Merge branch 'dev' of https://gitlab.com/iker_martin/malleability_benchmark into dev

Analysis/.ipynb_checkpoints/MallTimes-checkpoint.py

+import sys
+import glob
+import numpy as np
+import pandas as pd
+
+def getData(lineS, outData, tp, hasIter = False):
+  for data in lineS:
+    k_v = data.split('=')
+    if k_v[0] == "time":
+      time = float(k_v[1])
+    elif k_v[0] == "iters" and hasIter:
+      iters = int(k_v[1])
+
+  outData[tp] = time
+  if hasIter:
+    outData[tp+1] = iters
+
+#-----------------------------------------------
+def record(f, observation, line):
+  # Record first line - General info
+  lineS = line.split()
+  for j in range(1,7):
+    observation[j] = int(lineS[j].split('=')[1])
+
+  # Record procces number
+  line = next(f)
+  lineS = line.split()
+  j = 7
+  for key_values in lineS:
+    k_v = key_values.split('=')
+    observation[j] = int(k_v[1])
+    j+=1
+
+  # Record data
+  j = 9
+  for j in range(9, 13):
+    line = next(f)
+    lineS = line.split()  
+    getData(lineS, observation, j)
+
+  line = next(f)
+  lineS = line.split()  
+  #if observation[0] == "A":
+  getData(lineS, observation, 13, True)
+  #else:
+   # getData(lineS, observation, 13)
+
+
+#-----------------------------------------------
+def read_file(f, dataA, dataB, it):
+  recording = False
+  resizes = 0
+  timer = 0
+  previousNP = 0
+
+  for line in f: 
+    lineS = line.split()
+
+    if len(lineS) > 0:
+      if lineS[0] == "Config": # CONFIG LINE
+        recording = True
+        it += 1
+        dataA.append([None]*8)
+        dataB.append([None]*11)
+        resizes = int(lineS[2].split('=')[1].split(',')[0])
+        matrix = int(lineS[3].split('=')[1].split(',')[0])
+        sdr = int(lineS[4].split('=')[1].split(',')[0])
+        adr = int(lineS[5].split('=')[1].split(',')[0]) #TODO Que lo tome como porcentaje
+        time = float(lineS[7].split('=')[1])
+
+        dataB[it][5] = matrix
+        dataB[it][0] = sdr
+        dataB[it][1] = adr 
+        dataB[it][6] = time
+        dataB[it][4] = ""
+
+        dataA[it][4] = matrix
+        dataA[it][0] = sdr
+        dataA[it][1] = adr 
+        dataA[it][5] = time
+        dataA[it][3] = ""
+
+      elif recording and resizes != 0: # RESIZE LINE
+        iters = int(lineS[2].split('=')[1].split(',')[0])
+        npr = int(lineS[3].split('=')[1].split(',')[0])
+        dist = lineS[5].split('=')[1]
+
+        dataB[it][7] = iters
+        dataA[it][6] = iters
+        resizes = resizes - 1
+        if resizes == 0:
+          dataB[it][3] = npr
+          dataB[it][4] += dist
+
+          dataA[it][3] += dist
+          dataA[it][2] = str(previousNP) + "," + str(npr)
+          timer = 3
+        else:
+          dataB[it][2] = npr
+          dataB[it][4] += dist + ","
+
+          dataA[it][3] += dist + ","
+          previousNP = npr
+
+      else: # SAVE TIMES
+        if timer == 3:
+          dataB[it][8] = float(lineS[1])
+        elif timer == 2:
+          dataB[it][9] = float(lineS[1])
+        elif timer == 1:
+          dataB[it][10] = float(lineS[1])
+        else:
+          dataA[it][7] = float(lineS[1])
+        timer = timer - 1
+          
+  return it
+#columnsA1 = ["N", "%Async", "Groups", "Dist", "Matrix", "Time", "Iters", "TE"] #7
+#columnsB1 = ["N", "%Async", "NP", "NS", "Dist", "Matrix", "Time", "Iters", "TC", "TS", "TA"] #10
+#Config loaded: resizes=2, matrix=1000, sdr=1000000000, adr=0, aib=0, time=2.000000 || grp=1
+#Resize 0: Iters=100, Procs=2, Factors=1.000000, Phy=2
+#Resize 1: Iters=100, Procs=4, Factors=0.500000, Phy=2
+#Tspawn: 0.249393 
+#Tsync: 0.330391 
+#Tasync: 0
+#Tex: 301.428615
+#-----------------------------------------------
+
+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:
+    BaseDir = sys.argv[2]
+
+if len(sys.argv) >= 4:
+  print("Csv name will be: " + sys.argv[3] + ".csv")
+  name = sys.argv[3]
+else:
+  name = "data"
+
+insideDir = "Run"
+lista = glob.glob("./" + BaseDir + insideDir + "*/" + sys.argv[1]+ "*Global.o*")
+print("Number of files found: "+ str(len(lista)));
+
+it = -1
+dataA = []
+dataB = []
+columnsA = ["N", "%Async", "Groups", "Dist", "Matrix", "Time", "Iters", "TE"] #7
+columnsB = ["N", "%Async", "NP", "NS", "Dist", "Matrix", "Time", "Iters", "TC", "TS", "TA"] #10
+
+for elem in lista:
+  f = open(elem, "r")
+  it = read_file(f, dataA, dataB, it)
+  f.close()
+
+#print(data)
+dfA = pd.DataFrame(dataA, columns=columnsA)
+dfA.to_csv(name + '_G.csv')
+
+dfB = pd.DataFrame(dataB, columns=columnsB)
+dfB.to_csv(name + '_M.csv')

