576 lines
No EOL
40 KiB
HTML
576 lines
No EOL
40 KiB
HTML
|
||
<!DOCTYPE html>
|
||
|
||
|
||
|
||
|
||
<html lang="en">
|
||
<head>
|
||
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
|
||
|
||
<title>7.5. Barriers — Computer Systems Fundamentals</title>
|
||
|
||
<link rel="stylesheet" href="_static/css/bootstrap.min.css" integrity="sha384-ggOyR0iXCbMQv3Xipma34MD+dH/1fQ784/j6cY/iJTQUOhcWr7x9JvoRxT2MZw1T" crossorigin="anonymous" />
|
||
<link rel="stylesheet" href="_static/css/pygments.css" type="text/css" />
|
||
<link rel="stylesheet" href="_static/css/normalize.css" type="text/css" />
|
||
<link rel="stylesheet" href="../../../JSAV/css/JSAV.css" type="text/css" />
|
||
<link rel="stylesheet" href="../../../lib/odsaMOD-min.css" type="text/css" />
|
||
<link rel="stylesheet" href="_static/css/jquery-1.11.4-smoothness-ui.css" type="text/css" />
|
||
<link rel="stylesheet" href="../../../lib/odsaStyle-min.css" type="text/css" />
|
||
<link rel="stylesheet" href="_static/css/csf.css" type="text/css" />
|
||
|
||
<style>
|
||
.underline { text-decoration: underline; }
|
||
</style>
|
||
|
||
<script type="text/javascript">
|
||
var DOCUMENTATION_OPTIONS = {
|
||
URL_ROOT: './',
|
||
VERSION: '0.4.1',
|
||
COLLAPSE_INDEX: false,
|
||
FILE_SUFFIX: '.html',
|
||
HAS_SOURCE: true
|
||
};
|
||
</script>
|
||
|
||
<script type="text/x-mathjax-config">
|
||
MathJax.Hub.Config({
|
||
tex2jax: {
|
||
inlineMath: [['$','$'], ['\\(','\\)']],
|
||
displayMath: [ ['$$','$$'], ["\\[","\\]"] ],
|
||
processEscapes: true
|
||
},
|
||
"HTML-CSS": {
|
||
scale: "80"
|
||
}
|
||
});
|
||
</script>
|
||
<link rel="shortcut icon" href="_static/favicon.ico"/>
|
||
<link rel="index" title="Index" href="genindex.html" />
|
||
<link rel="search" title="Search" href="search.html" />
|
||
<link rel="index" title="Computer Systems Fundamentals" href="index.html" />
|
||
<link rel="next" title="6. Condition Variables" href="Condvars.html" />
|
||
<link rel="prev" title="4. Semaphores" href="Semaphores.html" />
|
||
|
||
</head><body>
|
||
|
||
<nav class="navbar navbar-expand-md navbar-dark navbar-custom fixed-top">
|
||
|
||
<a class="navbar-brand py-0" href="index.html"><img src="_static/CSF-Logo-Square-Text.png" alt="OpenCSF Logo" height="40em" class="py-1 px-2 mb-0 align-center rounded-lg bg-white" /></a>
|
||
<!-- Show a navbar toggler on mobile -->
|
||
<button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#defaultNavbars" aria-controls="defaultNavbars" aria-expanded="false" aria-label="Toggle navigation">
|
||
<span class="navbar-toggler-icon"></span>
|
||
</button>
|
||
<div class="collapse navbar-collapse" id="defaultNavbars">
|
||
<ul class="navbar-nav mr-auto">
|
||
<li class="nav-item dropdown">
|
||
<a class="nav-link dropdown-toggle jmu-gold rounded" href="Barriers.html#" id="navbarDropdownChapters" role="button" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">Contents</a>
|
||
<div class="dropdown-menu scrollable-menu" role="menu" aria-labelledby="navbarDropdownChapters">
|
||
<a class="dropdown-item" tabindex="-1" href="Barriers.html#"><b>Chapter 1</b></a>
|
||
<a class="dropdown-item" href="IntroConcSysOverview.html"> 1.1. Introduction to Concurrent Systems</a>
|
||
<a class="dropdown-item" href="SysAndModels.html"> 1.2. Systems and Models</a>
|
||
<a class="dropdown-item" href="Themes.html"> 1.3. Themes and Guiding Principles</a>
|
||
<a class="dropdown-item" href="Architectures.html"> 1.4. System Architectures</a>
|
||
<a class="dropdown-item" href="StateModels.html"> 1.5. State Models in UML</a>
|
||
<a class="dropdown-item" href="SequenceModels.html"> 1.6. Sequence Models in UML</a>
|
||
<a class="dropdown-item" href="StateModelImplementation.html"> 1.7. Extended Example: State Model Implementation</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 2</b></a>
|
||
<a class="dropdown-item" href="ProcessesOverview.html"> 2.1. Processes and OS Basics</a>
|
||
<a class="dropdown-item" href="Multiprogramming.html"> 2.2. Processes and Multiprogramming</a>
|
||
<a class="dropdown-item" href="KernelMechanics.html"> 2.3. Kernel Mechanics</a>
|
||
<a class="dropdown-item" href="Syscall.html"> 2.4. System Call Interface</a>
|
||
<a class="dropdown-item" href="ProcessCycle.html"> 2.5. Process Life Cycle</a>
|
||
<a class="dropdown-item" href="UnixFile.html"> 2.6. The UNIX File Abstraction</a>
|
||
<a class="dropdown-item" href="EventsSignals.html"> 2.7. Events and Signals</a>
|
||
<a class="dropdown-item" href="Extended2Processes.html"> 2.8. Extended Example: Listing Files with Processes</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 3</b></a>
|
||
<a class="dropdown-item" href="IPCOverview.html"> 3.1. Concurrency with IPC</a>
|
||
