages[i].real_bytes, user_config->stages[i].t_capped);
}
printf("Group %zu: Iters=%d, Procs=%d, Factors=%f, Dist=%d, AT=%d, SM=%d, SS=%d, parents=%d, children=%d\n",
printf("Group %zu: Iters=%d, Procs=%d, Factors=%f, Dist=%d, RM=%d, RS=%d, SM=%d, SS=%d, parents=%d, children=%d\n",
grp, user_config->groups[grp].iters, user_config->groups[grp].procs, user_config->groups[grp].factor,
user_config->groups[grp].phy_dist, user_config->groups[grp].at, user_config->groups[grp].sm,
user_config->groups[grp].ss, parents, sons);
user_config->groups[grp].phy_dist, user_config->groups[grp].rm, user_config->groups[grp].rs,
user_config->groups[grp].sm, user_config->groups[grp].ss, parents, sons);
}
}
......@@ -270,17 +267,17 @@ void def_struct_config_file(configuration *config_file) {
MPI_Datatype types[counts];
// Rellenar vector types
types[0] = types[1] = MPI_UNSIGNED_LONG;
types[2] = types[3] = types[4] = types[5] = MPI_INT;
types[0] = types[1] = types[2] = types[3] = MPI_UNSIGNED_LONG;
types[4] = types[5] = MPI_INT;
// Rellenar vector displs
MPI_Get_address(config_file, &dir);
MPI_Get_address(&(config_file->n_groups), &displs[0]);
MPI_Get_address(&(config_file->n_stages), &displs[1]);
MPI_Get_address(&(config_file->granularity), &displs[2]);
MPI_Get_address(&(config_file->sdr), &displs[3]);
MPI_Get_address(&(config_file->adr), &displs[4]);
MPI_Get_address(&(config_file->sdr), &displs[2]);
MPI_Get_address(&(config_file->adr), &displs[3]);
MPI_Get_address(&(config_file->granularity), &displs[4]);
MPI_Get_address(&(config_file->rigid_times), &displs[5]);
for(i=0;i<counts;i++) displs[i] -= dir;
......@@ -295,15 +292,15 @@ void def_struct_config_file(configuration *config_file) {
* en una sola comunicacion.
*/
void def_struct_groups(configuration *config_file) {
int i, counts = 7;
int blocklengths[7] = {1, 1, 1, 1, 1, 1, 1};
int i, counts = 8;
int blocklengths[8] = {1, 1, 1, 1, 1, 1, 1, 1};
MPI_Aint displs[counts], dir;
MPI_Datatype aux, types[counts];
group_config_t *groups = config_file->groups;
// Rellenar vector types
types[0] = types[1] = types[2] = types[3] = types[4] = types[5] = MPI_INT;
types[6] = MPI_FLOAT;
types[0] = types[1] = types[2] = types[3] = types[4] = types[5] = types[6] = MPI_INT;
types[7] = MPI_FLOAT;
// Rellenar vector displs
MPI_Get_address(groups, &dir);
......@@ -313,8 +310,9 @@ void def_struct_groups(configuration *config_file) {
MPI_Get_address(&(groups->sm), &displs[2]);
MPI_Get_address(&(groups->ss), &displs[3]);
MPI_Get_address(&(groups->phy_dist), &displs[4]);
MPI_Get_address(&(groups->at), &displs[5]);
MPI_Get_address(&(groups->factor), &displs[6]);
MPI_Get_address(&(groups->rm), &displs[5]);
MPI_Get_address(&(groups->rs), &displs[6]);
MPI_Get_address(&(groups->factor), &displs[7]);
for(i=0;i<counts;i++) displs[i] -= dir;
......@@ -326,7 +324,7 @@ void def_struct_groups(configuration *config_file) {
// Tipo derivado para enviar N elementos de la estructura
MPI_Type_create_resized(aux, 0, sizeof(group_config_t), &(config_file->group_type));
MPI_Type_commit(&(config_file->group_type));
// MPI_Type_free(&aux); //FIXME It should be freed
MPI_Type_free(&aux);
}
}
......@@ -364,6 +362,6 @@ void def_struct_iter_stage(configuration *config_file) {
// Tipo derivado para enviar N elementos de la estructura
MPI_Type_create_resized(aux, 0, sizeof(iter_stage_t), &(config_file->iter_stage_type));
MPI_Type_commit(&(config_file->iter_stage_type));
// MPI_Type_free(&aux); //FIXME It should be freed
MPI_Type_free(&aux);
}
}
Codes/Makefile
View file @ f1511cb4
CC = gcc
MCC = mpicc
#C_FLAGS_ALL = -Wconversion -Wpedantic
C_FLAGS = -Wall -Wextra -Wshadow -Wfatal-errors -g
C_FLAGS = -Wall -Wextra -Wshadow -Wfatal-errors
LD_FLAGS = -lm -pthread
DEF =
......
Codes/malleability/CommDist.c
View file @ f1511cb4
......@@ -5,16 +5,22 @@
#include "distribution_methods/block_distribution.h"
#include "CommDist.h"
void send_sync_arrays(struct Dist_data dist_data, char *array, int numP_child, struct Counts counts);
void recv_sync_arrays(struct Dist_data dist_data, char *array, int numP_parents, struct Counts counts);
void prepare_redistribution(int qty, int myId, int numP, int numO, int is_children_group, int is_intercomm, char **recv, struct Counts *s_counts, struct Counts *r_counts);
void check_requests(struct Counts s_counts, struct Counts r_counts, int red_strategies, MPI_Request **requests, size_t *request_qty);
void send_async_arrays(struct Dist_data dist_data, char *array, int numP_child, struct Counts counts, MPI_Request *comm_req);
void recv_async_arrays(struct Dist_data dist_data, char *array, int numP_parents, struct Counts counts, MPI_Request *comm_req);
void sync_point2point(char *send, char *recv, int is_intercomm, int myId, struct Counts s_counts, struct Counts r_counts, MPI_Comm comm);
void sync_rma(char *send, char *recv, struct Counts r_counts, int tamBl, MPI_Comm comm, int red_method);
void sync_rma_lock(char *recv, struct Counts r_counts, MPI_Win win);
void sync_rma_lockall(char *recv, struct Counts r_counts, MPI_Win win);
void send_async_point_arrays(struct Dist_data dist_data, char *array, int numP_child, struct Counts counts, MPI_Request *comm_req);
void recv_async_point_arrays(struct Dist_data dist_data, char *array, int numP_parents, struct Counts counts, MPI_Request *comm_req);
void getIds_intercomm(struct Dist_data dist_data, int numP_other, int **idS);
void async_point2point(char *send, char *recv, struct Counts s_counts, struct Counts r_counts, MPI_Comm comm, MPI_Request *requests);
void async_rma(char *send, char *recv, struct Counts r_counts, int tamBl, MPI_Comm comm, int red_method, MPI_Request *requests, MPI_Win *win);
void async_rma_lock(char *recv, struct Counts r_counts, MPI_Win win, MPI_Request *requests);
void async_rma_lockall(char *recv, struct Counts r_counts, MPI_Win win, MPI_Request *requests);
void perform_manual_communication(char *send, char *recv, int myId, struct Counts s_counts, struct Counts r_counts);
/*
* Reserva memoria para un vector de hasta "qty" elementos.
* Los "qty" elementos se disitribuyen entre los "numP" procesos
......@@ -24,7 +30,7 @@ void malloc_comm_array(char **array, int qty, int myId, int numP) {
struct Dist_data dist_data;
get_block_dist(qty, myId, numP, &dist_data);
if( (*array = malloc(dist_data.tamBl * sizeof(char))) == NULL) {
if( (*array = calloc(dist_data.tamBl, sizeof(char))) == NULL) {
printf("Memory Error (Malloc Arrays(%d))\n", dist_data.tamBl);
exit(1);
}
......@@ -46,102 +52,200 @@ void malloc_comm_array(char **array, int qty, int myId, int numP) {
//================================================================================
/*
* Realiza un envio síncrono del vector array desde este grupo de procesos al grupo
* enlazado por el intercomunicador intercomm.
* Performs a communication to redistribute an array in a block distribution.
* In the redistribution is differenciated parent group from the children and the values each group indicates can be
* different.
*
* El vector array no se modifica en esta funcion.
* - send (IN): Array with the data to send. This data can not be null for parents.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - qty (IN): Sum of elements shared by all processes that will send data.
* - myId (IN): Rank of the MPI process in the local communicator. For the parents is not the rank obtained from "comm".
* - numP (IN): Size of the local group. If it is a children group, this parameter must correspond to using
* "MPI_Comm_size(comm)". For the parents is not always the size obtained from "comm".
* - numO (IN): Amount of processes in the remote group. For the parents is the target quantity of processes after the
* resize, while for the children is the amount of parents.
* - is_children_group (IN): Indicates wether this MPI rank is a children(TRUE) or a parent(FALSE).
* - comm (IN): Communicator to use to perform the redistribution.
*
* returns: An integer indicating if the operation has been completed(TRUE) or not(FALSE). //FIXME In this case is always true...
*/
int send_sync(char *array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_child) {
int *idS = NULL;
struct Counts counts;
int sync_communication(char *send, char **recv, int qty, int myId, int numP, int numO, int is_children_group, int red_method, MPI_Comm comm) {
int is_intercomm, aux_comm_used = 0;
struct Counts s_counts, r_counts;
struct Dist_data dist_data;
MPI_Comm aux_comm = MPI_COMM_NULL;
get_block_dist(qty, myId, numP, &dist_data); // Distribucion de este proceso en su grupo
dist_data.intercomm = intercomm;
// Create arrays which contains info about how many elements will be send to each created process
mallocCounts(&counts, numP_child);
/* PREPARE COMMUNICATION */
MPI_Comm_test_inter(comm, &is_intercomm);
prepare_redistribution(qty, myId, numP, numO, is_children_group, is_intercomm, recv, &s_counts, &r_counts);
getIds_intercomm(dist_data, numP_child, &idS); // Obtener rango de Id hijos a los que este proceso manda datos
/* PERFORM COMMUNICATION */
switch(red_method) {
send_sync_arrays(dist_data, array, numP_child, counts);
freeCounts(&counts);
free(idS);
case MALL_RED_RMA_LOCKALL:
case MALL_RED_RMA_LOCK:
if(is_children_group) {
dist_data.tamBl = 0;
} else {
get_block_dist(qty, myId, numO, &dist_data);
}
if(is_intercomm) {
MPI_Intercomm_merge(comm, is_children_group, &aux_comm);
aux_comm_used = 1;
} else { aux_comm = comm; }
sync_rma(send, *recv, r_counts, dist_data.tamBl, aux_comm, red_method);
break;
case MALL_RED_POINT:
sync_point2point(send, *recv, is_intercomm, myId, s_counts, r_counts, comm);
break;
case MALL_RED_BASELINE:
default:
MPI_Alltoallv(send, s_counts.counts, s_counts.displs, MPI_CHAR, *recv, r_counts.counts, r_counts.displs, MPI_CHAR, comm);
break;
}
return 1;
if(aux_comm_used) {
MPI_Comm_free(&aux_comm);
}
freeCounts(&s_counts);
freeCounts(&r_counts);
return 1; //FIXME In this case is always true...