Analysis/.ipynb_checkpoints/iterTimes-checkpoint.py

+import sys
+import glob
+import numpy as np
+import pandas as pd
+
+#-----------------------------------------------
+def read_file(f, data, it):
+  matrix = 0
+  sdr = 0
+  adr = 0
+  time = 0
+  recording = False
+  it_line = 0
+  aux_it = 0
+  iters = 0
+  np = 0
+  np_par = 0
+  ns = 0
+
+  for line in f: 
+    lineS = line.split()
+
+    if len(lineS) > 1:
+        if recording:
+          aux_it = 0
+          if it_line==0:
+            lineS.pop(0)
+            for observation in lineS:
+              data.append([None]*11)
+              data[it+aux_it][0] = sdr
+              data[it+aux_it][1] = adr 
+              data[it+aux_it][2] = np
+              data[it+aux_it][3] = np_par
+              data[it+aux_it][4] = ns
+              data[it+aux_it][5] = matrix
+              data[it+aux_it][6] = time
+              data[it+aux_it][7] = iters
+              data[it+aux_it][8] = float(observation)
+              aux_it+=1
+            it_line = it_line + 1
+
+          elif it_line==1:
+            lineS.pop(0)
+            for observation in lineS:
+              data[it+aux_it][9] = float(observation)           
+              aux_it+=1
+            it_line = it_line + 1
+          else:
+            lineS.pop(0)
+            for observation in lineS:
+              data[it+aux_it][10] = float(observation)
+              aux_it+=1
+            it = it + aux_it
+            recording = False
+            it_line = 0
+
+#TODO Que tome adr como porcentaje
+        if lineS[0] == "Config:":
+          matrix = int(lineS[1].split('=')[1].split(',')[0])
+          sdr = int(lineS[2].split('=')[1].split(',')[0])
+          adr = int(lineS[3].split('=')[1].split(',')[0]) 
+          time = float(lineS[5].split('=')[1])
+        elif lineS[0] == "Config":
+          recording = True
+          iters = int(lineS[2].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 it
+#-----------------------------------------------
+#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:
+    BaseDir = sys.argv[2]
+
+if len(sys.argv) >= 4:
+  print("Csv name will be: " + sys.argv[3] + ".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)));
+
+it = 0
+data = []  #0   #1        #2    #3       #4    #5        #6      #7       #8    #9    #10
+columns = ["N", "%Async", "NP", "N_par", "NS", "Matrix", "Time", "Iters", "Ti", "Tt", "To"] #11
+
+for elem in lista:
+  f = open(elem, "r")
+  it = read_file(f, data, it)
+  f.close()
+
+#print(data)
+df = pd.DataFrame(data, columns=columns)
+df.to_csv(name + '.csv')