<a class="dropdown-item" href="IPCModels.html"> 3.2. IPC Models</a>
|
||
<a class="dropdown-item" href="Pipes.html"> 3.3. Pipes and FIFOs</a>
|
||
<a class="dropdown-item" href="MMap.html"> 3.4. Shared Memory With Memory-mapped Files</a>
|
||
<a class="dropdown-item" href="POSIXvSysV.html"> 3.5. POSIX vs. System V IPC</a>
|
||
<a class="dropdown-item" href="MQueues.html"> 3.6. Message Passing With Message Queues</a>
|
||
<a class="dropdown-item" href="ShMem.html"> 3.7. Shared Memory</a>
|
||
<a class="dropdown-item" href="IPCSems.html"> 3.8. Semaphores</a>
|
||
<a class="dropdown-item" href="Extended3Bash.html"> 3.9. Extended Example: Bash-lite: A Simple Command-line Shell</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 4</b></a>
|
||
<a class="dropdown-item" href="SocketsOverview.html"> 4.1. Networked Concurrency</a>
|
||
<a class="dropdown-item" href="FiveLayer.html"> 4.2. The TCP/IP Internet Model</a>
|
||
<a class="dropdown-item" href="NetApps.html"> 4.3. Network Applications and Protocols</a>
|
||
<a class="dropdown-item" href="Sockets.html"> 4.4. The Socket Interface</a>
|
||
<a class="dropdown-item" href="TCPSockets.html"> 4.5. TCP Socket Programming: HTTP</a>
|
||
<a class="dropdown-item" href="UDPSockets.html"> 4.6. UDP Socket Programming: DNS</a>
|
||
<a class="dropdown-item" href="AppBroadcast.html"> 4.7. Application-Layer Broadcasting: DHCP</a>
|
||
<a class="dropdown-item" href="Extended4CGI.html"> 4.8. Extended Example: CGI Web Server</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 5</b></a>
|
||
<a class="dropdown-item" href="InternetOverview.html"> 5.1. The Internet and Connectivity</a>
|
||
<a class="dropdown-item" href="AppLayer.html"> 5.2. Application Layer: Overlay Networks</a>
|
||
<a class="dropdown-item" href="TransLayer.html"> 5.3. Transport Layer</a>
|
||
<a class="dropdown-item" href="NetSec.html"> 5.4. Network Security Fundamentals</a>
|
||
<a class="dropdown-item" href="NetLayer.html"> 5.5. Network Layer: IP</a>
|
||
<a class="dropdown-item" href="LinkLayer.html"> 5.6. Link Layer</a>
|
||
<a class="dropdown-item" href="Wireless.html"> 5.7. Wireless Connectivity: Wi-Fi, Bluetooth, and Zigbee</a>
|
||
<a class="dropdown-item" href="Extended5DNS.html"> 5.8. Extended Example: DNS client</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 6</b></a>
|
||
<a class="dropdown-item" href="ThreadsOverview.html"> 6.1. Concurrency with Multithreading</a>
|
||
<a class="dropdown-item" href="ProcVThreads.html"> 6.2. Processes vs. Threads</a>
|
||
<a class="dropdown-item" href="RaceConditions.html"> 6.3. Race Conditions and Critical Sections</a>
|
||
<a class="dropdown-item" href="POSIXThreads.html"> 6.4. POSIX Thread Library</a>
|
||
<a class="dropdown-item" href="ThreadArgs.html"> 6.5. Thread Arguments and Return Values</a>
|
||
<a class="dropdown-item" href="ImplicitThreads.html"> 6.6. Implicit Threading and Language-based Threads</a>
|
||
<a class="dropdown-item" href="Extended6Input.html"> 6.7. Extended Example: Keyboard Input Listener</a>
|
||
<a class="dropdown-item" href="Extended6Primes.html"> 6.8. Extended Example: Concurrent Prime Number Search</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 7</b></a>
|
||
<a class="dropdown-item" href="SynchOverview.html"> 7.1. Synchronization Primitives</a>
|
||
<a class="dropdown-item" href="CritSect.html"> 7.2. Critical Sections and Peterson's Solution</a>
|
||
<a class="dropdown-item" href="Locks.html"> 7.3. Locks</a>
|
||
<a class="dropdown-item" href="Semaphores.html"> 7.4. Semaphores</a>
|
||
<a class="dropdown-item" href="Barriers.html"> 7.5. Barriers</a>
|
||
<a class="dropdown-item" href="Condvars.html"> 7.6. Condition Variables</a>
|
||
<a class="dropdown-item" href="Deadlock.html"> 7.7. Deadlock</a>
|
||
<a class="dropdown-item" href="Extended7Events.html"> 7.8. Extended Example: Event Log File</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 8</b></a>
|
||
<a class="dropdown-item" href="SynchProblemsOverview.html"> 8.1. Synchronization Patterns and Problems</a>
|
||
<a class="dropdown-item" href="SynchDesign.html"> 8.2. Basic Synchronization Design Patterns</a>
|
||
<a class="dropdown-item" href="ProdCons.html"> 8.3. Producer-Consumer Problem</a>
|
||
<a class="dropdown-item" href="ReadWrite.html"> 8.4. Readers-Writers Problem</a>
|
||
<a class="dropdown-item" href="DiningPhil.html"> 8.5. Dining Philosophers Problem and Deadlock</a>
|
||
<a class="dropdown-item" href="CigSmokers.html"> 8.6. Cigarette Smokers Problem and the Limits of Semaphores and Locks</a>
|
||
<a class="dropdown-item" href="Extended8ModExp.html"> 8.7. Extended Example: Parallel Modular Exponentiation</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Chapter 9</b></a>
|
||
