<?xml version="1.0" encoding="utf-8"?>
<html><head><title>JEMALLOC</title><meta name="generator" content="DocBook XSL Stylesheets V1.78.1"/></head><body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"><div class="refentry"><a name="idm140689766630352"/><div class="titlepage"/><div class="refnamediv"><h2>Name</h2><p>jemalloc — general purpose memory allocation functions</p></div><div class="refsect1"><a name="library"/><h2>LIBRARY</h2><p>This manual describes jemalloc 5.1.0-0-g61efbda7098de6fe64c362d309824864308c36d4. More information
can be found at the <a class="ulink" href="http://jemalloc.net/" target="_top">jemalloc website</a>.</p></div><div class="refsynopsisdiv"><h2>SYNOPSIS</h2><div class="funcsynopsis"><pre class="funcsynopsisinfo">#include <<code class="filename">jemalloc/jemalloc.h</code>></pre><div class="refsect2"><a name="idm140689767408528"/><h3>Standard API</h3><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void *<b class="fsfunc">malloc</b>(</code></td><td>size_t <var class="pdparam">size</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void *<b class="fsfunc">calloc</b>(</code></td><td>size_t <var class="pdparam">number</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">int <b class="fsfunc">posix_memalign</b>(</code></td><td>void **<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">alignment</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void *<b class="fsfunc">aligned_alloc</b>(</code></td><td>size_t <var class="pdparam">alignment</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void *<b class="fsfunc">realloc</b>(</code></td><td>void *<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void <b class="fsfunc">free</b>(</code></td><td>void *<var class="pdparam">ptr</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="refsect2"><a name="idm140689770688880"/><h3>Non-standard API</h3><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void *<b class="fsfunc">mallocx</b>(</code></td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void *<b class="fsfunc">rallocx</b>(</code></td><td>void *<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">size_t <b class="fsfunc">xallocx</b>(</code></td><td>void *<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">extra</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">size_t <b class="fsfunc">sallocx</b>(</code></td><td>void *<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void <b class="fsfunc">dallocx</b>(</code></td><td>void *<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void <b class="fsfunc">sdallocx</b>(</code></td><td>void *<var class="pdparam">ptr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">size_t <b class="fsfunc">nallocx</b>(</code></td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>int <var class="pdparam">flags</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">int <b class="fsfunc">mallctl</b>(</code></td><td>const char *<var class="pdparam">name</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">oldp</var>, </td></tr><tr><td> </td><td>size_t *<var class="pdparam">oldlenp</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">newp</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">newlen</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">int <b class="fsfunc">mallctlnametomib</b>(</code></td><td>const char *<var class="pdparam">name</var>, </td></tr><tr><td> </td><td>size_t *<var class="pdparam">mibp</var>, </td></tr><tr><td> </td><td>size_t *<var class="pdparam">miblenp</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">int <b class="fsfunc">mallctlbymib</b>(</code></td><td>const size_t *<var class="pdparam">mib</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">miblen</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">oldp</var>, </td></tr><tr><td> </td><td>size_t *<var class="pdparam">oldlenp</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">newp</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">newlen</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void <b class="fsfunc">malloc_stats_print</b>(</code></td><td>void <var class="pdparam">(*write_cb)</var>
<code>(</code>void *, const char *<code>)</code>
, </td></tr><tr><td> </td><td>void *<var class="pdparam">cbopaque</var>, </td></tr><tr><td> </td><td>const char *<var class="pdparam">opts</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">size_t <b class="fsfunc">malloc_usable_size</b>(</code></td><td>const void *<var class="pdparam">ptr</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">void <b class="fsfunc">(*malloc_message)</b>(</code></td><td>void *<var class="pdparam">cbopaque</var>, </td></tr><tr><td> </td><td>const char *<var class="pdparam">s</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div><p><span class="type">const char *</span><code class="varname">malloc_conf</code>;</p></div></div></div><div class="refsect1"><a name="description"/><h2>DESCRIPTION</h2><div class="refsect2"><a name="idm140689765494448"/><h3>Standard API</h3><p>The <code class="function">malloc()</code> function allocates
