src/analyzer/tunerloganalyzer.py

   1 import re
   2 import argparse
   3 import sys
   4 import os
   5 import plot as pl
   6
   7 class AutoTunerArgParser:
   8         def __init__(self):
   9                 self.parser = argparse.ArgumentParser(description='Parsing the output log of the CSolver auto tuner ...')
  10                 self.parser.add_argument('--folder', '-f', metavar='bin', type=str, nargs=1,help='output log of running the autotuner ...')
  11                 self.parser.add_argument('--number', '-n', metavar='122', type=int, nargs=1,help='Number of runs ...')
  12                 self.args = self.parser.parse_args()
  13
  14         def getFolder(self):
  15                 return self.args.folder[0]
  16
  17         def getRunNumber(self):
  18                 return self.args.number[0]
  19
  20 TUNABLEHEADER = ["DECOMPOSEORDER", "MUSTREACHGLOBAL", "MUSTREACHLOCAL", "MUSTREACHPRUNE", "OPTIMIZEORDERSTRUCTURE",
  21                 "ORDERINTEGERENCODING", "PREPROCESS", "NODEENCODING", "EDGEENCODING", "MUSTEDGEPRUNE", "ELEMENTOPT",
  22                 "ENCODINGGRAPHOPT", "ELEMENTOPTSETS", "PROXYVARIABLE", "MUSTVALUE", "NAIVEENCODER", "VARIABLEORDER",
  23                 "PROBLEM","SATTIME", "EXECTIME","TUNERNUMBER"]
  24
  25 configs = {"EXECTIME": "-",
  26                 "SATTIME":"-",
  27                 "TESTCASE":"-",
  28                 "PREPROCESS" : "-",
  29                 "ELEMENTOPT" : "-",
  30                 "ELEMENTOPTSETS" : "-",
  31                 "PROXYVARIABLE" : "-",
  32                 "#SubGraph" : "-",
  33                 "NODEENCODING" : "-",
  34                 "EDGEENCODING" : "-",
  35                 "NAIVEENCODER" :"-",
  36                 "ENCODINGGRAPHOPT" : "-"
  37                 }
  38
  39 REGEXES = {"EXECTIME": "CSOLVER solve time: (.*)",
  40                 "SATTIME":"SAT Solving time: (.*)",
  41                 "TESTCASE": "deserializing (.+) ...",
  42                 "PREPROCESS" : "Param PREPROCESS = (.*)range=\[0,1\]",
  43                 "ELEMENTOPT" : "Param ELEMENTOPT = (.*)range=\[0,1\]",
  44                 "ELEMENTOPTSETS" : "Param ELEMENTOPTSETS = (.*)range=\[0,1\]",
  45                 "PROXYVARIABLE" : "Param PROXYVARIABLE = (.*)range=\[1,5\]",
  46                 "#SubGraph" : "#SubGraph = (.*)",
  47                 "NODEENCODING" : "Param NODEENCODING = (.*)range=\[0,3\](.*)",
  48                 "EDGEENCODING" : "Param EDGEENCODING = (.*)range=\[0,2\](.*)",
  49                 "NAIVEENCODER" : "Param NAIVEENCODER = (.*)range=\[1,3\](.*)",
  50                 "ENCODINGGRAPHOPT" : "Param ENCODINGGRAPHOPT = (.*)range=\[0,1\]"
  51                 }
  52
  53 def printHeader(file):
  54         global TUNABLEHEADER
  55         mystr=""
  56         for header in TUNABLEHEADER:
  57                  mystr+=str(header)+","
  58         print >>file, mystr
  59
  60 def dump(file, row):
  61         global TUNABLEHEADER
  62         mystr=""
  63         data = []
  64         for i in range(len(TUNABLEHEADER)):
  65                 mystr += row[TUNABLEHEADER[i]]+ ","
  66                 data.append(row[TUNABLEHEADER[i]])
  67         print ("mystr is:"+ mystr)
  68         print >>file, mystr
  69         return data
  70
  71 def loadTunerInfo(row, filename):
  72         with open(filename) as f:
  73                 for line in f:
  74                         numbers = re.findall('\d+',line)
  75                         numbers = map(int,numbers)
  76                         row[TUNABLEHEADER[numbers[3]]] = row[TUNABLEHEADER[numbers[3]]] + str(numbers[7])
  77
  78 def loadSolverTime(row, filename):
  79         global REGEXES
  80         global configs
  81         with open(filename) as f:
  82                 for line in f:
  83                         for regex in REGEXES:
  84                                 p = re.compile(REGEXES[regex])
  85                                 token = p.search(line)