<a class="dropdown-item" href="ParallelDistributedOverview.html"> 9.1. Parallel and Distributed Systems</a>
|
||
<a class="dropdown-item" href="ParVConc.html"> 9.2. Parallelism vs. Concurrency</a>
|
||
<a class="dropdown-item" href="ParallelDesign.html"> 9.3. Parallel Design Patterns</a>
|
||
<a class="dropdown-item" href="Scaling.html"> 9.4. Limits of Parallelism and Scaling</a>
|
||
<a class="dropdown-item" href="DistTiming.html"> 9.5. Timing in Distributed Environments</a>
|
||
<a class="dropdown-item" href="DistDataStorage.html"> 9.6. Reliable Data Storage and Location</a>
|
||
<a class="dropdown-item" href="DistConsensus.html"> 9.7. Consensus in Distributed Systems</a>
|
||
<a class="dropdown-item" href="Extended9Blockchain.html"> 9.8. Extended Example: Blockchain Proof-of-Work</a>
|
||
<div class="dropdown-divider"></div>
|
||
<a class="dropdown-item disabled"><b>Appendix A</b></a>
|
||
<a class="dropdown-item" href="CLangOverview.html"> A.1. C Language Reintroduction</a>
|
||
<a class="dropdown-item" href="Debugging.html"> A.2. Documentation and Debugging</a>
|
||
<a class="dropdown-item" href="BasicTypes.html"> A.3. Basic Types and Pointers</a>
|
||
<a class="dropdown-item" href="Arrays.html"> A.4. Arrays, Structs, Enums, and Type Definitions</a>
|
||
<a class="dropdown-item" href="Functions.html"> A.5. Functions and Scope</a>
|
||
<a class="dropdown-item" href="Pointers.html"> A.6. Pointers and Dynamic Allocation</a>
|
||
<a class="dropdown-item" href="Strings.html"> A.7. Strings</a>
|
||
<a class="dropdown-item" href="FunctionPointers.html"> A.8. Function Pointers</a>
|
||
<a class="dropdown-item" href="Files.html"> A.9. Files</a>
|
||
</div>
|
||
</li>
|
||
|
||
|
||
|
||
</ul>
|
||
</div>
|
||
|
||
<ul class="navbar-nav flex-row ml-md-auto d-none d-md-flex">
|
||
<li class="nav-item"><a class="nav-link jmu-gold" href="https://w3.cs.jmu.edu/kirkpams/OpenCSF/Books/csf/source/Barriers.rst"
|
||
target="_blank" rel="nofollow">Show Source</a></li>
|
||
|
||
</ul>
|
||
</nav>
|
||
|
||
|
||
<div class="container center">
|
||
«  <a id="prevmod" href="Semaphores.html">7.4. Semaphores</a>
|
||
  ::  
|
||
<a class="uplink" href="index.html">Contents</a>
|
||
  ::  
|
||
<a id="nextmod" href="Condvars.html">7.6. Condition Variables</a>  »
|
||
|
||
</div>
|
||
<br />
|
||
|
||
|
||
|
||
<script type="text/javascript" src="_static/js/jquery-2.1.4.min.js"></script>
|
||
<script type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.1/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
|
||
<script type="text/javascript" src="_static/js/jquery-1.11.4-ui.min.js"></script>
|
||
<script type="text/javascript" src="_static/js/forge-0.7.0.min.js"></script>
|
||
<script type="text/javascript" src="../../../JSAV/lib/jquery.transit.js"></script>
|
||
<script type="text/javascript" src="../../../JSAV/lib/raphael.js"></script>
|
||
<script type="text/javascript" src="../../../JSAV/build/JSAV-min.js"></script>
|
||
<script type="text/javascript" src="_static/js/config.js"></script>
|
||
<script type="text/javascript" src="../../../lib/odsaUtils-min.js"></script>
|
||
<script type="text/javascript" src="../../../lib/odsaMOD-min.js"></script>
|
||
<script type="text/javascript" src="_static/js/d3-4.13.0.min.js"></script>
|
||
<script type="text/javascript" src="_static/js/d3-selection-multi.v1.min.js"></script>
|
||
<script type="text/javascript" src="../../../lib/dataStructures.js"></script>
|
||
|
||
|
||
|
||
|
||
|
||
|
||
<div class="container">
|
||
|
||
<script>ODSA.SETTINGS.DISP_MOD_COMP = true;ODSA.SETTINGS.MODULE_NAME = "Barriers";ODSA.SETTINGS.MODULE_LONG_NAME = "Barriers";ODSA.SETTINGS.MODULE_CHAPTER = "Synchronization Primitives"; ODSA.SETTINGS.BUILD_DATE = "2021-06-01 15:31:50"; ODSA.SETTINGS.BUILD_CMAP = false;JSAV_OPTIONS['lang']='en';JSAV_EXERCISE_OPTIONS['code']='java';</script><div class="section" id="id1">
|
||
<h1>7.5. Barriers<a class="headerlink" href="Barriers.html#id1" title="Permalink to this headline">¶</a></h1>
|
||
<p>Semaphores provide a general mechanism that allows one thread to indicate that a
|
||
particular event has happened. One type of event that arises quite commonly in
|
||
concurrent applications is the need to make sure all threads reach some common
|
||
point before any of them continues on. That is, multiple threads may perform
|
||
some initial computation then pause; at that point some manager thread checks
|
||
these initial results and allows the threads to continue once their results have been approved.</p>
|
||
<p>As an example, consider the case of an autonomous car or flying drone. These
|
||
types of systems require accurate measurements of real-world phenomena like
|
||
speed, position, and acceleration. However, these measurements involve
|
||
