mirrors/emacs
1
Fork 0
mirror of git://git.sv.gnu.org/emacs.git synced 2025-12-26 07:11:34 -08:00
Code Issues Projects Releases Wiki Activity
emacs/mps/manual/html/topic/cache.html
Gareth Rees eaad4493d4 Bring html up to date. 
Copied from Perforce
 Change: 181731
 ServerID: perforce.ravenbrook.com
2013-05-11 20:07:28 +01:00

455 lines
No EOL
36 KiB
HTML
Raw Blame History

<!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>15. Segregated allocation caches &mdash; Memory Pool System 1.111.0 documentation</title>
<link rel="stylesheet" href="../_static/mps.css" type="text/css" />
<link rel="stylesheet" href="../_static/pygments.css" type="text/css" />
<script type="text/javascript">
var DOCUMENTATION_OPTIONS = {
URL_ROOT: '../',
VERSION: '1.111.0',
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="copyright" title="Copyright" href="../copyright.html" />
<link rel="top" title="Memory Pool System 1.111.0 documentation" href="../index.html" />
<link rel="up" title="Reference" href="index.html" />
<link rel="next" title="16. Allocation patterns" href="pattern.html" />
<link rel="prev" title="14. Location dependency" href="location.html" />
</head>
<body>
<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="pattern.html" title="16. Allocation patterns"
accesskey="N">next</a> |</li>
<li class="right" >
<a href="location.html" title="14. Location dependency"
accesskey="P">previous</a> |</li>
<li><a href="../index.html">Memory Pool System 1.111.0 documentation</a> &raquo;</li>
<li><a href="index.html" accesskey="U">Reference</a> &raquo;</li>
</ul>
</div>
<div class="document">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body">
<div class="section" id="segregated-allocation-caches">
<span id="topic-cache"></span><span id="index-0"></span><h1>15. Segregated allocation caches<a class="headerlink" href="#segregated-allocation-caches" title="Permalink to this headline"></a></h1>
<p>A <em class="dfn">segregated allocation cache</em> is a data structure that can be
attached to any <a class="reference internal" href="../glossary/m.html#term-manual-memory-management"><em class="xref std std-term">manually managed</em></a>
<a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a>, that maintains a <a class="reference internal" href="../glossary/s.html#term-segregated-free-list"><em class="xref std std-term">segregated free list</em></a>, that is,
a reserve of free blocks segregated by size.</p>
<p>Create a segregated allocation cache by preparing an array of
structures of type <a class="reference internal" href="#mps_sac_class_s" title="mps_sac_class_s"><tt class="xref c c-type docutils literal"><span class="pre">mps_sac_class_s</span></tt></a> and passing them to
<a class="reference internal" href="#mps_sac_create" title="mps_sac_create"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_create()</span></tt></a>. The values in these structures are hints as
to the size of the blocks, the number of blocks of each size, and the
relative frequency of allocations and deallocations at that size.</p>
<p>For example, suppose we have a pool where we expect to allocate a
small number of relatively long-lived 128-byte objects, and a large
number of relatively short-lived 8-byte objects, we might create a
cache as follows:</p>
<div class="highlight-c"><div class="highlight"><pre><span class="n">mps_sac_class_s</span> <span class="n">classes</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span> <span class="o">=</span> <span class="p">{{</span><span class="mi">8</span><span class="p">,</span> <span class="mi">100</span><span class="p">,</span> <span class="mi">10</span><span class="p">},</span> <span class="p">{</span><span class="mi">128</span><span class="p">,</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">1</span><span class="p">}};</span>
<span class="n">mps_sac_t</span> <span class="n">sac</span><span class="p">;</span>
<span class="n">res</span> <span class="o">=</span> <span class="n">mps_sac_create</span><span class="p">(</span><span class="o">&amp;</span><span class="n">sac</span><span class="p">,</span> <span class="n">pool</span><span class="p">,</span> <span class="k">sizeof</span> <span class="n">classes</span> <span class="o">/</span> <span class="k">sizeof</span> <span class="n">classes</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">classes</span><span class="p">);</span>