}
/*
* Realiza una recepcion síncrona del vector array a este grupo de procesos desde el grupo
* enlazado por el intercomunicador intercomm.
* Performs a series of blocking point2point communications to redistribute an array in a block distribution.
* It should be called after calculating how data should be redistributed.
*
* - send (IN): Array with the data to send. This value can not be NULL for parents.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to
* receive data. If the process receives data and is NULL, the behaviour is undefined.
* - is_intercomm (IN): Indicates wether the communicator is an intercommunicator (TRUE) or an
* intracommunicator (FALSE).
* - myId (IN): Rank of the MPI process in the local communicator. For the parents is not the rank obtained from "comm".
* - s_counts (IN): Struct which describes how many elements will send this process to each children and
* the displacements.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent
* and the displacements.
* - comm (IN): Communicator to use to perform the redistribution.
*
* El vector array se reserva dentro de la funcion y se devuelve en el mismo argumento.
* Tiene que ser liberado posteriormente por el usuario.
*/
void recv_sync(char **array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_parents) {
int *idS = NULL;
struct Counts counts;
struct Dist_data dist_data;
void sync_point2point(char *send, char *recv, int is_intercomm, int myId, struct Counts s_counts, struct Counts r_counts, MPI_Comm comm) {
int i, j, init, end, total_sends;
MPI_Request *sends;
// Obtener distribución para este hijo
get_block_dist(qty, myId, numP, &dist_data);
*array = malloc(dist_data.tamBl * sizeof(char));
//(*array)[dist_data.tamBl] = '\0';
dist_data.intercomm = intercomm;
init = s_counts.idI;
end = s_counts.idE;
if(!is_intercomm && (s_counts.idI == myId || s_counts.idE == myId + 1)) {
perform_manual_communication(send, recv, myId, s_counts, r_counts);
/* PREPARAR DATOS DE RECEPCION SOBRE VECTOR*/
mallocCounts(&counts, numP_parents);
if(s_counts.idI == myId) init = s_counts.idI+1;
else end = s_counts.idE-1;
}
getIds_intercomm(dist_data, numP_parents, &idS); // Obtener el rango de Ids de padres del que este proceso recibira datos
total_sends = end - init;
j = 0;
if(total_sends > 0) {
sends = (MPI_Request *) malloc(total_sends * sizeof(MPI_Request));
}
for(i=init; i<end; i++) {
sends[j] = MPI_REQUEST_NULL;
MPI_Isend(send+s_counts.displs[i], s_counts.counts[i], MPI_CHAR, i, 99, comm, &(sends[j]));
j++;
}
recv_sync_arrays(dist_data, *array, numP_parents, counts);
//printf("S%d Tam %d String: %s END\n", myId, dist_data.tamBl, *array);
init = r_counts.idI;
end = r_counts.idE;
if(!is_intercomm) {
if(r_counts.idI == myId) init = r_counts.idI+1;
else if(r_counts.idE == myId + 1) end = r_counts.idE-1;
}
freeCounts(&counts);
free(idS);
for(i=init; i<end; i++) {
MPI_Recv(recv+r_counts.displs[i], r_counts.counts[i], MPI_CHAR, i, 99, comm, MPI_STATUS_IGNORE);
}
if(total_sends > 0) {
MPI_Waitall(total_sends, sends, MPI_STATUSES_IGNORE);
}
}
/*
* Envia a los hijos un vector que es redistribuido a los procesos
* hijos. Antes de realizar la comunicacion, cada proceso padre calcula sobre que procesos
* del otro grupo se transmiten elementos.
* Performs synchronous MPI-RMA operations to redistribute an array in a block distribution. Is should be called after calculating
* how data should be redistributed
*
* - send (IN): Array with the data to send. This value can be NULL for children.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent and the
* displacements.
* - tamBl (IN): How many elements are stored in the parameter "send".
* - comm (IN): Communicator to use to perform the redistribution. Must be an intracommunicator as MPI-RMA requirements.
* - red_method (IN): Type of data redistribution to use. In this case indicates the RMA operation(Lock or LockAll).
*
*/
void send_sync_arrays(struct Dist_data dist_data, char *array, int numP_child, struct Counts counts) {
void sync_rma(char *send, char *recv, struct Counts r_counts, int tamBl, MPI_Comm comm, int red_method) {
MPI_Win win;
MPI_Win_create(send, (MPI_Aint)tamBl, sizeof(char), MPI_INFO_NULL, comm, &win);
switch(red_method) {
case MALL_RED_RMA_LOCKALL:
sync_rma_lockall(recv, r_counts, win);
break;
case MALL_RED_RMA_LOCK:
sync_rma_lock(recv, r_counts, win);
break;
}
prepare_comm_alltoall(dist_data.myId, dist_data.numP, numP_child, dist_data.qty, &counts);
/* COMUNICACION DE DATOS */
MPI_Alltoallv(array, counts.counts, counts.displs, MPI_CHAR, NULL, counts.zero_arr, counts.zero_arr, MPI_CHAR, dist_data.intercomm);
MPI_Win_free(&win);
}
/*
* Recibe de los padres un vector que es redistribuido a los procesos
* de este grupo. Antes de realizar la comunicacion cada hijo calcula sobre que procesos
* del otro grupo se transmiten elementos.
* Performs a passive MPI-RMA data redistribution for a single array using the passive epochs Lock/Unlock.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent and the
* displacements.
* - win (IN): Window to use to perform the redistribution.
*
*/
void recv_sync_arrays(struct Dist_data dist_data, char *array, int numP_parents, struct Counts counts) {
void sync_rma_lock(char *recv, struct Counts r_counts, MPI_Win win) {
int i, target_displs;
target_displs = r_counts.first_target_displs;
for(i=r_counts.idI; i<r_counts.idE; i++) {
MPI_Win_lock(MPI_LOCK_SHARED, i, MPI_MODE_NOCHECK, win);
MPI_Get(recv+r_counts.displs[i], r_counts.counts[i], MPI_CHAR, i, target_displs, r_counts.counts[i], MPI_CHAR, win);
MPI_Win_unlock(i, win);
target_displs=0;
}
}
char aux;
prepare_comm_alltoall(dist_data.myId, dist_data.numP, numP_parents, dist_data.qty, &counts);
// Ajustar los valores de recepcion
/*
if(idI == 0) {
set_counts(0, numP_parents, dist_data, counts.counts);
idI++;
/*
* Performs a passive MPI-RMA data redistribution for a single array using the passive epochs Lockall/Unlockall.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent and the
* displacements.
* - win (IN): Window to use to perform the redistribution.
*
*/
void sync_rma_lockall(char *recv, struct Counts r_counts, MPI_Win win) {
int i, target_displs;
target_displs = r_counts.first_target_displs;
MPI_Win_lock_all(MPI_MODE_NOCHECK, win);
for(i=r_counts.idI; i<r_counts.idE; i++) {
MPI_Get(recv+r_counts.displs[i], r_counts.counts[i], MPI_CHAR, i, target_displs, r_counts.counts[i], MPI_CHAR, win);
target_displs=0;
}
for(i=idI; i<idE; i++) {
set_counts(i, numP_parents, dist_data, counts.counts);
counts.displs[i] = counts.displs[i-1] + counts.counts[i-1];
}*/
//print_counts(dist_data, counts.counts, counts.displs, numP_parents, "Hijos");
/* COMUNICACION DE DATOS */
MPI_Alltoallv(&aux, counts.zero_arr, counts.zero_arr, MPI_CHAR, array, counts.counts, counts.displs, MPI_CHAR, dist_data.intercomm);
MPI_Win_unlock_all(win);
}
//================================================================================
//================================================================================
//========================ASYNCHRONOUS FUNCTIONS==================================
......@@ -149,187 +253,270 @@ void recv_sync_arrays(struct Dist_data dist_data, char *array, int numP_parents,
//================================================================================
/*
* Realiza un envio asincrono del vector array desde este grupo de procesos al grupo
* enlazado por el intercomunicador intercomm.
* Performs a communication to redistribute an array in a block distribution with non-blocking MPI functions.
* In the redistribution is differenciated parent group from the children and the values each group indicates can be
* different.
*
* El objeto MPI_Request se devuelve con el manejador para comprobar si la comunicacion
* ha terminado.
* - send (IN): Array with the data to send. This data can not be null for parents.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - qty (IN): Sum of elements shared by all processes that will send data.
* - myId (IN): Rank of the MPI process in the local communicator. For the parents is not the rank obtained from "comm".
* - numP (IN): Size of the local group. If it is a children group, this parameter must correspond to using
* "MPI_Comm_size(comm)". For the parents is not always the size obtained from "comm".
* - numO (IN): Amount of processes in the remote group. For the parents is the target quantity of processes after the
* resize, while for the children is the amount of parents.
* - is_children_group (IN): Indicates wether this MPI rank is a children(TRUE) or a parent(FALSE).
* - comm (IN): Communicator to use to perform the redistribution.
* - requests (OUT): Pointer to array of requests to be used to determine if the communication has ended. If the pointer
* is null or not enough space has been reserved the pointer is allocated/reallocated.
* - request_qty (OUT): Quantity of requests to be used. If a process sends and receives data, this value will be
* modified to the expected value.
*
* El vector array no se modifica en esta funcion.
* returns: An integer indicating if the operation has been completed(TRUE) or not(FALSE). //FIXME In this case is always false...