floating-point calculations that may encounter rounding errors. To improve the
|
||
accuracy of the vehicle’s autonomous controls, multiple threads may be
|
||
simultaneously and independently performing these calculations. The manager
|
||
control thread will occasionally require these threads to reach a common point
|
||
in time where each thread reports on its best guess of the measurements. The
|
||
manager may then correct errors in some threads, then allow them to continue.</p>
|
||
<p>Achieving this kind of synchronization with semaphores is complex and
|
||
error-prone. One approach would be to use two semaphores: the manager thread
|
||
would repeatedly wait on one of them as the threads were finishing their
|
||
calculations; the manager would then signal on a different semaphore that each
|
||
thread was waiting on. However, if all threads agree but one crashes, then the
|
||
entire system would fail; the manager thread would continue to wait on that last
|
||
thread to finish (which will never happen).</p>
|
||
<p>As an alternative, <a class="reference internal" href="Glossary.html#term-barrier"><span class="xref std std-term">synchronization barriers</span></a>
|
||
allow programs to specify a <em>minimum</em> number of threads that must reach a common
|
||
point before any can continue. For instance, if the autonomous vehicle manager
|
||
thread receives reports from 7 out of 10 threads about the current position,
|
||
then the system may be designed to consider this measurement <em>good enough</em> and
|
||
let the threads continue rather than requiring the last three threads to
|
||
finish.</p>
|
||
<p>The base functionality of barriers is specified with thread functions. The
|
||
barrier is initialized with <code class="docutils literal notranslate"><span class="pre">pthread_barrier_init()</span></code>, with the last parameter
|
||
indicating the <em>minimum</em> number of threads that must reach the barrier before
|
||
any can be allowed to continue. Threads use the <code class="docutils literal notranslate"><span class="pre">pthread_barrier_wait()</span></code>
|
||
function to indicate that they have reached the common point. If the minimum
|
||
number of threads have not yet reached the barrier, then the current thread will
|
||
become blocked. Once the minimum number has been reached, all threads waiting at
|
||
the barrier become unblocked; any future thread that calls
|
||
<code class="docutils literal notranslate"><span class="pre">pthread_barrier_wait()</span></code> will pass right through the barrier without any
|
||
delay. Finally, <code class="docutils literal notranslate"><span class="pre">pthread_barrier_destroy()</span></code> is used to clean up any resources
|
||
associated with the barrier.</p>
|
||
<div class="topic border border-dark rounded-lg bg-light px-2 mb-3">
|
||
<div class="figure align-left">
|
||
<a class="reference internal image-reference" href="_images/CSF-Images-Library.png"><img alt="Decorative C library image" src="_images/CSF-Images-Library.png" style="width: 100%;" /></a>
|
||
</div>
|
||
<p class="topic-title first pt-2 mb-1">C library functions – <pthread.h></p><hr class="mt-1" />
|
||
<dl class="docutils">
|
||
<dt><code class="docutils literal notranslate"><span class="pre">int</span> <span class="pre">pthread_barrier_init</span> <span class="pre">(pthread_barrier_t</span> <span class="pre">*barrier,</span> <span class="pre">const</span> <span class="pre">pthread_barrierattr_t</span> <span class="pre">*attr,</span> <span class="pre">unsigned</span> <span class="pre">count);</span></code></dt>
|
||
<dd>Initialize a synchronization barrier with the specified attributes.</dd>
|
||
<dt><code class="docutils literal notranslate"><span class="pre">int</span> <span class="pre">pthread_barrier_destroy</span> <span class="pre">(pthread_barrier_t</span> <span class="pre">*barrier);</span></code></dt>
|
||
<dd>Delete a synchronization barrier.</dd>
|
||
<dt><code class="docutils literal notranslate"><span class="pre">int</span> <span class="pre">pthread_barrier_wait</span> <span class="pre">(pthread_barrier_t</span> <span class="pre">*barrier);</span></code></dt>
|
||
<dd>Make a thread wait until enough have reached the barrier.</dd>
|
||
</dl>
|
||
</div>
|
||
<div class="section" id="concurrent-calculations-with-barriers">
|
||
<h2>7.5.1. Concurrent Calculations with Barriers<a class="headerlink" href="Barriers.html#concurrent-calculations-with-barriers" title="Permalink to this headline">¶</a></h2>
|
||
<p>To show how barriers work, consider using two threads to determine which
|
||
function grows faster: the Fibonnacci sequence or the exponential function. That
|
||