  86                                 if token is not None:
  87                                         if regex == "TESTCASE":
  88                                                 configs[regex] = re.search(REGEXES[regex], line).group(1)
  89                                         else:
  90                                                 configs[regex] = re.findall("\d+\.?\d*", line)[0]
  91         row["SATTIME"] = configs["SATTIME"]
  92         row["EXECTIME"] = configs["EXECTIME"]
  93
  94 def loadProblemName(row,filename):
  95         with open(filename) as f:
  96                 row["PROBLEM"] = f.readline().replace("\n","")
  97 def loadTunerNumber(row, filename):
  98         with open(filename) as f:
  99                 row["TUNERNUMBER"] = f.readline().replace("\n","")
 100 def analyzeLogs(file):
 101         global configs
 102         argprocess = AutoTunerArgParser()
 103         printHeader(file)
 104         rows = []
 105         data = []
 106         for i in range(argprocess.getRunNumber()):
 107                 row = {"DECOMPOSEORDER" : "",
 108                         "MUSTREACHGLOBAL" : "",
 109                         "MUSTREACHLOCAL" : "",
 110                         "MUSTREACHPRUNE" : "",
 111                         "OPTIMIZEORDERSTRUCTURE" : "",
 112                         "ORDERINTEGERENCODING" : "",
 113                         "PREPROCESS" : "",
 114                         "NODEENCODING" : "",
 115                         "EDGEENCODING" : "",
 116                         "MUSTEDGEPRUNE" : "",
 117                         "ELEMENTOPT" : "",
 118                         "ENCODINGGRAPHOPT" : "",
 119                         "ELEMENTOPTSETS" : "",
 120                         "PROXYVARIABLE" : "",
 121                         "MUSTVALUE" : "",
 122                         "NAIVEENCODER" : "",
 123                         "VARIABLEORDER" : "",
 124                         "PROBLEM":"",
 125                         "SATTIME":"",
 126                         "EXECTIME": "",
 127                         "TUNERNUMBER":""
 128                 }
 129                 loadTunerNumber(row, argprocess.getFolder() + "/tunernum" + str(i))
 130                 loadTunerInfo(row, argprocess.getFolder()+"/tuner"+str(i)+"used")
 131                 loadSolverTime(row, argprocess.getFolder()+"/log"+str(i))
 132                 loadProblemName(row, argprocess.getFolder()+"/problem"+str(i))
 133                 data.append(dump(file, row))
 134                 rows.append(row)
 135         return rows, data
 136
 137 def tunerNumberAnalysis(file, rows):
 138         global TUNABLEHEADER
 139         tunercount = {}
 140         tunernumber = {}
 141         for row in rows:
 142                 mystr=""
 143                 for i in range(18):
 144                         mystr+=row[TUNABLEHEADER[i]]
 145                 if mystr not in tunercount:
 146                         tunercount.update({mystr : 1})
 147                         tunernumber.update({mystr : str(row["TUNERNUMBER"])})
 148                 else :
 149                         tunercount[mystr] += 1
 150                         tunernumber[mystr] += "-" + str(row["TUNERNUMBER"])
 151
 152         problems = set(map(lambda x: x["PROBLEM"], rows))
 153         print ("Number of repititive tuners")
 154         for key in tunercount:
 155                 if tunercount[key] > 1:
 156                         print (key + "(ids:" + tunernumber[key]  + ") = #" + str(tunercount[key]))
 157
 158 def transformDataset(rows):
 159         print(rows)
 160 def main():
 161         global TUNABLEHEADER
 162         file = open("tuner.csv", "w")
 163         rows, data = analyzeLogs(file)
 164         tunerNumberAnalysis(file, rows)
 165         file.close()
 166         #transformDataset(data)
 167         pl.plot(data, TUNABLEHEADER)
 168         return
 169
 170 if __name__ == "__main__":
 171         main()