railsbench 0.9.2 → 0.9.8
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- data/CHANGELOG +1808 -451
- data/GCPATCH +73 -0
- data/INSTALL +5 -0
- data/Manifest.txt +23 -13
- data/PROBLEMS +0 -0
- data/README +23 -7
- data/Rakefile +1 -2
- data/bin/railsbench +7 -1
- data/config/benchmarking.rb +0 -0
- data/config/benchmarks.rb +3 -2
- data/config/benchmarks.yml +0 -0
- data/images/empty.png +0 -0
- data/images/minus.png +0 -0
- data/images/plus.png +0 -0
- data/install.rb +1 -1
- data/latest_changes.txt +18 -0
- data/lib/benchmark.rb +0 -0
- data/lib/railsbench/benchmark.rb +576 -0
- data/lib/railsbench/benchmark_specs.rb +63 -63
- data/lib/railsbench/gc_info.rb +38 -3
- data/lib/railsbench/perf_info.rb +1 -1
- data/lib/railsbench/perf_utils.rb +202 -179
- data/lib/railsbench/railsbenchmark.rb +213 -55
- data/lib/railsbench/version.rb +9 -9
- data/lib/railsbench/write_headers_only.rb +15 -15
- data/postinstall.rb +0 -0
- data/ruby185gc.patch +56 -29
- data/ruby186gc.patch +564 -0
- data/ruby19gc.patch +2425 -0
- data/script/convert_raw_data_files +49 -49
- data/script/generate_benchmarks +14 -4
- data/script/perf_bench +12 -8
- data/script/perf_comp +1 -1
- data/script/perf_comp_gc +9 -1
- data/script/perf_diff +2 -2
- data/script/perf_diff_gc +2 -2
- data/script/perf_html +1 -1
- data/script/perf_plot +192 -75
- data/script/perf_plot_gc +213 -74
- data/script/perf_prof +29 -10
- data/script/perf_run +2 -2
- data/script/perf_run_gc +2 -2
- data/script/perf_table +2 -2
- data/script/perf_tex +1 -1
- data/script/perf_times +6 -6
- data/script/perf_times_gc +14 -2
- data/script/run_urls +16 -10
- data/setup.rb +0 -0
- data/test/railsbench_test.rb +0 -0
- data/test/test_helper.rb +2 -0
- metadata +77 -55
data/GCPATCH
ADDED
@@ -0,0 +1,73 @@
|
|
1
|
+
The garbage collector distributed with ruby tries to adapt memory
|
2
|
+
usage to the amount of memory used by the program, dynamically growing
|
3
|
+
or shrinking the allocated heap as it sees fit.
|
4
|
+
|
5
|
+
For long running server apps this doesn't really work very well. The
|
6
|
+
performance very much depends on the ratio heap size/program size. It
|
7
|
+
behaves somewhat erratic: adding code can actually make your program
|
8
|
+
run faster.
|
9
|
+
|
10
|
+
The supplied patch fixes this problem: it allows one to specify the
|
11
|
+
initial heap size to start with. Heap size will never drop below the
|
12
|
+
initial size. By carefully selecting the initial heap size one can
|
13
|
+
decrease startup time and increase throughput of server apps.
|
14
|
+
|
15
|
+
There are 4 patches for different versions of Ruby:
|
16
|
+
rubygc184.patch Ruby 1.8.4 (but will apply to 1.82 as well)
|
17
|
+
rubygc185.patch Ruby 1.8.5
|
18
|
+
rubygc186.patch Ruby 1.8.6
|
19
|
+
rubygc19.patch Ruby 1.9
|
20
|
+
|
21
|
+
Heap size and other options are controlled through environment
|
22
|
+
variables:
|
23
|
+
|
24
|
+
RUBY_HEAP_MIN_SLOTS
|
25
|
+
- the initial heap size in number of slots used
|
26
|
+
|
27
|
+
RUBY_HEAP_SLOTS_INCREMENT
|
28
|
+
- how many additional slots to allocate when Ruby allocates
|
29
|
+
new heap slots
|
30
|
+
|
31
|
+
RUBY_HEAP_SLOTS_GROWTH_FACTOR
|
32
|
+
- multiplicator used to increase heap block size for the next
|
33
|
+
allocation.
|
34
|
+
|
35
|
+
RUBY_GC_MALLOC_LIMIT
|
36
|
+
- the amount of C data structures which can be allocated
|
37
|
+
without triggering a GC (if this is set too low, GC will be
|
38
|
+
started even if there are empty slots available)
|
39
|
+
|
40
|
+
RUBY_HEAP_FREE_MIN
|
41
|
+
- number of free slots that should be available after GC
|
42
|
+
if fewer slots are available, additional heap will be allocated
|
43
|
+
(Ruby >= 1.8.5 ensures the freelist has at least heapsize*0.2 entries)
|
44
|
+
|
45
|
+
The following values make the patched GC behave like the unpatched GC:
|
46
|
+
|
47
|
+
RUBY_HEAP_MIN_SLOTS=10000
|
48
|
+
RUBY_HEAP_SLOTS_INCREMENT=10000
|
49
|
+
RUBY_HEAP_SLOTS_GROWTH_FACTOR=1.8
|
50
|
+
RUBY_GC_MALLOC_LIMIT=8000000
|
51
|
+
RUBY_HEAP_FREE_MIN=4096
|
52
|
+
|
53
|
+
Try experimenting with these values. You can use perf_run_gc to find
|
54
|
+
out how many slots you need.
|
55
|
+
|
56
|
+
Memory usage of the ruby interpreter can be observed by setting
|
57
|
+
RUBY_GC_STATS=1, before you invoke any of the railsbench commands.
|
58
|
+
|
59
|
+
Per default, GC data gets written to stderr. You can change this
|
60
|
+
behavior by setting environment variable RUBY_GC_DATA_FILE.
|
61
|
+
|
62
|
+
Additionally, the Ruby module GC gets some new methods:
|
63
|
+
|
64
|
+
GC.enable_stats - enable GC statistics collection
|
65
|
+
GC.disable_stats - disable GC statistics collection
|
66
|
+
GC.clear_stats - reset GC statistics
|
67
|
+
GC.collections - number of collections since stats have been enabled
|
68
|
+
GC.time - GC time used since stats have been enabled (miro seconds)
|
69
|
+
GC.dump - dumps current heap topology to the GC log file
|
70
|
+
GC.log - log given string to the GC log file
|
71
|
+
|
72
|
+
railsbench detects whether the patch has been applied and provides
|
73
|
+
GC statistics only in this case.
|
data/INSTALL
CHANGED
@@ -36,6 +36,11 @@ base directory, in order for railsbench to work properly.
|
|
36
36
|
If you obtained railsbench as a svn checkout, add the railsbench
|
37
37
|
script directory to your search path.
|
38
38
|
|
39
|
+
For bash shells you can use railsbench command completion by adding
|
40
|
+
this line code to your .bashrc file:
|
41
|
+
|
42
|
+
complete -W "`railsbench commands`" -o default railsbench
|
43
|
+
|
39
44
|
|
40
45
|
STEP 2: prepare your application for benchmarking
|
41
46
|
-------------------------------------------------
|
data/Manifest.txt
CHANGED
@@ -1,27 +1,37 @@
|
|
1
1
|
BUGS
|
2
2
|
CHANGELOG
|
3
|
+
GCPATCH
|
3
4
|
INSTALL
|
4
|
-
install.rb
|
5
|
-
postinstall.rb
|
6
5
|
LICENSE
|
7
6
|
Manifest.txt
|
8
7
|
PROBLEMS
|
9
8
|
README
|
10
9
|
Rakefile
|
11
|
-
setup.rb
|
12
|
-
ruby184gc.patch
|
13
|
-
ruby185gc.patch
|
14
10
|
bin/railsbench
|
15
|
-
|
11
|
+
config/benchmarking.rb
|
12
|
+
config/benchmarks.rb
|
13
|
+
config/benchmarks.yml
|
14
|
+
images/empty.png
|
15
|
+
images/minus.png
|
16
|
+
images/plus.png
|
17
|
+
install.rb
|
18
|
+
latest_changes.txt
|
19
|
+
lib/benchmark.rb
|
20
|
+
lib/railsbench/benchmark.rb
|
21
|
+
lib/railsbench/benchmark_specs.rb
|
16
22
|
lib/railsbench/gc_info.rb
|
17
23
|
lib/railsbench/perf_info.rb
|
18
24
|
lib/railsbench/perf_utils.rb
|
19
|
-
lib/railsbench/write_headers_only.rb
|
20
25
|
lib/railsbench/railsbenchmark.rb
|
21
|
-
lib/railsbench/
|
22
|
-
lib/
|
23
|
-
|
26
|
+
lib/railsbench/version.rb
|
27
|
+
lib/railsbench/write_headers_only.rb
|
28
|
+
postinstall.rb
|
29
|
+
ruby184gc.patch
|
30
|
+
ruby185gc.patch
|
31
|
+
ruby186gc.patch
|
32
|
+
ruby19gc.patch
|
24
33
|
script/convert_raw_data_files
|
34
|
+
script/generate_benchmarks
|
25
35
|
script/perf_bench
|
26
36
|
script/perf_comp
|
27
37
|
script/perf_comp_gc
|
@@ -38,6 +48,6 @@ script/perf_tex
|
|
38
48
|
script/perf_times
|
39
49
|
script/perf_times_gc
|
40
50
|
script/run_urls
|
41
|
-
|
42
|
-
|
43
|
-
|
51
|
+
setup.rb
|
52
|
+
test/railsbench_test.rb
|
53
|
+
test/test_helper.rb
|
data/PROBLEMS
CHANGED
File without changes
|
data/README
CHANGED
@@ -10,7 +10,7 @@ from the various scripts (and some won't run at all without the
|
|
10
10
|
patch).
|
11
11
|
|
12
12
|
This software was written and conceived by Stefan Kaes. The author can
|
13
|
-
be reached via email: <skaes@
|
13
|
+
be reached via email: <skaes@railsexpress.de>. Please send comments, bug
|
14
14
|
reports, patches or feature requests to this address.
|
15
15
|
|
16
16
|
|
@@ -21,7 +21,7 @@ FILES
|
|
21
21
|
switches:
|
22
22
|
-gcXXX : perform gc after XXX requests
|
23
23
|
-log[=level] : turn on rails logging (at given level)
|
24
|
-
-
|
24
|
+
-nocache : turn off rails caching
|
25
25
|
-path : print $: after loading rails and exit
|
26
26
|
|
27
27
|
config/bechmarks.yml
|
@@ -83,7 +83,7 @@ FILES
|
|
83
83
|
perf_run_gc n [ option-string [ config-name ] ]
|
84
84
|
- run a given benchmark, store performance data in a file
|
85
85
|
in directory $RAILS_PERF_DATA and print results
|
86
|
-
- requires Ruby GC patch
|
86
|
+
- requires Ruby GC patch (or Ruby Enterprise Edition)
|
87
87
|
|
88
88
|
perf_times_gc file
|
89
89
|
- analyse and print garbage collection statistics, which you
|
@@ -103,20 +103,36 @@ FILES
|
|
103
103
|
store profile data in a HTML file in directory $RAILS_PERF_DATA
|
104
104
|
file name is computed from date and benchmark name as described above
|
105
105
|
but has a .html extension instead of .txt
|
106
|
+
- a number of options to steer ruby-prof are available:
|
107
|
+
-ruby_prof=number[/number]
|
108
|
+
sets threshold and min_percent for ruby-prof (defaults to 0.1/1)
|
109
|
+
-profile_type=stack|grind|flat|graph|multi
|
110
|
+
selects the profile format (defaults to stack)
|
111
|
+
-measure_mode=process_time|wall_time|cpu_time|allocations|memory
|
112
|
+
selects what to measure (default to wall_time)
|
106
113
|
|
107
114
|
perf_plot [ options ] file1 file2 ...
|
108
|
-
- plot performance data from raw performance files using gruff
|
115
|
+
- plot performance data from raw performance files using gruff or gnuplot
|
116
|
+
see source for options
|
117
|
+
|
118
|
+
perf_plot_gc [ options ] file1 file2 ...
|
119
|
+
- plot data points from GC performance data stored in raw GC log files using gruff or gnuplot
|
120
|
+
this basically shows object type distribution across garbage collections
|
109
121
|
see source for options
|
110
122
|
|
111
123
|
perf_table [ options ] file1 file2 ...
|
112
124
|
- produces a tabular overview of perf data from given raw data files
|
113
125
|
see source for options
|
114
126
|
|
127
|
+
analyze_heap_dump file
|
128
|
+
- produces a html representation of a ruby heap dump.
|
129
|
+
useful for finding memory leaks.
|
115
130
|
|
116
131
|
ENVIRONMENT
|
117
132
|
|
118
133
|
RAILS_ROOT
|
119
|
-
-
|
134
|
+
- can be set to point to your rails app. if not set, railsbench can only
|
135
|
+
be called from the top level directory of your rails app
|
120
136
|
|
121
137
|
RAILS_PERF_DATA
|
122
138
|
- performance data sets will be stored into this directory
|
@@ -167,7 +183,7 @@ USAGE
|
|
167
183
|
|
168
184
|
will store benchmark data in file
|
169
185
|
|
170
|
-
$RAILS_PERF_DATA
|
186
|
+
$RAILS_PERF_DATA/<current-date>.index.mail.txt.
|
171
187
|
|
172
188
|
You can get nicely formatted output of benchmark data by running
|
173
189
|
|
@@ -306,7 +322,7 @@ CONFIGURATION
|
|
306
322
|
for the following benchmarks:
|
307
323
|
|
308
324
|
a) an entry for each named route
|
309
|
-
b) an entry for each route generated
|
325
|
+
b) an entry for each route generated by unnamed route definitions
|
310
326
|
c) an entry for each controller, combining all actions (named xyz_controller)
|
311
327
|
d) an entry combining all controllers (all_controllers), combining all
|
312
328
|
benchmarks generated by step c.
|
data/Rakefile
CHANGED
@@ -12,14 +12,13 @@ include FileUtils
|
|
12
12
|
require File.join(File.dirname(__FILE__), 'lib', 'railsbench', 'version')
|
13
13
|
|
14
14
|
AUTHOR = "Stefan Kaes" # can also be an array of Authors
|
15
|
-
EMAIL = "skaes@
|
15
|
+
EMAIL = "skaes@railsexpress.de"
|
16
16
|
DESCRIPTION = "rails benchmarking tools"
|
17
17
|
GEM_NAME = "railsbench" # what ppl will type to install your gem
|
18
18
|
RUBYFORGE_PROJECT = "railsbench" # The unix name for your project
|
19
19
|
HOMEPATH = "http://#{RUBYFORGE_PROJECT}.rubyforge.org"
|
20
20
|
RELEASE_TYPES = %w(gem tar zip) # can use: gem, tar, zip
|
21
21
|
|
22
|
-
|
23
22
|
NAME = "railsbench"
|
24
23
|
REV = nil # UNCOMMENT IF REQUIRED: File.read(".svn/entries")[/committed-rev="(d+)"/, 1] rescue nil
|
25
24
|
VERS = (ENV['VERSION'] ||= (Railsbench::VERSION::STRING + (REV ? ".#{REV}" : "")))
|
data/bin/railsbench
CHANGED
@@ -1,8 +1,10 @@
|
|
1
1
|
#!/usr/bin/env ruby
|
2
2
|
|
3
3
|
RAILSBENCH_CMDS = %w(
|
4
|
+
analyze_heap_dump
|
4
5
|
base
|
5
6
|
generate_benchmarks
|
7
|
+
commands
|
6
8
|
convert_raw_data_files
|
7
9
|
help
|
8
10
|
install
|
@@ -38,6 +40,8 @@ end
|
|
38
40
|
RAILSBENCH_BASE = File.expand_path(File.dirname(__FILE__) + '/..')
|
39
41
|
|
40
42
|
case real_cmd
|
43
|
+
when 'commands'
|
44
|
+
puts RAILSBENCH_CMDS.keys.uniq.sort
|
41
45
|
when 'base'
|
42
46
|
puts "railsbench is installed in: #{RAILSBENCH_BASE}"
|
43
47
|
exit
|
@@ -49,12 +53,14 @@ when 'path'
|
|
49
53
|
exit
|
50
54
|
when 'install', 'postinstall'
|
51
55
|
load "#{RAILSBENCH_BASE}/#{real_cmd}.rb"
|
56
|
+
when 'analyze_heap_dump', /plot/
|
57
|
+
load "#{RAILSBENCH_BASE}/script/#{real_cmd}"
|
52
58
|
else
|
53
59
|
unless ENV['RAILS_ROOT']
|
54
60
|
if File.exists? 'config/environment.rb'
|
55
61
|
ENV['RAILS_ROOT'] = File.expand_path '.'
|
56
62
|
else
|
57
|
-
$stderr.puts "railsbench: RAILS_ROOT not set"
|
63
|
+
$stderr.puts "railsbench: RAILS_ROOT not set and could not be configured automatically"
|
58
64
|
exit 1
|
59
65
|
end
|
60
66
|
end
|
data/config/benchmarking.rb
CHANGED
File without changes
|
data/config/benchmarks.rb
CHANGED
@@ -13,8 +13,9 @@ RAILS_BENCHMARKER = RailsBenchmark.new
|
|
13
13
|
# especially if you use page caching.
|
14
14
|
# RAILS_BENCHMARKER.relative_url_root = '/blog'
|
15
15
|
|
16
|
-
#
|
17
|
-
#
|
16
|
+
# Customize the code below if your benchmark needs session data
|
17
|
+
# and/or your application specifies a custom session key.
|
18
18
|
|
19
19
|
# require 'user'
|
20
20
|
# RAILS_BENCHMARKER.session_data = {'account' => User.find_first("name='stefan'")}
|
21
|
+
# RAILS_BENCHMARKER.session_key = "my_secret_cookie_name"
|
data/config/benchmarks.yml
CHANGED
File without changes
|
data/images/empty.png
ADDED
Binary file
|
data/images/minus.png
ADDED
Binary file
|
data/images/plus.png
ADDED
Binary file
|
data/install.rb
CHANGED
@@ -53,7 +53,7 @@ __END__
|
|
53
53
|
### mode:ruby ***
|
54
54
|
### End: ***
|
55
55
|
|
56
|
-
# Copyright (C) 2005
|
56
|
+
# Copyright (C) 2005-2008 Stefan Kaes
|
57
57
|
#
|
58
58
|
# This program is free software; you can redistribute it and/or modify
|
59
59
|
# it under the terms of the GNU General Public License as published by
|
data/latest_changes.txt
ADDED
@@ -0,0 +1,18 @@
|
|
1
|
+
* added support for generating call trees using ruby-prof
|
2
|
+
* show length of freelist in gc plots
|
3
|
+
* make it possible to plot very large gc logs by artificially reducing the data set size
|
4
|
+
* changed logger handling to no longer set logger to nil as most apps don't test for logger being nil
|
5
|
+
* rails 2.2 compatibility
|
6
|
+
* support gnuplot as a plotting engine
|
7
|
+
* allow post and query data to be stored in raw form in benchmarks file
|
8
|
+
* railsbench command completion: complete -W "`railsbench commands`" -o default railsbench
|
9
|
+
* individual scripts now try to autoconfigure RAILS_ROOT too
|
10
|
+
* make it possible to add cookie data in benchmarks
|
11
|
+
* support for memory leak checking under OS X
|
12
|
+
* print detailed comparison of total garbage allocated per object type
|
13
|
+
* provide an interface to summed garbage per object type via :garbage_totals
|
14
|
+
* railsbench now requires ruby-prof version >= 0.6
|
15
|
+
* support for XMLHttpRequest. patch by Vít Ondruch
|
16
|
+
* enable specification of session key. patch by Vít Ondruch
|
17
|
+
* exit benchmarking when an action raises an exception
|
18
|
+
* new GC patches for 1.8.6 and 1.9
|
data/lib/benchmark.rb
CHANGED
File without changes
|
@@ -0,0 +1,576 @@
|
|
1
|
+
=begin
|
2
|
+
#
|
3
|
+
# benchmark.rb - a performance benchmarking library
|
4
|
+
#
|
5
|
+
# $Id$
|
6
|
+
#
|
7
|
+
# Created by Gotoken (gotoken@notwork.org).
|
8
|
+
#
|
9
|
+
# Documentation by Gotoken (original RD), Lyle Johnson (RDoc conversion), and
|
10
|
+
# Gavin Sinclair (editing).
|
11
|
+
#
|
12
|
+
=end
|
13
|
+
|
14
|
+
# == Overview
|
15
|
+
#
|
16
|
+
# The Benchmark module provides methods for benchmarking Ruby code, giving
|
17
|
+
# detailed reports on the time taken for each task.
|
18
|
+
#
|
19
|
+
|
20
|
+
# The Benchmark module provides methods to measure and report the time
|
21
|
+
# used to execute Ruby code.
|
22
|
+
#
|
23
|
+
# * Measure the time to construct the string given by the expression
|
24
|
+
# <tt>"a"*1_000_000</tt>:
|
25
|
+
#
|
26
|
+
# require 'benchmark'
|
27
|
+
#
|
28
|
+
# puts Benchmark.measure { "a"*1_000_000 }
|
29
|
+
#
|
30
|
+
# On my machine (FreeBSD 3.2 on P5, 100MHz) this generates:
|
31
|
+
#
|
32
|
+
# 1.166667 0.050000 1.216667 ( 0.571355)
|
33
|
+
#
|
34
|
+
# This report shows the user CPU time, system CPU time, the sum of
|
35
|
+
# the user and system CPU times, and the elapsed real time. The unit
|
36
|
+
# of time is seconds.
|
37
|
+
#
|
38
|
+
# * Do some experiments sequentially using the #bm method:
|
39
|
+
#
|
40
|
+
# require 'benchmark'
|
41
|
+
#
|
42
|
+
# n = 50000
|
43
|
+
# Benchmark.bm do |x|
|
44
|
+
# x.report { for i in 1..n; a = "1"; end }
|
45
|
+
# x.report { n.times do ; a = "1"; end }
|
46
|
+
# x.report { 1.upto(n) do ; a = "1"; end }
|
47
|
+
# end
|
48
|
+
#
|
49
|
+
# The result:
|
50
|
+
#
|
51
|
+
# user system total real
|
52
|
+
# 1.033333 0.016667 1.016667 ( 0.492106)
|
53
|
+
# 1.483333 0.000000 1.483333 ( 0.694605)
|
54
|
+
# 1.516667 0.000000 1.516667 ( 0.711077)
|
55
|
+
#
|
56
|
+
# * Continuing the previous example, put a label in each report:
|
57
|
+
#
|
58
|
+
# require 'benchmark'
|
59
|
+
#
|
60
|
+
# n = 50000
|
61
|
+
# Benchmark.bm(7) do |x|
|
62
|
+
# x.report("for:") { for i in 1..n; a = "1"; end }
|
63
|
+
# x.report("times:") { n.times do ; a = "1"; end }
|
64
|
+
# x.report("upto:") { 1.upto(n) do ; a = "1"; end }
|
65
|
+
# end
|
66
|
+
#
|
67
|
+
# The result:
|
68
|
+
#
|
69
|
+
# user system total real
|
70
|
+
# for: 1.050000 0.000000 1.050000 ( 0.503462)
|
71
|
+
# times: 1.533333 0.016667 1.550000 ( 0.735473)
|
72
|
+
# upto: 1.500000 0.016667 1.516667 ( 0.711239)
|
73
|
+
#
|
74
|
+
#
|
75
|
+
# * The times for some benchmarks depend on the order in which items
|
76
|
+
# are run. These differences are due to the cost of memory
|
77
|
+
# allocation and garbage collection. To avoid these discrepancies,
|
78
|
+
# the #bmbm method is provided. For example, to compare ways to
|
79
|
+
# sort an array of floats:
|
80
|
+
#
|
81
|
+
# require 'benchmark'
|
82
|
+
#
|
83
|
+
# array = (1..1000000).map { rand }
|
84
|
+
#
|
85
|
+
# Benchmark.bmbm do |x|
|
86
|
+
# x.report("sort!") { array.dup.sort! }
|
87
|
+
# x.report("sort") { array.dup.sort }
|
88
|
+
# end
|
89
|
+
#
|
90
|
+
# The result:
|
91
|
+
#
|
92
|
+
# Rehearsal -----------------------------------------
|
93
|
+
# sort! 11.928000 0.010000 11.938000 ( 12.756000)
|
94
|
+
# sort 13.048000 0.020000 13.068000 ( 13.857000)
|
95
|
+
# ------------------------------- total: 25.006000sec
|
96
|
+
#
|
97
|
+
# user system total real
|
98
|
+
# sort! 12.959000 0.010000 12.969000 ( 13.793000)
|
99
|
+
# sort 12.007000 0.000000 12.007000 ( 12.791000)
|
100
|
+
#
|
101
|
+
#
|
102
|
+
# * Report statistics of sequential experiments with unique labels,
|
103
|
+
# using the #benchmark method:
|
104
|
+
#
|
105
|
+
# require 'benchmark'
|
106
|
+
#
|
107
|
+
# n = 50000
|
108
|
+
# Benchmark.benchmark(" "*7 + CAPTION, 7, FMTSTR, ">total:", ">avg:") do |x|
|
109
|
+
# tf = x.report("for:") { for i in 1..n; a = "1"; end }
|
110
|
+
# tt = x.report("times:") { n.times do ; a = "1"; end }
|
111
|
+
# tu = x.report("upto:") { 1.upto(n) do ; a = "1"; end }
|
112
|
+
# [tf+tt+tu, (tf+tt+tu)/3]
|
113
|
+
# end
|
114
|
+
#
|
115
|
+
# The result:
|
116
|
+
#
|
117
|
+
# user system total real
|
118
|
+
# for: 1.016667 0.016667 1.033333 ( 0.485749)
|
119
|
+
# times: 1.450000 0.016667 1.466667 ( 0.681367)
|
120
|
+
# upto: 1.533333 0.000000 1.533333 ( 0.722166)
|
121
|
+
# >total: 4.000000 0.033333 4.033333 ( 1.889282)
|
122
|
+
# >avg: 1.333333 0.011111 1.344444 ( 0.629761)
|
123
|
+
|
124
|
+
module Benchmark
|
125
|
+
|
126
|
+
BENCHMARK_VERSION = "2002-04-25" #:nodoc"
|
127
|
+
|
128
|
+
OUTPUT = STDOUT unless defined?(OUTPUT)
|
129
|
+
SYNC = true unless defined?(SYNC)
|
130
|
+
|
131
|
+
def Benchmark::times() # :nodoc:
|
132
|
+
Process::times()
|
133
|
+
end
|
134
|
+
|
135
|
+
|
136
|
+
# Invokes the block with a <tt>Benchmark::Report</tt> object, which
|
137
|
+
# may be used to collect and report on the results of individual
|
138
|
+
# benchmark tests. Reserves <i>label_width</i> leading spaces for
|
139
|
+
# labels on each line. Prints _caption_ at the top of the
|
140
|
+
# report, and uses _fmt_ to format each line.
|
141
|
+
# If the block returns an array of
|
142
|
+
# <tt>Benchmark::Tms</tt> objects, these will be used to format
|
143
|
+
# additional lines of output. If _label_ parameters are
|
144
|
+
# given, these are used to label these extra lines.
|
145
|
+
#
|
146
|
+
# _Note_: Other methods provide a simpler interface to this one, and are
|
147
|
+
# suitable for nearly all benchmarking requirements. See the examples in
|
148
|
+
# Benchmark, and the #bm and #bmbm methods.
|
149
|
+
#
|
150
|
+
# Example:
|
151
|
+
#
|
152
|
+
# require 'benchmark'
|
153
|
+
# include Benchmark # we need the CAPTION and FMTSTR constants
|
154
|
+
#
|
155
|
+
# n = 50000
|
156
|
+
# Benchmark.benchmark(" "*7 + CAPTION, 7, FMTSTR, ">total:", ">avg:") do |x|
|
157
|
+
# tf = x.report("for:") { for i in 1..n; a = "1"; end }
|
158
|
+
# tt = x.report("times:") { n.times do ; a = "1"; end }
|
159
|
+
# tu = x.report("upto:") { 1.upto(n) do ; a = "1"; end }
|
160
|
+
# [tf+tt+tu, (tf+tt+tu)/3]
|
161
|
+
# end
|
162
|
+
#
|
163
|
+
# <i>Generates:</i>
|
164
|
+
#
|
165
|
+
# user system total real
|
166
|
+
# for: 1.016667 0.016667 1.033333 ( 0.485749)
|
167
|
+
# times: 1.450000 0.016667 1.466667 ( 0.681367)
|
168
|
+
# upto: 1.533333 0.000000 1.533333 ( 0.722166)
|
169
|
+
# >total: 4.000000 0.033333 4.033333 ( 1.889282)
|
170
|
+
# >avg: 1.333333 0.011111 1.344444 ( 0.629761)
|
171
|
+
#
|
172
|
+
|
173
|
+
def benchmark(caption = "", label_width = nil, fmtstr = nil, *labels) # :yield: report
|
174
|
+
if SYNC
|
175
|
+
sync = OUTPUT.sync
|
176
|
+
OUTPUT.sync = true
|
177
|
+
end
|
178
|
+
label_width ||= 0
|
179
|
+
fmtstr ||= FMTSTR
|
180
|
+
raise ArgumentError, "no block" unless iterator?
|
181
|
+
OUTPUT.print caption
|
182
|
+
results = yield(Report.new(label_width, fmtstr))
|
183
|
+
Array === results and results.grep(Tms).each {|t|
|
184
|
+
OUTPUT.print((labels.shift || t.label || "").ljust(label_width),
|
185
|
+
t.format(fmtstr))
|
186
|
+
}
|
187
|
+
OUTPUT.sync = sync if SYNC
|
188
|
+
end
|
189
|
+
|
190
|
+
|
191
|
+
# A simple interface to the #benchmark method, #bm is generates sequential reports
|
192
|
+
# with labels. The parameters have the same meaning as for #benchmark.
|
193
|
+
#
|
194
|
+
# require 'benchmark'
|
195
|
+
#
|
196
|
+
# n = 50000
|
197
|
+
# Benchmark.bm(7) do |x|
|
198
|
+
# x.report("for:") { for i in 1..n; a = "1"; end }
|
199
|
+
# x.report("times:") { n.times do ; a = "1"; end }
|
200
|
+
# x.report("upto:") { 1.upto(n) do ; a = "1"; end }
|
201
|
+
# end
|
202
|
+
#
|
203
|
+
# <i>Generates:</i>
|
204
|
+
#
|
205
|
+
# user system total real
|
206
|
+
# for: 1.050000 0.000000 1.050000 ( 0.503462)
|
207
|
+
# times: 1.533333 0.016667 1.550000 ( 0.735473)
|
208
|
+
# upto: 1.500000 0.016667 1.516667 ( 0.711239)
|
209
|
+
#
|
210
|
+
|
211
|
+
def bm(label_width = 0, *labels, &blk) # :yield: report
|
212
|
+
benchmark(" "*label_width + CAPTION, label_width, FMTSTR, *labels, &blk)
|
213
|
+
end
|
214
|
+
|
215
|
+
|
216
|
+
# Sometimes benchmark results are skewed because code executed
|
217
|
+
# earlier encounters different garbage collection overheads than
|
218
|
+
# that run later. #bmbm attempts to minimize this effect by running
|
219
|
+
# the tests twice, the first time as a rehearsal in order to get the
|
220
|
+
# runtime environment stable, the second time for
|
221
|
+
# real. <tt>GC.start</tt> is executed before the start of each of
|
222
|
+
# the real timings; the cost of this is not included in the
|
223
|
+
# timings. In reality, though, there's only so much that #bmbm can
|
224
|
+
# do, and the results are not guaranteed to be isolated from garbage
|
225
|
+
# collection and other effects.
|
226
|
+
#
|
227
|
+
# Because #bmbm takes two passes through the tests, it can
|
228
|
+
# calculate the required label width.
|
229
|
+
#
|
230
|
+
# require 'benchmark'
|
231
|
+
#
|
232
|
+
# array = (1..1000000).map { rand }
|
233
|
+
#
|
234
|
+
# Benchmark.bmbm do |x|
|
235
|
+
# x.report("sort!") { array.dup.sort! }
|
236
|
+
# x.report("sort") { array.dup.sort }
|
237
|
+
# end
|
238
|
+
#
|
239
|
+
# <i>Generates:</i>
|
240
|
+
#
|
241
|
+
# Rehearsal -----------------------------------------
|
242
|
+
# sort! 11.928000 0.010000 11.938000 ( 12.756000)
|
243
|
+
# sort 13.048000 0.020000 13.068000 ( 13.857000)
|
244
|
+
# ------------------------------- total: 25.006000sec
|
245
|
+
#
|
246
|
+
# user system total real
|
247
|
+
# sort! 12.959000 0.010000 12.969000 ( 13.793000)
|
248
|
+
# sort 12.007000 0.000000 12.007000 ( 12.791000)
|
249
|
+
#
|
250
|
+
# #bmbm yields a Benchmark::Job object and returns an array of
|
251
|
+
# Benchmark::Tms objects.
|
252
|
+
#
|
253
|
+
def bmbm(width = 0, &blk) # :yield: job
|
254
|
+
job = Job.new(width)
|
255
|
+
yield(job)
|
256
|
+
width = job.width
|
257
|
+
if SYNC
|
258
|
+
sync = OUTPUT.sync
|
259
|
+
OUTPUT.sync = true
|
260
|
+
end
|
261
|
+
|
262
|
+
# rehearsal
|
263
|
+
OUTPUT.print "Rehearsal "
|
264
|
+
puts '-'*(width+CAPTION.length - "Rehearsal ".length)
|
265
|
+
list = []
|
266
|
+
job.list.each{|label,item|
|
267
|
+
OUTPUT.print(label.ljust(width))
|
268
|
+
res = Benchmark::measure(&item)
|
269
|
+
OUTPUT.print res.format()
|
270
|
+
list.push res
|
271
|
+
}
|
272
|
+
sum = Tms.new; list.each{|i| sum += i}
|
273
|
+
ets = sum.format("total: %tsec")
|
274
|
+
OUTPUT.printf("%s %s\n\n",
|
275
|
+
"-"*(width+CAPTION.length-ets.length-1), ets)
|
276
|
+
|
277
|
+
# take
|
278
|
+
OUTPUT.print ' '*width, CAPTION
|
279
|
+
list = []
|
280
|
+
ary = []
|
281
|
+
job.list.each{|label,item|
|
282
|
+
GC::start
|
283
|
+
OUTPUT.print label.ljust(width)
|
284
|
+
res = Benchmark::measure(&item)
|
285
|
+
OUTPUT.print res.format()
|
286
|
+
ary.push res
|
287
|
+
list.push [label, res]
|
288
|
+
}
|
289
|
+
|
290
|
+
OUTPUT.sync = sync if SYNC
|
291
|
+
ary
|
292
|
+
end
|
293
|
+
|
294
|
+
#
|
295
|
+
# Returns the time used to execute the given block as a
|
296
|
+
# Benchmark::Tms object.
|
297
|
+
#
|
298
|
+
def measure(label = "") # :yield:
|
299
|
+
t0, r0 = Benchmark.times, Time.now
|
300
|
+
yield
|
301
|
+
t1, r1 = Benchmark.times, Time.now
|
302
|
+
Benchmark::Tms.new(t1.utime - t0.utime,
|
303
|
+
t1.stime - t0.stime,
|
304
|
+
t1.cutime - t0.cutime,
|
305
|
+
t1.cstime - t0.cstime,
|
306
|
+
r1.to_f - r0.to_f,
|
307
|
+
label)
|
308
|
+
end
|
309
|
+
|
310
|
+
#
|
311
|
+
# Returns the elapsed real time used to execute the given block.
|
312
|
+
#
|
313
|
+
def realtime(&blk) # :yield:
|
314
|
+
Benchmark::measure(&blk).real
|
315
|
+
end
|
316
|
+
|
317
|
+
|
318
|
+
|
319
|
+
#
|
320
|
+
# A Job is a sequence of labelled blocks to be processed by the
|
321
|
+
# Benchmark.bmbm method. It is of little direct interest to the user.
|
322
|
+
#
|
323
|
+
class Job # :nodoc:
|
324
|
+
#
|
325
|
+
# Returns an initialized Job instance.
|
326
|
+
# Usually, one doesn't call this method directly, as new
|
327
|
+
# Job objects are created by the #bmbm method.
|
328
|
+
# _width_ is a initial value for the label offset used in formatting;
|
329
|
+
# the #bmbm method passes its _width_ argument to this constructor.
|
330
|
+
#
|
331
|
+
def initialize(width)
|
332
|
+
@width = width
|
333
|
+
@list = []
|
334
|
+
end
|
335
|
+
|
336
|
+
#
|
337
|
+
# Registers the given label and block pair in the job list.
|
338
|
+
#
|
339
|
+
def item(label = "", &blk) # :yield:
|
340
|
+
raise ArgmentError, "no block" unless block_given?
|
341
|
+
label.concat ' '
|
342
|
+
w = label.length
|
343
|
+
@width = w if @width < w
|
344
|
+
@list.push [label, blk]
|
345
|
+
self
|
346
|
+
end
|
347
|
+
|
348
|
+
alias report item
|
349
|
+
|
350
|
+
# An array of 2-element arrays, consisting of label and block pairs.
|
351
|
+
attr_reader :list
|
352
|
+
|
353
|
+
# Length of the widest label in the #list, plus one.
|
354
|
+
attr_reader :width
|
355
|
+
end
|
356
|
+
|
357
|
+
module_function :benchmark, :measure, :realtime, :bm, :bmbm
|
358
|
+
|
359
|
+
|
360
|
+
|
361
|
+
#
|
362
|
+
# This class is used by the Benchmark.benchmark and Benchmark.bm methods.
|
363
|
+
# It is of little direct interest to the user.
|
364
|
+
#
|
365
|
+
class Report # :nodoc:
|
366
|
+
#
|
367
|
+
# Returns an initialized Report instance.
|
368
|
+
# Usually, one doesn't call this method directly, as new
|
369
|
+
# Report objects are created by the #benchmark and #bm methods.
|
370
|
+
# _width_ and _fmtstr_ are the label offset and
|
371
|
+
# format string used by Tms#format.
|
372
|
+
#
|
373
|
+
def initialize(width = 0, fmtstr = nil)
|
374
|
+
@width, @fmtstr = width, fmtstr
|
375
|
+
end
|
376
|
+
|
377
|
+
#
|
378
|
+
# Prints the _label_ and measured time for the block,
|
379
|
+
# formatted by _fmt_. See Tms#format for the
|
380
|
+
# formatting rules.
|
381
|
+
#
|
382
|
+
def item(label = "", *fmt, &blk) # :yield:
|
383
|
+
OUTPUT.print label.ljust(@width)
|
384
|
+
res = Benchmark::measure(&blk)
|
385
|
+
OUTPUT.print res.format(@fmtstr, *fmt)
|
386
|
+
res
|
387
|
+
end
|
388
|
+
|
389
|
+
alias report item
|
390
|
+
end
|
391
|
+
|
392
|
+
|
393
|
+
|
394
|
+
#
|
395
|
+
# A data object, representing the times associated with a benchmark
|
396
|
+
# measurement.
|
397
|
+
#
|
398
|
+
class Tms
|
399
|
+
CAPTION = " user system total real\n"
|
400
|
+
FMTSTR = "%10.6u %10.6y %10.6t %10.6r\n"
|
401
|
+
|
402
|
+
# User CPU time
|
403
|
+
attr_reader :utime
|
404
|
+
|
405
|
+
# System CPU time
|
406
|
+
attr_reader :stime
|
407
|
+
|
408
|
+
# User CPU time of children
|
409
|
+
attr_reader :cutime
|
410
|
+
|
411
|
+
# System CPU time of children
|
412
|
+
attr_reader :cstime
|
413
|
+
|
414
|
+
# Elapsed real time
|
415
|
+
attr_reader :real
|
416
|
+
|
417
|
+
# Total time, that is _utime_ + _stime_ + _cutime_ + _cstime_
|
418
|
+
attr_reader :total
|
419
|
+
|
420
|
+
# Label
|
421
|
+
attr_reader :label
|
422
|
+
|
423
|
+
#
|
424
|
+
# Returns an initialized Tms object which has
|
425
|
+
# _u_ as the user CPU time, _s_ as the system CPU time,
|
426
|
+
# _cu_ as the children's user CPU time, _cs_ as the children's
|
427
|
+
# system CPU time, _real_ as the elapsed real time and _l_
|
428
|
+
# as the label.
|
429
|
+
#
|
430
|
+
def initialize(u = 0.0, s = 0.0, cu = 0.0, cs = 0.0, real = 0.0, l = nil)
|
431
|
+
@utime, @stime, @cutime, @cstime, @real, @label = u, s, cu, cs, real, l
|
432
|
+
@total = @utime + @stime + @cutime + @cstime
|
433
|
+
end
|
434
|
+
|
435
|
+
#
|
436
|
+
# Returns a new Tms object whose times are the sum of the times for this
|
437
|
+
# Tms object, plus the time required to execute the code block (_blk_).
|
438
|
+
#
|
439
|
+
def add(&blk) # :yield:
|
440
|
+
self + Benchmark::measure(&blk)
|
441
|
+
end
|
442
|
+
|
443
|
+
#
|
444
|
+
# An in-place version of #add.
|
445
|
+
#
|
446
|
+
def add!
|
447
|
+
t = Benchmark::measure(&blk)
|
448
|
+
@utime = utime + t.utime
|
449
|
+
@stime = stime + t.stime
|
450
|
+
@cutime = cutime + t.cutime
|
451
|
+
@cstime = cstime + t.cstime
|
452
|
+
@real = real + t.real
|
453
|
+
self
|
454
|
+
end
|
455
|
+
|
456
|
+
#
|
457
|
+
# Returns a new Tms object obtained by memberwise summation
|
458
|
+
# of the individual times for this Tms object with those of the other
|
459
|
+
# Tms object.
|
460
|
+
# This method and #/() are useful for taking statistics.
|
461
|
+
#
|
462
|
+
def +(other); memberwise(:+, other) end
|
463
|
+
|
464
|
+
#
|
465
|
+
# Returns a new Tms object obtained by memberwise subtraction
|
466
|
+
# of the individual times for the other Tms object from those of this
|
467
|
+
# Tms object.
|
468
|
+
#
|
469
|
+
def -(other); memberwise(:-, other) end
|
470
|
+
|
471
|
+
#
|
472
|
+
# Returns a new Tms object obtained by memberwise multiplication
|
473
|
+
# of the individual times for this Tms object by _x_.
|
474
|
+
#
|
475
|
+
def *(x); memberwise(:*, x) end
|
476
|
+
|
477
|
+
#
|
478
|
+
# Returns a new Tms object obtained by memberwise division
|
479
|
+
# of the individual times for this Tms object by _x_.
|
480
|
+
# This method and #+() are useful for taking statistics.
|
481
|
+
#
|
482
|
+
def /(x); memberwise(:/, x) end
|
483
|
+
|
484
|
+
#
|
485
|
+
# Returns the contents of this Tms object as
|
486
|
+
# a formatted string, according to a format string
|
487
|
+
# like that passed to Kernel.format. In addition, #format
|
488
|
+
# accepts the following extensions:
|
489
|
+
#
|
490
|
+
# <tt>%u</tt>:: Replaced by the user CPU time, as reported by Tms#utime.
|
491
|
+
# <tt>%y</tt>:: Replaced by the system CPU time, as reported by #stime (Mnemonic: y of "s*y*stem")
|
492
|
+
# <tt>%U</tt>:: Replaced by the children's user CPU time, as reported by Tms#cutime
|
493
|
+
# <tt>%Y</tt>:: Replaced by the children's system CPU time, as reported by Tms#cstime
|
494
|
+
# <tt>%t</tt>:: Replaced by the total CPU time, as reported by Tms#total
|
495
|
+
# <tt>%r</tt>:: Replaced by the elapsed real time, as reported by Tms#real
|
496
|
+
# <tt>%n</tt>:: Replaced by the label string, as reported by Tms#label (Mnemonic: n of "*n*ame")
|
497
|
+
#
|
498
|
+
# If _fmtstr_ is not given, FMTSTR is used as default value, detailing the
|
499
|
+
# user, system and real elapsed time.
|
500
|
+
#
|
501
|
+
def format(arg0 = nil, *args)
|
502
|
+
fmtstr = (arg0 || FMTSTR).dup
|
503
|
+
fmtstr.gsub!(/(%[-+\.\d]*)n/){"#{$1}s" % label}
|
504
|
+
fmtstr.gsub!(/(%[-+\.\d]*)u/){"#{$1}f" % utime}
|
505
|
+
fmtstr.gsub!(/(%[-+\.\d]*)y/){"#{$1}f" % stime}
|
506
|
+
fmtstr.gsub!(/(%[-+\.\d]*)U/){"#{$1}f" % cutime}
|
507
|
+
fmtstr.gsub!(/(%[-+\.\d]*)Y/){"#{$1}f" % cstime}
|
508
|
+
fmtstr.gsub!(/(%[-+\.\d]*)t/){"#{$1}f" % total}
|
509
|
+
fmtstr.gsub!(/(%[-+\.\d]*)r/){"(#{$1}f)" % real}
|
510
|
+
arg0 ? Kernel::format(fmtstr, *args) : fmtstr
|
511
|
+
end
|
512
|
+
|
513
|
+
#
|
514
|
+
# Same as #format.
|
515
|
+
#
|
516
|
+
def to_s
|
517
|
+
format
|
518
|
+
end
|
519
|
+
|
520
|
+
#
|
521
|
+
# Returns a new 6-element array, consisting of the
|
522
|
+
# label, user CPU time, system CPU time, children's
|
523
|
+
# user CPU time, children's system CPU time and elapsed
|
524
|
+
# real time.
|
525
|
+
#
|
526
|
+
def to_a
|
527
|
+
[@label, @utime, @stime, @cutime, @cstime, @real]
|
528
|
+
end
|
529
|
+
|
530
|
+
protected
|
531
|
+
def memberwise(op, x)
|
532
|
+
case x
|
533
|
+
when Benchmark::Tms
|
534
|
+
Benchmark::Tms.new(utime.__send__(op, x.utime),
|
535
|
+
stime.__send__(op, x.stime),
|
536
|
+
cutime.__send__(op, x.cutime),
|
537
|
+
cstime.__send__(op, x.cstime),
|
538
|
+
real.__send__(op, x.real)
|
539
|
+
)
|
540
|
+
else
|
541
|
+
Benchmark::Tms.new(utime.__send__(op, x),
|
542
|
+
stime.__send__(op, x),
|
543
|
+
cutime.__send__(op, x),
|
544
|
+
cstime.__send__(op, x),
|
545
|
+
real.__send__(op, x)
|
546
|
+
)
|
547
|
+
end
|
548
|
+
end
|
549
|
+
end
|
550
|
+
|
551
|
+
# The default caption string (heading above the output times).
|
552
|
+
CAPTION = Benchmark::Tms::CAPTION
|
553
|
+
|
554
|
+
# The default format string used to display times. See also Benchmark::Tms#format.
|
555
|
+
FMTSTR = Benchmark::Tms::FMTSTR
|
556
|
+
end
|
557
|
+
|
558
|
+
if __FILE__ == $0
|
559
|
+
include Benchmark
|
560
|
+
|
561
|
+
n = ARGV[0].to_i.nonzero? || 50000
|
562
|
+
puts %Q([#{n} times iterations of `a = "1"'])
|
563
|
+
benchmark(" " + CAPTION, 7, FMTSTR) do |x|
|
564
|
+
x.report("for:") {for i in 1..n; a = "1"; end} # Benchmark::measure
|
565
|
+
x.report("times:") {n.times do ; a = "1"; end}
|
566
|
+
x.report("upto:") {1.upto(n) do ; a = "1"; end}
|
567
|
+
end
|
568
|
+
|
569
|
+
benchmark do
|
570
|
+
[
|
571
|
+
measure{for i in 1..n; a = "1"; end}, # Benchmark::measure
|
572
|
+
measure{n.times do ; a = "1"; end},
|
573
|
+
measure{1.upto(n) do ; a = "1"; end}
|
574
|
+
]
|
575
|
+
end
|
576
|
+
end
|