1
0
mirror of https://git.yoctoproject.org/poky synced 2026-05-09 17:39:31 +00:00

scripts: add oe-build-perf-report script

A new tool for pretty-printing build perf test results stored in a Git
repository. The scripts is able to produce either simple plaintext
report showing the difference between two commits, or, an html report
that also displays trendcharts of the test results. The script uses
Jinja2 templates for generating HTML reports so it requires
python3-jinja2 to be installed on the system.

[YOCTO #10931]

(From OE-Core rev: 3b25404f0f99b72f222bdca815929be1cf1cee35)

Signed-off-by: Markus Lehtonen <markus.lehtonen@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This commit is contained in:
Markus Lehtonen
2017-03-31 17:07:29 +03:00
committed by Richard Purdie
parent 5a85d39c9d
commit 9f299876f7
6 changed files with 1182 additions and 0 deletions
+31
View File
@@ -0,0 +1,31 @@
#
# Copyright (c) 2017, Intel Corporation.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms and conditions of the GNU General Public License,
# version 2, as published by the Free Software Foundation.
#
# This program is distributed in the hope it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
# more details.
#
"""Build performance test library functions"""
def print_table(rows, row_fmt=None):
"""Print data table"""
if not rows:
return
if not row_fmt:
row_fmt = ['{:{wid}} '] * len(rows[0])
# Go through the data to get maximum cell widths
num_cols = len(row_fmt)
col_widths = [0] * num_cols
for row in rows:
for i, val in enumerate(row):
col_widths[i] = max(col_widths[i], len(str(val)))
for row in rows:
print(*[row_fmt[i].format(col, wid=col_widths[i]) for i, col in enumerate(row)])
+19
View File
@@ -0,0 +1,19 @@
#
# Copyright (c) 2017, Intel Corporation.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms and conditions of the GNU General Public License,
# version 2, as published by the Free Software Foundation.
#
# This program is distributed in the hope it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
# more details.
#
"""Helper module for HTML reporting"""
from jinja2 import Environment, PackageLoader
env = Environment(loader=PackageLoader('build_perf', 'html'))
template = env.get_template('report.html')
@@ -0,0 +1,50 @@
<script type="text/javascript">
google.charts.setOnLoadCallback(drawChart_{{ chart_elem_id }});
function drawChart_{{ chart_elem_id }}() {
var data = new google.visualization.DataTable();
// Chart options
var options = {
theme : 'material',
legend: 'none',
hAxis: { format: '', title: 'Commit number',
minValue: {{ chart_opts.haxis.min }},
maxValue: {{ chart_opts.haxis.max }} },
{% if measurement.type == 'time' %}
vAxis: { format: 'h:mm:ss' },
{% else %}
vAxis: { format: '' },
{% endif %}
pointSize: 5,
chartArea: { left: 80, right: 15 },
};
// Define data columns
data.addColumn('number', 'Commit');
data.addColumn('{{ measurement.value_type.gv_data_type }}',
'{{ measurement.value_type.quantity }}');
// Add data rows
data.addRows([
{% for sample in measurement.samples %}
[{{ sample.commit_num }}, {{ sample.mean.gv_value() }}],
{% endfor %}
]);
// Finally, draw the chart
chart_div = document.getElementById('{{ chart_elem_id }}');
var chart = new google.visualization.LineChart(chart_div);
google.visualization.events.addListener(chart, 'ready', function () {
//chart_div = document.getElementById('{{ chart_elem_id }}');
//chart_div.innerHTML = '<img src="' + chart.getImageURI() + '">';
png_div = document.getElementById('{{ chart_elem_id }}_png');
png_div.outerHTML = '<a id="{{ chart_elem_id }}_png" href="' + chart.getImageURI() + '">PNG</a>';
console.log("CHART READY: {{ chart_elem_id }}");
