gem5/util/stats/output.py

189 lines
7.1 KiB
Python
Raw Normal View History

# Copyright (c) 2005 The Regents of The University of Michigan
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met: redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer;
# redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution;
# neither the name of the copyright holders nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
# Authors: Nathan Binkert
class StatOutput(object):
Major cleanup of the statistics handling code util/stats/db.py: Build a result object as the result of a query operation so it is easier to populate and contains a bit more information than just a big dict. Also change the next level data into a matrix instead of a dict of dicts. Move the "get" function into the Database object. (The get function is used by the output parsing function as the interface for accessing backend storage, same interface for profile stuff.) Change the old get variable to the method variable, it describes how the get works, (whether using sum, stdev, etc.) util/stats/display.py: Clean up the display functions, mostly formatting. Handle values the way they should be now. util/stats/info.py: Totally re-work how values are accessed from their data store. Access individual values on demand instead of calculating everything and passing up a huge result from the bottom. This impacts the way that proxying works, and in general, everything is now esentially a proxy for the lower level database. Provide new operators: unproxy, scalar, vector, value, values, total, and len which retrieve the proper result from the object they are called on. Move the ProxyGroup stuff (proxies of proxies!) here from the now gone proxy.py file and integrate the shared parts of the code. The ProxyGroup stuff allows you to write formulas without specifying the statistics until evaluation time. Get rid of global variables! util/stats/output.py: Move the dbinfo stuff into the Database itself. Each source should have it's own get() function for accessing it's data. This get() function behaves a bit differently than before in that it can return vectors as well, deal with these vectors and with no result conditions better. util/stats/stats.py: the info module no longer has the source global variable, just create the database source and pass it around as necessary --HG-- extra : convert_revision : 8e5aa228e5d3ae8068ef9c40f65b3a2f9e7c0cff
2005-10-21 22:29:13 +02:00
def __init__(self, name, jobfile, info, stat=None, binstats=None):
self.name = name
self.jobfile = jobfile
self.stat = stat
self.binstats = None
self.label = self.name
self.invert = False
self.info = info
def printdata(self, bin = None, printmode = 'G'):
import info
if bin:
print '%s %s stats' % (self.name, bin)
if self.binstats:
for stat in self.binstats:
stat.bins = bin
if printmode == 'G':
valformat = '%g'
elif printmode != 'F' and value > 1e6:
valformat = '%0.5e'
else:
valformat = '%f'
for job in self.jobfile.jobs():
value = self.info.get(job, self.stat)
if value is None:
return
if not isinstance(value, list):
value = [ value ]
if self.invert:
for i,val in enumerate(value):
if val != 0.0:
value[i] = 1 / val
valstring = ', '.join([ valformat % val for val in value ])
print '%-50s %s' % (job.name + ':', valstring)
def display(self, binned = False, printmode = 'G'):
if binned and self.binstats:
self.printdata('kernel', printmode)
self.printdata('idle', printmode)
self.printdata('user', printmode)
self.printdata('interrupt', printmode)
print '%s total stats' % self.name
self.printdata(printmode=printmode)
def graph(self, graphdir):
from os.path import expanduser, isdir, join as joinpath
from barchart import BarChart
from matplotlib.numerix import Float, array, zeros
import os, re
confgroups = self.jobfile.groups()
ngroups = len(confgroups)
skiplist = [ False ] * ngroups
groupopts = None
baropts = None
groups = []
for i,group in enumerate(confgroups):
if group.flags.graph_group:
if groupopts is not None:
raise AttributeError, \
'Two groups selected for graph group'
groupopts = group.subopts()
skiplist[i] = True
elif group.flags.graph_bars:
if baropts is not None:
raise AttributeError, \
'Two groups selected for graph bars'
baropts = group.subopts()
skiplist[i] = True
else:
groups.append(group)
if groupopts is None:
raise AttributeError, 'No group selected for graph group'
if baropts is None:
raise AttributeError, 'No group selected for graph bars'
directory = expanduser(graphdir)
if not isdir(directory):
os.mkdir(directory)
html = file(joinpath(directory, '%s.html' % self.name), 'w')
print >>html, '<html>'
print >>html, '<title>Graphs for %s</title>' % self.name
print >>html, '<body>'
for options in self.jobfile.options(groups):