Analysis/MallTimes.py

+import sys
+import glob
+import numpy as np
+import pandas as pd
+
+def getData(lineS, outData, tp, hasIter = False):
+  for data in lineS:
+    k_v = data.split('=')
+    if k_v[0] == "time":
+      time = float(k_v[1])
+    elif k_v[0] == "iters" and hasIter:
+      iters = int(k_v[1])
+
+  outData[tp] = time
+  if hasIter:
+    outData[tp+1] = iters
+
+#-----------------------------------------------
+def record(f, observation, line):
+  # Record first line - General info
+  lineS = line.split()
+  for j in range(1,7):
+    observation[j] = int(lineS[j].split('=')[1])
+
+  # Record procces number
+  line = next(f)
+  lineS = line.split()
+  j = 7
+  for key_values in lineS:
+    k_v = key_values.split('=')
+    observation[j] = int(k_v[1])
+    j+=1
+
+  # Record data
+  j = 9
+  for j in range(9, 13):
+    line = next(f)
+    lineS = line.split()  
+    getData(lineS, observation, j)
+
+  line = next(f)
+  lineS = line.split()  
+  #if observation[0] == "A":
+  getData(lineS, observation, 13, True)
+  #else:
+   # getData(lineS, observation, 13)
+
+
+#-----------------------------------------------
+def read_file(f, dataA, dataB, it):
+  recording = False
+  resizes = 0
+  timer = 0
+  previousNP = 0
+
+  for line in f: 
+    lineS = line.split()
+
+    if len(lineS) > 0:
+      if lineS[0] == "Config": # CONFIG LINE
+        recording = True
+        it += 1
+        dataA.append([None]*9)
+        dataB.append([None]*12)
+        resizes = int(lineS[2].split('=')[1].split(',')[0])
+        compute_tam = int(lineS[3].split('=')[1].split(',')[0])
+        comm_tam = int(lineS[4].split('=')[1].split(',')[0])
+        sdr = int(lineS[5].split('=')[1].split(',')[0])
+        adr = int(lineS[6].split('=')[1].split(',')[0]) #TODO Que lo tome como porcentaje
+        # TODO Que obtenga Aib
+        time = float(lineS[8].split('=')[1])
+
+        dataB[it][0] = sdr
+        dataB[it][1] = adr 
+        dataB[it][4] = "" 
+        dataB[it][5] = compute_tam
+        dataB[it][6] = comm_tam
+        dataB[it][7] = time
+        dataB[it][8] = "" 
+
+        dataA[it][0] = sdr
+        dataA[it][1] = adr 
+        dataA[it][3] = ""
+        dataA[it][4] = compute_tam
+        dataA[it][5] = comm_tam
+        dataA[it][6] = time
+        dataA[it][7] = ""
+
+      elif recording and resizes != 0: # RESIZE LINE
+        iters = int(lineS[2].split('=')[1].split(',')[0])
+        npr = int(lineS[3].split('=')[1].split(',')[0])
+        dist = lineS[5].split('=')[1]
+
+        resizes = resizes - 1
+        if resizes == 0:
+          dataB[it][3] = npr
+          dataB[it][4] += dist
+          dataB[it][8] += str(iters)
+
+          dataA[it][2] = str(previousNP) + "," + str(npr)
+          dataA[it][3] += dist
+          dataA[it][7] += str(iters)
+          timer = 3
+        else:
+          dataB[it][2] = npr
+          dataB[it][4] += dist + ","
+          dataB[it][8] += str(iters) + ","
+
+          dataA[it][3] += dist + ","
+          dataA[it][7] += str(iters) + ","
+          previousNP = npr
+
+      else: # SAVE TIMES
+        if timer == 3:
+          dataB[it][9] = float(lineS[1])
+        elif timer == 2:
+          dataB[it][10] = float(lineS[1])
+        elif timer == 1:
+          dataB[it][11] = float(lineS[1])
+        else:
+          dataA[it][8] = float(lineS[1])
+        timer = timer - 1
+          
+  return it
+#columnsA1 = ["N", "%Async", "Groups", "Dist", "Matrix", "CommTam", "Time", "Iters", "TE"] #8
+#columnsB1 = ["N", "%Async", "NP", "NS", "Dist", "Matrix", "CommTam", "Time", "Iters", "TC", "TS", "TA"] #11
+#Config loaded: resizes=2, matrix=1000, sdr=1000000000, adr=0, aib=0, time=2.000000 || grp=1
+#Resize 0: Iters=100, Procs=2, Factors=1.000000, Phy=2
+#Resize 1: Iters=100, Procs=4, Factors=0.500000, Phy=2
+#Tspawn: 0.249393 
+#Tsync: 0.330391 
+#Tasync: 0
+#Tex: 301.428615
+#-----------------------------------------------
+
+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:
+    BaseDir = sys.argv[2]
+
+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"
+
+insideDir = "Run"
+lista = glob.glob("./" + BaseDir + insideDir + "*/" + sys.argv[1]+ "*Global.o*")
+print("Number of files found: "+ str(len(lista)));
+
+it = -1
+dataA = []
+dataB = []
+columnsA = ["N", "%Async", "Groups", "Dist", "Matrix", "CommTam", "Time", "Iters", "TE"] #8
+columnsB = ["N", "%Async", "NP", "NS", "Dist", "Matrix", "CommTam", "Time", "Iters", "TC", "TS", "TA"] #11
+
+for elem in lista:
+  f = open(elem, "r")
+  it = read_file(f, dataA, dataB, it)
+  f.close()
+
+#print(data)
+dfA = pd.DataFrame(dataA, columns=columnsA)
+dfA.to_csv(name + 'G.csv')
+
+dfB = pd.DataFrame(dataB, columns=columnsB)
+dfB.to_csv(name + 'M.csv')

