profile.py

来自「M5,一个功能强大的多处理器系统模拟器.很多针对处理器架构,性能的研究都使用它作」· Python 代码 · 共 504 行 · 第 1/2 页

        for root,dirs,files in os.walk(directory):
            for name in files:
                match = label_ex.match(name)
                if not match:
                    continue

                filename = joinpath(root, name)
                prefix = os.path.commonprefix([root, directory])
                dirname = root[len(prefix)+1:]
                data = self.datatype(filename, self.categorize)
                self.setdata(dirname, match.group(1), data)

    def setdata(self, run, cpu, data):
        if run not in self.data:
            self.data[run] = {}

        if cpu in self.data[run]:
            raise AttributeError, \
                  'data already stored for run %s and cpu %s' % (run, cpu)

        self.data[run][cpu] = data

    def getdata(self, run, cpu):
        try:
            return self.data[run][cpu]
        except KeyError:
            print run, cpu
            return None

    def alldata(self):
        for run,cpus in self.data.iteritems():
            for cpu,data in cpus.iteritems():
                yield run,cpu,data

    def get(self, job, stat, system=None):
        if system is None and hasattr('system', job):
            system = job.system

        if system is None:
            raise AttributeError, 'The job must have a system set'

        cpu = '%s.run%d' % (system, self.cpu)

        data = self.getdata(str(job), cpu)
        if not data:
            return None

        values = []
        for category in self.categories:
            val = float(data.get(category, 0.0))
            if val < 0.0:
                raise ValueError, 'value is %f' % val
            values.append(val)
        total = sum(values)
        return [ v / total * 100.0 for v in values ]

    def dump(self):
        for run,cpu,data in self.alldata():
            print 'run %s, cpu %s' % (run, cpu)
            data.dump()
            print

    def write_dot(self, threshold, jobfile=None, jobs=None):
        import pydot

        if jobs is None:
            jobs = [ job for job in jobfile.jobs() ]

        for job in jobs:
            cpu =  '%s.run%d' % (job.system, self.cpu)
            symbols = self.getdata(job.name, cpu)
            if not symbols:
                continue

            dot = pydot.Dot()
            symbols.tree.dot(dot, threshold=threshold)
            dot.write(symbols.filename[:-3] + 'dot')

    def write_txt(self, jobfile=None, jobs=None, limit=None):
        if jobs is None:
            jobs = [ job for job in jobfile.jobs() ]

        for job in jobs:
            cpu =  '%s.run%d' % (job.system, self.cpu)
            symbols = self.getdata(job.name, cpu)
            if not symbols:
                continue

            output = file(symbols.filename[:-3] + 'txt', 'w')
            symbols.display(output, limit)

    def display(self, jobfile=None, jobs=None, limit=None):
        if jobs is None:
            jobs = [ job for job in jobfile.jobs() ]

        maxsymlen = 0

        thejobs = []
        for job in jobs:
            cpu =  '%s.run%d' % (job.system, self.cpu)
            symbols = self.getdata(job.name, cpu)
            if symbols:
                thejobs.append(job)
                maxsymlen = max(maxsymlen, symbols.maxsymlen)

        for job in thejobs:
            cpu =  '%s.run%d' % (job.system, self.cpu)
            symbols = self.getdata(job.name, cpu)
            print job.name
            symbols.display(limit=limit, maxsymlen=maxsymlen)
            print


from categories import func_categorize, pc_categorize
class PCProfile(Profile):
    def __init__(self, categorize=pc_categorize):
        super(PCProfile, self).__init__(PCData, categorize)


class FuncProfile(Profile):
    def __init__(self, categorize=func_categorize):
        super(FuncProfile, self).__init__(FuncData, categorize)

def usage(exitcode = None):
    print '''\
Usage: %s [-bc] [-g <dir>] [-j <jobfile>] [-n <num>]

    -c           groups symbols into categories
    -b           dumps data for bar charts
    -d           generate dot output
    -g <d>       draw graphs and send output to <d>
    -j <jobfile> specify a different jobfile (default is Test.py)
    -n <n>       selects number of top symbols to print (default 5)
''' % sys.argv[0]

    if exitcode is not None:
        sys.exit(exitcode)

if __name__ == '__main__':
    import getopt, re, sys
    from os.path import expanduser
    from output import StatOutput

    # default option values
    numsyms = 10
    graph = None
    cpus = [ 0 ]
    categorize = False
    showidle = True
    funcdata = True
    jobfilename = 'Test.py'
    dodot = False
    dotfile = None
    textout = False
    threshold = 0.01
    inputfile = None

    try:
        opts, args = getopt.getopt(sys.argv[1:], 'C:cdD:f:g:ij:n:pT:t')
    except getopt.GetoptError:
        usage(2)

    for o,a in opts:
        if o == '-C':
            cpus = [ int(x) for x in a.split(',') ]
        elif o == '-c':
            categorize = True
        elif o == '-D':
            dotfile = a
        elif o == '-d':
            dodot = True
        elif o == '-f':
            inputfile = expanduser(a)
        elif o == '-g':
            graph = a
        elif o == '-i':
            showidle = False
        elif o == '-j':
            jobfilename = a
        elif o == '-n':
            numsyms = int(a)
        elif o == '-p':
            funcdata = False
        elif o == '-T':
            threshold = float(a)
        elif o == '-t':
            textout = True

    if args:
        print "'%s'" % args, len(args)
        usage(1)

    if inputfile:
        catfunc = None
        if categorize:
            catfunc = func_categorize
        data = FuncData(inputfile, categorize=catfunc)

        if dodot:
            import pydot
            dot = pydot.Dot()
            data.tree.dot(dot, threshold=threshold)
            #dot.orientation = 'landscape'
            #dot.ranksep='equally'
            #dot.rank='samerank'
            dot.write(dotfile, format='png')
        else:
            data.display(limit=numsyms)

    else:
        from jobfile import JobFile
        jobfile = JobFile(jobfilename)

        if funcdata:
            profile = FuncProfile()
        else:
            profile = PCProfile()

        if not categorize:
            profile.categorize = None
        profile.inputdir(jobfile.rootdir)

        if graph:
            for cpu in cpus:
                profile.cpu = cpu
                if funcdata:
                    name = 'funcstacks%d' % cpu
                else:
                    name = 'stacks%d' % cpu
                output = StatOutput(jobfile, info=profile)
                output.xlabel = 'System Configuration'
                output.ylabel = '% CPU utilization'
                output.stat = name
                output.graph(name, graph)

        if dodot:
            for cpu in cpus:
                profile.cpu = cpu
                profile.write_dot(jobfile=jobfile, threshold=threshold)

        if textout:
            for cpu in cpus:
                profile.cpu = cpu
                profile.write_txt(jobfile=jobfile)

        if not graph and not textout and not dodot:
            for cpu in cpus:
                if not categorize:
                    profile.categorize = None
                profile.cpu = cpu
                profile.display(jobfile=jobfile, limit=numsyms)