*/
int send_async(char *array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_child, MPI_Request **comm_req, int parents_wait) {
int i;
int *idS = NULL;
struct Counts counts;
int async_communication_start(char *send, char **recv, int qty, int myId, int numP, int numO, int is_children_group, int red_method, int red_strategies, MPI_Comm comm, MPI_Request **requests, size_t *request_qty, MPI_Win *win) {
int is_intercomm, aux_comm_used = 0;
struct Counts s_counts, r_counts;
struct Dist_data dist_data;
MPI_Comm aux_comm = MPI_COMM_NULL;
/* PREPARE COMMUNICATION */
MPI_Comm_test_inter(comm, &is_intercomm);
prepare_redistribution(qty, myId, numP, numO, is_children_group, is_intercomm, recv, &s_counts, &r_counts);
check_requests(s_counts, r_counts, red_strategies, requests, request_qty);
/* PERFORM COMMUNICATION */
switch(red_method) {
case MALL_RED_RMA_LOCKALL:
case MALL_RED_RMA_LOCK:
if(is_children_group) {
dist_data.tamBl = 0;
} else {
get_block_dist(qty, myId, numO, &dist_data);
}
if(is_intercomm) {
MPI_Intercomm_merge(comm, is_children_group, &aux_comm);
aux_comm_used = 1;
} else { aux_comm = comm; }
async_rma(send, *recv, r_counts, dist_data.tamBl, aux_comm, red_method, *requests, win);
break;
case MALL_RED_POINT:
async_point2point(send, *recv, s_counts, r_counts, comm, *requests);
break;
case MALL_RED_BASELINE:
default:
MPI_Ialltoallv(send, s_counts.counts, s_counts.displs, MPI_CHAR, *recv, r_counts.counts, r_counts.displs, MPI_CHAR, comm, &((*requests)[0]));
break;
}
get_block_dist(qty, myId, numP, &dist_data); // Distribucion de este proceso en su grupo
dist_data.intercomm = intercomm;
// Create arrays which contains info about how many elements will be send to each created process
mallocCounts(&counts, numP_child);
getIds_intercomm(dist_data, numP_child, &idS); // Obtener rango de Id hijos a los que este proceso manda datos
// MAL_USE_THREAD sigue el camino sincrono
if(parents_wait == MAL_USE_NORMAL) {
//*comm_req = (MPI_Request *) malloc(sizeof(MPI_Request));
*comm_req[0] = MPI_REQUEST_NULL;
send_async_arrays(dist_data, array, numP_child, counts, &(*comm_req[0]));
} else if (parents_wait == MAL_USE_IBARRIER){
//*comm_req = (MPI_Request *) malloc(2 * sizeof(MPI_Request));
*comm_req[0] = MPI_REQUEST_NULL;
*comm_req[1] = MPI_REQUEST_NULL;
send_async_arrays(dist_data, array, numP_child, counts, &((*comm_req)[1]));
MPI_Ibarrier(intercomm, &((*comm_req)[0]) );
} else if (parents_wait == MAL_USE_POINT){
//*comm_req = (MPI_Request *) malloc(numP_child * sizeof(MPI_Request));
for(i=0; i<numP_child; i++){
(*comm_req)[i] = MPI_REQUEST_NULL;
/* POST REQUESTS CHECKS */
if(malleability_red_contains_strat(red_strategies, MALL_RED_IBARRIER, NULL)) {
if(!is_children_group && (is_intercomm || myId >= numO)) {
MPI_Ibarrier(comm, &((*requests)[*request_qty-1]) ); //FIXME Not easy to read...
}
send_async_point_arrays(dist_data, array, numP_child, counts, *comm_req);
} else if (parents_wait == MAL_USE_THREAD) { //TODO
}
freeCounts(&counts);
free(idS);
if(aux_comm_used) {
MPI_Comm_free(&aux_comm);
}
return 1;
freeCounts(&s_counts);
freeCounts(&r_counts);
return 0; //FIXME In this case is always false...
}
/*
* Realiza una recepcion asincrona del vector array a este grupo de procesos desde el grupo
* enlazado por el intercomunicador intercomm.
* Checks if a set of requests have been completed (1) or not (0).
*
* El vector array se reserva dentro de la funcion y se devuelve en el mismo argumento.
* Tiene que ser liberado posteriormente por el usuario.
* - myId (IN): Rank of the MPI process in the local communicator. For the parents is not the rank obtained from "comm".
* - is_children_group (IN): Indicates wether this MPI rank is a children(TRUE) or a parent(FALSE).
* - red_strategies (IN):
* - requests (IN): Pointer to array of requests to be used to determine if the communication has ended.
* - request_qty (IN): Quantity of requests in "requests".
*
* El argumento "parents_wait" sirve para indicar si se usará la versión en la los padres
* espera a que terminen de enviar, o en la que esperan a que los hijos acaben de recibir.
* returns: An integer indicating if the operation has been completed(TRUE) or not(FALSE).
*/
void recv_async(char **array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_parents, int parents_wait) {
int *idS = NULL;
int wait_err, i;
struct Counts counts;
struct Dist_data dist_data;
MPI_Request *comm_req, aux;
// Obtener distribución para este hijo
get_block_dist(qty, myId, numP, &dist_data);
*array = malloc( dist_data.tamBl * sizeof(char));
dist_data.intercomm = intercomm;
/* PREPARAR DATOS DE RECEPCION SOBRE VECTOR*/
mallocCounts(&counts, numP_parents);
getIds_intercomm(dist_data, numP_parents, &idS); // Obtener el rango de Ids de padres del que este proceso recibira datos
// MAL_USE_THREAD sigue el camino sincrono
if(parents_wait == MAL_USE_POINT) {
comm_req = (MPI_Request *) malloc(numP_parents * sizeof(MPI_Request));
for(i=0; i<numP_parents; i++){
comm_req[i] = MPI_REQUEST_NULL;
int async_communication_check(int myId, int is_children_group, int red_strategies, MPI_Comm comm, MPI_Request *requests, size_t request_qty) {
int completed, req_completed, all_req_null, test_err, aux_condition;
size_t i;
completed = 1;
all_req_null = 1;
test_err = MPI_SUCCESS;
if (is_children_group) return 1;
if(malleability_red_contains_strat(red_strategies, MALL_RED_IBARRIER, NULL)) {
// The Ibarrier should only be posted at this point if the process
// has other requests which has not confirmed as completed yet,
// but are confirmed now.
if (requests[request_qty-1] == MPI_REQUEST_NULL) {
for(i=0; i<request_qty; i++) {
aux_condition = requests[i] == MPI_REQUEST_NULL;
all_req_null = all_req_null && aux_condition;
test_err = MPI_Test(&(requests[i]), &req_completed, MPI_STATUS_IGNORE);
completed = completed && req_completed;
}
recv_async_point_arrays(dist_data, *array, numP_parents, counts, comm_req);
wait_err = MPI_Waitall(numP_parents, comm_req, MPI_STATUSES_IGNORE);
} else if (parents_wait == MAL_USE_NORMAL || parents_wait == MAL_USE_IBARRIER) {
comm_req = (MPI_Request *) malloc(sizeof(MPI_Request));
*comm_req = MPI_REQUEST_NULL;
recv_async_arrays(dist_data, *array, numP_parents, counts, comm_req);
wait_err = MPI_Wait(comm_req, MPI_STATUS_IGNORE);
} else if (parents_wait == MAL_USE_THREAD) { //TODO
if(completed && !all_req_null) { MPI_Ibarrier(comm, &(requests[request_qty-1])); }
}
test_err = MPI_Test(&(requests[request_qty-1]), &completed, MPI_STATUS_IGNORE);
if(wait_err != MPI_SUCCESS) {
MPI_Abort(MPI_COMM_WORLD, wait_err);
} else {
for(i=0; i<request_qty; i++) {
test_err = MPI_Test(&(requests[i]), &req_completed, MPI_STATUS_IGNORE);
completed = completed && req_completed;
}
// test_err = MPI_Testall(request_qty, requests, &completed, MPI_STATUSES_IGNORE); //FIXME Some kind of bug with Mpich.
}
if(parents_wait == MAL_USE_IBARRIER) { //MAL USE IBARRIER END
MPI_Ibarrier(intercomm, &aux);
MPI_Wait(&aux, MPI_STATUS_IGNORE); //Es necesario comprobar que la comunicación ha terminado para desconectar los grupos de procesos
if (test_err != MPI_SUCCESS && test_err != MPI_ERR_PENDING) {
printf("P%d aborting -- Test Async\n", myId);
MPI_Abort(MPI_COMM_WORLD, test_err);
}
//printf("S%d Tam %d String: %s END\n", myId, dist_data.tamBl, *array);
freeCounts(&counts);
free(idS);
free(comm_req);
return completed;
}
/*
* Envia a los hijos un vector que es redistribuido a los procesos
* hijos. Antes de realizar la comunicacion, cada proceso padre calcula sobre que procesos
* del otro grupo se transmiten elementos.
* Waits until the completion of a set of requests. If the Ibarrier strategy
* is being used, the corresponding ibarrier is posted.
*
* El envio se realiza a partir de una comunicación colectiva.
* - red_strategies (IN):
* - comm (IN): Communicator to use to confirm finalizations of redistribution
* - requests (IN): Pointer to array of requests to be used to determine if the communication has ended.
* - request_qty (IN): Quantity of requests in "requests".
*/
void send_async_arrays(struct Dist_data dist_data, char *array, int numP_child, struct Counts counts, MPI_Request *comm_req) {
void async_communication_wait(int red_strategies, MPI_Comm comm, MPI_Request *requests, size_t request_qty) {
MPI_Waitall(request_qty, requests, MPI_STATUSES_IGNORE);
if(malleability_red_contains_strat(red_strategies, MALL_RED_IBARRIER, NULL)) {
MPI_Ibarrier(comm, &(requests[request_qty-1]) );
MPI_Wait(&(requests[request_qty-1]), MPI_STATUS_IGNORE); //TODO Is it really needed? It will be ensured later
}
}
prepare_comm_alltoall(dist_data.myId, dist_data.numP, numP_child, dist_data.qty, &counts);
/*
* Frees Requests/Windows associated to a particular redistribution.
* Should be called for each output result of calling "async_communication_start".
*
* - red_method (IN):
* - red_strategies (IN):
* - requests (IN): Pointer to array of requests to be used to determine if the communication has ended.
* - request_qty (IN): Quantity of requests in "requests".
* - win (IN): Window to free.
*/
void async_communication_end(int red_method, int red_strategies, MPI_Request *requests, size_t request_qty, MPI_Win *win) {
/* COMUNICACION DE DATOS */
MPI_Ialltoallv(array, counts.counts, counts.displs, MPI_CHAR, NULL, counts.zero_arr, counts.zero_arr, MPI_CHAR, dist_data.intercomm, comm_req);
//Para la desconexión de ambos grupos de procesos es necesario indicar a MPI que esta comm
//ha terminado, aunque solo se pueda llegar a este punto cuando ha terminado
if(malleability_red_contains_strat(red_strategies, MALL_RED_IBARRIER, NULL)) { MPI_Waitall(request_qty, requests, MPI_STATUSES_IGNORE); }
if(red_method == MALL_RED_RMA_LOCKALL || red_method == MALL_RED_RMA_LOCK) { MPI_Win_free(win); }
}
/*
* Envia a los hijos un vector que es redistribuido a los procesos
* hijos. Antes de realizar la comunicacion, cada proceso padre calcula sobre que procesos
* del otro grupo se transmiten elementos.
