This file is indexed.

/usr/share/doc/llvm-3.5-doc/html/BlockFrequencyTerminology.html is in llvm-3.5-doc 1:3.5-4ubuntu2~trusty2.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

  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
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
<!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>LLVM Block Frequency Terminology &mdash; LLVM 3.5 documentation</title>
    
    <link rel="stylesheet" href="_static/llvm-theme.css" type="text/css" />
    <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
    
    <script type="text/javascript">
      var DOCUMENTATION_OPTIONS = {
        URL_ROOT:    './',
        VERSION:     '3.5',
        COLLAPSE_INDEX: false,
        FILE_SUFFIX: '.html',
        HAS_SOURCE:  true
      };
    </script>
    <script type="text/javascript" src="_static/jquery.js"></script>
    <script type="text/javascript" src="_static/underscore.js"></script>
    <script type="text/javascript" src="_static/doctools.js"></script>
    <link rel="top" title="LLVM 3.5 documentation" href="index.html" />
    <link rel="next" title="LLVM Branch Weight Metadata" href="BranchWeightMetadata.html" />
    <link rel="prev" title="LLVM Bitcode File Format" href="BitCodeFormat.html" />
<style type="text/css">
  table.right { float: right; margin-left: 20px; }
  table.right td { border: 1px solid #ccc; }
</style>

  </head>
  <body>
<div class="logo">
  <a href="index.html">
    <img src="_static/logo.png"
         alt="LLVM Logo" width="250" height="88"/></a>
</div>

    <div class="related">
      <h3>Navigation</h3>
      <ul>
        <li class="right" style="margin-right: 10px">
          <a href="genindex.html" title="General Index"
             accesskey="I">index</a></li>
        <li class="right" >
          <a href="BranchWeightMetadata.html" title="LLVM Branch Weight Metadata"
             accesskey="N">next</a> |</li>
        <li class="right" >
          <a href="BitCodeFormat.html" title="LLVM Bitcode File Format"
             accesskey="P">previous</a> |</li>
  <li><a href="http://llvm.org/">LLVM Home</a>&nbsp;|&nbsp;</li>
  <li><a href="index.html">Documentation</a>&raquo;</li>
 
      </ul>
    </div>


    <div class="document">
      <div class="documentwrapper">
          <div class="body">
            