<span class="k">if</span> <span class="p">(</span><span class="n">res</span> <span class="o">!=</span> <span class="n">MPS_RES_OK</span><span class="p">)</span>
<span class="n">error</span><span class="p">(</span><span class="s">&quot;failed to create allocation cache&quot;</span><span class="p">);</span>
</pre></div>
</div>
<p>Allocations through the cache (using <a class="reference internal" href="#mps_sac_alloc" title="mps_sac_alloc"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_alloc()</span></tt></a> or
<a class="reference internal" href="#MPS_SAC_ALLOC_FAST" title="MPS_SAC_ALLOC_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_ALLOC_FAST()</span></tt></a>) are serviced from the cache if possible,
otherwise from the pool. Similarly, deallocations through the cache
(using <a class="reference internal" href="#mps_sac_free" title="mps_sac_free"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_free()</span></tt></a> or <a class="reference internal" href="#MPS_SAC_FREE_FAST" title="MPS_SAC_FREE_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_FREE_FAST()</span></tt></a>) return
the block to the appopriate free list for its size. For example:</p>
<div class="highlight-c"><div class="highlight"><pre><span class="n">Foo</span> <span class="o">*</span><span class="n">foo</span><span class="p">;</span>
<span class="n">mps_addr_t</span> <span class="n">p</span><span class="p">;</span>
<span class="n">mps_res_t</span> <span class="n">res</span><span class="p">;</span>
<span class="n">res</span> <span class="o">=</span> <span class="n">mps_sac_alloc</span><span class="p">(</span><span class="o">&amp;</span><span class="n">p</span><span class="p">,</span> <span class="n">sac</span><span class="p">,</span> <span class="k">sizeof</span> <span class="o">*</span><span class="n">foo</span><span class="p">,</span> <span class="nb">false</span><span class="p">);</span>
<span class="k">if</span> <span class="p">(</span><span class="n">res</span> <span class="o">!=</span> <span class="n">MPS_RES_OK</span><span class="p">)</span>
<span class="n">error</span><span class="p">(</span><span class="s">&quot;failed to alloc foo&quot;</span><span class="p">);</span>
<span class="n">foo</span> <span class="o">=</span> <span class="n">p</span><span class="p">;</span>
<span class="cm">/* use &#39;foo&#39; */</span>
<span class="n">mps_sac_free</span><span class="p">(</span><span class="n">sac</span><span class="p">,</span> <span class="n">p</span><span class="p">,</span> <span class="k">sizeof</span> <span class="o">*</span><span class="n">foo</span><span class="p">);</span>
</pre></div>
</div>
<p>The macros <a class="reference internal" href="#MPS_SAC_ALLOC_FAST" title="MPS_SAC_ALLOC_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_ALLOC_FAST()</span></tt></a> and
<a class="reference internal" href="#MPS_SAC_FREE_FAST" title="MPS_SAC_FREE_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_FREE_FAST()</span></tt></a> allow allocation and deallocation to be
inlined in the calling functions, in the case where a free block is
found in the cache.</p>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p>It is recommended that you deallocate a block via the same
segregated allocation cache that you allocated it from. However,
the system is more general than that, and in fact a block that was
allocated from cache A can be deallocated via cache B, provided
that:</p>
<ol class="last arabic simple">
<li>the two caches are attached to the same pool; and</li>
<li>the two caches have the same <em class="dfn">class structure</em>, that is,
they were created by passing identical arrays of <a class="reference internal" href="../glossary/s.html#term-size-class"><em class="xref std std-term">size
classes</em></a>.</li>
</ol>
</div>
<div class="admonition warning">
<p class="first admonition-title">Warning</p>
<p class="last">Segregated allocation caches work poorly with debugging pool
classes: the debugging checks only happen when blocks are moved
between the cache and the pool.</p>
</div>
<div class="section" id="cache-interface">
<span id="index-1"></span><h2>15.1. Cache interface<a class="headerlink" href="#cache-interface" title="Permalink to this headline"></a></h2>