* Performs a series of non-blocking point2point communications to redistribute an array in a block distribution.
* It should be called after calculating how data should be redistributed.
*
* - send (IN): Array with the data to send. This value can not be NULL for parents.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to
* receive data. If the process receives data and is NULL, the behaviour is undefined.
* - s_counts (IN): Struct which describes how many elements will send this process to each children and
* the displacements.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent
* and the displacements.
* - comm (IN): Communicator to use to perform the redistribution.
* - requests (OUT): Pointer to array of requests to be used to determine if the communication has ended.
*
* El envio se realiza a partir de varias comunicaciones punto a punto.
*/
void send_async_point_arrays(struct Dist_data dist_data, char *array, int numP_child, struct Counts counts, MPI_Request *comm_req) {
int i;
// PREPARAR ENVIO DEL VECTOR
prepare_comm_alltoall(dist_data.myId, dist_data.numP, numP_child, dist_data.qty, &counts);
void async_point2point(char *send, char *recv, struct Counts s_counts, struct Counts r_counts, MPI_Comm comm, MPI_Request *requests) {
int i, j = 0;
for(i=0; i<numP_child; i++) { //TODO Esta propuesta ya no usa el IdI y Ide
if(counts.counts[0] != 0) {
MPI_Isend(array+counts.displs[i], counts.counts[i], MPI_CHAR, i, 99, dist_data.intercomm, &(comm_req[i]));
for(i=s_counts.idI; i<s_counts.idE; i++) {
MPI_Isend(send+s_counts.displs[i], s_counts.counts[i], MPI_CHAR, i, 99, comm, &(requests[j]));
j++;
}
for(i=r_counts.idI; i<r_counts.idE; i++) {
MPI_Irecv(recv+r_counts.displs[i], r_counts.counts[i], MPI_CHAR, i, 99, comm, &(requests[j]));
j++;
}
//print_counts(dist_data, counts.counts, counts.displs, numP_child, "Padres");
}
/*
* Recibe de los padres un vector que es redistribuido a los procesos
* de este grupo. Antes de realizar la comunicacion cada hijo calcula sobre que procesos
* del otro grupo se transmiten elementos.
* Performs asynchronous MPI-RMA operations to redistribute an array in a block distribution. Is should be called after calculating
* how data should be redistributed.
*
* - send (IN): Array with the data to send. This value can be NULL for children.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent and the
* displacements.
* - tamBl (IN): How many elements are stored in the parameter "send".
* - comm (IN): Communicator to use to perform the redistribution. Must be an intracommunicator as MPI-RMA requirements.
* - red_method (IN): Type of data redistribution to use. In this case indicates the RMA operation(Lock or LockAll).
* - window (OUT): Pointer to a window object used for the RMA operations.
* - requests (OUT): Pointer to array of requests to be used to determine if the communication has ended.
*
* La recepcion se realiza a partir de una comunicacion colectiva.
*/
void recv_async_arrays(struct Dist_data dist_data, char *array, int numP_parents, struct Counts counts, MPI_Request *comm_req) {
char *aux = malloc(1);
// Ajustar los valores de recepcion
prepare_comm_alltoall(dist_data.myId, dist_data.numP, numP_parents, dist_data.qty, &counts);
//print_counts(dist_data, counts.counts, counts.displs, numP_parents, "Hijos");
/* COMUNICACION DE DATOS */
MPI_Ialltoallv(aux, counts.zero_arr, counts.zero_arr, MPI_CHAR, array, counts.counts, counts.displs, MPI_CHAR, dist_data.intercomm, comm_req);
free(aux);
void async_rma(char *send, char *recv, struct Counts r_counts, int tamBl, MPI_Comm comm, int red_method, MPI_Request *requests, MPI_Win *win) {
MPI_Win_create(send, (MPI_Aint)tamBl, sizeof(char), MPI_INFO_NULL, comm, win);
switch(red_method) {
case MALL_RED_RMA_LOCKALL:
async_rma_lockall(recv, r_counts, *win, requests);
break;
case MALL_RED_RMA_LOCK:
async_rma_lock(recv, r_counts, *win, requests);
break;
}
}
/*
* Recibe de los padres un vector que es redistribuido a los procesos
* de este grupo. Antes de realizar la comunicacion cada hijo calcula sobre que procesos
* del otro grupo se transmiten elementos.
* Performs an asynchronous and passive MPI-RMA data redistribution for a single array using the passive epochs Lock/Unlock.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent and the
* displacements.
* - win (IN): Window to use to perform the redistribution.
* - requests (OUT): Pointer to array of requests to be used to determine if the communication has ended.
*
* La recepcion se realiza a partir de varias comunicaciones punto a punto.
*/
void recv_async_point_arrays(struct Dist_data dist_data, char *array, int numP_parents, struct Counts counts, MPI_Request *comm_req) {
int i;
// Ajustar los valores de recepcion
prepare_comm_alltoall(dist_data.myId, dist_data.numP, numP_parents, dist_data.qty, &counts);
for(i=0; i<numP_parents; i++) { //TODO Esta propuesta ya no usa el IdI y Ide
if(counts.counts[0] != 0) {
MPI_Irecv(array+counts.displs[i], counts.counts[i], MPI_CHAR, i, 99, dist_data.intercomm, &(comm_req[i])); //FIXME BUffer recv
void async_rma_lock(char *recv, struct Counts r_counts, MPI_Win win, MPI_Request *requests) {
int i, target_displs, j = 0;
target_displs = r_counts.first_target_displs;
for(i=r_counts.idI; i<r_counts.idE; i++) {
MPI_Win_lock(MPI_LOCK_SHARED, i, MPI_MODE_NOCHECK, win);
MPI_Rget(recv+r_counts.displs[i], r_counts.counts[i], MPI_CHAR, i, target_displs, r_counts.counts[i], MPI_CHAR, win, &(requests[j]));
MPI_Win_unlock(i, win);
target_displs=0;
j++;
}
}
/*
* Performs an asynchronous and passive MPI-RMA data redistribution for a single array using the passive epochs Lockall/Unlockall.
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* If the process receives data and is NULL, the behaviour is undefined.
* - r_counts (IN): Structure which describes how many elements will receive this process from each parent and the
* displacements.
* - win (IN): Window to use to perform the redistribution.
* - requests (OUT): Pointer to array of requests to be used to determine if the communication has ended.
*
*/
void async_rma_lockall(char *recv, struct Counts r_counts, MPI_Win win, MPI_Request *requests) {
int i, target_displs, j = 0;
target_displs = r_counts.first_target_displs;
MPI_Win_lock_all(MPI_MODE_NOCHECK, win);
for(i=r_counts.idI; i<r_counts.idE; i++) {
MPI_Rget(recv+r_counts.displs[i], r_counts.counts[i], MPI_CHAR, i, target_displs, r_counts.counts[i], MPI_CHAR, win, &(requests[j]));
target_displs=0;
j++;
}
//print_counts(dist_data, counts.counts, counts.displs, numP_parents, "Hijos");
MPI_Win_unlock_all(win);
}
/*
......@@ -341,40 +528,141 @@ void recv_async_point_arrays(struct Dist_data dist_data, char *array, int numP_p
*/
/*
* Obtiene para un proceso de un grupo a que rango procesos de
* otro grupo tiene que enviar o recibir datos.
* Performs a communication to redistribute an array in a block distribution. For each process calculates
* how many elements sends/receives to other processes for the new group.
*
* - qty (IN): Sum of elements shared by all processes that will send data.
* - myId (IN): Rank of the MPI process in the local communicator. For the parents is not the rank obtained from "comm".
* - numP (IN): Size of the local group. If it is a children group, this parameter must correspond to using
* "MPI_Comm_size(comm)". For the parents is not always the size obtained from "comm".
* - numO (IN): Amount of processes in the remote group. For the parents is the target quantity of processes after the
* resize, while for the children is the amount of parents.
* - is_children_group (IN): Indicates wether this MPI rank is a children(TRUE) or a parent(FALSE).
* - is_intercomm (IN): Indicates wether the used communicator is a intercomunicator(TRUE) or intracommunicator(FALSE).
* - recv (OUT): Array where data will be written. A NULL value is allowed if the process is not going to receive data.
* process receives data and is NULL, the behaviour is undefined.
* - s_counts (OUT): Struct where is indicated how many elements sends this process to processes in the new group.
* - r_counts (OUT): Struct where is indicated how many elements receives this process from other processes in the previous group.
*
* Devuelve el primer identificador y el último (Excluido) con el que
* comunicarse.
*/
void getIds_intercomm(struct Dist_data dist_data, int numP_other, int **idS) {
int idI, idE;
int tamOther = dist_data.qty / numP_other;
int remOther = dist_data.qty % numP_other;
// Indica el punto de corte del grupo de procesos externo que
// divide entre los procesos que tienen
// un tamaño tamOther + 1 y un tamaño tamOther
int middle = (tamOther + 1) * remOther;
// Calcular idI teniendo en cuenta si se comunica con un
// proceso con tamano tamOther o tamOther+1
if(middle > dist_data.ini) { // First subgroup (tamOther+1)
idI = dist_data.ini / (tamOther + 1);
} else { // Second subgroup (tamOther)
idI = ((dist_data.ini - middle) / tamOther) + remOther;
}
// Calcular idR teniendo en cuenta si se comunica con un
// proceso con tamano tamOther o tamOther+1
if(middle >= dist_data.fin) { // First subgroup (tamOther +1)
idE = dist_data.fin / (tamOther + 1);
idE = (dist_data.fin % (tamOther + 1) > 0 && idE+1 <= numP_other) ? idE+1 : idE;
} else { // Second subgroup (tamOther)
idE = ((dist_data.fin - middle) / tamOther) + remOther;
idE = ((dist_data.fin - middle) % tamOther > 0 && idE+1 <= numP_other) ? idE+1 : idE;
}
*idS = malloc(2 * sizeof(int));
(*idS)[0] = idI;
(*idS)[1] = idE;
void prepare_redistribution(int qty, int myId, int numP, int numO, int is_children_group, int is_intercomm, char **recv, struct Counts *s_counts, struct Counts *r_counts) {
int array_size = numO;
int offset_ids = 0;
struct Dist_data dist_data;
if(is_intercomm) {
//offset_ids = !is_children_group ? numP : 0; //FIXME Modify only if active?