data = zeros((len(groupopts), len(baropts)), Float)
data = [ [ None ] * len(baropts) for i in xrange(len(groupopts)) ]
enabled = False
Major cleanup of the statistics handling code util/stats/db.py: Build a result object as the result of a query operation so it is easier to populate and contains a bit more information than just a big dict. Also change the next level data into a matrix instead of a dict of dicts. Move the "get" function into the Database object. (The get function is used by the output parsing function as the interface for accessing backend storage, same interface for profile stuff.) Change the old get variable to the method variable, it describes how the get works, (whether using sum, stdev, etc.) util/stats/display.py: Clean up the display functions, mostly formatting. Handle values the way they should be now. util/stats/info.py: Totally re-work how values are accessed from their data store. Access individual values on demand instead of calculating everything and passing up a huge result from the bottom. This impacts the way that proxying works, and in general, everything is now esentially a proxy for the lower level database. Provide new operators: unproxy, scalar, vector, value, values, total, and len which retrieve the proper result from the object they are called on. Move the ProxyGroup stuff (proxies of proxies!) here from the now gone proxy.py file and integrate the shared parts of the code. The ProxyGroup stuff allows you to write formulas without specifying the statistics until evaluation time. Get rid of global variables! util/stats/output.py: Move the dbinfo stuff into the Database itself. Each source should have it's own get() function for accessing it's data. This get() function behaves a bit differently than before in that it can return vectors as well, deal with these vectors and with no result conditions better. util/stats/stats.py: the info module no longer has the source global variable, just create the database source and pass it around as necessary --HG-- extra : convert_revision : 8e5aa228e5d3ae8068ef9c40f65b3a2f9e7c0cff
2005-10-21 22:29:13 +02:00
stacked = 0
for g,gopt in enumerate(groupopts):
for b,bopt in enumerate(baropts):
job = self.jobfile.job(options + [ gopt, bopt ])
if not job:
continue
val = self.info.get(job, self.stat)
Major cleanup of the statistics handling code util/stats/db.py: Build a result object as the result of a query operation so it is easier to populate and contains a bit more information than just a big dict. Also change the next level data into a matrix instead of a dict of dicts. Move the "get" function into the Database object. (The get function is used by the output parsing function as the interface for accessing backend storage, same interface for profile stuff.) Change the old get variable to the method variable, it describes how the get works, (whether using sum, stdev, etc.) util/stats/display.py: Clean up the display functions, mostly formatting. Handle values the way they should be now. util/stats/info.py: Totally re-work how values are accessed from their data store. Access individual values on demand instead of calculating everything and passing up a huge result from the bottom. This impacts the way that proxying works, and in general, everything is now esentially a proxy for the lower level database. Provide new operators: unproxy, scalar, vector, value, values, total, and len which retrieve the proper result from the object they are called on. Move the ProxyGroup stuff (proxies of proxies!) here from the now gone proxy.py file and integrate the shared parts of the code. The ProxyGroup stuff allows you to write formulas without specifying the statistics until evaluation time. Get rid of global variables! util/stats/output.py: Move the dbinfo stuff into the Database itself. Each source should have it's own get() function for accessing it's data. This get() function behaves a bit differently than before in that it can return vectors as well, deal with these vectors and with no result conditions better. util/stats/stats.py: the info module no longer has the source global variable, just create the database source and pass it around as necessary --HG-- extra : convert_revision : 8e5aa228e5d3ae8068ef9c40f65b3a2f9e7c0cff
2005-10-21 22:29:13 +02:00
if isinstance(val, (list, tuple)):
if len(val) == 1:
val = val[0]
else:
stacked = len(val)
data[g][b] = val