{% if last_chart == true %}
console.log("ALL CHARTS READY");
{% endif %}
//console.log(chart_div.innerHTML);
});
chart.draw(data, options);
}
</script>
+209
View File
@@ -0,0 +1,209 @@
<!DOCTYPE html>
<html lang="en">
<head>
{# Scripts, for visualization#}
<!--START-OF-SCRIPTS-->
<script type="text/javascript" src="https://www.gstatic.com/charts/loader.js"></script>
<script type="text/javascript">
google.charts.load('current', {'packages':['corechart']});
</script>
{# Render measurement result charts #}
{% for test in test_data %}
{% set test_loop = loop %}
{% if test.status == 'SUCCESS' %}
{% for measurement in test.measurements %}
{% set chart_elem_id = test.name + '_' + measurement.name + '_chart' %}
{% if test_loop.last and loop.last %}
{% set last_chart = true %}
{% endif %}
{% include 'measurement_chart.html' %}
{% endfor %}
{% endif %}
{% endfor %}
<!--END-OF-SCRIPTS-->
{# Styles #}
<style>
.meta-table {
font-size: 14px;
text-align: left;
border-collapse: collapse;
}
.meta-table tr:nth-child(even){background-color: #f2f2f2}
meta-table th, .meta-table td {
padding: 4px;
}
.summary {
margin: 0;
font-size: 14px;
text-align: left;
border-collapse: collapse;
}
summary th, .meta-table td {
padding: 4px;
}
.measurement {
padding: 8px 0px 8px 8px;
border: 2px solid #f0f0f0;
margin-bottom: 10px;
}
.details {
margin: 0;
font-size: 12px;
text-align: left;
border-collapse: collapse;
}
.details th {
font-weight: normal;
padding-right: 8px;
}
.preformatted {
font-family: monospace;
white-space: pre-wrap;
background-color: #f0f0f0;
margin-left: 10px;
}
hr {
color: #f0f0f0;
}
h2 {
font-size: 20px;
margin-bottom: 0px;
color: #707070;
}
h3 {
font-size: 16px;
margin: 0px;
color: #707070;
}
</style>
<title>{{ title }}</title>
</head>
{% macro poky_link(commit) -%}
<a href="http://git.yoctoproject.org/cgit/cgit.cgi/poky/log/?id={{ commit }}">{{ commit[0:11] }}</a>
{%- endmacro %}
<body><div style="width: 700px">
{# Test metadata #}
<h2>General</h2>
<hr>
<table class="meta-table" style="width: 100%">
<tr>
<th></th>
<th>Current commit</th>
<th>Comparing with</th>
</tr>
{% for key, item in metadata.items() %}
<tr>
<th>{{ item.title }}</th>
{%if key == 'commit' %}
<td>{{ poky_link(item.value) }}</td>
<td>{{ poky_link(item.value_old) }}</td>
{% else %}
<td>{{ item.value }}</td>
<td>{{ item.value_old }}</td>
{% endif %}
</tr>
{% endfor %}
</table>
{# Test result summary #}
<h2>Test result summary</h2>
<hr>
<table class="summary" style="width: 100%">
{% for test in test_data %}
{% if loop.index is even %}
{% set row_style = 'style="background-color: #f2f2f2"' %}
{% else %}
{% set row_style = 'style="background-color: #ffffff"' %}
{% endif %}
<tr {{ row_style }}><td>{{ test.name }}: {{ test.description }}</td>
{% if test.status == 'SUCCESS' %}
{% for measurement in test.measurements %}
{# add empty cell in place of the test name#}
{% if loop.index > 1 %}<td></td>{% endif %}
{% if measurement.absdiff > 0 %}
{% set result_style = "color: red" %}
{% elif measurement.absdiff == measurement.absdiff %}
{% set result_style = "color: green" %}
{% else %}
{% set result_style = "color: orange" %}
{%endif %}
<td>{{ measurement.description }}</td>
<td style="font-weight: bold">{{ measurement.value.mean }}</td>
<td style="{{ result_style }}">{{ measurement.absdiff_str }}</td>
<td style="{{ result_style }}">{{ measurement.reldiff }}</td>
</tr><tr {{ row_style }}>
{% endfor %}
{% else %}
<td style="font-weight: bold; color: red;">{{test.status }}</td>
<td></td> <td></td> <td></td> <td></td>
{% endif %}
</tr>
{% endfor %}
</table>
{# Detailed test results #}
{% for test in test_data %}
<h2>{{ test.name }}: {{ test.description }}</h2>
<hr>