} else {
array_size = numP > numO ? numP : numO;
}
mallocCounts(s_counts, array_size+offset_ids);
mallocCounts(r_counts, array_size+offset_ids);
if(is_children_group) {
prepare_comm_alltoall(myId, numP, numO, qty, offset_ids, r_counts);
// Obtener distribución para este hijo
get_block_dist(qty, myId, numP, &dist_data);
*recv = malloc(dist_data.tamBl * sizeof(char));
//get_block_dist(qty, myId, numP, &dist_data);
//print_counts(dist_data, r_counts->counts, r_counts->displs, numO+offset_ids, 0, "Children C ");
} else {
//get_block_dist(qty, myId, numP, &dist_data);
prepare_comm_alltoall(myId, numP, numO, qty, offset_ids, s_counts);
if(!is_intercomm && myId < numO) {
prepare_comm_alltoall(myId, numO, numP, qty, offset_ids, r_counts);
// Obtener distribución para este hijo y reservar vector de recibo
get_block_dist(qty, myId, numO, &dist_data);
*recv = malloc(dist_data.tamBl * sizeof(char));
//print_counts(dist_data, r_counts->counts, r_counts->displs, array_size, 0, "Children P ");
}
//print_counts(dist_data, s_counts->counts, s_counts->displs, numO+offset_ids, 0, "Parents ");
}
}
/*
* Ensures that the array of request of a process has an amount of elements equal to the amount of communication
* functions the process will perform. In case the array is not initialized or does not have enough space it is
* allocated/reallocated to the minimum amount of space needed.
*
* - s_counts (IN): Struct where is indicated how many elements sends this process to processes in the new group.
* - r_counts (IN): Struct where is indicated how many elements receives this process from other processes in the previous group.
* - requests (IN/OUT): Pointer to array of requests to be used to determine if the communication has ended. If the pointer
* is null or not enough space has been reserved the pointer is allocated/reallocated.
* - request_qty (IN/OUT): Quantity of requests to be used. If the value is smaller than the amount of communication
* functions to perform, it is modified to the minimum value.
*/
void check_requests(struct Counts s_counts, struct Counts r_counts, int red_strategies, MPI_Request **requests, size_t *request_qty) {
size_t i, sum;
MPI_Request *aux;
sum = (size_t) s_counts.idE - s_counts.idI;
sum += (size_t) r_counts.idE - r_counts.idI;
if(malleability_red_contains_strat(red_strategies, MALL_RED_IBARRIER, NULL)) {
sum++;
}
if (*requests != NULL && sum <= *request_qty) return; // Expected amount of requests
if (*requests == NULL) {
*requests = (MPI_Request *) malloc(sum * sizeof(MPI_Request));
} else { // Array exists, but is too small
aux = (MPI_Request *) realloc(*requests, sum * sizeof(MPI_Request));
*requests = aux;
}
if (*requests == NULL) {
fprintf(stderr, "Fatal error - It was not possible to allocate/reallocate memory for the MPI_Requests before the redistribution\n");
MPI_Abort(MPI_COMM_WORLD, 1);
}
for(i=0; i < sum; i++) {
(*requests)[i] = MPI_REQUEST_NULL;
}
*request_qty = sum;
}
/*
* Special case to perform a manual copy of data when a process has to send data to itself. Only used
* when the MPI communication is not able to hand this situation. An example is when using point to point
* communications and the process has to perform a Send and Recv to itself
* - send (IN): Array with the data to send. This value can not be NULL.
* - recv (OUT): Array where data will be written. This value can not be NULL.
* - myId (IN): Rank of the MPI process in the local communicator. For the parents is not the rank obtained from "comm".
* - s_counts (IN): Struct where is indicated how many elements sends this process to processes in the new group.
* - r_counts (IN): Struct where is indicated how many elements receives this process from other processes in the previous group.
*/
void perform_manual_communication(char *send, char *recv, int myId, struct Counts s_counts, struct Counts r_counts) {
int i;
for(i=0; i<s_counts.counts[myId];i++) {
recv[i+r_counts.displs[myId]] = send[i+s_counts.displs[myId]];
}
}
/*
* Función para obtener si entre las estrategias elegidas, se utiliza
* la estrategia pasada como segundo argumento.
*
* Devuelve en "result" 1(Verdadero) si utiliza la estrategia, 0(Falso) en caso
* contrario.
*/
int malleability_red_contains_strat(int comm_strategies, int strategy, int *result) {
int value = comm_strategies % strategy ? 0 : 1;
if(result != NULL) *result = value;
return value;
}
/*
* Función para anyadir una estrategia a un conjunto.
*
* Devuelve en "result" 1(Verdadero) si se ha anyadido, 0(Falso) en caso
* contrario.
*/
int malleability_red_add_strat(int *comm_strategies, int strategy) {
if(malleability_red_contains_strat(*comm_strategies, strategy, NULL)) return 1;
*comm_strategies = *comm_strategies * strategy;
return 1;
}
Codes/malleability/CommDist.h
View file @ f1511cb4
......@@ -16,13 +16,18 @@
//#define MAL_USE_POINT 2
//#define MAL_USE_THREAD 3
int send_sync(char *array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_child);
void recv_sync(char **array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_parents);
int sync_communication(char *send, char **recv, int qty, int myId, int numP, int numO, int is_children_group, int comm_type, MPI_Comm comm);
//int async_communication(char *send, char **recv, int qty, int myId, int numP, int numO, int is_children_group, int red_method, int red_strategies, MPI_Comm comm, MPI_Request **requests, size_t *request_qty);
int send_async(char *array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_child, MPI_Request **comm_req, int parents_wait);
void recv_async(char **array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_parents, int parents_wait);
int async_communication_start(char *send, char **recv, int qty, int myId, int numP, int numO, int is_children_group, int red_method, int red_strategies, MPI_Comm comm, MPI_Request **requests, size_t *request_qty, MPI_Win *win);
int async_communication_check(int myId, int is_children_group, int red_strategies, MPI_Comm comm, MPI_Request *requests, size_t request_qty);
void async_communication_wait(int red_strategies, MPI_Comm comm, MPI_Request *requests, size_t request_qty);
void async_communication_end(int red_method, int red_strategies, MPI_Request *requests, size_t request_qty, MPI_Win *win);
//int send_async(char *array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_child, MPI_Request **comm_req, int red_method, int red_strategies);
//void recv_async(char **array, int qty, int myId, int numP, MPI_Comm intercomm, int numP_parents, int red_method, int red_strategies);
void malloc_comm_array(char **array, int qty, int myId, int numP);
int malleability_red_contains_strat(int comm_strategies, int strategy, int *result);
int malleability_red_add_strat(int *comm_strategies, int strategy);
#endif
Codes/malleability/distribution_methods/block_distribution.c
View file @ f1511cb4
......@@ -3,7 +3,7 @@
#include <mpi.h>
#include "block_distribution.h"
void set_interblock_counts(int id, int numP, struct Dist_data data_dist, int *sendcounts);
void set_interblock_counts(int id, int numP, struct Dist_data data_dist, int offset_ids, int *sendcounts);
void get_util_ids(struct Dist_data dist_data, int numP_other, int **idS);
/*
......@@ -13,22 +13,43 @@ void get_util_ids(struct Dist_data dist_data, int numP_other, int **idS);
*
* The struct should be freed with freeCounts
*/
void prepare_comm_alltoall(int myId, int numP, int numP_other, int n, struct Counts *counts) {
int i, *idS;
struct Dist_data dist_data;
void prepare_comm_alltoall(int myId, int numP, int numP_other, int n, int offset_ids, struct Counts *counts) {
int i, *idS, first_id = 0;
struct Dist_data dist_data, dist_target;
if(counts == NULL) {
fprintf(stderr, "Counts is NULL for rank %d/%d ", myId, numP);
MPI_Abort(MPI_COMM_WORLD, -3);
}
get_block_dist(n, myId, numP, &dist_data);
mallocCounts(counts, numP_other);
get_util_ids(dist_data, numP_other, &idS);
if(idS[0] == 0) {
set_interblock_counts(0, numP_other, dist_data, counts->counts);
idS[0]++;
counts->idI = idS[0] + offset_ids;
counts->idE = idS[1] + offset_ids;
get_block_dist(n, idS[0], numP_other, &dist_target); // RMA Specific operation -- uses idS[0], not idI
counts->first_target_displs = dist_data.ini - dist_target.ini; // RMA Specific operation
if(idS[0] == 0) { // Uses idS[0], not idI
set_interblock_counts(counts->idI, numP_other, dist_data, offset_ids, counts->counts);
first_id++;
}
for(i=idS[0]; i<idS[1]; i++) {
set_interblock_counts(i, numP_other, dist_data, counts->counts);
for(i=counts->idI + first_id; i<counts->idE; i++) {
set_interblock_counts(i, numP_other, dist_data, offset_ids, counts->counts);
counts->displs[i] = counts->displs[i-1] + counts->counts[i-1];
}
free(idS);
for(i=0; i<numP_other; i++) {
if(counts->counts[i] < 0) {
fprintf(stderr, "Counts value [i=%d/%d] is negative for rank %d/%d ", i, numP_other, myId, numP);
MPI_Abort(MPI_COMM_WORLD, -3);
}
if(counts->displs[i] < 0) {
fprintf(stderr, "Displs value [i=%d/%d] is negative for rank %d/%d ", i, numP_other, myId, numP);
MPI_Abort(MPI_COMM_WORLD, -3);
}
}
}
/*
......@@ -83,12 +104,8 @@ void get_block_dist(int qty, int id, int numP, struct Dist_data *dist_data) {
dist_data->fin = (id+1) * dist_data->tamBl + rem;
}
if(dist_data->fin > qty) {
dist_data->fin = qty;
}
if(dist_data->ini > dist_data->fin) {
dist_data->ini = dist_data->fin;
}
if(dist_data->fin > qty) { dist_data->fin = qty; }
if(dist_data->ini > dist_data->fin) { dist_data->ini = dist_data->fin; }
dist_data->tamBl = dist_data->fin - dist_data->ini;
}
......@@ -98,11 +115,11 @@ void get_block_dist(int qty, int id, int numP, struct Dist_data *dist_data) {
* Obtiene para el Id de un proceso dado, cuantos elementos
* enviara o recibira desde el proceso indicado en Dist_data.