<dl class="type">
<dt id="mps_sac_t">
<tt class="descname">mps_sac_t</tt><a class="headerlink" href="#mps_sac_t" title="Permalink to this definition"></a></dt>
<dd><p>The type of <a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation caches</em></a>.</p>
</dd></dl>
<dl class="macro">
<dt id="MPS_SAC_CLASS_LIMIT">
<tt class="descname">MPS_SAC_CLASS_LIMIT</tt><a class="headerlink" href="#MPS_SAC_CLASS_LIMIT" title="Permalink to this definition"></a></dt>
<dd><p>The number of <a class="reference internal" href="../glossary/s.html#term-size-class"><em class="xref std std-term">size classes</em></a> that <a class="reference internal" href="#mps_sac_create" title="mps_sac_create"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_create()</span></tt></a>
is guaranteed to accept.</p>
<p>More might be accepted: in fact, there might not be any limit in
the implementation on the maximum number of size classes, but if
you specify more than this many, you should be prepared to handle
the <a class="reference internal" href="../glossary/r.html#term-result-code"><em class="xref std std-term">result code</em></a> <a class="reference internal" href="error.html#MPS_RES_LIMIT" title="MPS_RES_LIMIT"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_LIMIT</span></tt></a>.</p>
</dd></dl>
<dl class="type">
<dt id="mps_sac_class_s">
<tt class="descname">mps_sac_class_s</tt><a class="headerlink" href="#mps_sac_class_s" title="Permalink to this definition"></a></dt>
<dd><p>The type of the structure describing a <a class="reference internal" href="../glossary/s.html#term-size-class"><em class="xref std std-term">size class</em></a> in a
<a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation cache</em></a>.</p>
<div class="highlight-c"><div class="highlight"><pre><span class="k">typedef</span> <span class="k">struct</span> <span class="n">mps_sac_class_s</span> <span class="p">{</span>
<span class="kt">size_t</span> <span class="n">mps_block_size</span><span class="p">;</span>
<span class="kt">size_t</span> <span class="n">mps_cached_count</span><span class="p">;</span>
<span class="kt">unsigned</span> <span class="n">mps_frequency</span><span class="p">;</span>
<span class="p">}</span> <span class="n">mps_sac_class_s</span><span class="p">;</span>
</pre></div>
</div>
<p>An array of these structures must be passed to
<a class="reference internal" href="#mps_sac_create" title="mps_sac_create"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_create()</span></tt></a> when creating a segregated allocation
cache.</p>
<p><tt class="docutils literal"><span class="pre">mps_block_size</span></tt> is the maximum <a class="reference internal" href="../glossary/s.html#term-size"><em class="xref std std-term">size</em></a> of any <a class="reference internal" href="../glossary/b.html#term-block"><em class="xref std std-term">block</em></a>
in this size class. It must be a multiple of the alignment of the
<a class="reference internal" href="../glossary/a.html#term-alignment"><em class="xref std std-term">alignment</em></a> of the <a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a> to which the cache belongs.</p>
<p><tt class="docutils literal"><span class="pre">mps_cached_count</span></tt> is the number of blocks of this size class to
cache. It is advice to the MPS on how many blocks to cache, not an
absolute limit. The cache policy tries to accommodate fluctuations
in the population and minimize the cost of responding to client
requests; the purpose of this parameter is to limit how much
memory the <a class="reference internal" href="../glossary/c.html#term-client-program"><em class="xref std std-term">client program</em></a> is willing to set aside for this
purpose. However, a <tt class="docutils literal"><span class="pre">cached_count</span></tt> of zero prevents any caching of
blocks falling into that size class.</p>
<p><tt class="docutils literal"><span class="pre">mps_frequency</span></tt> is a number that describes the frequency of
requests (allocation and deallocation combined) in this size class
relative to the other size classes in the cache.</p>
</dd></dl>
<dl class="function">
<dt id="mps_sac_create">