is, is the 100th Fibonacci number greater than $c^{100}$ for some constant $c$? <a class="footnote-reference" href="Barriers.html#f43" id="id2">[1]</a>
|
||
It can be shown that each Fibonacci number is slightly less than double
|
||
the previous value, so the answer would be no for any constant integer greater
|
||
than 1. But what if the constant is 1.6? <a class="footnote-reference" href="Barriers.html#f44" id="id3">[2]</a> Then the answer is not obvious,
|
||
but it is easy enough to calculate.</p>
|
||
<p><a class="reference external" href="Barriers.html#cl7-13">Code Listing 7.13</a> shows how barriers can address this problem. The
|
||
<code class="docutils literal notranslate"><span class="pre">fibonacci()</span></code> and <code class="docutils literal notranslate"><span class="pre">exponential()</span></code> threads take a struct that contains a
|
||
shared <code class="docutils literal notranslate"><span class="pre">pthread_barrier_t</span></code>, a digit that specifies how far into the sequence
|
||
and what exponent to calculate, and a base value that specifies the base of the
|
||
exponent. That is, if digit is specified as 30 and base is 1.6, <code class="docutils literal notranslate"><span class="pre">fibonacci()</span></code>
|
||
will calculate the 30th Fibonacci number, while <code class="docutils literal notranslate"><span class="pre">exponent()</span></code> will calculate
|
||
(approximately) $1.6^{30}$.</p>
|
||
<div class="highlight-c border border-dark rounded-lg bg-light px-0 mb-3 notranslate" id="cl7-13"><table class="highlighttable"><tr><td class="linenos px-0 mx-0"><div class="linenodiv"><pre class="mb-0"> 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</pre></div></td><td class="code"><div class="highlight bg-light"><pre class="mb-0"><span></span><span class="cm">/* Code Listing 7.13:</span>
|
||
<span class="cm"> Using a barrier to pause multiple threads</span>
|
||
<span class="cm"> */</span>
|
||
|
||
<span class="cm">/* Use a common struct with the barrier and parameters */</span>
|
||
<span class="k">struct</span> <span class="n">args</span> <span class="p">{</span>
|
||
<span class="n">pthread_barrier_t</span> <span class="o">*</span><span class="n">barrier</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">int</span> <span class="n">digit</span><span class="p">;</span>
|
||
<span class="kt">double</span> <span class="n">base</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">long</span> <span class="n">fib</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">long</span> <span class="n">exp</span><span class="p">;</span>
|
||
<span class="p">};</span>
|
||
|
||
<span class="cm">/* Iterative version of the Fibonacci calculation */</span>
|
||
<span class="kt">void</span> <span class="o">*</span>
|
||
<span class="nf">fibonacci</span> <span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">_args</span><span class="p">)</span>
|
||
<span class="p">{</span>
|
||
<span class="k">struct</span> <span class="n">args</span> <span class="o">*</span><span class="n">args</span> <span class="o">=</span> <span class="p">(</span><span class="k">struct</span> <span class="n">args</span> <span class="o">*</span><span class="p">)</span> <span class="n">_args</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">long</span> <span class="n">previous</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">long</span> <span class="n">current</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">int</span> <span class="n">i</span><span class="p">;</span>
|
||
|
||
<span class="k">for</span> <span class="p">(</span><span class="n">i</span> <span class="o">=</span> <span class="mi">2</span><span class="p">;</span> <span class="n">i</span> <span class="o"><</span> <span class="n">args</span><span class="o">-></span><span class="n">digit</span><span class="p">;</span> <span class="n">i</span><span class="o">++</span><span class="p">)</span>
|
||
<span class="p">{</span>
|
||
<span class="n">current</span> <span class="o">+=</span> <span class="n">previous</span><span class="p">;</span>
|
||
<span class="n">previous</span> <span class="o">=</span> <span class="n">current</span> <span class="o">-</span> <span class="n">previous</span><span class="p">;</span>
|
||
<span class="p">}</span>
|
||
<span class="cm">/* current is the i-th Fibonacci number */</span>
|
||
<span class="n">args</span><span class="o">-></span><span class="n">fib</span> <span class="o">=</span> <span class="n">current</span><span class="p">;</span>
|
||
|
||
<span class="cm">/* Wait at the barrier and compare results */</span>
|
||
<span class="n">pthread_barrier_wait</span> <span class="p">(</span><span class="n">args</span><span class="o">-></span><span class="n">barrier</span><span class="p">);</span>
|
||