*/
void set_interblock_counts(int id, int numP, struct Dist_data data_dist, int *sendcounts) {
void set_interblock_counts(int id, int numP, struct Dist_data data_dist, int offset_ids, int *sendcounts) {
struct Dist_data other;
int biggest_ini, smallest_end;
get_block_dist(data_dist.qty, id, numP, &other);
get_block_dist(data_dist.qty, id - offset_ids, numP, &other);
// Si el rango de valores no coincide, se pasa al siguiente proceso
if(data_dist.ini >= other.fin || data_dist.fin <= other.ini) {
......@@ -110,18 +127,10 @@ void set_interblock_counts(int id, int numP, struct Dist_data data_dist, int *se
}
// Obtiene el proceso con mayor ini entre los dos procesos
if(data_dist.ini > other.ini) {
biggest_ini = data_dist.ini;
} else {
biggest_ini = other.ini;
}
biggest_ini = (data_dist.ini > other.ini) ? data_dist.ini : other.ini;
// Obtiene el proceso con menor fin entre los dos procesos
if(data_dist.fin < other.fin) {
smallest_end = data_dist.fin;
} else {
smallest_end = other.fin;
}
smallest_end = (data_dist.fin < other.fin) ? data_dist.fin : other.fin;
sendcounts[id] = smallest_end - biggest_ini; // Numero de elementos a enviar/recibir del proceso Id
}
......@@ -184,18 +193,19 @@ void get_util_ids(struct Dist_data dist_data, int numP_other, int **idS) {
* El vector displs indica los desplazamientos necesarios para cada comunicacion
* con el proceso "i" del otro grupo.
*
* El vector zero_arr se utiliza cuando se quiere indicar un vector incializado
* a 0 en todos sus elementos. Sirve para indicar que no hay comunicacion.
*/
void mallocCounts(struct Counts *counts, size_t numP) {
counts->counts = calloc(numP, sizeof(int));
if(counts->counts == NULL) { MPI_Abort(MPI_COMM_WORLD, -2);}
counts->displs = calloc(numP, sizeof(int));
if(counts->displs == NULL) { MPI_Abort(MPI_COMM_WORLD, -2);}
counts->zero_arr = calloc(numP, sizeof(int));
if(counts->zero_arr == NULL) { MPI_Abort(MPI_COMM_WORLD, -2);}
counts->len = numP;
counts->idI = -1;
counts->idE = -1;
counts->first_target_displs = -1;
}
......@@ -206,12 +216,18 @@ void mallocCounts(struct Counts *counts, size_t numP) {
* de forma dinamica.
*/
void freeCounts(struct Counts *counts) {
if(counts == NULL) {
return;
}
if(counts->counts != NULL) {
free(counts->counts);
free(counts->displs);
free(counts->zero_arr);
counts->counts = NULL;
}
if(counts->displs != NULL) {
free(counts->displs);
counts->displs = NULL;
counts->zero_arr = NULL;
}
}
/*
......
Codes/malleability/distribution_methods/block_distribution.h
View file @ f1511cb4
......@@ -18,12 +18,13 @@ struct Dist_data {
};
struct Counts {
int len, idI, idE;
int first_target_displs; // RMA. Indicates displacement for first target when performing a Get.
int *counts;
int *displs;
int *zero_arr;
};
void prepare_comm_alltoall(int myId, int numP, int numP_other, int n, struct Counts *counts);
void prepare_comm_alltoall(int myId, int numP, int numP_other, int n, int offset_ids, struct Counts *counts);
void prepare_comm_allgatherv(int numP, int n, struct Counts *counts);
void get_block_dist(int qty, int id, int numP, struct Dist_data *dist_data);
......
Codes/malleability/malleabilityManager.c
View file @ f1511cb4
......@@ -30,15 +30,14 @@ int thread_check();
void* thread_async_work();
void print_comms_state();
void malleability_comms_update(MPI_Comm comm);
typedef struct {
int spawn_method;
int spawn_dist;
int spawn_strategies;
//int spawn_is_single;
//int spawn_threaded;
int comm_type;
int comm_threaded;
int red_method;
int red_strategies;
int grp;
configuration *config_file;
......@@ -51,6 +50,7 @@ typedef struct { //FIXME numC_spawned no se esta usando
MPI_Comm comm, thread_comm;
MPI_Comm intercomm;
MPI_Comm user_comm;
int dup_user_comm;
char *name_exec, *nodelist;
int num_cpus, num_nodes, nodelist_len;
......@@ -86,6 +86,7 @@ int init_malleability(int myId, int numP, int root, MPI_Comm comm, char *name_ex
rep_a_data = (malleability_data_t *) malloc(sizeof(malleability_data_t));
dist_a_data = (malleability_data_t *) malloc(sizeof(malleability_data_t));
mall->dup_user_comm = 0;
MPI_Comm_dup(comm, &dup_comm);
MPI_Comm_dup(comm, &thread_comm);
MPI_Comm_set_name(dup_comm, "MPI_COMM_MALL");
......@@ -181,6 +182,7 @@ int malleability_checkpoint() {
break;
case MALL_NOT_STARTED:
// Comprobar si se tiene que realizar un redimensionado
mall_conf->results->malleability_time[mall_conf->grp] = MPI_Wtime();
//if(CHECK_RMS()) {return MALL_DENIED;}
state = spawn_step();
......@@ -208,7 +210,7 @@ int malleability_checkpoint() {
break;
case MALL_DIST_PENDING:
if(mall_conf->comm_type == MAL_USE_THREAD) {
if(malleability_red_contains_strat(mall_conf->red_strategies, MALL_RED_THREAD, NULL)) {
state = thread_check();
} else {
state = check_redistribution();
......@@ -225,6 +227,7 @@ int malleability_checkpoint() {
if(!malleability_spawn_contains_strat(mall_conf->spawn_strategies, MALL_SPAWN_PTHREAD, NULL)) {
mall_conf->results->spawn_time[mall_conf->grp] = MPI_Wtime() - mall_conf->results->spawn_start;
malleability_checkpoint();
}
break;
......@@ -234,6 +237,7 @@ int malleability_checkpoint() {
break;
case MALL_DIST_COMPLETED: //TODO No es esto muy feo?
mall_conf->results->malleability_end = MPI_Wtime();
state = MALL_COMPLETED;
break;
}
......@@ -263,12 +267,17 @@ void get_benchmark_results(results_data **results) {
}
//-------------------------------------------------------------------------------------------------------------
void set_malleability_configuration(int spawn_method, int spawn_strategies, int spawn_dist, int comm_type, int comm_threaded) {
void set_malleability_configuration(int spawn_method, int spawn_strategies, int spawn_dist, int red_method, int red_strategies) {
mall_conf->spawn_method = spawn_method;
mall_conf->spawn_strategies = spawn_strategies;
mall_conf->spawn_dist = spawn_dist;
mall_conf->comm_type = comm_type;
mall_conf->comm_threaded = comm_threaded;
mall_conf->red_method = red_method;
mall_conf->red_strategies = red_strategies;
if(!malleability_red_contains_strat(mall_conf->red_strategies, MALL_RED_IBARRIER, NULL) &&
(mall_conf->red_method == MALL_RED_RMA_LOCK || mall_conf->red_method == MALL_RED_RMA_LOCKALL)) {
malleability_red_add_strat(&(mall_conf->red_strategies), MALL_RED_IBARRIER);
}
}
/*
......@@ -294,6 +303,12 @@ void set_children_number(int numC){
* TODO
*/
void get_malleability_user_comm(MPI_Comm *comm) {
if(mall->dup_user_comm) {
if(mall->user_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->user_comm));
MPI_Comm_dup(mall->comm, &(mall->user_comm));
MPI_Comm_set_name(mall->user_comm, "MPI_COMM_MALL_USER");
mall->dup_user_comm = 0;
}
*comm = mall->user_comm;
}
......@@ -304,28 +319,30 @@ void get_malleability_user_comm(MPI_Comm *comm) {
* Los datos variables se tienen que anyadir cuando quieran ser mandados, no antes
*
* Mas informacion en la funcion "add_data".
*
* //FIXME Si es constante se debería ir a asincrono, no sincrono
*/
void malleability_add_data(void *data, size_t total_qty, int type, int is_replicated, int is_constant) {
size_t total_reqs = 0;
if(is_constant) {
if(is_replicated) {
add_data(data, total_qty, type, 0, rep_s_data); //FIXME Numero magico
add_data(data, total_qty, type, total_reqs, rep_s_data);
} else {
add_data(data, total_qty, type, 0, dist_s_data); //FIXME Numero magico
add_data(data, total_qty, type, total_reqs, dist_s_data);
}
} else {
if(is_replicated) {
add_data(data, total_qty, type, 0, rep_a_data); //FIXME Numero magico || Un request?
add_data(data, total_qty, type, total_reqs, rep_a_data); //FIXME total_reqs==0 ???
} else {
size_t total_reqs = 0;
if(mall_conf->comm_type == MAL_USE_NORMAL) {
if(mall_conf->red_method == MALL_RED_BASELINE) {
total_reqs = 1;
} else if(mall_conf->comm_type == MAL_USE_IBARRIER) {
total_reqs = 2;
} else if(mall_conf->comm_type == MAL_USE_POINT) {
} else if(mall_conf->red_method == MALL_RED_POINT || mall_conf->red_method == MALL_RED_RMA_LOCK || mall_conf->red_method == MALL_RED_RMA_LOCKALL) {
total_reqs = mall->numC;
}
if(malleability_red_contains_strat(mall_conf->red_strategies, MALL_RED_IBARRIER, NULL)) {
total_reqs++;
}
add_data(data, total_qty, type, total_reqs, dist_a_data);
}
......@@ -339,29 +356,31 @@ void malleability_add_data(void *data, size_t total_qty, int type, int is_replic
* Los datos variables se tienen que modificar cuando quieran ser mandados, no antes
*
* Mas informacion en la funcion "modify_data".
* //FIXME Si es constante se debería ir a asincrono, no sincrono
*/
void malleability_modify_data(void *data, size_t index, size_t total_qty, int type, int is_replicated, int is_constant) {
size_t total_reqs = 0;
if(is_constant) {
if(is_replicated) {
modify_data(data, index, total_qty, type, 0, rep_s_data); //FIXME Numero magico
modify_data(data, index, total_qty, type, total_reqs, rep_s_data);
} else {
modify_data(data, index, total_qty, type, 0, dist_s_data); //FIXME Numero magico
modify_data(data, index, total_qty, type, total_reqs, dist_s_data);
}
} else {
if(is_replicated) {
modify_data(data, index, total_qty, type, 0, rep_a_data); //FIXME Numero magico || UN request?
modify_data(data, index, total_qty, type, total_reqs, rep_a_data); //FIXME total_reqs==0 ???