<a class="reference internal" href="error.html#mps_res_t" title="mps_res_t">mps_res_t</a> <tt class="descname">mps_sac_create</tt><big>(</big><a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;*sac_o</em>, <a class="reference internal" href="pool.html#mps_pool_t" title="mps_pool_t">mps_pool_t</a><em>&nbsp;pool</em>, size_t<em>&nbsp;classes_count</em>, <a class="reference internal" href="#mps_sac_class_s" title="mps_sac_class_s">mps_sac_class_s</a><em>&nbsp;*classes</em><big>)</big><a class="headerlink" href="#mps_sac_create" title="Permalink to this definition"></a></dt>
<dd><p>Create a <a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation cache</em></a> for a <a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a>.</p>
<p><tt class="docutils literal"><span class="pre">sac_o</span></tt> points to a location that will hold the address of the
segregated allocation cache.</p>
<p><tt class="docutils literal"><span class="pre">pool</span></tt> is the pool the cache is attached to.</p>
<p><tt class="docutils literal"><span class="pre">classes_count</span></tt> is the number of <a class="reference internal" href="../glossary/s.html#term-size-class"><em class="xref std std-term">size classes</em></a> in the
cache.</p>
<p><tt class="docutils literal"><span class="pre">classes</span></tt> points to an array describing the size classes in the
cache.</p>
<p>Returns <a class="reference internal" href="error.html#MPS_RES_OK" title="MPS_RES_OK"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_OK</span></tt></a> if the segregated allocation cache
is created successfully. Returns <a class="reference internal" href="error.html#MPS_RES_MEMORY" title="MPS_RES_MEMORY"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_MEMORY</span></tt></a> or
<a class="reference internal" href="error.html#MPS_RES_COMMIT_LIMIT" title="MPS_RES_COMMIT_LIMIT"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_COMMIT_LIMIT</span></tt></a> when it fails to allocate memory
for the internal cache structure. Returns <a class="reference internal" href="error.html#MPS_RES_LIMIT" title="MPS_RES_LIMIT"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_LIMIT</span></tt></a>
if you ask for too many size classes: in this case, combine some
small adjacent classes. Returns <a class="reference internal" href="error.html#MPS_RES_PARAM" title="MPS_RES_PARAM"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_PARAM</span></tt></a> if the
pool doesn&#8217;t support segregated allocation caches.</p>
<p>After this function returns, the array of size classes pointed to
be <tt class="docutils literal"><span class="pre">classes</span></tt> is no longer needed and may be discarded. The
segregated allocation cache pointed to by <tt class="docutils literal"><span class="pre">sac_o</span></tt> persists until
it is destroyed by calling <a class="reference internal" href="#mps_sac_destroy" title="mps_sac_destroy"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_destroy()</span></tt></a>.</p>
<p>This function creates an allocation cache whose <a class="reference internal" href="../glossary/f.html#term-free-list"><em class="xref std std-term">free list</em></a>
is segregated into the given size classes. The cache can get more
memory from the given pool, or return memory to it.</p>
<p>Segregated allocation caches can be associated with any pool that
supports <a class="reference internal" href="../glossary/m.html#term-manual-memory-management"><em class="xref std std-term">manual</em></a> allocation with
the functions <a class="reference internal" href="allocation.html#mps_alloc" title="mps_alloc"><tt class="xref c c-func docutils literal"><span class="pre">mps_alloc()</span></tt></a> and <a class="reference internal" href="allocation.html#mps_free" title="mps_free"><tt class="xref c c-func docutils literal"><span class="pre">mps_free()</span></tt></a>.</p>
<p>The size classes are described by an array of element type
<a class="reference internal" href="#mps_sac_class_s" title="mps_sac_class_s"><tt class="xref c c-type docutils literal"><span class="pre">mps_sac_class_s</span></tt></a>. This array is used to initialize the
segregated allocation cache, and is not needed
after:c:func:<cite>mps_sac_create</cite> returns. The following constraints
apply to the array:</p>
<ul class="simple">
<li>You must specify at least one size class.</li>