<em class="parameter"><code>size</code></em> bytes of uninitialized memory. The allocated
space is suitably aligned (after possible pointer coercion) for storage
of any type of object.</p><p>The <code class="function">calloc()</code> function allocates
space for <em class="parameter"><code>number</code></em> objects, each
<em class="parameter"><code>size</code></em> bytes in length. The result is identical to
calling <code class="function">malloc()</code> with an argument of
<em class="parameter"><code>number</code></em> * <em class="parameter"><code>size</code></em>, with the
exception that the allocated memory is explicitly initialized to zero
bytes.</p><p>The <code class="function">posix_memalign()</code> function
allocates <em class="parameter"><code>size</code></em> bytes of memory such that the
allocation's base address is a multiple of
<em class="parameter"><code>alignment</code></em>, and returns the allocation in the value
pointed to by <em class="parameter"><code>ptr</code></em>. The requested
<em class="parameter"><code>alignment</code></em> must be a power of 2 at least as large as
<code class="code">sizeof(<span class="type">void *</span>)</code>.</p><p>The <code class="function">aligned_alloc()</code> function
allocates <em class="parameter"><code>size</code></em> bytes of memory such that the
allocation's base address is a multiple of
<em class="parameter"><code>alignment</code></em>. The requested
<em class="parameter"><code>alignment</code></em> must be a power of 2. Behavior is
undefined if <em class="parameter"><code>size</code></em> is not an integral multiple of
<em class="parameter"><code>alignment</code></em>.</p><p>The <code class="function">realloc()</code> function changes the
size of the previously allocated memory referenced by
<em class="parameter"><code>ptr</code></em> to <em class="parameter"><code>size</code></em> bytes. The
contents of the memory are unchanged up to the lesser of the new and old
sizes. If the new size is larger, the contents of the newly allocated
portion of the memory are undefined. Upon success, the memory referenced
by <em class="parameter"><code>ptr</code></em> is freed and a pointer to the newly
allocated memory is returned. Note that
<code class="function">realloc()</code> may move the memory allocation,
resulting in a different return value than <em class="parameter"><code>ptr</code></em>.
If <em class="parameter"><code>ptr</code></em> is <code class="constant">NULL</code>, the
<code class="function">realloc()</code> function behaves identically to
<code class="function">malloc()</code> for the specified size.</p><p>The <code class="function">free()</code> function causes the
allocated memory referenced by <em class="parameter"><code>ptr</code></em> to be made
available for future allocations. If <em class="parameter"><code>ptr</code></em> is
<code class="constant">NULL</code>, no action occurs.</p></div><div class="refsect2"><a name="idm140689765469472"/><h3>Non-standard API</h3><p>The <code class="function">mallocx()</code>,
<code class="function">rallocx()</code>,
<code class="function">xallocx()</code>,
<code class="function">sallocx()</code>,
<code class="function">dallocx()</code>,
<code class="function">sdallocx()</code>, and
<code class="function">nallocx()</code> functions all have a
<em class="parameter"><code>flags</code></em> argument that can be used to specify
options. The functions only check the options that are contextually
relevant. Use bitwise or (<code class="code">|</code>) operations to
specify one or more of the following:
</p><div class="variablelist"><dl class="variablelist"><dt><a name="MALLOCX_LG_ALIGN"/><span class="term"><code class="constant">MALLOCX_LG_ALIGN(<em class="parameter"><code>la</code></em>)
</code></span></dt><dd><p>Align the memory allocation to start at an address
that is a multiple of <code class="code">(1 <<
<em class="parameter"><code>la</code></em>)</code>. This macro does not validate
that <em class="parameter"><code>la</code></em> is within the valid
range.</p></dd><dt><a name="MALLOCX_ALIGN"/><span class="term"><code class="constant">MALLOCX_ALIGN(<em class="parameter"><code>a</code></em>)
</code></span></dt><dd><p>Align the memory allocation to start at an address
that is a multiple of <em class="parameter"><code>a</code></em>, where
<em class="parameter"><code>a</code></em> is a power of two. This macro does not
validate that <em class="parameter"><code>a</code></em> is a power of 2.
