1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
|
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>Design</title><meta name="generator" content="DocBook XSL Stylesheets V1.74.3" /><meta name="keywords" content=" C++ , library , profile " /><meta name="keywords" content=" ISO C++ , library " /><link rel="home" href="../spine.html" title="The GNU C++ Library Documentation" /><link rel="up" href="profile_mode.html" title="Chapter 32. Profile Mode" /><link rel="prev" href="profile_mode.html" title="Chapter 32. Profile Mode" /><link rel="next" href="bk01pt12ch32s03.html" title="Extensions for Custom Containers" /></head><body><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Design</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="profile_mode.html">Prev</a> </td><th width="60%" align="center">Chapter 32. Profile Mode</th><td width="20%" align="right"> <a accesskey="n" href="bk01pt12ch32s03.html">Next</a></td></tr></table><hr /></div><div class="sect1" lang="en" xml:lang="en"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="manual.ext.profile_mode.design"></a>Design</h2></div></div></div><p>
</p><div class="table"><a id="id474437"></a><p class="title"><b>Table 32.1. Code Location</b></p><div class="table-contents"><table summary="Code Location" border="1"><colgroup><col align="left" /><col align="left" /></colgroup><thead><tr><th align="left">Code Location</th><th align="left">Use</th></tr></thead><tbody><tr><td align="left"><code class="code">libstdc++-v3/include/std/*</code></td><td align="left">Preprocessor code to redirect to profile extension headers.</td></tr><tr><td align="left"><code class="code">libstdc++-v3/include/profile/*</code></td><td align="left">Profile extension public headers (map, vector, ...).</td></tr><tr><td align="left"><code class="code">libstdc++-v3/include/profile/impl/*</code></td><td align="left">Profile extension internals. Implementation files are
only included from <code class="code">impl/profiler.h</code>, which is the only
file included from the public headers.</td></tr></tbody></table></div></div><br class="table-break" /><p>
</p><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.wrapper"></a>Wrapper Model</h3></div></div></div><p>
In order to get our instrumented library version included instead of the
release one,
we use the same wrapper model as the debug mode.
We subclass entities from the release version. Wherever
<code class="code">_GLIBCXX_PROFILE</code> is defined, the release namespace is
<code class="code">std::__norm</code>, whereas the profile namespace is
<code class="code">std::__profile</code>. Using plain <code class="code">std</code> translates
into <code class="code">std::__profile</code>.
</p><p>
Whenever possible, we try to wrap at the public interface level, e.g.,
in <code class="code">unordered_set</code> rather than in <code class="code">hashtable</code>,
in order not to depend on implementation.
</p><p>
Mixing object files built with and without the profile mode must
not affect the program execution. However, there are no guarantees to
the accuracy of diagnostics when using even a single object not built with
<code class="code">-D_GLIBCXX_PROFILE</code>.
Currently, mixing the profile mode with debug and parallel extensions is
not allowed. Mixing them at compile time will result in preprocessor errors.
Mixing them at link time is undefined.
</p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.instrumentation"></a>Instrumentation</h3></div></div></div><p>
Instead of instrumenting every public entry and exit point,
we chose to add instrumentation on demand, as needed
by individual diagnostics.
The main reason is that some diagnostics require us to extract bits of
internal state that are particular only to that diagnostic.
We plan to formalize this later, after we learn more about the requirements
of several diagnostics.
</p><p>
All the instrumentation points can be switched on and off using
<code class="code">-D[_NO]_GLIBCXX_PROFILE_<diagnostic></code> options.
With all the instrumentation calls off, there should be negligible
overhead over the release version. This property is needed to support
diagnostics based on timing of internal operations. For such diagnostics,
we anticipate turning most of the instrumentation off in order to prevent
profiling overhead from polluting time measurements, and thus diagnostics.
</p><p>
All the instrumentation on/off compile time switches live in
<code class="code">include/profile/profiler.h</code>.