{% if test.status == 'SUCCESS' %}
{% for measurement in test.measurements %}
<div class="measurement">
<h3>{{ measurement.description }}</h3>
<div style="font-weight:bold;">
<span style="font-size: 23px;">{{ measurement.value.mean }}</span>
<span style="font-size: 20px; margin-left: 12px">
{% if measurement.absdiff > 0 %}
<span style="color: red">
{% elif measurement.absdiff == measurement.absdiff %}
<span style="color: green">
{% else %}
<span style="color: orange">
{% endif %}
{{ measurement.absdiff_str }} ({{measurement.reldiff}})
</span></span>
</div>
<table style="width: 100%">
<tr>
<td style="width: 75%">
{# Linechart #}
<div id="{{ test.name }}_{{ measurement.name }}_chart"></div>
</td>
<td>
{# Measurement statistics #}
<table class="details">
<tr>
<th>Test runs</th><td>{{ measurement.value.sample_cnt }}</td>
</tr><tr>
<th>-/+</th><td>-{{ measurement.value.minus }} / +{{ measurement.value.plus }}</td>
</tr><tr>
<th>Min</th><td>{{ measurement.value.min }}</td>
</tr><tr>
<th>Max</th><td>{{ measurement.value.max }}</td>
</tr><tr>
<th>Stdev</th><td>{{ measurement.value.stdev }}</td>
</tr><tr>
<th><div id="{{ test.name }}_{{ measurement.name }}_chart_png"></div></th>
</tr>
</table>
</td>
</tr>
</table>
</div>
{% endfor %}
{# Unsuccessful test #}
{% else %}
<span style="font-size: 150%; font-weight: bold; color: red;">{{ test.status }}
{% if test.err_type %}<span style="font-size: 75%; font-weight: normal">({{ test.err_type }})</span>{% endif %}
</span>
<div class="preformatted">{{ test.message }}</div>
{% endif %}
{% endfor %}
</div></body>
</html>
+342
View File
@@ -0,0 +1,342 @@
#
# Copyright (c) 2017, Intel Corporation.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms and conditions of the GNU General Public License,
# version 2, as published by the Free Software Foundation.
#
# This program is distributed in the hope it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
# more details.
#
"""Handling of build perf test reports"""
from collections import OrderedDict, Mapping
from datetime import datetime, timezone
from numbers import Number
from statistics import mean, stdev, variance
def isofmt_to_timestamp(string):
"""Convert timestamp string in ISO 8601 format into unix timestamp"""
if '.' in string:
dt = datetime.strptime(string, '%Y-%m-%dT%H:%M:%S.%f')
else:
dt = datetime.strptime(string, '%Y-%m-%dT%H:%M:%S')
return dt.replace(tzinfo=timezone.utc).timestamp()
def metadata_xml_to_json(elem):
"""Convert metadata xml into JSON format"""
assert elem.tag == 'metadata', "Invalid metadata file format"
def _xml_to_json(elem):
"""Convert xml element to JSON object"""
out = OrderedDict()
for child in elem.getchildren():
key = child.attrib.get('name', child.tag)
if len(child):
out[key] = _xml_to_json(child)
else:
out[key] = child.text
return out
return _xml_to_json(elem)
def results_xml_to_json(elem):
"""Convert results xml into JSON format"""
rusage_fields = ('ru_utime', 'ru_stime', 'ru_maxrss', 'ru_minflt',
'ru_majflt', 'ru_inblock', 'ru_oublock', 'ru_nvcsw',
'ru_nivcsw')
iostat_fields = ('rchar', 'wchar', 'syscr', 'syscw', 'read_bytes',
'write_bytes', 'cancelled_write_bytes')
def _read_measurement(elem):
"""Convert measurement to JSON"""
data = OrderedDict()
data['type'] = elem.tag
data['name'] = elem.attrib['name']
data['legend'] = elem.attrib['legend']
values = OrderedDict()
# SYSRES measurement
if elem.tag == 'sysres':
for subel in elem:
if subel.tag == 'time':
values['start_time'] = isofmt_to_timestamp(subel.attrib['timestamp'])
values['elapsed_time'] = float(subel.text)
elif subel.tag == 'rusage':
rusage = OrderedDict()
for field in rusage_fields:
if 'time' in field:
rusage[field] = float(subel.attrib[field])