<li>All size classes must have different sizes.</li>
<li>The size classes must be given in the order of increasing size.</li>
<li>The smallest size must be at least as large as <tt class="docutils literal"><span class="pre">sizeof(void</span> <span class="pre">*)</span></tt>.</li>
<li>Each size must be a multiple of the <a class="reference internal" href="../glossary/a.html#term-alignment"><em class="xref std std-term">alignment</em></a> of the
pool.</li>
<li>There might be a limit on how many classes can be described, but
it will be at least <a class="reference internal" href="#MPS_SAC_CLASS_LIMIT" title="MPS_SAC_CLASS_LIMIT"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_SAC_CLASS_LIMIT</span></tt></a>.</li>
</ul>
<p>The MPS automatically provides an &#8220;overlarge&#8221; size class for
arbitrarily large allocations above the largest size class
described. Allocations falling into the overlarge size class are
not cached.</p>
<p>Any allocations whose size falls between two size classes are
allocated from the larger size class.</p>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">Too many size classes will slow down allocation; too few size
classes waste more space in internal fragmentation. It is
assumed that overlarge allocations are rare; otherwise, you
would add another size class for them, or even create separate
allocation caches or pools for them.</p>
</div>
</dd></dl>
<dl class="function">
<dt id="mps_sac_destroy">
void <tt class="descname">mps_sac_destroy</tt><big>(</big><a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;sac</em><big>)</big><a class="headerlink" href="#mps_sac_destroy" title="Permalink to this definition"></a></dt>
<dd><p>Destroy a <a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation cache</em></a>.</p>
<p><tt class="docutils literal"><span class="pre">sac</span></tt> is the segregated allocation cache to destroy.</p>
<p>Returns all memory in the cache to the associated <a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a>.
The pool might then return some memory to the <a class="reference internal" href="../glossary/a.html#term-arena"><em class="xref std std-term">arena</em></a>, but
that&#8217;s up to the pool&#8217;s usual policy.</p>
<p>Destroying the cache has no effect on blocks allocated through it.</p>
</dd></dl>
<dl class="function">
<dt id="mps_sac_flush">
void <tt class="descname">mps_sac_flush</tt><big>(</big><a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;sac</em><big>)</big><a class="headerlink" href="#mps_sac_flush" title="Permalink to this definition"></a></dt>
<dd><p>Flush a <a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation cache</em></a>, returning all memory
held in it to the associated <a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a>.</p>
<p><tt class="docutils literal"><span class="pre">sac</span></tt> is the segregated allocation cache to flush.</p>
<p>This is something that you&#8217;d typically do when you know you won&#8217;t
be using the segregated allocation cache for awhile, but want to
hold on to the cache itself. Destroying a cache has the effect of
flushing it.</p>
<p>Flushing the segregated allocation cache might well cause the pool
to return some memory to the <a class="reference internal" href="../glossary/a.html#term-arena"><em class="xref std std-term">arena</em></a>, but that&#8217;s up to the
pool&#8217;s usual policy.</p>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">The MPS might also decide to take memory from the segregated
allocation cache without the <a class="reference internal" href="../glossary/c.html#term-client-program"><em class="xref std std-term">client program</em></a> requesting
a flush.</p>
</div>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">The <a class="reference internal" href="../glossary/c.html#term-client-program"><em class="xref std std-term">client program</em></a> is responsible for synchronizing
the access to the cache, but if the cache decides to access
the pool, the MPS will properly synchronize with any other
<a class="reference internal" href="../glossary/t.html#term-thread"><em class="xref std std-term">threads</em></a> that might be accessing the same
pool.</p>
</div>
</dd></dl>
</div>
<div class="section" id="allocation-interface">