} else {
size_t total_reqs = 0;
if(mall_conf->comm_type == MAL_USE_NORMAL) {
if(mall_conf->red_method == MALL_RED_BASELINE) {
total_reqs = 1;
} else if(mall_conf->comm_type == MAL_USE_IBARRIER) {
total_reqs = 2;
} else if(mall_conf->comm_type == MAL_USE_POINT) {
} else if(mall_conf->red_method == MALL_RED_POINT || mall_conf->red_method == MALL_RED_RMA_LOCK || mall_conf->red_method == MALL_RED_RMA_LOCKALL) {
total_reqs = mall->numC;
}
if(malleability_red_contains_strat(mall_conf->red_strategies, MALL_RED_IBARRIER, NULL)) {
total_reqs++;
}
modify_data(data, index, total_qty, type, total_reqs, dist_a_data); //FIXME Numero magico
modify_data(data, index, total_qty, type, total_reqs, dist_a_data);
}
}
}
......@@ -369,6 +388,7 @@ void malleability_modify_data(void *data, size_t index, size_t total_qty, int ty
/*
* Devuelve el numero de entradas para la estructura de descripcion de
* datos elegida.
* //FIXME Si es constante se debería ir a asincrono, no sincrono
*/
void malleability_get_entries(size_t *entries, int is_replicated, int is_constant){
......@@ -393,8 +413,9 @@ void malleability_get_entries(size_t *entries, int is_replicated, int is_constan
* con la funcion "malleability_add_data()".
* Es tarea del usuario saber el tipo de esos datos.
* TODO Refactor a que sea automatico
* //FIXME Si es constante se debería ir a asincrono, no sincrono
*/
void malleability_get_data(void **data, int index, int is_replicated, int is_constant) {
void malleability_get_data(void **data, size_t index, int is_replicated, int is_constant) {
malleability_data_t *data_struct;
if(is_constant) {
......@@ -421,7 +442,6 @@ void malleability_get_data(void **data, int index, int is_replicated, int is_con
//======================================================||
//======================================================||
/*
* Funcion generalizada para enviar datos desde los hijos.
* La asincronizidad se refiere a si el hilo padre e hijo lo hacen
......@@ -429,17 +449,22 @@ void malleability_get_data(void **data, int index, int is_replicated, int is_con
*/
void send_data(int numP_children, malleability_data_t *data_struct, int is_asynchronous) {
size_t i;
char *aux;
char *aux_send, *aux_recv;
if(is_asynchronous) {
for(i=0; i < data_struct->entries; i++) {
aux = (char *) data_struct->arrays[i]; //TODO Comprobar que realmente es un char
send_async(aux, data_struct->qty[i], mall->myId, mall->numP, mall->intercomm, numP_children, data_struct->requests, mall_conf->comm_type);
aux_send = (char *) data_struct->arrays[i]; //TODO Comprobar que realmente es un char
aux_recv = NULL;
async_communication_start(aux_send, &aux_recv, data_struct->qty[i], mall->myId, mall->numP, numP_children, MALLEABILITY_NOT_CHILDREN, mall_conf->red_method, mall_conf->red_strategies,
mall->intercomm, &(data_struct->requests[i]), &(data_struct->request_qty[i]), &(data_struct->windows[i]));
if(aux_recv != NULL) data_struct->arrays[i] = (void *) aux_recv;
}
} else {
for(i=0; i < data_struct->entries; i++) {
aux = (char *) data_struct->arrays[i]; //TODO Comprobar que realmente es un char
send_sync(aux, data_struct->qty[i], mall->myId, mall->numP, mall->intercomm, numP_children);
aux_send = (char *) data_struct->arrays[i]; //TODO Comprobar que realmente es un char
aux_recv = NULL;
sync_communication(aux_send, &aux_recv, data_struct->qty[i], mall->myId, mall->numP, numP_children, MALLEABILITY_NOT_CHILDREN, mall_conf->red_method, mall->intercomm);
if(aux_recv != NULL) data_struct->arrays[i] = (void *) aux_recv;
}
}
}
......@@ -451,18 +476,19 @@ void send_data(int numP_children, malleability_data_t *data_struct, int is_async
*/
void recv_data(int numP_parents, malleability_data_t *data_struct, int is_asynchronous) {
size_t i;
char *aux;
char *aux, aux_s;
if(is_asynchronous) {
for(i=0; i < data_struct->entries; i++) {
aux = (char *) data_struct->arrays[i]; //TODO Comprobar que realmente es un char
recv_async(&aux, data_struct->qty[i], mall->myId, mall->numP, mall->intercomm, numP_parents, mall_conf->comm_type);
async_communication_start(&aux_s, &aux, data_struct->qty[i], mall->myId, mall->numP, numP_parents, MALLEABILITY_CHILDREN, mall_conf->red_method, mall_conf->red_strategies,
mall->intercomm, &(data_struct->requests[i]), &(data_struct->request_qty[i]), &(data_struct->windows[i]));
data_struct->arrays[i] = (void *) aux;
}
} else {
for(i=0; i < data_struct->entries; i++) {
aux = (char *) data_struct->arrays[i]; //TODO Comprobar que realmente es un char
recv_sync(&aux, data_struct->qty[i], mall->myId, mall->numP, mall->intercomm, numP_parents);
sync_communication(&aux_s, &aux, data_struct->qty[i], mall->myId, mall->numP, numP_parents, MALLEABILITY_CHILDREN, mall_conf->red_method, mall->intercomm);
data_struct->arrays[i] = (void *) aux;
}
}
......@@ -473,7 +499,6 @@ void recv_data(int numP_parents, malleability_data_t *data_struct, int is_asynch
//=====================CHILDREN=========================||
//======================================================||
//======================================================||
/*
* Inicializacion de los datos de los hijos.
* En la misma se reciben datos de los padres: La configuracion
......@@ -487,10 +512,15 @@ void Children_init() {
malleability_connect_children(mall->myId, mall->numP, mall->root, mall->comm, &numP_parents, &root_parents, &(mall->intercomm));
MPI_Comm_test_inter(mall->intercomm, &is_intercomm);
// TODO A partir de este punto tener en cuenta si es BASELINE o MERGE
if(!is_intercomm) { // For intracommunicators, these processes will be added
MPI_Comm_rank(mall->intercomm, &(mall->myId));
MPI_Comm_size(mall->intercomm, &(mall->numP));
}
recv_config_file(mall->root, mall->intercomm, &(mall_conf->config_file));
comm_node_data(root_parents, MALLEABILITY_CHILDREN);
MPI_Bcast(&(mall_conf->red_method), 1, MPI_INT, root_parents, mall->intercomm);
MPI_Bcast(&(mall_conf->red_strategies), 1, MPI_INT, root_parents, mall->intercomm);
mall_conf->results = (results_data *) malloc(sizeof(results_data));
init_results_data(mall_conf->results, mall_conf->config_file->n_resizes, mall_conf->config_file->n_stages, RESULTS_INIT_DATA_QTY);
......@@ -498,19 +528,25 @@ void Children_init() {
comm_data_info(rep_a_data, dist_a_data, MALLEABILITY_CHILDREN, mall->myId, root_parents, mall->intercomm);
if(dist_a_data->entries || rep_a_data->entries) { // Recibir datos asincronos
if(mall_conf->comm_type == MAL_USE_NORMAL || mall_conf->comm_type == MAL_USE_IBARRIER || mall_conf->comm_type == MAL_USE_POINT) {
recv_data(numP_parents, dist_a_data, 1);
if(malleability_red_contains_strat(mall_conf->red_strategies, MALL_RED_THREAD, NULL)) {
recv_data(numP_parents, dist_a_data, MALLEABILITY_USE_SYNCHRONOUS);
} else {
recv_data(numP_parents, dist_a_data, MALLEABILITY_USE_ASYNCHRONOUS);
} else if (mall_conf->comm_type == MAL_USE_THREAD) { //TODO Modificar uso para que tenga sentido comm_threaded
recv_data(numP_parents, dist_a_data, 0);
for(i=0; i<dist_a_data->entries; i++) {
async_communication_wait(mall_conf->red_strategies, mall->intercomm, dist_a_data->requests[i], dist_a_data->request_qty[i]);
}
for(i=0; i<dist_a_data->entries; i++) {
async_communication_end(mall_conf->red_method, mall_conf->red_strategies, dist_a_data->requests[i], dist_a_data->request_qty[i], &(dist_a_data->windows[i]));
}
}
mall_conf->results->async_end= MPI_Wtime(); // Obtener timestamp de cuando termina comm asincrona
}
comm_data_info(rep_s_data, dist_s_data, MALLEABILITY_CHILDREN, mall->myId, root_parents, mall->intercomm);
if(dist_s_data->entries || rep_s_data->entries) { // Recibir datos sincronos
recv_data(numP_parents, dist_s_data, 0);
recv_data(numP_parents, dist_s_data, MALLEABILITY_USE_SYNCHRONOUS);
mall_conf->results->sync_end = MPI_Wtime(); // Obtener timestamp de cuando termina comm sincrona
// TODO Crear funcion especifica y anyadir para Asinc
......@@ -525,20 +561,15 @@ void Children_init() {
MPI_Bcast(rep_s_data->arrays[i], rep_s_data->qty[i], datatype, root_parents, mall->intercomm);
}
}
mall_conf->results->malleability_end = MPI_Wtime(); // Obtener timestamp de cuando termina maleabilidad
// Guardar los resultados de esta transmision
comm_results(mall_conf->results, mall->root, mall_conf->config_file->n_resizes, mall->intercomm);
if(!is_intercomm) {
if(mall->thread_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->thread_comm));
if(mall->comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->comm));
if(mall->user_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->user_comm)); //TODO No es peligroso?
MPI_Comm_dup(mall->intercomm, &(mall->thread_comm));
MPI_Comm_dup(mall->intercomm, &(mall->comm));
MPI_Comm_dup(mall->intercomm, &(mall->user_comm));
malleability_comms_update(mall->intercomm);
}
MPI_Comm_disconnect(&(mall->intercomm));
MPI_Comm_disconnect(&(mall->intercomm)); //FIXME Error en OpenMPI + Merge
}
//======================================================||
......@@ -597,11 +628,14 @@ int start_redistribution() {
send_config_file(mall_conf->config_file, rootBcast, mall->intercomm);
comm_node_data(rootBcast, MALLEABILITY_NOT_CHILDREN);
MPI_Bcast(&(mall_conf->red_method), 1, MPI_INT, rootBcast, mall->intercomm);
MPI_Bcast(&(mall_conf->red_strategies), 1, MPI_INT, rootBcast, mall->intercomm);
comm_data_info(rep_a_data, dist_a_data, MALLEABILITY_NOT_CHILDREN, mall->myId, mall->root, mall->intercomm);
if(dist_a_data->entries || rep_a_data->entries) { // Enviar datos asincronos
//FIXME No se envian los datos replicados (rep_a_data)
mall_conf->results->async_time[mall_conf->grp] = MPI_Wtime();
if(mall_conf->comm_type == MAL_USE_THREAD) {
if(malleability_red_contains_strat(mall_conf->red_strategies, MALL_RED_THREAD, NULL)) {
return thread_creation();
} else {
send_data(mall->numC, dist_a_data, MALLEABILITY_USE_ASYNCHRONOUS);
......@@ -613,7 +647,6 @@ int start_redistribution() {
/*
* @deprecated
* Comprueba si la redistribucion asincrona ha terminado.