<span class="k">if</span> <span class="p">(</span><span class="n">args</span><span class="o">-></span><span class="n">fib</span> <span class="o">></span> <span class="n">args</span><span class="o">-></span><span class="n">exp</span><span class="p">)</span>
|
||
<span class="n">printf</span> <span class="p">(</span><span class="s">"Fibonacci wins: %lu > %lu</span><span class="se">\n</span><span class="s">"</span><span class="p">,</span> <span class="n">args</span><span class="o">-></span><span class="n">fib</span><span class="p">,</span> <span class="n">args</span><span class="o">-></span><span class="n">exp</span><span class="p">);</span>
|
||
|
||
<span class="n">pthread_exit</span> <span class="p">(</span><span class="nb">NULL</span><span class="p">);</span>
|
||
<span class="p">}</span>
|
||
|
||
<span class="cm">/* Calculate base^digit using repeated multiplication */</span>
|
||
<span class="kt">void</span> <span class="o">*</span>
|
||
<span class="nf">exponential</span> <span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">_args</span><span class="p">)</span>
|
||
<span class="p">{</span>
|
||
<span class="k">struct</span> <span class="n">args</span> <span class="o">*</span><span class="n">args</span> <span class="o">=</span> <span class="p">(</span><span class="k">struct</span> <span class="n">args</span> <span class="o">*</span><span class="p">)</span> <span class="n">_args</span><span class="p">;</span>
|
||
<span class="kt">double</span> <span class="n">result</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span>
|
||
<span class="kt">unsigned</span> <span class="kt">int</span> <span class="n">i</span><span class="p">;</span>
|
||
|
||
<span class="cm">/* WARNING: Don't actually do this in real code that needs</span>
|
||
<span class="cm"> Accurate calculations. Repeated floating point arithmetic</span>
|
||
<span class="cm"> inherently has rounding errors that will compound. */</span>
|
||
|
||
<span class="cm">/* Calculate base^digit; cast to unsigned long for comparison */</span>
|
||
<span class="k">for</span> <span class="p">(</span><span class="n">i</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span> <span class="n">i</span> <span class="o"><</span> <span class="n">args</span><span class="o">-></span><span class="n">digit</span><span class="p">;</span> <span class="n">i</span><span class="o">++</span><span class="p">)</span>
|
||
<span class="n">result</span> <span class="o">*=</span> <span class="n">args</span><span class="o">-></span><span class="n">base</span><span class="p">;</span>
|
||
<span class="n">args</span><span class="o">-></span><span class="n">exp</span> <span class="o">=</span> <span class="p">(</span><span class="kt">unsigned</span> <span class="kt">long</span><span class="p">)</span> <span class="n">result</span><span class="p">;</span>
|
||
|
||
<span class="cm">/* Wait on threads to reach the barrier and check the result */</span>
|
||
<span class="n">pthread_barrier_wait</span> <span class="p">(</span><span class="n">args</span><span class="o">-></span><span class="n">barrier</span><span class="p">);</span>
|
||
<span class="k">if</span> <span class="p">(</span><span class="n">args</span><span class="o">-></span><span class="n">exp</span> <span class="o">></span> <span class="n">args</span><span class="o">-></span><span class="n">fib</span><span class="p">)</span>
|
||
<span class="n">printf</span> <span class="p">(</span><span class="s">"Exponential wins: %lu > %lu</span><span class="se">\n</span><span class="s">"</span><span class="p">,</span> <span class="n">args</span><span class="o">-></span><span class="n">exp</span><span class="p">,</span> <span class="n">args</span><span class="o">-></span><span class="n">fib</span><span class="p">);</span>
|
||
|
||
<span class="n">pthread_exit</span> <span class="p">(</span><span class="nb">NULL</span><span class="p">);</span>
|
||
<span class="p">}</span>
|
||
</pre></div>
|
||
</td></tr></table></div>
|
||
<div class="topic border border-dark rounded-lg alert-danger px-2 mb-3">
|
||
<div class="figure align-left">
|
||
<a class="reference internal image-reference" href="_images/CSF-Images-BugWarning.png"><img alt="Decorative bug warning" src="_images/CSF-Images-BugWarning.png" style="width: 90%;" /></a>
|
||
</div>
|
||
<p class="topic-title first pt-2 mb-1">Bug Warning</p><hr class="mt-1" />
|
||
<p>As indicated in the code, you should not try to calculate the exponential value
|
||
as shown in this code. For one thing, this code is (intentionally) inefficient
|
||
and could be easily optimized. More importantly, though, the result is
|
||
guaranteed to be inaccurate. Floating-point arithmetic is susceptible to
|
||
rounding errors, particularly with repeated calculations and large values.