<span id="index-2"></span><h2>15.2. Allocation interface<a class="headerlink" href="#allocation-interface" title="Permalink to this headline"></a></h2>
<dl class="function">
<dt id="mps_sac_alloc">
<a class="reference internal" href="error.html#mps_res_t" title="mps_res_t">mps_res_t</a> <tt class="descname">mps_sac_alloc</tt><big>(</big><a class="reference internal" href="interface.html#mps_addr_t" title="mps_addr_t">mps_addr_t</a><em>&nbsp;*p_o</em>, <a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;sac</em>, size_t<em>&nbsp;size</em>, <a class="reference internal" href="interface.html#mps_bool_t" title="mps_bool_t">mps_bool_t</a><em>&nbsp;has_reservoir_permit</em><big>)</big><a class="headerlink" href="#mps_sac_alloc" title="Permalink to this definition"></a></dt>
<dd><p>Allocate a <a class="reference internal" href="../glossary/b.html#term-block"><em class="xref std std-term">block</em></a> using a <a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation
cache</em></a>. If no suitable block exists in the cache, ask for more
memory from the associated <a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a>.</p>
<p><tt class="docutils literal"><span class="pre">p_o</span></tt> points to a location that will hold the address of the
allocated block.</p>
<p><tt class="docutils literal"><span class="pre">sac</span></tt> is the segregated allocation cache.</p>
<p><tt class="docutils literal"><span class="pre">size</span></tt> is the <a class="reference internal" href="../glossary/s.html#term-size"><em class="xref std std-term">size</em></a> of the block to allocate. It does not
have to be one of the <a class="reference internal" href="../glossary/s.html#term-size-class"><em class="xref std std-term">size classes</em></a> of the cache; nor does
it have to be aligned.</p>
<p><tt class="docutils literal"><span class="pre">has_reservoir_permit</span></tt> should be false.</p>
<p>Returns <a class="reference internal" href="error.html#MPS_RES_OK" title="MPS_RES_OK"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_OK</span></tt></a> if successful: in this case the
address of the allocated block is <tt class="docutils literal"><span class="pre">*p_o</span></tt>. The allocated block
can be larger than requested. Blocks not matching any size class
are allocated from the next largest class, and blocks larger than
the largest size class are simply allocated at the requested size
(rounded up to alignment, as usual).</p>
<p>Returns <a class="reference internal" href="error.html#MPS_RES_MEMORY" title="MPS_RES_MEMORY"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_MEMORY</span></tt></a> if there wasn&#8217;t enough memory,
<a class="reference internal" href="error.html#MPS_RES_COMMIT_LIMIT" title="MPS_RES_COMMIT_LIMIT"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_COMMIT_LIMIT</span></tt></a> if the <a class="reference internal" href="../glossary/c.html#term-commit-limit"><em class="xref std std-term">commit limit</em></a> was
exceeded, or <a class="reference internal" href="error.html#MPS_RES_RESOURCE" title="MPS_RES_RESOURCE"><tt class="xref c c-macro docutils literal"><span class="pre">MPS_RES_RESOURCE</span></tt></a> if it ran out of
<a class="reference internal" href="../glossary/v.html#term-virtual-memory"><em class="xref std std-term">virtual memory</em></a>.</p>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">There&#8217;s also a macro <a class="reference internal" href="#MPS_SAC_ALLOC_FAST" title="MPS_SAC_ALLOC_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_ALLOC_FAST()</span></tt></a> that does
the same thing. The macro is faster, but generates more code
and does less checking.</p>
</div>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">The <a class="reference internal" href="../glossary/c.html#term-client-program"><em class="xref std std-term">client program</em></a> is responsible for synchronizing
the access to the cache, but if the cache decides to access
the pool, the MPS will properly synchronize with any other
<a class="reference internal" href="../glossary/t.html#term-thread"><em class="xref std std-term">threads</em></a> that might be accessing the same
pool.</p>
</div>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">Blocks allocated through a segregated allocation cache should
