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 PROBLEMS = []
  21
  22 TUNABLEHEADER = ["DECOMPOSEORDER", "MUSTREACHGLOBAL", "MUSTREACHLOCAL", "MUSTREACHPRUNE", "OPTIMIZEORDERSTRUCTURE",
  23                 "ORDERINTEGERENCODING", "PREPROCESS", "NODEENCODING", "EDGEENCODING", "MUSTEDGEPRUNE", "ELEMENTOPT",
  24                 "ENCODINGGRAPHOPT", "ELEMENTOPTSETS", "PROXYVARIABLE", "MUSTVALUE", "NAIVEENCODER", "VARIABLEORDER",
  25                 "PROBLEM","SATTIME", "EXECTIME","TUNERNUMBER"]
  26
  27 configs = {"EXECTIME": "-",
  28                 "SATTIME":"-",
  29                 "TESTCASE":"-",
  30                 "PREPROCESS" : "-",
  31                 "ELEMENTOPT" : "-",
  32                 "ELEMENTOPTSETS" : "-",
  33                 "PROXYVARIABLE" : "-",
  34                 "#SubGraph" : "-",
  35                 "NODEENCODING" : "-",
  36                 "EDGEENCODING" : "-",
  37                 "NAIVEENCODER" :"-",
  38                 "ENCODINGGRAPHOPT" : "-"
  39                 }
  40
  41 REGEXES = {"EXECTIME": "CSOLVER solve time: (.*)",
  42                 "SATTIME":"SAT Solving time: (.*)",
  43                 "TESTCASE": "deserializing (.+) ...",
  44                 "PREPROCESS" : "Param PREPROCESS = (.*)range=\[0,1\]",
  45                 "ELEMENTOPT" : "Param ELEMENTOPT = (.*)range=\[0,1\]",
  46                 "ELEMENTOPTSETS" : "Param ELEMENTOPTSETS = (.*)range=\[0,1\]",
  47                 "PROXYVARIABLE" : "Param PROXYVARIABLE = (.*)range=\[1,5\]",
  48                 "#SubGraph" : "#SubGraph = (.*)",
  49                 "NODEENCODING" : "Param NODEENCODING = (.*)range=\[0,3\](.*)",
  50                 "EDGEENCODING" : "Param EDGEENCODING = (.*)range=\[0,2\](.*)",
  51                 "NAIVEENCODER" : "Param NAIVEENCODER = (.*)range=\[1,3\](.*)",
  52                 "ENCODINGGRAPHOPT" : "Param ENCODINGGRAPHOPT = (.*)range=\[0,1\]"
  53                 }
  54
  55 def printHeader(file):
  56         global TUNABLEHEADER
  57         mystr=""
  58         for header in TUNABLEHEADER:
  59                  mystr+=str(header)+","
  60         print >>file, mystr
  61
  62 def dump(file, row):
  63         global TUNABLEHEADER
  64         mystr=""
  65         data = []
  66         for i in range(len(TUNABLEHEADER)):
  67                 mystr += row[TUNABLEHEADER[i]]+ ","
  68                 data.append(row[TUNABLEHEADER[i]])
  69         print ("mystr is:"+ mystr)
  70         print >>file, mystr
  71         return data
  72
  73 def loadTunerInfo(row, filename):
  74         with open(filename) as f:
  75                 for line in f:
  76                         numbers = re.findall('\d+',line)
  77                         numbers = map(int,numbers)
  78                         row[TUNABLEHEADER[numbers[3]]] = row[TUNABLEHEADER[numbers[3]]] + str(numbers[7])
  79
  80 def loadSolverTime(row, filename):
  81         global REGEXES
  82         global configs
  83         with open(filename) as f:
  84                 for line in f:
  85                         for regex in REGEXES:
  86                                 p = re.compile(REGEXES[regex])
  87                                 token = p.search(line)
  88                                 if token is not None:
  89                                         if regex == "TESTCASE":
  90                                                 configs[regex] = re.search(REGEXES[regex], line).group(1)
  91                                         else:
  92                                                 configs[regex] = re.findall("\d+\.?\d*", line)[0]
  93         row["SATTIME"] = configs["SATTIME"]
  94         row["EXECTIME"] = configs["EXECTIME"]
  95
  96 def loadProblemName(row,filename):
  97         global PROBLEMS
  98         with open(filename) as f:
  99                 problem = f.readline().replace("\n","")
 100                 probNumber = int(f.readline())
 101                 if probNumber >= len(PROBLEMS):
 102                         PROBLEMS.insert(probNumber,problem)
 103                 elif PROBLEMS[probNumber] != problem:
 104                         PROBLEMS[probNumber] = problem
 105                 row["PROBLEM"] = problem
 106
 107 def loadTunerNumber(row, filename):
 108         with open(filename) as f:
 109                 row["TUNERNUMBER"] = f.readline().replace("\n","")
 110 def analyzeLogs(file):
 111         global configs
 112         argprocess = AutoTunerArgParser()
 113         printHeader(file)
 114         rows = []
 115         data = []
 116         for i in range(argprocess.getRunNumber()):
 117                 row = {"DECOMPOSEORDER" : "",
 118                         "MUSTREACHGLOBAL" : "",
 119                         "MUSTREACHLOCAL" : "",
 120                         "MUSTREACHPRUNE" : "",
 121                         "OPTIMIZEORDERSTRUCTURE" : "",
 122                         "ORDERINTEGERENCODING" : "",
 123                         "PREPROCESS" : "",
 124                         "NODEENCODING" : "",
 125                         "EDGEENCODING" : "",
 126                         "MUSTEDGEPRUNE" : "",
 127                         "ELEMENTOPT" : "",
 128                         "ENCODINGGRAPHOPT" : "",
 129                         "ELEMENTOPTSETS" : "",
 130                         "PROXYVARIABLE" : "",
 131                         "MUSTVALUE" : "",
 132                         "NAIVEENCODER" : "",
 133                         "VARIABLEORDER" : "",
 134                         "PROBLEM":"",
 135                         "SATTIME":"",
 136                         "EXECTIME": "",
 137                         "TUNERNUMBER":""
 138                 }
 139                 loadTunerNumber(row, argprocess.getFolder() + "/tunernum" + str(i))
 140                 loadTunerInfo(row, argprocess.getFolder()+"/tuner"+str(i)+"used")
 141                 loadSolverTime(row, argprocess.getFolder()+"/log"+str(i))
 142                 loadProblemName(row, argprocess.getFolder()+"/problem"+str(i))
 143                 data.append(dump(file, row))
 144                 rows.append(row)
 145         return rows, data
 146
 147 def tunerCountAnalysis(file, rows):
 148         global TUNABLEHEADER
 149         global PROBLEMS
 150         tunercount = {}
 151         tunernumber = {}
 152         for row in rows:
 153                 mystr=""
 154                 for i in range(18):
 155                         if not row[TUNABLEHEADER[i]]:
 156                                 mystr += "."
 157                         else:
 158                                 mystr+=row[TUNABLEHEADER[i]]
 159                 if mystr not in tunercount:
 160                         tunercount.update({mystr : 1})
 161                         tunernumber.update({mystr : str(row["TUNERNUMBER"])})
 162                 else :
 163                         tunercount[mystr] += 1
 164                         tunernumber[mystr] += "-" + str(row["TUNERNUMBER"])
 165
 166         problems = set(map(lambda x: x["PROBLEM"], rows))
 167         print ("Number of repititive tuners")
 168         for key in tunercount:
 169                 if tunercount[key] > 1:
 170                         print key + "(ids:" + tunernumber[key]  + ") = #" + str(tunercount[key])
 171
 172 def combineRowForEachTuner(rows):
 173         global PROBLEMS
 174         newRows = []
 175         combined = None
 176         for row in rows:
 177                 if row["PROBLEM"] == PROBLEMS[0]:
 178                         combined = row
 179                 for key in row:
 180                         if row[key]:
 181                                 combined[key] = row[key]
 182                 if row["PROBLEM"] == PROBLEMS[len(PROBLEMS)-1]:
 183                         newRows.append(combined)
 184         return newRows
 185
 186 def transformDataset(rows):
 187         print(rows)
 188
 189
 190 def main():
 191         global TUNABLEHEADER
 192         file = open("tuner.csv", "w")
 193         rows, data = analyzeLogs(file)
 194         tunerCountAnalysis(file, combineRowForEachTuner(rows) )
 195         file.close()
 196         #transformDataset(data)
 197         pl.plot(data, TUNABLEHEADER)
 198
 199
 200 if __name__ == "__main__":
 201         main()