Analysis/README.md

+Esta carpeta contiene códigos para poder analizar los resultados obtenidos.
+Para utilizar los códigos es necesario Python con los módulos Numpy y Pandas.
+El código analyser.ipynb necesita además de la aplicación JupyterLab.
+
+Los códigos son los siguientes:
+
+ - Malltimes.py: Recoge los tiempos globales de maleabilidad y ejecución de todos los ficheros pasados como argumento y
+     los almacena en dos ficheros CSV para ser utilizados en analyser.ipynb
+ - Itertimes.py: Recoge los tiempos locales de iteraciones de un grupo de procesos de todos los ficheros pasados como
+     argumento y los almacena en un fichero CSV para ser utilizado en analyser.ipynb
+ + Ejemplo de uso de ambos códigos (Esperan los mismos argumentos):
+   python3 Malltimes.py NombreFicheros DirectorioFicheros/ NombreCSV
+     NombreFicheros: La parte común de los ficheros, los códigos buscan solo aquellos nombres que empiecen por esta cadena.
+                     Por defecto, con poner "R" es suficiente.
+     DirectorioFicheros/: Nombre del directorio donde se encuentran todos los resultados. Esta pensado para que busque
+                     en todos las subdirectorios que tenga en el primer nivel, pero no en segundos niveles o más.
+     NombreCSV: Nombre del fichero CSV en el que escribir la recopilación de resultados.
+
+ - analyser.ipynb: Código para ser ejecutado por JupyterNotebook. Dentro del mismo hay que indicar los nombres de los
+     ficheros CSV a analizar y tras ellos ejecutar las celdas. Como resultado se obtienen tres ficheros XLSX e imagenes
+     en el directorio "Images", y además varios resultados sobre T-test entre varios resultados que se reflejan como
+     output en la salida estandar de JupyterNotebook.

Analysis/iterTimes.py

+import sys
+import glob
+import numpy as np
+import pandas as pd
+
+#-----------------------------------------------
+def read_file(f, data, it):
+  compute_tam = 0
+  comm_tam = 0
+  sdr = 0
+  adr = 0
+  dist = 0
+  time = 0
+  recording = False
+  it_line = 0
+  aux_it = 0
+  iters = 0
+  np = 0
+  np_par = 0
+  ns = 0
+  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_it = 0
+          lineS.pop(0)
+
+          if it_line==0:
+            for observation in lineS:
+              data.append([None]*13)
+              data[it+aux_it][0] = sdr
+              data[it+aux_it][1] = adr
+              data[it+aux_it][2] = np
+              data[it+aux_it][3] = np_par
+              data[it+aux_it][4] = ns
+              data[it+aux_it][5] = dist
+              data[it+aux_it][6] = compute_tam
+              data[it+aux_it][7] = comm_tam
+              data[it+aux_it][8] = time
+              data[it+aux_it][9] = iters
+              data[it+aux_it][10] = float(observation)
+              aux_it+=1
+          elif it_line==1:
+            for observation in lineS:
+              data[it+aux_it][11] = float(observation)
+              aux_it+=1
+          else:
+            for observation in lineS:
+              data[it+aux_it][12] = float(observation)
+              aux_it+=1
+
+          it_line += 1
+          if(it_line % 3 == 0): # Comprobar si se ha terminado de mirar esta ejecucion
+            recording = False
+            it_line = 0
+            it = it + aux_it
+
+        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]) 
+          time = float(lineS[6].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 it
+#-----------------------------------------------
+#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:
+    BaseDir = sys.argv[2]
+
+if len(sys.argv) >= 4:
+  print("Csv name will be: " + sys.argv[3] + ".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)));
+
+it = 0
+data = []  #0   #1        #2    #3       #4     #5      #6             #7          #8       #9   #10    #11    #12
+columns = ["N", "%Async", "NP", "N_par", "NS", "Dist", "Compute_tam", "Comm_tam", "Time", "Iters", "Ti", "Tt", "To"] #13
+
+for elem in lista:
+  f = open(elem, "r")
+  it = read_file(f, data, it)
+  f.close()
+
+#print(data)
+df = pd.DataFrame(data, columns=columns)
+
+df['N'] += df['%Async']
+df['%Async'] = (df['%Async'] / df['N']) * 100
+df.to_csv(name + '.csv')