only be freed through a segregated allocation cache with the
same class structure. Calling <a class="reference internal" href="allocation.html#mps_free" title="mps_free"><tt class="xref c c-func docutils literal"><span class="pre">mps_free()</span></tt></a> on them can
cause <a class="reference internal" href="../glossary/m.html#term-memory-leak"><em class="xref std std-term">memory leaks</em></a>, because the size of
the block might be larger than you think. Naturally, the cache
must also be attached to the same pool.</p>
</div>
</dd></dl>
<dl class="function">
<dt id="MPS_SAC_ALLOC_FAST">
<tt class="descname">MPS_SAC_ALLOC_FAST</tt><big>(</big><a class="reference internal" href="error.html#mps_res_t" title="mps_res_t">mps_res_t</a><em>&nbsp;res_v</em>, <a class="reference internal" href="interface.html#mps_addr_t" title="mps_addr_t">mps_addr_t</a><em>&nbsp;*p_v</em>, <a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;sac</em>, size_t<em>&nbsp;size</em>, <a class="reference internal" href="interface.html#mps_bool_t" title="mps_bool_t">mps_bool_t</a><em>&nbsp;has_reservoir_permit</em><big>)</big><a class="headerlink" href="#MPS_SAC_ALLOC_FAST" title="Permalink to this definition"></a></dt>
<dd><p>A macro alternative to <a class="reference internal" href="#mps_sac_alloc" title="mps_sac_alloc"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_alloc()</span></tt></a>. It is faster than
the function, but generates more code, does less checking.</p>
<p>It takes an lvalue <tt class="docutils literal"><span class="pre">p_v</span></tt> which is assigned the address of the
allocated block (instead of a pointer to a location to store
it). It takes an additional first argument, the lvalue <tt class="docutils literal"><span class="pre">res_v</span></tt>,
which is assigned the <a class="reference internal" href="../glossary/r.html#term-result-code"><em class="xref std std-term">result code</em></a>.</p>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last"><a class="reference internal" href="#MPS_SAC_ALLOC_FAST" title="MPS_SAC_ALLOC_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_ALLOC_FAST()</span></tt></a> may evaluate its arguments
multiple times, except for <tt class="docutils literal"><span class="pre">has_reservoir_permit</span></tt>, which it
evaluates at most once, and only if it decides to access the
pool.</p>
</div>
</dd></dl>
<dl class="function">
<dt id="mps_sac_free">
void <tt class="descname">mps_sac_free</tt><big>(</big><a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;sac</em>, <a class="reference internal" href="interface.html#mps_addr_t" title="mps_addr_t">mps_addr_t</a><em>&nbsp;p</em>, size_t<em>&nbsp;size</em><big>)</big><a class="headerlink" href="#mps_sac_free" title="Permalink to this definition"></a></dt>
<dd><p>Free a <a class="reference internal" href="../glossary/b.html#term-block"><em class="xref std std-term">block</em></a> using a <a class="reference internal" href="../glossary/s.html#term-segregated-allocation-cache"><em class="xref std std-term">segregated allocation
cache</em></a>. If the cache would become too full, some blocks may be
returned to the associated <a class="reference internal" href="../glossary/p.html#term-pool"><em class="xref std std-term">pool</em></a>.</p>
<p><tt class="docutils literal"><span class="pre">sac</span></tt> is the segregated allocation cache.</p>
<p><tt class="docutils literal"><span class="pre">p</span></tt> points to the block to be freed. This block must have been
allocated through a segregated allocation cache with the same
class structure, attached to the same pool. (Usually, you&#8217;d use
the same cache to allocate and deallocate a block, but the MPS is
more flexible.)</p>
<p><tt class="docutils literal"><span class="pre">size</span></tt> is the <a class="reference internal" href="../glossary/s.html#term-size"><em class="xref std std-term">size</em></a> of the block. It should be the size
that was specified when the block was allocated (the cache knows
what the real size of the block is).</p>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">The <a class="reference internal" href="../glossary/c.html#term-client-program"><em class="xref std std-term">client program</em></a> is responsible for synchronizing