else:
rusage[field] = int(subel.attrib[field])
values['rusage'] = rusage
elif subel.tag == 'iostat':
values['iostat'] = OrderedDict([(f, int(subel.attrib[f]))
for f in iostat_fields])
elif subel.tag == 'buildstats_file':
values['buildstats_file'] = subel.text
else:
raise TypeError("Unknown sysres value element '{}'".format(subel.tag))
# DISKUSAGE measurement
elif elem.tag == 'diskusage':
values['size'] = int(elem.find('size').text)
else:
raise Exception("Unknown measurement tag '{}'".format(elem.tag))
data['values'] = values
return data
def _read_testcase(elem):
"""Convert testcase into JSON"""
assert elem.tag == 'testcase', "Expecting 'testcase' element instead of {}".format(elem.tag)
data = OrderedDict()
data['name'] = elem.attrib['name']
data['description'] = elem.attrib['description']
data['status'] = 'SUCCESS'
data['start_time'] = isofmt_to_timestamp(elem.attrib['timestamp'])
data['elapsed_time'] = float(elem.attrib['time'])
measurements = OrderedDict()
for subel in elem.getchildren():
if subel.tag == 'error' or subel.tag == 'failure':
data['status'] = subel.tag.upper()
data['message'] = subel.attrib['message']
data['err_type'] = subel.attrib['type']
data['err_output'] = subel.text
elif subel.tag == 'skipped':
data['status'] = 'SKIPPED'
data['message'] = subel.text
else:
measurements[subel.attrib['name']] = _read_measurement(subel)
data['measurements'] = measurements
return data
def _read_testsuite(elem):
"""Convert suite to JSON"""
assert elem.tag == 'testsuite', \
"Expecting 'testsuite' element instead of {}".format(elem.tag)
data = OrderedDict()
if 'hostname' in elem.attrib:
data['tester_host'] = elem.attrib['hostname']
data['start_time'] = isofmt_to_timestamp(elem.attrib['timestamp'])
data['elapsed_time'] = float(elem.attrib['time'])
tests = OrderedDict()
for case in elem.getchildren():
tests[case.attrib['name']] = _read_testcase(case)
data['tests'] = tests
return data
# Main function
assert elem.tag == 'testsuites', "Invalid test report format"
assert len(elem) == 1, "Too many testsuites"
return _read_testsuite(elem.getchildren()[0])
def aggregate_metadata(metadata):
"""Aggregate metadata into one, basically a sanity check"""
mutable_keys = ('pretty_name', 'version_id')
def aggregate_obj(aggregate, obj, assert_str=True):
"""Aggregate objects together"""
assert type(aggregate) is type(obj), \
"Type mismatch: {} != {}".format(type(aggregate), type(obj))
if isinstance(obj, Mapping):
assert set(aggregate.keys()) == set(obj.keys())
for key, val in obj.items():
aggregate_obj(aggregate[key], val, key not in mutable_keys)
elif isinstance(obj, list):
assert len(aggregate) == len(obj)
for i, val in enumerate(obj):
aggregate_obj(aggregate[i], val)
elif not isinstance(obj, str) or (isinstance(obj, str) and assert_str):
assert aggregate == obj, "Data mismatch {} != {}".format(aggregate, obj)
if not metadata:
return {}
# Do the aggregation
aggregate = metadata[0].copy()
for testrun in metadata[1:]:
aggregate_obj(aggregate, testrun)
aggregate['testrun_count'] = len(metadata)
return aggregate
def aggregate_data(data):
"""Aggregate multiple test results JSON structures into one"""
mutable_keys = ('status', 'message', 'err_type', 'err_output')
class SampleList(list):
"""Container for numerical samples"""
pass
def new_aggregate_obj(obj):
"""Create new object for aggregate"""
if isinstance(obj, Number):
new_obj = SampleList()
new_obj.append(obj)
elif isinstance(obj, str):
new_obj = obj
else:
# Lists and and dicts are kept as is
new_obj = obj.__class__()
aggregate_obj(new_obj, obj)
return new_obj
def aggregate_obj(aggregate, obj, assert_str=True):
"""Recursive "aggregation" of JSON objects"""
if isinstance(obj, Number):