Major cleanup of the statistics handling code util/stats/db.py: Build a result object as the result of a query operation so it is easier to populate and contains a bit more information than just a big dict. Also change the next level data into a matrix instead of a dict of dicts. Move the "get" function into the Database object. (The get function is used by the output parsing function as the interface for accessing backend storage, same interface for profile stuff.) Change the old get variable to the method variable, it describes how the get works, (whether using sum, stdev, etc.) util/stats/display.py: Clean up the display functions, mostly formatting. Handle values the way they should be now. util/stats/info.py: Totally re-work how values are accessed from their data store. Access individual values on demand instead of calculating everything and passing up a huge result from the bottom. This impacts the way that proxying works, and in general, everything is now esentially a proxy for the lower level database. Provide new operators: unproxy, scalar, vector, value, values, total, and len which retrieve the proper result from the object they are called on. Move the ProxyGroup stuff (proxies of proxies!) here from the now gone proxy.py file and integrate the shared parts of the code. The ProxyGroup stuff allows you to write formulas without specifying the statistics until evaluation time. Get rid of global variables! util/stats/output.py: Move the dbinfo stuff into the Database itself. Each source should have it's own get() function for accessing it's data. This get() function behaves a bit differently than before in that it can return vectors as well, deal with these vectors and with no result conditions better. util/stats/stats.py: the info module no longer has the source global variable, just create the database source and pass it around as necessary --HG-- extra : convert_revision : 8e5aa228e5d3ae8068ef9c40f65b3a2f9e7c0cff
2005-10-21 22:29:13 +02:00
if stacked == 0:
for i in xrange(len(groupopts)):
for j in xrange(len(baropts)):
if data[i][j] is None:
data[i][j] = 0.0
else:
for i in xrange(len(groupopts)):
for j in xrange(len(baropts)):
val = data[i][j]
if val is None:
data[i][j] = [] * stacked
elif len(val) != stacked:
raise ValueError, "some stats stacked, some not"
data = array(data)
if data.sum() == 0:
continue
bar_descs = [ opt.desc for opt in baropts ]
group_descs = [ opt.desc for opt in groupopts ]
if stacked:
Major cleanup of the statistics handling code util/stats/db.py: Build a result object as the result of a query operation so it is easier to populate and contains a bit more information than just a big dict. Also change the next level data into a matrix instead of a dict of dicts. Move the "get" function into the Database object. (The get function is used by the output parsing function as the interface for accessing backend storage, same interface for profile stuff.) Change the old get variable to the method variable, it describes how the get works, (whether using sum, stdev, etc.) util/stats/display.py: Clean up the display functions, mostly formatting. Handle values the way they should be now. util/stats/info.py: Totally re-work how values are accessed from their data store. Access individual values on demand instead of calculating everything and passing up a huge result from the bottom. This impacts the way that proxying works, and in general, everything is now esentially a proxy for the lower level database. Provide new operators: unproxy, scalar, vector, value, values, total, and len which retrieve the proper result from the object they are called on. Move the ProxyGroup stuff (proxies of proxies!) here from the now gone proxy.py file and integrate the shared parts of the code. The ProxyGroup stuff allows you to write formulas without specifying the statistics until evaluation time. Get rid of global variables! util/stats/output.py: Move the dbinfo stuff into the Database itself. Each source should have it's own get() function for accessing it's data. This get() function behaves a bit differently than before in that it can return vectors as well, deal with these vectors and with no result conditions better. util/stats/stats.py: the info module no longer has the source global variable, just create the database source and pass it around as necessary --HG-- extra : convert_revision : 8e5aa228e5d3ae8068ef9c40f65b3a2f9e7c0cff
2005-10-21 22:29:13 +02:00
try:
legend = self.info.rcategories
except:
legend = [ str(i) for i in xrange(stacked) ]
else:
legend = bar_descs
chart = BarChart(data=data, xlabel='Benchmark', ylabel=self.label,
legend=legend, xticks=group_descs)
chart.graph()
names = [ opt.name for opt in options ]
descs = [ opt.desc for opt in options ]
filename = '%s-%s.png' % (self.name, ':'.join(names))
desc = ' '.join(descs)
filepath = joinpath(directory, filename)
chart.savefig(filepath)
filename = re.sub(':', '%3A', filename)
print >>html, '''%s<br><img src="%s"><br>''' % (desc, filename)
print >>html, '</body>'
print >>html, '</html>'
html.close()