* Si no ha terminado la funcion termina indicandolo, en caso contrario,
* se continua con la comunicacion sincrona, el envio de resultados y
......@@ -625,38 +658,34 @@ int start_redistribution() {
* terminada cuando los padres terminan de enviar.
* Si se utiliza el modo "MAL_USE_IBARRIER", se considera terminada cuando
* los hijos han terminado de recibir.
* //FIXME Modificar para que se tenga en cuenta rep_a_data
*/
int check_redistribution() {
int completed, all_completed, test_err;
int is_intercomm, completed, local_completed, all_completed;
size_t i, req_qty;
MPI_Request *req_completed;
//dist_a_data->requests[0][X] //FIXME Numero magico 0 -- Modificar para que sea un for?
if (mall_conf->comm_type == MAL_USE_POINT) {
test_err = MPI_Testall(mall->numC, dist_a_data->requests[0], &completed, MPI_STATUSES_IGNORE);
} else {
if(mall_conf->comm_type == MAL_USE_NORMAL) {
req_completed = &(dist_a_data->requests[0][0]);
} else if (mall_conf->comm_type == MAL_USE_IBARRIER) {
req_completed = &(dist_a_data->requests[0][1]);
}
test_err = MPI_Test(req_completed, &completed, MPI_STATUS_IGNORE);
}
MPI_Win window;
local_completed = 1;
if (test_err != MPI_SUCCESS && test_err != MPI_ERR_PENDING) {
printf("P%d aborting -- Test Async\n", mall->myId);
MPI_Abort(MPI_COMM_WORLD, test_err);
for(i=0; i<dist_a_data->entries; i++) {
req_completed = dist_a_data->requests[i];
req_qty = dist_a_data->request_qty[i];
completed = async_communication_check(mall->myId, MALLEABILITY_NOT_CHILDREN, mall_conf->red_strategies, mall->intercomm, req_completed, req_qty);
local_completed = local_completed && completed;
}
MPI_Allreduce(&completed, &all_completed, 1, MPI_INT, MPI_MIN, mall->comm);
MPI_Allreduce(&local_completed, &all_completed, 1, MPI_INT, MPI_MIN, mall->comm);
if(!all_completed) return MALL_DIST_PENDING; // Continue only if asynchronous send has ended
if(mall_conf->comm_type == MAL_USE_IBARRIER) {
MPI_Wait(&(dist_a_data->requests[0][0]), MPI_STATUS_IGNORE); // Indicar como completado el envio asincrono
//Para la desconexión de ambos grupos de procesos es necesario indicar a MPI que esta comm
//ha terminado, aunque solo se pueda llegar a este punto cuando ha terminado
for(i=0; i<dist_a_data->entries; i++) {
req_completed = dist_a_data->requests[i];
req_qty = dist_a_data->request_qty[i];
window = dist_a_data->windows[i];
async_communication_end(mall_conf->red_method, mall_conf->red_strategies, req_completed, req_qty, &window);
}
MPI_Comm_test_inter(mall->intercomm, &is_intercomm);
if(!is_intercomm) mall_conf->results->async_end = MPI_Wtime(); // Merge method only
return end_redistribution();
}
......@@ -673,14 +702,7 @@ int end_redistribution() {
size_t i;
int is_intercomm, rootBcast, local_state;
is_intercomm = 0;
if(mall->intercomm != MPI_COMM_NULL) {
MPI_Comm_test_inter(mall->intercomm, &is_intercomm);
} else {
// Si no tiene comunicador creado, se debe a que se ha pospuesto el Spawn
// y se trata del spawn Merge Shrink
mall->intercomm = mall->comm;
}
if(is_intercomm) {
rootBcast = mall->myId == mall->root ? MPI_ROOT : MPI_PROC_NULL;
} else {
......@@ -691,6 +713,7 @@ int end_redistribution() {
if(dist_s_data->entries || rep_s_data->entries) { // Enviar datos sincronos
mall_conf->results->sync_time[mall_conf->grp] = MPI_Wtime();
send_data(mall->numC, dist_s_data, MALLEABILITY_USE_SYNCHRONOUS);
if(!is_intercomm) mall_conf->results->sync_end = MPI_Wtime(); // Merge method only
// TODO Crear funcion especifica y anyadir para Asinc
// TODO Tener en cuenta el tipo
......@@ -710,25 +733,14 @@ int end_redistribution() {
local_state = MALL_DIST_COMPLETED;
if(!is_intercomm) { // Merge Spawn
if(mall->numP < mall->numC) { // Expand
if(mall->thread_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->thread_comm));
if(mall->comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->comm));
if(mall->user_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->user_comm)); //TODO No es peligroso?
MPI_Comm_dup(mall->intercomm, &(mall->thread_comm));
MPI_Comm_dup(mall->intercomm, &(mall->comm));
MPI_Comm_dup(mall->intercomm, &(mall->user_comm));
MPI_Comm_set_name(mall->thread_comm, "MPI_COMM_MALL_THREAD");
MPI_Comm_set_name(mall->comm, "MPI_COMM_MALL");
MPI_Comm_set_name(mall->user_comm, "MPI_COMM_MALL_USER");
malleability_comms_update(mall->intercomm);
} else { // Shrink || Merge Shrink requiere de mas tareas
local_state = MALL_SPAWN_ADAPT_PENDING;
}
}
if(mall->intercomm != MPI_COMM_NULL && mall->intercomm != MPI_COMM_WORLD) {
MPI_Comm_disconnect(&(mall->intercomm));
MPI_Comm_disconnect(&(mall->intercomm)); //FIXME Error en OpenMPI + Merge
}
return local_state;
......@@ -742,21 +754,19 @@ int end_redistribution() {
int shrink_redistribution() {
double time_extra = MPI_Wtime();
//TODO REFACTOR -- Que solo la llamada de collect iters este fuera de los hilos
zombies_collect_suspended(mall->comm, mall->myId, mall->numP, mall->numC, mall->root, (void *) mall_conf->results, mall_conf->config_file->n_stages);
//TODO Create new state before collecting zombies. Processes can perform tasks before that. Then call again Malleability to commit the change
zombies_collect_suspended(mall->user_comm, mall->myId, mall->numP, mall->numC, mall->root, (void *) mall_conf->results, mall_conf->config_file->n_stages);
if(mall->myId < mall->numC) {
if(mall->thread_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->thread_comm));
if(mall->thread_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->thread_comm)); //FIXME Modificar a que se pida pro el usuario el cambio y se llama a comms_update
if(mall->comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->comm));
if(mall->user_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->user_comm)); //TODO No es peligroso?
mall->dup_user_comm = 1;
MPI_Comm_dup(mall->intercomm, &(mall->thread_comm));
MPI_Comm_dup(mall->intercomm, &(mall->comm));
MPI_Comm_dup(mall->intercomm, &(mall->user_comm));
MPI_Comm_set_name(mall->thread_comm, "MPI_COMM_MALL_THREAD");
MPI_Comm_set_name(mall->comm, "MPI_COMM_MALL");
MPI_Comm_set_name(mall->user_comm, "MPI_COMM_MALL_USER");
MPI_Comm_free(&(mall->intercomm));
......@@ -821,16 +831,19 @@ void def_nodeinfo_type(MPI_Datatype *node_type) {
//======================================================||
//======================================================||
int comm_state; //FIXME Usar un handler
/*
* Crea una hebra para ejecutar una comunicación en segundo plano.
*/
int thread_creation() {
comm_state = MALL_DIST_PENDING;
if(pthread_create(&(mall->async_thread), NULL, thread_async_work, NULL)) {
printf("Error al crear el hilo\n");
MPI_Abort(MPI_COMM_WORLD, -1);
return -1;
}
return MALL_DIST_PENDING;
return comm_state;
}
/*
......@@ -840,10 +853,10 @@ int thread_creation() {
* El estado de la comunicación es devuelto al finalizar la función.
*/
int thread_check() {
int all_completed = 0;
int all_completed = 0, is_intercomm;
// Comprueba que todos los hilos han terminado la distribucion (Mismo valor en commAsync)
MPI_Allreduce(&state, &all_completed, 1, MPI_INT, MPI_MAX, mall->comm);
MPI_Allreduce(&comm_state, &all_completed, 1, MPI_INT, MPI_MAX, mall->comm);
if(all_completed != MALL_DIST_COMPLETED) return MALL_DIST_PENDING; // Continue only if asynchronous send has ended
//FIXME No se tiene en cuenta el estado MALL_APP_ENDED
......@@ -852,6 +865,8 @@ int thread_check() {
MPI_Abort(MPI_COMM_WORLD, -1);
return -2;
}
MPI_Comm_test_inter(mall->intercomm, &is_intercomm);
if(!is_intercomm) mall_conf->results->async_end = MPI_Wtime(); // Merge method only
return end_redistribution();
}
......@@ -866,7 +881,7 @@ int thread_check() {
*/
void* thread_async_work() {
send_data(mall->numC, dist_a_data, MALLEABILITY_USE_SYNCHRONOUS);
state = MALL_DIST_COMPLETED;
comm_state = MALL_DIST_COMPLETED;
pthread_exit(NULL);
}
......@@ -889,3 +904,17 @@ void print_comms_state() {
}
free(test);
}
void malleability_comms_update(MPI_Comm comm) {
if(mall->thread_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->thread_comm));
if(mall->comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->comm));
if(mall->user_comm != MPI_COMM_WORLD) MPI_Comm_free(&(mall->user_comm)); //TODO No es peligroso?
MPI_Comm_dup(comm, &(mall->thread_comm));
MPI_Comm_dup(comm, &(mall->comm));
MPI_Comm_dup(comm, &(mall->user_comm));
MPI_Comm_set_name(mall->thread_comm, "MPI_COMM_MALL_THREAD");
MPI_Comm_set_name(mall->comm, "MPI_COMM_MALL");
MPI_Comm_set_name(mall->user_comm, "MPI_COMM_MALL_USER");
}
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment