rdp-ruby-prof 0.7.4
Sign up to get free protection for your applications and to get access to all the features.
- data/CHANGES +202 -0
- data/LICENSE +23 -0
- data/README +445 -0
- data/Rakefile +123 -0
- data/bin/ruby-prof +207 -0
- data/examples/flat.txt +55 -0
- data/examples/graph.html +823 -0
- data/examples/graph.txt +170 -0
- data/ext/#ruby_prof.c# +1679 -0
- data/ext/Makefile +180 -0
- data/ext/extconf.rb +40 -0
- data/ext/measure_allocations.h +58 -0
- data/ext/measure_cpu_time.h +152 -0
- data/ext/measure_gc_runs.h +76 -0
- data/ext/measure_gc_time.h +57 -0
- data/ext/measure_memory.h +101 -0
- data/ext/measure_process_time.h +52 -0
- data/ext/measure_wall_time.h +53 -0
- data/ext/mingw/Rakefile +23 -0
- data/ext/mingw/build.rake +38 -0
- data/ext/ruby_prof.c +1707 -0
- data/ext/ruby_prof.e +19984 -0
- data/ext/ruby_prof.h +188 -0
- data/ext/vc/ruby_prof.sln +20 -0
- data/ext/vc/ruby_prof.vcproj +241 -0
- data/ext/version.h +4 -0
- data/lib/ruby-prof.rb +48 -0
- data/lib/ruby-prof/abstract_printer.rb +41 -0
- data/lib/ruby-prof/aggregate_call_info.rb +62 -0
- data/lib/ruby-prof/call_info.rb +47 -0
- data/lib/ruby-prof/call_tree_printer.rb +84 -0
- data/lib/ruby-prof/flat_printer.rb +79 -0
- data/lib/ruby-prof/graph_html_printer.rb +256 -0
- data/lib/ruby-prof/graph_html_printer.rb.orig +256 -0
- data/lib/ruby-prof/graph_html_printer.rb.rej +34 -0
- data/lib/ruby-prof/graph_printer.rb +164 -0
- data/lib/ruby-prof/graph_printer.rb.orig +164 -0
- data/lib/ruby-prof/method_info.rb +111 -0
- data/lib/ruby-prof/task.rb +146 -0
- data/lib/ruby-prof/test.rb +148 -0
- data/lib/unprof.rb +8 -0
- data/rails/environment/profile.rb +24 -0
- data/rails/example/example_test.rb +9 -0
- data/rails/profile_test_helper.rb +21 -0
- data/test/aggregate_test.rb +121 -0
- data/test/basic_test.rb +283 -0
- data/test/duplicate_names_test.rb +32 -0
- data/test/exceptions_test.rb +15 -0
- data/test/exclude_threads_test.rb +54 -0
- data/test/line_number_test.rb +73 -0
- data/test/measurement_test.rb +121 -0
- data/test/module_test.rb +54 -0
- data/test/no_method_class_test.rb +13 -0
- data/test/prime.rb +58 -0
- data/test/prime_test.rb +13 -0
- data/test/printers_test.rb +71 -0
- data/test/recursive_test.rb +254 -0
- data/test/singleton_test.rb +37 -0
- data/test/stack_test.rb +138 -0
- data/test/start_stop_test.rb +95 -0
- data/test/test_suite.rb +23 -0
- data/test/thread_test.rb +159 -0
- data/test/unique_call_path_test.rb +206 -0
- metadata +124 -0
data/CHANGES
ADDED
@@ -0,0 +1,202 @@
|
|
1
|
+
0.7.3 (2008-12-09)
|
2
|
+
========================
|
3
|
+
* Fixed compile error with new x86_64 code using GCC.
|
4
|
+
|
5
|
+
|
6
|
+
0.7.2 (2008-12-08)
|
7
|
+
========================
|
8
|
+
* Fixed major bug in printing child methods in graph reports.
|
9
|
+
|
10
|
+
* Fixes for supporting x86_64 machines (Diego Pettenò)
|
11
|
+
|
12
|
+
|
13
|
+
0.7.1 (2008-11-28)
|
14
|
+
========================
|
15
|
+
* Added new AggregateCallInfo class for printers to
|
16
|
+
make results easier to read. Take this call sequence
|
17
|
+
for example:
|
18
|
+
|
19
|
+
A B C
|
20
|
+
| | |
|
21
|
+
Z A A
|
22
|
+
| |
|
23
|
+
Z Z
|
24
|
+
|
25
|
+
By default, ruby-prof will show that Z was called by 3 separate
|
26
|
+
instances of A. In an IDE that is helpful but in a text report
|
27
|
+
it is not since it makes the report much harder to read.
|
28
|
+
As a result, printers now aggregate together callers (and children),
|
29
|
+
matching ruby-prof's output from versions prior to 0.7.0.
|
30
|
+
|
31
|
+
* Fixes for supporting x86_64 machines (Matt Sanford)
|
32
|
+
|
33
|
+
|
34
|
+
0.7.0 (2008-11-04)
|
35
|
+
========================
|
36
|
+
|
37
|
+
Features
|
38
|
+
--------
|
39
|
+
* Added two new methods - RubyProf.resume and RubyProf.pause.
|
40
|
+
RubyProf.resume takes an optional block, which ensures that
|
41
|
+
RubyProf.pause is called. For example:
|
42
|
+
|
43
|
+
10.times do |i|
|
44
|
+
RubyProf.resume do
|
45
|
+
# Some long process
|
46
|
+
end
|
47
|
+
end
|
48
|
+
|
49
|
+
result = RubyProf.stop
|
50
|
+
|
51
|
+
* Added support for profiling tests that use Ruby's built-in
|
52
|
+
unit test framework (ie, test derived from
|
53
|
+
Test::Unit::TestCase). To enable profiling simply add
|
54
|
+
the following line of code to your test class:
|
55
|
+
|
56
|
+
include RubyProf::Test
|
57
|
+
|
58
|
+
By default, profiling results are written to the current
|
59
|
+
processes working directory. To change this, or other
|
60
|
+
profiling options, simply modify the PROFILE_OPTIONS hash
|
61
|
+
table as needed.
|
62
|
+
|
63
|
+
* Used the new support for profiling test cases to revamp
|
64
|
+
the way that Rails profiling works. For more information
|
65
|
+
please refer to RubyProf's documentation.
|
66
|
+
|
67
|
+
* Keep track of call stack for each method to enable more
|
68
|
+
powerful profile visualizations in Integrated Development
|
69
|
+
Environments (Hin Boean, work supported by CodeGear).
|
70
|
+
|
71
|
+
* Expose measurements to Ruby (Jeremy Kemper).
|
72
|
+
|
73
|
+
* Add support for additional memory measurements modes in Ruby 1.9 (Jeremy Kemper).
|
74
|
+
|
75
|
+
* Add support for Lloyd Hilaiel's Ruby patch for measuring total heap size.
|
76
|
+
See http://lloydforge.org/projects/ruby. (Jeremy Kemper).
|
77
|
+
|
78
|
+
|
79
|
+
Fixes
|
80
|
+
-------
|
81
|
+
* RubyProf.profile no longer crashes if an exception is
|
82
|
+
thrown during a profiling run.
|
83
|
+
|
84
|
+
* Measure memory in fractional kilobytes rather than rounding down (Jeremy Kemper)
|
85
|
+
|
86
|
+
|
87
|
+
0.6.0 (2008-02-03)
|
88
|
+
========================
|
89
|
+
|
90
|
+
ruby-prof 0.6.0 adds support for Ruby 1.9 and memory profiling.
|
91
|
+
|
92
|
+
Features
|
93
|
+
--------
|
94
|
+
* Added support for ruby 1.9 (Shugo Maeda)
|
95
|
+
* Added support for outputting printer results to a String, Array or IO
|
96
|
+
object (Michael Granger)
|
97
|
+
* Add new memory profiling mode. Note this mode depends on a
|
98
|
+
patched Ruby interpreter (Alexander Dymo)
|
99
|
+
|
100
|
+
Fixes
|
101
|
+
-------
|
102
|
+
* Improvements to GraphHtmlPrinter including updated documentation,
|
103
|
+
fixes for min_time support, ability to specify templates using
|
104
|
+
strings or filenames, and table layout fixes (Makoto Kuwata)
|
105
|
+
* Fixes to scaling factor for calltrees so that precision is not lost
|
106
|
+
due to the conversion to doubles (Sylvain Joyeux)
|
107
|
+
* Changed constant ALLOCATED_OBJECTS to ALLOCATIONS in the C code to
|
108
|
+
match the Ruby code (Sylvain Joyeux)
|
109
|
+
* Added support for calltree printer to ruby-prof binary script (Sylvain Joyeux)
|
110
|
+
* Fix support for the allocator measure mode to extconf.rb (Sylvain Joyeux)
|
111
|
+
* Honor measure mode when specified on the command line (Sylvain Joyeux)
|
112
|
+
* Sorting of methods by total time was incorrect (Dan Fitch, Charlie Savage)
|
113
|
+
* Fix ruby-prof to work with the latest version of GEMS (Alexander Dymo)
|
114
|
+
* Always define MEASURE_CPU_TIME and MEASURE_ALLOCATIONS in Ruby code, but
|
115
|
+
set their values to nil if the functionality is not available.
|
116
|
+
|
117
|
+
|
118
|
+
0.5.2 (2007-07-19)
|
119
|
+
========================
|
120
|
+
|
121
|
+
ruby-prof 0.5.2 is a bug fix release.
|
122
|
+
|
123
|
+
Fixes
|
124
|
+
-------
|
125
|
+
* Include missing rails plugin
|
126
|
+
|
127
|
+
|
128
|
+
0.5.1 (2007-07-18)
|
129
|
+
========================
|
130
|
+
|
131
|
+
ruby-prof 0.5.1 is a bug fix and performance release.
|
132
|
+
|
133
|
+
Performance
|
134
|
+
--------
|
135
|
+
* Significantly reduced the number of thread lookups by
|
136
|
+
caching the last executed thread.
|
137
|
+
|
138
|
+
Fixes
|
139
|
+
-------
|
140
|
+
* Properly escape method names in HTML reports
|
141
|
+
* Fix use of -m and --min-percent command line switches
|
142
|
+
* Default source file information to ruby_runtime#0 for c calls
|
143
|
+
* Moved rails_plugin to top level so it is more obvious
|
144
|
+
* Updated rails_plugin to write reports to the current
|
145
|
+
Rails log directory
|
146
|
+
* Added additional tests
|
147
|
+
|
148
|
+
|
149
|
+
0.5.0 (2007-07-09)
|
150
|
+
========================
|
151
|
+
|
152
|
+
Features
|
153
|
+
--------
|
154
|
+
* Added support for timing multi-threaded applications
|
155
|
+
* Added support for 64 bit systems (patch from Diego 'Flameeyes' Petten)
|
156
|
+
* Added suport for outputting data in the format used by
|
157
|
+
KCacheGrind (patch from Carl Shimer)
|
158
|
+
* Add filename and line numbers to call tree information (patch from Carl Shimer)
|
159
|
+
* Added Visual Studio 2005 project file.
|
160
|
+
* Added replace-progname switch, als rcov.
|
161
|
+
* Added better support for recursive methods
|
162
|
+
* Added better support for profiling Rails applications
|
163
|
+
|
164
|
+
Fixes
|
165
|
+
-------
|
166
|
+
* Fixes bug when the type of an attached object (singleton) is inherited
|
167
|
+
from T_OBJECT as opposed to being a T_OBJECT (identified by Francis Cianfrocca)
|
168
|
+
* ruby-prof now works in IRB.
|
169
|
+
* Fix sort order in reports.
|
170
|
+
* Fixed rdoc compile error.
|
171
|
+
* Fix tabs in erb template for graph html report on windows.
|
172
|
+
|
173
|
+
0.4.1 (2006-06-26)
|
174
|
+
========================
|
175
|
+
|
176
|
+
Features
|
177
|
+
--------
|
178
|
+
* Added a RubyProf.running? method to indicate whether a profile is in progress.
|
179
|
+
* Added tgz and zip archives to release
|
180
|
+
|
181
|
+
Fixes
|
182
|
+
-------
|
183
|
+
* Duplicate method names are now allowed
|
184
|
+
* The documentation has been updated to show the correct API usage is RubyProf.stop not RubyProf.end
|
185
|
+
|
186
|
+
|
187
|
+
0.4.0 (2006-06-16)
|
188
|
+
========================
|
189
|
+
Features
|
190
|
+
--------
|
191
|
+
* added support for call graphs
|
192
|
+
* added support for printers. Currently there is a FlatPrinter,
|
193
|
+
GraphPrinter and GraphHtmlPrinter.
|
194
|
+
* added support for recursive methods
|
195
|
+
* added Windows support
|
196
|
+
* now packaged as a RubyGem
|
197
|
+
|
198
|
+
Fixes
|
199
|
+
-------
|
200
|
+
* Fixes bug where RubyProf would crash depending on the
|
201
|
+
way it was invoked - for example, it did not run when
|
202
|
+
used with Arachno Ruby's customized version of Ruby.
|
data/LICENSE
ADDED
@@ -0,0 +1,23 @@
|
|
1
|
+
Copyright (C) 2005 Shugo Maeda <shugo@ruby-lang.org>
|
2
|
+
All rights reserved.
|
3
|
+
*
|
4
|
+
Redistribution and use in source and binary forms, with or without
|
5
|
+
modification, are permitted provided that the following conditions
|
6
|
+
are met:
|
7
|
+
1. Redistributions of source code must retain the above copyright
|
8
|
+
notice, this list of conditions and the following disclaimer.
|
9
|
+
2. Redistributions in binary form must reproduce the above copyright
|
10
|
+
notice, this list of conditions and the following disclaimer in the
|
11
|
+
documentation and/or other materials provided with the distribution.
|
12
|
+
*
|
13
|
+
THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
14
|
+
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
15
|
+
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
16
|
+
ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
17
|
+
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
18
|
+
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
19
|
+
OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
20
|
+
HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
21
|
+
LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
22
|
+
OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
23
|
+
SUCH DAMAGE.
|
data/README
ADDED
@@ -0,0 +1,445 @@
|
|
1
|
+
= ruby-prof
|
2
|
+
|
3
|
+
== Overview
|
4
|
+
|
5
|
+
ruby-prof is a fast code profiler for Ruby. Its features include:
|
6
|
+
|
7
|
+
* Speed - it is a C extension and therefore many times faster than the standard Ruby profiler.
|
8
|
+
* Modes - Ruby prof can measure a number of different parameters, including
|
9
|
+
call times, memory usage and object allocations.
|
10
|
+
* Reports - can generate text and cross-referenced html reports
|
11
|
+
- Flat Profiles - similar to the reports generated by the standard Ruby profiler
|
12
|
+
- Graph profiles - similar to GProf, these show how long a method runs, which methods call it and which methods it calls.
|
13
|
+
- Call tree profiles - outputs results in the calltree format suitable for the KCacheGrind profiling tool.
|
14
|
+
* Threads - supports profiling multiple threads simultaneously
|
15
|
+
* Recursive calls - supports profiling recursive method calls
|
16
|
+
|
17
|
+
|
18
|
+
== Requirements
|
19
|
+
|
20
|
+
ruby-prof requires Ruby 1.8.4 or higher.
|
21
|
+
|
22
|
+
If you are running Linux or Unix you'll need a C compiler so the extension
|
23
|
+
can be compiled when it is installed.
|
24
|
+
|
25
|
+
If you are running Windows, then install the Windows specific RubyGem which
|
26
|
+
includes an already built extension.
|
27
|
+
|
28
|
+
|
29
|
+
== Install
|
30
|
+
|
31
|
+
The easiest way to install ruby-prof is by using Ruby Gems. To install:
|
32
|
+
|
33
|
+
<tt>gem install ruby-prof</tt>
|
34
|
+
|
35
|
+
If you are running Windows, make sure to install the Win32 RubyGem which
|
36
|
+
includes a pre-built binary. Due to a bug in ruby-gems, you cannot
|
37
|
+
install the gem to a path that contains spaces
|
38
|
+
(see http://rubyforge.org/tracker/?func=detail&aid=23003&group_id=126&atid=577).
|
39
|
+
|
40
|
+
If you on windows mswin [not mingw] (check via ruby -v)
|
41
|
+
please install v0.7.3 which has a prebuilt binary
|
42
|
+
C:> gem install ruby-prof -v0.7.3
|
43
|
+
|
44
|
+
If you're on mingw, please install the devkit first.
|
45
|
+
|
46
|
+
ruby-prof is also available as a tarred gzip archive and zip archive.
|
47
|
+
|
48
|
+
== Usage
|
49
|
+
|
50
|
+
There are three ways of running ruby-prof.
|
51
|
+
|
52
|
+
|
53
|
+
=== ruby-prof executable
|
54
|
+
|
55
|
+
The first is to use ruby-prof to run the Ruby program
|
56
|
+
you want to profile. For more information refer to
|
57
|
+
the ruby-prof documentation[link:files/bin/ruby-prof.html].
|
58
|
+
|
59
|
+
|
60
|
+
=== ruby-prof API
|
61
|
+
|
62
|
+
The second way is to use the ruby-prof API to profile
|
63
|
+
particular segments of code.
|
64
|
+
|
65
|
+
require 'ruby-prof'
|
66
|
+
|
67
|
+
# Profile the code
|
68
|
+
RubyProf.start
|
69
|
+
...
|
70
|
+
[code to profile]
|
71
|
+
...
|
72
|
+
result = RubyProf.stop
|
73
|
+
|
74
|
+
# Print a flat profile to text
|
75
|
+
printer = RubyProf::FlatPrinter.new(result)
|
76
|
+
printer.print(STDOUT, 0)
|
77
|
+
|
78
|
+
Alternatively, you can use a block to tell ruby-prof what
|
79
|
+
to profile:
|
80
|
+
|
81
|
+
require 'ruby-prof'
|
82
|
+
|
83
|
+
# Profile the code
|
84
|
+
result = RubyProf.profile do
|
85
|
+
...
|
86
|
+
[code to profile]
|
87
|
+
...
|
88
|
+
end
|
89
|
+
|
90
|
+
# Print a graph profile to text
|
91
|
+
printer = RubyProf::GraphPrinter.new(result)
|
92
|
+
printer.print(STDOUT, 0)
|
93
|
+
|
94
|
+
Starting with the 0.6.1 release, ruby-prof also supports pausing and resuming
|
95
|
+
profiling runs.
|
96
|
+
|
97
|
+
require 'ruby-prof'
|
98
|
+
|
99
|
+
# Profile the code
|
100
|
+
RubyProf.start
|
101
|
+
[code to profile]
|
102
|
+
RubyProf.pause
|
103
|
+
[other code]
|
104
|
+
RubyProf.resume
|
105
|
+
[code to profile]
|
106
|
+
result = RubyProf.stop
|
107
|
+
|
108
|
+
Note that resume will automatically call start if a profiling run
|
109
|
+
has not yet started. In addition, resume can also take a block:
|
110
|
+
|
111
|
+
require 'ruby-prof'
|
112
|
+
|
113
|
+
# Profile the code
|
114
|
+
RubyProf.resume do
|
115
|
+
[code to profile]
|
116
|
+
end
|
117
|
+
|
118
|
+
data = RubyProf.stop
|
119
|
+
|
120
|
+
With this usage, resume will automatically call pause at the
|
121
|
+
end of the block.
|
122
|
+
|
123
|
+
|
124
|
+
=== require unprof
|
125
|
+
|
126
|
+
The third way of using ruby-prof is by requiring unprof.rb:
|
127
|
+
|
128
|
+
require 'unprof'
|
129
|
+
|
130
|
+
This will start profiling immediately and will output the results
|
131
|
+
using a flat profile report.
|
132
|
+
|
133
|
+
This method is provided for backwards compatibility. Using
|
134
|
+
{ruby-prof}[link:files/bin/ruby-prof.html] provides more flexibility.
|
135
|
+
|
136
|
+
|
137
|
+
== Profiling Tests
|
138
|
+
|
139
|
+
Starting with the 0.6.1 release, ruby-prof supports profiling tests cases
|
140
|
+
written using Ruby's built-in unit test framework (ie, test derived from
|
141
|
+
Test::Unit::TestCase). To enable profiling simply add the following line
|
142
|
+
of code to your test class:
|
143
|
+
|
144
|
+
include RubyProf::Test
|
145
|
+
|
146
|
+
Each test method is profiled separately. ruby-prof will run each test method
|
147
|
+
once as a warmup and then ten additional times to gather profile data.
|
148
|
+
Note that the profile data will *not* include the class's setup or
|
149
|
+
teardown methods.
|
150
|
+
|
151
|
+
Separate reports are generated for each method and saved, by default,
|
152
|
+
in the test process's working directory. To change this, or other profiling
|
153
|
+
options, modify your test class's PROFILE_OPTIONS hash table. To globally
|
154
|
+
change test profiling options, modify RubyProf::Test::PROFILE_OPTIONS.
|
155
|
+
|
156
|
+
|
157
|
+
== Profiling Rails
|
158
|
+
|
159
|
+
To profile a Rails application it is vital to run it using production like
|
160
|
+
settings (cache classes, cache view lookups, etc.). Otherwise, Rail's
|
161
|
+
dependency loading code will overwhelm any time spent in the application
|
162
|
+
itself (our tests show that Rails dependency loading causes a roughly 6x
|
163
|
+
slowdown). The best way to do this is create a new Rails environment,
|
164
|
+
profile.rb.
|
165
|
+
|
166
|
+
So to profile Rails:
|
167
|
+
|
168
|
+
1. Create a new profile.rb environment - or simply copy the example file
|
169
|
+
in ruby-prof/rails/environment/profile.rb
|
170
|
+
|
171
|
+
2. Copy the file:
|
172
|
+
|
173
|
+
ruby-prof/rails/profile_test_helper.rb
|
174
|
+
|
175
|
+
To:
|
176
|
+
|
177
|
+
your_rails_app/test/profile_test_helper.rb
|
178
|
+
|
179
|
+
3. Create a new test directory for profiling:
|
180
|
+
|
181
|
+
your_rails_app/test/profile
|
182
|
+
|
183
|
+
|
184
|
+
4. Write unit, functional or integration tests specifically designed
|
185
|
+
to profile some part of your Rails application. At the top
|
186
|
+
of each test, replace this line:
|
187
|
+
|
188
|
+
require File.dirname(__FILE__) + '/../test_helper'
|
189
|
+
|
190
|
+
With:
|
191
|
+
|
192
|
+
require File.dirname(__FILE__) + '/../profile_test_helper'
|
193
|
+
|
194
|
+
For example:
|
195
|
+
|
196
|
+
require File.dirname(__FILE__) + '/../profile_test_helper'
|
197
|
+
|
198
|
+
class ExampleTest < Test::Unit::TestCase
|
199
|
+
include RubyProf::Test
|
200
|
+
fixtures ....
|
201
|
+
|
202
|
+
def test_stuff
|
203
|
+
puts "Test method"
|
204
|
+
end
|
205
|
+
end
|
206
|
+
|
207
|
+
5. Now run your tests. Results will be written to:
|
208
|
+
|
209
|
+
your_rails_app/tmp/profile
|
210
|
+
|
211
|
+
|
212
|
+
== Reports
|
213
|
+
|
214
|
+
ruby-prof can generate a number of different reports:
|
215
|
+
|
216
|
+
* Flat Reports
|
217
|
+
* Graph Reports
|
218
|
+
* HTML Graph Reports
|
219
|
+
* Call graphs
|
220
|
+
|
221
|
+
Flat profiles show the overall time spent in each method. They
|
222
|
+
are a good of quickly identifying which methods take the most time.
|
223
|
+
An example of a flat profile and an explanation can be found in
|
224
|
+
{examples/flat.txt}[link:files/examples/flat_txt.html].
|
225
|
+
|
226
|
+
Graph profiles also show the overall time spent in each method.
|
227
|
+
In addition, they also show which methods call the current
|
228
|
+
method and which methods its calls. Thus they are good for
|
229
|
+
understanding how methods gets called and provide insight into
|
230
|
+
the flow of your program. An example text graph profile
|
231
|
+
is located at {examples/graph.txt}[link:files/examples/graph_txt.html].
|
232
|
+
|
233
|
+
HTML Graph profiles are the same as graph profiles, except
|
234
|
+
output is generated in hyper-linked HTML. Since graph profiles
|
235
|
+
can be quite large, the embedded links make it much easier to
|
236
|
+
navigate the results. An example html graph profile
|
237
|
+
is located at {examples/graph.html}[link:files/examples/graph_html.html].
|
238
|
+
|
239
|
+
HTML Graph profiles are the same as graph profiles, except
|
240
|
+
output is generated in hyper-linked HTML. Since graph profiles
|
241
|
+
can be quite large, the embedded links make it much easier to
|
242
|
+
navigate the results. An example html graph profile
|
243
|
+
is located at {examples/graph.html}[link:files/examples/graph_html.html].
|
244
|
+
|
245
|
+
Call graphs output results in the calltree profile format which is used
|
246
|
+
by KCachegrind. Call graph support was generously donated by Carl Shimer.
|
247
|
+
More information about the format can be found at
|
248
|
+
the {KCachegrind}[link:http://kcachegrind.sourceforge.net/cgi-bin/show.cgi/KcacheGrindCalltreeFormat] site.
|
249
|
+
|
250
|
+
|
251
|
+
== Printers
|
252
|
+
|
253
|
+
Reports are created by printers. Supported printers include:
|
254
|
+
|
255
|
+
* RubyProf::FlatPrinter - Creates a flat report in text format
|
256
|
+
* RubyProf::GraphPrinter - Creates a call graph report in text format
|
257
|
+
* RubyProf::GraphHtmlPrinter - Creates a call graph report in HTML (separate files per thread)
|
258
|
+
* RubyProf::CallTreePrinter - Creates a call tree report compatible with KCachegrind.
|
259
|
+
|
260
|
+
To use a printer:
|
261
|
+
|
262
|
+
result = RubyProf.end
|
263
|
+
printer = RubyProf::GraphPrinter.new(result)
|
264
|
+
printer.print(STDOUT, 0)
|
265
|
+
|
266
|
+
The first parameter is any writable IO object such as STDOUT or a file.
|
267
|
+
The second parameter, which has a default value of 0, specifies
|
268
|
+
the minimum percentage a method must take to be printed. Percentages
|
269
|
+
should be specified as integers in the range 0 to 100. For more
|
270
|
+
information please see the documentation for the different printers.
|
271
|
+
|
272
|
+
|
273
|
+
== Measurements
|
274
|
+
|
275
|
+
Depending on the mode and platform, ruby-prof can measure various
|
276
|
+
aspects of a Ruby program. Supported measurements include:
|
277
|
+
|
278
|
+
* process time (RubyProf::PROCESS_TIME)
|
279
|
+
* wall time (RubyProf::WALL_TIME)
|
280
|
+
* cpu time (RubyProf::CPU_TIME)
|
281
|
+
* object allocations (RubyProf::ALLOCATIONS)
|
282
|
+
* memory usage (RubyProf::MEMORY)
|
283
|
+
* garbage collections runs (RubyProf::GC_RUNS)
|
284
|
+
* garbage collection time (RubyProf::GC_TIME)
|
285
|
+
|
286
|
+
|
287
|
+
Process time measures the time used by a process between any two moments.
|
288
|
+
It is unaffected by other processes concurrently running
|
289
|
+
on the system. Note that Windows does not support measuring process
|
290
|
+
times - therefore, all measurements on Windows use wall time.
|
291
|
+
|
292
|
+
Wall time measures the real-world time elapsed between any two moments.
|
293
|
+
If there are other processes concurrently running on the system
|
294
|
+
that use significant CPU or disk time during a profiling run
|
295
|
+
then the reported results will be too large.
|
296
|
+
|
297
|
+
CPU time uses the CPU clock counter to measure time. The returned
|
298
|
+
values are dependent on the correctly setting the CPU's frequency.
|
299
|
+
This mode is only supported on Pentium or PowerPC platforms.
|
300
|
+
|
301
|
+
Object allocation reports show how many objects each method in
|
302
|
+
a program allocates. This support was added by Sylvain Joyeux
|
303
|
+
and requires a patched Ruby interpreter. For more information and
|
304
|
+
the patch, please see:
|
305
|
+
http://rubyforge.org/tracker/index.php?func=detail&aid=11497&group_id=426&atid=1700
|
306
|
+
|
307
|
+
Memory usage reports show how much memory each method in a program
|
308
|
+
uses. This support was added by Alexander Dymo and requires a
|
309
|
+
patched Ruby interpreter. For more information, see:
|
310
|
+
http://rubyforge.org/tracker/index.php?func=detail&aid=17676&group_id=1814&atid=7062
|
311
|
+
|
312
|
+
Garbage collection runs report how many times Ruby's garbage collector
|
313
|
+
is invoked during a profiling session. This support was added by Jeremy
|
314
|
+
Kemper and requires a patched Ruby interpreter. For more information, see:
|
315
|
+
http://rubyforge.org/tracker/index.php?func=detail&aid=17676&group_id=1814&atid=7062
|
316
|
+
|
317
|
+
Garbage collection time reports how much time is spent in Ruby's garbage collector
|
318
|
+
during a profiling session. This support was added by Jeremy Kemper
|
319
|
+
and requires a patched Ruby interpreter. For more information, see:
|
320
|
+
http://rubyforge.org/tracker/index.php?func=detail&aid=17676&group_id=1814&atid=7062
|
321
|
+
|
322
|
+
To set the measurement:
|
323
|
+
|
324
|
+
* RubyProf.measure_mode = RubyProf::PROCESS_TIME
|
325
|
+
* RubyProf.measure_mode = RubyProf::WALL_TIME
|
326
|
+
* RubyProf.measure_mode = RubyProf::CPU_TIME
|
327
|
+
* RubyProf.measure_mode = RubyProf::ALLOCATIONS
|
328
|
+
* RubyProf.measure_mode = RubyProf::MEMORY
|
329
|
+
* RubyProf.measure_mode = RubyProf::GC_RUNS
|
330
|
+
* RubyProf.measure_mode = RubyProf::GC_TIME
|
331
|
+
|
332
|
+
The default value is RubyProf::PROCESS_TIME.
|
333
|
+
|
334
|
+
You may also specify the measure_mode by using the RUBY_PROF_MEASURE_MODE
|
335
|
+
environment variable:
|
336
|
+
|
337
|
+
* export RUBY_PROF_MEASURE_MODE=process
|
338
|
+
* export RUBY_PROF_MEASURE_MODE=wall
|
339
|
+
* export RUBY_PROF_MEASURE_MODE=cpu
|
340
|
+
* export RUBY_PROF_MEASURE_MODE=allocations
|
341
|
+
* export RUBY_PROF_MEASURE_MODE=memory
|
342
|
+
* export RUBY_PROF_MEASURE_MODE=gc_runs
|
343
|
+
* export RUBY_PROF_MEASURE_MODE=gc_time
|
344
|
+
|
345
|
+
Note that these values have changed since ruby-prof-0.3.0.
|
346
|
+
|
347
|
+
On Linux, process time is measured using the clock method provided
|
348
|
+
by the C runtime library. Note that the clock method does not
|
349
|
+
report time spent in the kernel or child processes and therefore
|
350
|
+
does not measure time spent in methods such as Kernel.sleep method.
|
351
|
+
If you need to measure these values, then use wall time. Wall time
|
352
|
+
is measured using the gettimeofday kernel method.
|
353
|
+
|
354
|
+
On Windows, timings are always wall times. If you set the clock
|
355
|
+
mode to PROCESS_TIME, then timing are read using the clock method
|
356
|
+
provided by the C runtime library. Note though, these values are
|
357
|
+
wall times on Windows and not process times like on Linux.
|
358
|
+
Wall time is measured using the GetLocalTime API.
|
359
|
+
|
360
|
+
If you use wall time, the results will be affected by other
|
361
|
+
processes running on your computer, network delays, disk access,
|
362
|
+
etc. As result, for the best results, try to make sure your
|
363
|
+
computer is only performing your profiling run and is
|
364
|
+
otherwise quiescent.
|
365
|
+
|
366
|
+
On both platforms, cpu time is measured using the RDTSC assembly
|
367
|
+
function provided by the Pentium and PowerPC platforms. CPU time
|
368
|
+
is dependent on the cpu's frequency. On Linux, ruby-prof attempts
|
369
|
+
to read this value from "/proc/cpuinfo." On Windows, you must
|
370
|
+
specify the clock frequency. This can be done using the
|
371
|
+
RUBY_PROF_CPU_FREQUENCY environment variable:
|
372
|
+
|
373
|
+
export RUBY_PROF_CPU_FREQUENCY=<value>
|
374
|
+
|
375
|
+
You can also directly set the cpu frequency by calling:
|
376
|
+
|
377
|
+
RubyProf.cpu_frequency = <value>
|
378
|
+
|
379
|
+
|
380
|
+
== Recursive Calls
|
381
|
+
|
382
|
+
Recursive calls occur when method A calls method A and cycles
|
383
|
+
occur when method A calls method B calls method C calls method A.
|
384
|
+
ruby-prof detects both direct recursive calls and cycles. Both
|
385
|
+
are indicated in reports by a dash and number following a method
|
386
|
+
name. For example, here is a flat profile from the test method
|
387
|
+
RecursiveTest#test_recursive:
|
388
|
+
|
389
|
+
|
390
|
+
%self total self wait child calls name
|
391
|
+
100.00 2.00 2.00 0.00 0.00 2 Kernel#sleep
|
392
|
+
0.00 2.00 0.00 0.00 2.00 0 RecursiveTest#test_cycle
|
393
|
+
0.00 0.00 0.00 0.00 0.00 2 Fixnum#==
|
394
|
+
0.00 0.00 0.00 0.00 0.00 2 Fixnum#-
|
395
|
+
0.00 1.00 0.00 0.00 1.00 1 Object#sub_cycle-1
|
396
|
+
0.00 2.00 0.00 0.00 2.00 1 Object#sub_cycle
|
397
|
+
0.00 2.00 0.00 0.00 2.00 1 Object#cycle
|
398
|
+
0.00 1.00 0.00 0.00 1.00 1 Object#cycle-1
|
399
|
+
|
400
|
+
Notice the presence of Object#cycle and Object#cycle-1. The -1 means
|
401
|
+
the method was either recursively called (directly or indirectly).
|
402
|
+
|
403
|
+
However, the self time values for recursive calls should always
|
404
|
+
be accurate. It is also believed that the total times are
|
405
|
+
accurate, but these should be carefully analyzed to verify their veracity.
|
406
|
+
|
407
|
+
|
408
|
+
== Multi-threaded Applications
|
409
|
+
|
410
|
+
Unfortunately, Ruby does not provide an internal api
|
411
|
+
for detecting thread context switches. As a result, the
|
412
|
+
timings ruby-prof reports for each thread may be slightly
|
413
|
+
inaccurate. In particular, this will happen for newly
|
414
|
+
spanned threads that immediately go to sleep. For instance,
|
415
|
+
if you use Ruby's timeout library to wait for 2 seconds,
|
416
|
+
the 2 seconds will be assigned to the foreground thread
|
417
|
+
and not the newly created background thread. These errors
|
418
|
+
can largely be avoided if the background thread performs an
|
419
|
+
operation before going to sleeep.
|
420
|
+
|
421
|
+
|
422
|
+
== Performance
|
423
|
+
|
424
|
+
Significant effort has been put into reducing ruby-prof's overhead
|
425
|
+
as much as possible. Our tests show that the overhead associated
|
426
|
+
with profiling code varies considerably with the code being
|
427
|
+
profiled. Most programs will run approximately twice as slow
|
428
|
+
while highly recursive programs (like the fibonacci series test)
|
429
|
+
will run three times slower.
|
430
|
+
|
431
|
+
|
432
|
+
== Windows Binary
|
433
|
+
|
434
|
+
The Windows binary is built with the latest version of MinGW. The source
|
435
|
+
repository also includes a Microsoft VC++ 2005 solution. If you wish to run
|
436
|
+
a debug version of ruby-prof on Windows, then it is highly recommended
|
437
|
+
you use VC++.
|
438
|
+
|
439
|
+
== License
|
440
|
+
|
441
|
+
See LICENSE for license information.
|
442
|
+
|
443
|
+
== Development
|
444
|
+
|
445
|
+
Code is located at http://github.com/rdp/ruby-prof
|