</p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.rtlib"></a>Run Time Behavior</h3></div></div></div><p>
For practical reasons, the instrumentation library processes the trace
partially
rather than dumping it to disk in raw form. Each event is processed when
it occurs. It is usually attached a cost and it is aggregated into
the database of a specific diagnostic class. The cost model
is based largely on the standard performance guarantees, but in some
cases we use knowledge about GCC's standard library implementation.
</p><p>
Information is indexed by (1) call stack and (2) instance id or address
to be able to understand and summarize precise creation-use-destruction
dynamic chains. Although the analysis is sensitive to dynamic instances,
the reports are only sensitive to call context. Whenever a dynamic instance
is destroyed, we accumulate its effect to the corresponding entry for the
call stack of its constructor location.
</p><p>
For details, see
<a class="ulink" href="http://dx.doi.org/10.1109/CGO.2009.36" target="_top">paper presented at
CGO 2009</a>.
</p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.analysis"></a>Analysis and Diagnostics</h3></div></div></div><p>
Final analysis takes place offline, and it is based entirely on the
generated trace and debugging info in the application binary.
See section Diagnostics for a list of analysis types that we plan to support.
</p><p>
The input to the analysis is a table indexed by profile type and call stack.
The data type for each entry depends on the profile type.
</p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.cost-model"></a>Cost Model</h3></div></div></div><p>
While it is likely that cost models become complex as we get into
more sophisticated analysis, we will try to follow a simple set of rules
at the beginning.
</p><div class="itemizedlist"><ul type="disc"><li><p><span class="emphasis"><em>Relative benefit estimation:</em></span>
The idea is to estimate or measure the cost of all operations
in the original scenario versus the scenario we advise to switch to.
For instance, when advising to change a vector to a list, an occurrence
of the <code class="code">insert</code> method will generally count as a benefit.
Its magnitude depends on (1) the number of elements that get shifted
and (2) whether it triggers a reallocation.
</p></li><li><p><span class="emphasis"><em>Synthetic measurements:</em></span>
We will measure the relative difference between similar operations on
different containers. We plan to write a battery of small tests that
compare the times of the executions of similar methods on different
containers. The idea is to run these tests on the target machine.
If this training phase is very quick, we may decide to perform it at
library initialization time. The results can be cached on disk and reused
across runs.
</p></li><li><p><span class="emphasis"><em>Timers:</em></span>
We plan to use timers for operations of larger granularity, such as sort.
For instance, we can switch between different sort methods on the fly
and report the one that performs best for each call context.
</p></li><li><p><span class="emphasis"><em>Show stoppers:</em></span>
We may decide that the presence of an operation nullifies the advice.
For instance, when considering switching from <code class="code">set</code> to
<code class="code">unordered_set</code>, if we detect use of operator <code class="code">++</code>,
we will simply not issue the advice, since this could signal that the use
care require a sorted container.</p></li></ul></div></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.reports"></a>Reports</h3></div></div></div><p>
There are two types of reports. First, if we recognize a pattern for which
we have a substitute that is likely to give better performance, we print
the advice and estimated performance gain. The advice is usually associated
to a code position and possibly a call stack.
</p><p>
Second, we report performance characteristics for which we do not have
a clear solution for improvement. For instance, we can point to the user
the top 10 <code class="code">multimap</code> locations
which have the worst data locality in actual traversals.
Although this does not offer a solution,
it helps the user focus on the key problems and ignore the uninteresting ones.
</p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.profile_mode.design.testing"></a>Testing</h3></div></div></div><p>
First, we want to make sure we preserve the behavior of the release mode.
You can just type <code class="code">"make check-profile"</code>, which
builds and runs the whole test suite in profile mode.
</p><p>
Second, we want to test the correctness of each diagnostic.
We created a <code class="code">profile</code> directory in the test suite.
Each diagnostic must come with at least two tests, one for false positives
and one for false negatives.
</p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="profile_mode.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="profile_mode.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="bk01pt12ch32s03.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Chapter 32. Profile Mode </td><td width="20%" align="center"><a accesskey="h" href="../spine.html">Home</a></td><td width="40%" align="right" valign="top"> Extensions for Custom Containers</td></tr></table></div></body></html>
|