|
||
Consequently, this example should primarily be considered for its use of
|
||
barriers rather than its accuracy regarding the comparison of Fibonacci and
|
||
exponential growth.</p>
|
||
</div>
|
||
<p>The key lines to note regarding these threads are the calls to
|
||
<code class="docutils literal notranslate"><span class="pre">pthread_barrier_wait()</span></code> and the references to the fields <code class="docutils literal notranslate"><span class="pre">fib</span></code> and <code class="docutils literal notranslate"><span class="pre">exp</span></code>,
|
||
which store the results of each calculation. Specifically, note that there is no
|
||
race condition regarding these fields because the threads only <em>read</em> the values
|
||
after the barrier. That is, nothing changes after the barrier, so there is no
|
||
concern that the values checked are inaccurate. This guarantee arises from the
|
||
fact that both threads must reach the barrier (i.e., the
|
||
<code class="docutils literal notranslate"><span class="pre">pthread_barrier_wait()</span></code> call) before either can proceed.</p>
|
||
<p>One question that cannot be answered with this code is which thread runs faster.
|
||
With barriers, all we can say for sure is that both threads reach the barrier
|
||
before either proceeds. We have no information about which one arrived first.
|
||
From the perspective of the barrier, such a question is irrelevant. The only
|
||
thing that matters is that both finish before they each check the results.</p>
|
||
<p><a class="reference external" href="Barriers.html#cl7-14">Code Listing 7.14</a> shows the initialization required. The main
|
||
thread initializes the barrier so that 2 threads are required, then creates the
|
||
threads. Notice that both threads receive pointers to the same <code class="docutils literal notranslate"><span class="pre">struct</span></code>
|
||
instance, so they share the same <code class="docutils literal notranslate"><span class="pre">barrier</span></code>, <code class="docutils literal notranslate"><span class="pre">fib</span></code>, and <code class="docutils literal notranslate"><span class="pre">exp</span></code> fields. The
|
||
main thread joins both of the helper threads, but the join must occur after the
|
||
threads have called <code class="docutils literal notranslate"><span class="pre">pthread_exit()</span></code>. Since both threads call
|
||
<code class="docutils literal notranslate"><span class="pre">pthread_barrier_wait()</span></code> before that, the main thread is guaranteed to clean
|
||
up the barrier only <em>after</em> it is no longer needed.</p>
|
||
<div class="highlight-c border border-dark rounded-lg bg-light px-0 mb-3 notranslate" id="cl7-14"><table class="highlighttable"><tr><td class="linenos px-0 mx-0"><div class="linenodiv"><pre class="mb-0"> 1
|
||
2
|
||
3
|
||
4
|
||
5
|
||
6
|
||
7
|
||
8
|
||
9
|
||
10
|
||
11
|
||
12
|
||
13
|
||
14
|
||
15
|
||
16
|
||
17
|
||
18
|
||
19
|
||
20
|
||
21
|
||
22
|
||
23
|
||
24</pre></div></td><td class="code"><div class="highlight bg-light"><pre class="mb-0"><span></span><span class="cm">/* Code Listing 7.14:</span>
|
||
<span class="cm"> Initializing the barrier and data structures for Code Listing 7.13
</span>
|
||
<span class="cm"> */</span>
|
||
|
||
<span class="cm">/* Initialize the barrier so that both threads must wait */</span>
|
||
<span class="n">pthread_barrier_t</span> <span class="n">barrier</span><span class="p">;</span>
|
||
<span class="n">pthread_barrier_init</span> <span class="p">(</span><span class="o">&</span><span class="n">barrier</span><span class="p">,</span> <span class="nb">NULL</span><span class="p">,</span> <span class="mi">2</span><span class="p">);</span>
|
||
|
||
<span class="cm">/* Set up parameters to get the 30th Fibonacci and 1.6^30 */</span>
|
||
<span class="k">struct</span> <span class="n">args</span> <span class="n">args</span><span class="p">;</span>
|
||
<span class="n">args</span><span class="p">.</span><span class="n">barrier</span> <span class="o">=</span> <span class="o">&</span><span class="n">barrier</span><span class="p">;</span>
|
||
<span class="n">args</span><span class="p">.</span><span class="n">digit</span> <span class="o">=</span> <span class="mi">30</span><span class="p">;</span>
|
||
<span class="n">args</span><span class="p">.</span><span class="n">base</span> <span class="o">=</span> <span class="mf">1.6</span><span class="p">;</span>
|
||
<span class="n">args</span><span class="p">.</span><span class="n">fib</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
|
||
<span class="n">args</span><span class="p">.</span><span class="n">exp</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