the access to the cache, but if the cache decides to access
the pool, the MPS will properly synchronize with any other
<a class="reference internal" href="../glossary/t.html#term-thread"><em class="xref std std-term">threads</em></a> that might be accessing the same
pool.</p>
</div>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last">There&#8217;s also a macro <a class="reference internal" href="#MPS_SAC_FREE_FAST" title="MPS_SAC_FREE_FAST"><tt class="xref c c-func docutils literal"><span class="pre">MPS_SAC_FREE_FAST()</span></tt></a> that does the
same thing. The macro is faster, but generates more code and
does no checking.</p>
</div>
<div class="admonition-note admonition">
<p class="first admonition-title">Note</p>
<p class="last"><a class="reference internal" href="#mps_sac_free" title="mps_sac_free"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_free()</span></tt></a> does very little checking: it&#8217;s
optimized for speed. <a class="reference internal" href="../glossary/d.html#term-double-free"><em class="xref std std-term">Double frees</em></a> and
other mistakes will only be detected when the cache is flushed
(either by calling <a class="reference internal" href="#mps_sac_flush" title="mps_sac_flush"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_flush()</span></tt></a> or automatically),
and may not be detected at all, if intervening operations have
obscured symptoms.</p>
</div>
</dd></dl>
<dl class="function">
<dt id="MPS_SAC_FREE_FAST">
<tt class="descname">MPS_SAC_FREE_FAST</tt><big>(</big><a class="reference internal" href="#mps_sac_t" title="mps_sac_t">mps_sac_t</a><em>&nbsp;sac</em>, <a class="reference internal" href="interface.html#mps_addr_t" title="mps_addr_t">mps_addr_t</a><em>&nbsp;p</em>, size_t<em>&nbsp;size</em><big>)</big><a class="headerlink" href="#MPS_SAC_FREE_FAST" title="Permalink to this definition"></a></dt>
<dd><p>A macro alternative to <a class="reference internal" href="#mps_sac_free" title="mps_sac_free"><tt class="xref c c-func docutils literal"><span class="pre">mps_sac_free()</span></tt></a> that is faster than
the function but does no checking. The arguments are identical to
the function.</p>
</dd></dl>
</div>
</div>
</div>
</div>
</div>
<div class="sphinxsidebar">
<div class="sphinxsidebarwrapper">
<p class="logo"><a href="../index.html">
<img class="logo" src="../_static/logo.png" alt="Logo"/>
</a></p>
<h3><a href="../index.html">Table Of Contents</a></h3>
<ul>
<li><a class="reference internal" href="#">15. Segregated allocation caches</a><ul>
<li><a class="reference internal" href="#cache-interface">15.1. Cache interface</a></li>
<li><a class="reference internal" href="#allocation-interface">15.2. Allocation interface</a></li>
</ul>
</li>
</ul>
<h4>Previous topic</h4>
<p class="topless"><a href="location.html"
title="previous chapter">14. Location dependency</a></p>
<h4>Next topic</h4>
<p class="topless"><a href="pattern.html"
title="next chapter">16. Allocation patterns</a></p><h4>Downloads</h4>
<p class="topless">
<a href="http://www.ravenbrook.com/project/mps/release/1.111.0/">MPS Kit release 1.111.0</a><br>
<a href="http://www.ravenbrook.com/project/mps/release/">All MPS Kit releases</a>
</p>
<h4>Issues</h4>
<p class="topless">
<a href="http://www.ravenbrook.com/project/mps/issue/?action=list&amp;view=status%3dopen&amp;display=Job:Priority:Title&amp;sort=Priority">Known issues</a><br>
<a href="http://www.ravenbrook.com/project/mps/issue/?action=fixed&release_fixed=1.111.0">Issues fixed in release 1.111.0</a>
</p><h4>Contact us</h4>
<p class="topless"><a href="mailto:mps-questions@ravenbrook.com">mps-questions@ravenbrook.com</a></p>
</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="pattern.html" title="16. Allocation patterns"
>next</a> |</li>
<li class="right" >
<a href="location.html" title="14. Location dependency"
>previous</a> |</li>
<li><a href="../index.html">Memory Pool System 1.111.0 documentation</a> &raquo;</li>
<li><a href="index.html" >Reference</a> &raquo;</li>
</ul>
</div>
<div class="footer">
&copy; <a href="../copyright.html">Copyright</a> 2013, Ravenbrook Limited.
Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.1.3.
</div>
</body>
</html>
Reference in a new issue View git blame Copy permalink