  <div class="section" id="llvm-block-frequency-terminology">
<h1>LLVM Block Frequency Terminology<a class="headerlink" href="#llvm-block-frequency-terminology" title="Permalink to this headline"></a></h1>
<div class="contents local topic" id="contents">
<ul class="simple">
<li><a class="reference internal" href="#introduction" id="id1">Introduction</a></li>
<li><a class="reference internal" href="#branch-probability" id="id2">Branch Probability</a></li>
<li><a class="reference internal" href="#branch-weight" id="id3">Branch Weight</a></li>
<li><a class="reference internal" href="#block-frequency" id="id4">Block Frequency</a></li>
<li><a class="reference internal" href="#implementation-a-series-of-dags" id="id5">Implementation: a series of DAGs</a></li>
<li><a class="reference internal" href="#block-mass" id="id6">Block Mass</a></li>
<li><a class="reference internal" href="#loop-scale" id="id7">Loop Scale</a></li>
<li><a class="reference internal" href="#implementation-getting-from-mass-and-scale-to-frequency" id="id8">Implementation: Getting from mass and scale to frequency</a></li>
<li><a class="reference internal" href="#block-bias" id="id9">Block Bias</a></li>
</ul>
</div>
<div class="section" id="introduction">
<h2><a class="toc-backref" href="#id1">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline"></a></h2>
<p>Block Frequency is a metric for estimating the relative frequency of different
basic blocks.  This document describes the terminology that the
<tt class="docutils literal"><span class="pre">BlockFrequencyInfo</span></tt> and <tt class="docutils literal"><span class="pre">MachineBlockFrequencyInfo</span></tt> analysis passes use.</p>
</div>
<div class="section" id="branch-probability">
<h2><a class="toc-backref" href="#id2">Branch Probability</a><a class="headerlink" href="#branch-probability" title="Permalink to this headline"></a></h2>
<p>Blocks with multiple successors have probabilities associated with each
outgoing edge.  These are called branch probabilities.  For a given block, the
sum of its outgoing branch probabilities should be 1.0.</p>
</div>
<div class="section" id="branch-weight">
<h2><a class="toc-backref" href="#id3">Branch Weight</a><a class="headerlink" href="#branch-weight" title="Permalink to this headline"></a></h2>
<p>Rather than storing fractions on each edge, we store an integer weight.
Weights are relative to the other edges of a given predecessor block.  The
branch probability associated with a given edge is its own weight divided by
the sum of the weights on the predecessor&#8217;s outgoing edges.</p>
<p>For example, consider this IR:</p>
<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="kt">void</span> <span class="vg">@foo</span><span class="p">()</span> <span class="p">{</span>
    <span class="c">; ...</span>
    <span class="nl">A:</span>
        <span class="k">br</span> <span class="k">i1</span> <span class="nv">%cond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%B</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%C</span><span class="p">,</span> <span class="nv">!prof</span> <span class="nv-Anonymous">!0</span>
    <span class="c">; ...</span>
<span class="p">}</span>
<span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="kt">metadata</span> <span class="p">!{</span><span class="kt">metadata</span> <span class="nv">!&quot;branch_weights&quot;</span><span class="p">,</span> <span class="k">i32</span> <span class="m">7</span><span class="p">,</span> <span class="k">i32</span> <span class="m">8</span><span class="p">}</span>
</pre></div>
</div>
<p>and this simple graph representation:</p>
<div class="highlight-python"><div class="highlight"><pre>A -&gt; B  (edge-weight: 7)
A -&gt; C  (edge-weight: 8)
</pre></div>
</div>
<p>The probability of branching from block A to block B is 7/15, and the
probability of branching from block A to block C is 8/15.</p>
<p>See <a class="reference internal" href="BranchWeightMetadata.html"><em>LLVM Branch Weight Metadata</em></a> for details about the branch weight IR
representation.</p>
</div>
<div class="section" id="block-frequency">
<h2><a class="toc-backref" href="#id4">Block Frequency</a><a class="headerlink" href="#block-frequency" title="Permalink to this headline"></a></h2>
<p>Block frequency is a relative metric that represents the number of times a
block executes.  The ratio of a block frequency to the entry block frequency is
the expected number of times the block will execute per entry to the function.</p>
<p>Block frequency is the main output of the <tt class="docutils literal"><span class="pre">BlockFrequencyInfo</span></tt> and
<tt class="docutils literal"><span class="pre">MachineBlockFrequencyInfo</span></tt> analysis passes.</p>
</div>
<div class="section" id="implementation-a-series-of-dags">
<h2><a class="toc-backref" href="#id5">Implementation: a series of DAGs</a><a class="headerlink" href="#implementation-a-series-of-dags" title="Permalink to this headline"></a></h2>