|
||
|
||
<span class="cm">/* Create and join the threads, then destroy the barrier */</span>
|
||
<span class="n">assert</span> <span class="p">(</span><span class="n">pthread_create</span> <span class="p">(</span><span class="o">&</span><span class="n">threads</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="nb">NULL</span><span class="p">,</span> <span class="n">fibonacci</span><span class="p">,</span> <span class="o">&</span><span class="n">args</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">);</span>
|
||
<span class="n">assert</span> <span class="p">(</span><span class="n">pthread_create</span> <span class="p">(</span><span class="o">&</span><span class="n">threads</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="nb">NULL</span><span class="p">,</span> <span class="n">exponential</span><span class="p">,</span> <span class="o">&</span><span class="n">args</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">);</span>
|
||
|
||
<span class="n">pthread_join</span> <span class="p">(</span><span class="n">threads</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="nb">NULL</span><span class="p">);</span>
|
||
<span class="n">pthread_join</span> <span class="p">(</span><span class="n">threads</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="nb">NULL</span><span class="p">);</span>
|
||
|
||
<span class="n">pthread_barrier_destroy</span> <span class="p">(</span><span class="o">&</span><span class="n">barrier</span><span class="p">);</span>
|
||
</pre></div>
|
||
</td></tr></table></div>
|
||
<table class="docutils footnote" frame="void" id="f43" rules="none">
|
||
<colgroup><col class="label" /><col /></colgroup>
|
||
<tbody valign="top">
|
||
<tr><td class="label"><a class="fn-backref" href="Barriers.html#id2">[1]</a></td><td>To be clear, the <code class="docutils literal notranslate"><span class="pre">exponential()</span></code> function is truly the exponential
|
||
$c^n$ rather than the polynomial $n^c$. In both <code class="docutils literal notranslate"><span class="pre">exponential()</span></code> and
|
||
<code class="docutils literal notranslate"><span class="pre">fibonacci(),</span></code> we are using $n = 100$.</td></tr>
|
||
</tbody>
|
||
</table>
|
||
<table class="docutils footnote" frame="void" id="f44" rules="none">
|
||
<colgroup><col class="label" /><col /></colgroup>
|
||
<tbody valign="top">
|
||
<tr><td class="label"><a class="fn-backref" href="Barriers.html#id3">[2]</a></td><td>The selection of 1.6 as the value was not an arbitrary choice. There
|
||
is a closed-form solution known as Binet’s formula, and calculations based on
|
||
this formula can be done to show that each Fibonacci number is approximately
|
||
1.6 times the value of the previous number.</td></tr>
|
||
</tbody>
|
||
</table>
|
||
<div
|
||
id="SynchBarrierSumm"
|
||
class="embedContainer"
|
||
data-exer-name="SynchBarrierSumm"
|
||
data-long-name="Barrier questions"
|
||
data-short-name="SynchBarrierSumm"
|
||
data-frame-src="../../../Exercises/Synch/SynchBarrierSumm.html?selfLoggingEnabled=false&localMode=true&module=Barriers"
|
||
data-frame-width="950"
|
||
data-frame-height="550"
|
||
data-external="false"
|
||
data-points="0"
|
||
data-required="False"
|
||
data-showhide="show"
|
||
data-threshold="2"
|
||
data-type="ka"
|
||
data-exer-id="">
|
||
|
||
<div class="center">
|
||
<div id="SynchBarrierSumm_iframe"></div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
|
||
|
||
</div>
|
||
|
||
|
||
|
||
<div class="container">
|
||
|
||
<div class="mt-4 container center">
|
||
«  <a id="prevmod1" href="Semaphores.html">7.4. Semaphores</a>
|
||
  ::  
|
||
<a class="uplink" href="index.html">Contents</a>
|
||
  ::  
|
||
<a id="nextmod1" href="Condvars.html">7.6. Condition Variables</a>  »
|
||
</div>
|
||
|
||
|
||
</div>
|
||
|
||
<br />
|
||
|
||
<div class="row jmu-dark-purple-bg">
|
||
<div class="col-md-12">
|
||
<center>
|
||
<a id="contact_us" class="btn button-link-no-blue jmu-gold" rel="nofollow" href="mailto:webmaster@opencsf.org" role="button">Contact Us</a>
|
||
<a id="license" class="btn button-link-no-blue jmu-gold" rel="nofollow" href="https://w3.cs.jmu.edu/kirkpams/OpenCSF/lib/license.html" target="_blank">License</a>
|
||
</center>
|
||
</div>
|
||
</div>
|
||
|
||
|
||
|
||
|
||
|
||
|
||
<script src="_static/js/popper.js-1.14.7-min.js" integrity="sha384-UO2eT0CpHqdSJQ6hJty5KVphtPhzWj9WO1clHTMGa3JDZwrnQq4sF86dIHNDz0W1" crossorigin="anonymous"></script>
|
||
<script src="_static/js/bootstrap.min.js" integrity="sha384-JjSmVgyd0p3pXB1rRibZUAYoIIy6OrQ6VrjIEaFf/nJGzIxFDsf4x0xIM+B07jRM" crossorigin="anonymous"></script>
|
||
</body>
|
||
</html> |