assert isinstance(aggregate, SampleList)
aggregate.append(obj)
return
assert type(aggregate) == type(obj), \
"Type mismatch: {} != {}".format(type(aggregate), type(obj))
if isinstance(obj, Mapping):
for key, val in obj.items():
if not key in aggregate:
aggregate[key] = new_aggregate_obj(val)
else:
aggregate_obj(aggregate[key], val, key not in mutable_keys)
elif isinstance(obj, list):
for i, val in enumerate(obj):
if i >= len(aggregate):
aggregate[key] = new_aggregate_obj(val)
else:
aggregate_obj(aggregate[i], val)
elif isinstance(obj, str):
# Sanity check for data
if assert_str:
assert aggregate == obj, "Data mismatch {} != {}".format(aggregate, obj)
else:
raise Exception("BUG: unable to aggregate '{}' ({})".format(type(obj), str(obj)))
if not data:
return {}
# Do the aggregation
aggregate = data[0].__class__()
for testrun in data:
aggregate_obj(aggregate, testrun)
return aggregate
class MeasurementVal(float):
"""Base class representing measurement values"""
gv_data_type = 'number'
def gv_value(self):
"""Value formatting for visualization"""
if self != self:
return "null"
else:
return self
class TimeVal(MeasurementVal):
"""Class representing time values"""
quantity = 'time'
gv_title = 'elapsed time'
gv_data_type = 'timeofday'
def hms(self):
"""Split time into hours, minutes and seconeds"""
hhh = int(abs(self) / 3600)
mmm = int((abs(self) % 3600) / 60)
sss = abs(self) % 60
return hhh, mmm, sss
def __str__(self):
if self != self:
return "nan"
hh, mm, ss = self.hms()
sign = '-' if self < 0 else ''
if hh > 0:
return '{}{:d}:{:02d}:{:02.0f}'.format(sign, hh, mm, ss)
elif mm > 0:
return '{}{:d}:{:04.1f}'.format(sign, mm, ss)
elif ss > 1:
return '{}{:.1f} s'.format(sign, ss)
else:
return '{}{:.2f} s'.format(sign, ss)
def gv_value(self):
"""Value formatting for visualization"""
if self != self:
return "null"
hh, mm, ss = self.hms()
return [hh, mm, int(ss), int(ss*1000) % 1000]
class SizeVal(MeasurementVal):
"""Class representing time values"""
quantity = 'size'
gv_title = 'size in MiB'
gv_data_type = 'number'
def __str__(self):
if self != self:
return "nan"
if abs(self) < 1024:
return '{:.1f} kiB'.format(self)
elif abs(self) < 1048576:
return '{:.2f} MiB'.format(self / 1024)
else:
return '{:.2f} GiB'.format(self / 1048576)
def gv_value(self):
"""Value formatting for visualization"""
if self != self:
return "null"
return self / 1024
def measurement_stats(meas, prefix=''):
"""Get statistics of a measurement"""
if not meas:
return {prefix + 'sample_cnt': 0,
prefix + 'mean': MeasurementVal('nan'),
prefix + 'stdev': MeasurementVal('nan'),
prefix + 'variance': MeasurementVal('nan'),
prefix + 'min': MeasurementVal('nan'),
prefix + 'max': MeasurementVal('nan'),
prefix + 'minus': MeasurementVal('nan'),
prefix + 'plus': MeasurementVal('nan')}
stats = {'name': meas['name']}
if meas['type'] == 'sysres':
val_cls = TimeVal
values = meas['values']['elapsed_time']
elif meas['type'] == 'diskusage':
val_cls = SizeVal
values = meas['values']['size']
else:
raise Exception("Unknown measurement type '{}'".format(meas['type']))
stats['val_cls'] = val_cls
stats['quantity'] = val_cls.quantity
stats[prefix + 'sample_cnt'] = len(values)
mean_val = val_cls(mean(values))
min_val = val_cls(min(values))
max_val = val_cls(max(values))
stats[prefix + 'mean'] = mean_val
if len(values) > 1:
stats[prefix + 'stdev'] = val_cls(stdev(values))
stats[prefix + 'variance'] = val_cls(variance(values))
else:
stats[prefix + 'stdev'] = float('nan')
stats[prefix + 'variance'] = float('nan')
stats[prefix + 'min'] = min_val
stats[prefix + 'max'] = max_val
stats[prefix + 'minus'] = val_cls(mean_val - min_val)
stats[prefix + 'plus'] = val_cls(max_val - mean_val)
return stats