<p>The implementation of the block frequency calculation analyses each loop,
bottom-up, ignoring backedges; i.e., as a DAG.  After each loop is processed,
it&#8217;s packaged up to act as a pseudo-node in its parent loop&#8217;s (or the
function&#8217;s) DAG analysis.</p>
</div>
<div class="section" id="block-mass">
<h2><a class="toc-backref" href="#id6">Block Mass</a><a class="headerlink" href="#block-mass" title="Permalink to this headline"></a></h2>
<p>For each DAG, the entry node is assigned a mass of <tt class="docutils literal"><span class="pre">UINT64_MAX</span></tt> and mass is
distributed to successors according to branch weights.  Block Mass uses a
fixed-point representation where <tt class="docutils literal"><span class="pre">UINT64_MAX</span></tt> represents <tt class="docutils literal"><span class="pre">1.0</span></tt> and <tt class="docutils literal"><span class="pre">0</span></tt>
represents a number just above <tt class="docutils literal"><span class="pre">0.0</span></tt>.</p>
<p>After mass is fully distributed, in any cut of the DAG that separates the exit
nodes from the entry node, the sum of the block masses of the nodes succeeded
by a cut edge should equal <tt class="docutils literal"><span class="pre">UINT64_MAX</span></tt>.  In other words, mass is conserved
as it &#8220;falls&#8221; through the DAG.</p>
<p>If a function&#8217;s basic block graph is a DAG, then block masses are valid block
frequencies.  This works poorly in practise though, since downstream users rely
on adding block frequencies together without hitting the maximum.</p>
</div>
<div class="section" id="loop-scale">
<h2><a class="toc-backref" href="#id7">Loop Scale</a><a class="headerlink" href="#loop-scale" title="Permalink to this headline"></a></h2>
<p>Loop scale is a metric that indicates how many times a loop iterates per entry.
As mass is distributed through the loop&#8217;s DAG, the (otherwise ignored) backedge
mass is collected.  This backedge mass is used to compute the exit frequency,
and thus the loop scale.</p>
</div>
<div class="section" id="implementation-getting-from-mass-and-scale-to-frequency">
<h2><a class="toc-backref" href="#id8">Implementation: Getting from mass and scale to frequency</a><a class="headerlink" href="#implementation-getting-from-mass-and-scale-to-frequency" title="Permalink to this headline"></a></h2>
<p>After analysing the complete series of DAGs, each block has a mass (local to
its containing loop, if any), and each loop pseudo-node has a loop scale and
its own mass (from its parent&#8217;s DAG).</p>
<p>We can get an initial frequency assignment (with entry frequency of 1.0) by
multiplying these masses and loop scales together.  A given block&#8217;s frequency
is the product of its mass, the mass of containing loops&#8217; pseudo nodes, and the
containing loops&#8217; loop scales.</p>
<p>Since downstream users need integers (not floating point), this initial
frequency assignment is shifted as necessary into the range of <tt class="docutils literal"><span class="pre">uint64_t</span></tt>.</p>
</div>
<div class="section" id="block-bias">
<h2><a class="toc-backref" href="#id9">Block Bias</a><a class="headerlink" href="#block-bias" title="Permalink to this headline"></a></h2>
<p>Block bias is a proposed <em>absolute</em> metric to indicate a bias toward or away
from a given block during a function&#8217;s execution.  The idea is that bias can be
used in isolation to indicate whether a block is relatively hot or cold, or to
compare two blocks to indicate whether one is hotter or colder than the other.</p>
<p>The proposed calculation involves calculating a <em>reference</em> block frequency,
where:</p>
<ul class="simple">
<li>every branch weight is assumed to be 1 (i.e., every branch probability
distribution is even) and</li>
<li>loop scales are ignored.</li>
</ul>
<p>This reference frequency represents what the block frequency would be in an
unbiased graph.</p>
<p>The bias is the ratio of the block frequency to this reference block frequency.</p>
</div>
</div>


          </div>
      </div>
      <div class="clearer"></div>
    </div>
    <div class="related">
      <h3>Navigation</h3>
      <ul>
        <li class="right" style="margin-right: 10px">
          <a href="genindex.html" title="General Index"
             >index</a></li>
        <li class="right" >
          <a href="BranchWeightMetadata.html" title="LLVM Branch Weight Metadata"
             >next</a> |</li>
        <li class="right" >
          <a href="BitCodeFormat.html" title="LLVM Bitcode File Format"
             >previous</a> |</li>
  <li><a href="http://llvm.org/">LLVM Home</a>&nbsp;|&nbsp;</li>
  <li><a href="index.html">Documentation</a>&raquo;</li>
 
      </ul>
    </div>
    <div class="footer">
        &copy; Copyright 2003-2014, LLVM Project.
      Last updated on 2015-02-12.
      Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.2.2.
    </div>
  </body>
</html>