</p></dd><dt><a name="MALLOCX_ZERO"/><span class="term"><code class="constant">MALLOCX_ZERO</code></span></dt><dd><p>Initialize newly allocated memory to contain zero
bytes. In the growing reallocation case, the real size prior to
reallocation defines the boundary between untouched bytes and those
that are initialized to contain zero bytes. If this macro is
absent, newly allocated memory is uninitialized.</p></dd><dt><a name="MALLOCX_TCACHE"/><span class="term"><code class="constant">MALLOCX_TCACHE(<em class="parameter"><code>tc</code></em>)
</code></span></dt><dd><p>Use the thread-specific cache (tcache) specified by
the identifier <em class="parameter"><code>tc</code></em>, which must have been
acquired via the <a class="link" href="#tcache.create"><quote><code class="mallctl">tcache.create</code></quote></a>
mallctl. This macro does not validate that
<em class="parameter"><code>tc</code></em> specifies a valid
identifier.</p></dd><dt><a name="MALLOC_TCACHE_NONE"/><span class="term"><code class="constant">MALLOCX_TCACHE_NONE</code></span></dt><dd><p>Do not use a thread-specific cache (tcache). Unless
<code class="constant">MALLOCX_TCACHE(<em class="parameter"><code>tc</code></em>)</code> or
<code class="constant">MALLOCX_TCACHE_NONE</code> is specified, an
automatically managed tcache will be used under many circumstances.
This macro cannot be used in the same <em class="parameter"><code>flags</code></em>
argument as
<code class="constant">MALLOCX_TCACHE(<em class="parameter"><code>tc</code></em>)</code>.</p></dd><dt><a name="MALLOCX_ARENA"/><span class="term"><code class="constant">MALLOCX_ARENA(<em class="parameter"><code>a</code></em>)
</code></span></dt><dd><p>Use the arena specified by the index
<em class="parameter"><code>a</code></em>. This macro has no effect for regions that
were allocated via an arena other than the one specified. This
macro does not validate that <em class="parameter"><code>a</code></em> specifies an
arena index in the valid range.</p></dd></dl></div><p>
</p><p>The <code class="function">mallocx()</code> function allocates at
least <em class="parameter"><code>size</code></em> bytes of memory, and returns a pointer
to the base address of the allocation. Behavior is undefined if
<em class="parameter"><code>size</code></em> is <code class="constant">0</code>.</p><p>The <code class="function">rallocx()</code> function resizes the
allocation at <em class="parameter"><code>ptr</code></em> to be at least
<em class="parameter"><code>size</code></em> bytes, and returns a pointer to the base
address of the resulting allocation, which may or may not have moved from
its original location. Behavior is undefined if
<em class="parameter"><code>size</code></em> is <code class="constant">0</code>.</p><p>The <code class="function">xallocx()</code> function resizes the
allocation at <em class="parameter"><code>ptr</code></em> in place to be at least
<em class="parameter"><code>size</code></em> bytes, and returns the real size of the
allocation. If <em class="parameter"><code>extra</code></em> is non-zero, an attempt is
made to resize the allocation to be at least <code class="code">(<em class="parameter"><code>size</code></em> +
<em class="parameter"><code>extra</code></em>)</code> bytes, though inability to allocate
the extra byte(s) will not by itself result in failure to resize.
Behavior is undefined if <em class="parameter"><code>size</code></em> is
<code class="constant">0</code>, or if <code class="code">(<em class="parameter"><code>size</code></em> + <em class="parameter"><code>extra</code></em>
> <code class="constant">SIZE_T_MAX</code>)</code>.</p><p>The <code class="function">sallocx()</code> function returns the
real size of the allocation at <em class="parameter"><code>ptr</code></em>.</p><p>The <code class="function">dallocx()</code> function causes the
memory referenced by <em class="parameter"><code>ptr</code></em> to be made available for
future allocations.</p><p>The <code class="function">sdallocx()</code> function is an
extension of <code class="function">dallocx()</code> with a
<em class="parameter"><code>size</code></em> parameter to allow the caller to pass in the
allocation size as an optimization. The minimum valid input size is the
original requested size of the allocation, and the maximum valid input
size is the corresponding value returned by
<code class="function">nallocx()</code> or
<code class="function">sallocx()</code>.</p><p>The <code class="function">nallocx()</code> function allocates no
memory, but it performs the same size computation as the
<code class="function">mallocx()</code> function, and returns the real
size of the allocation that would result from the equivalent
<code class="function">mallocx()</code> function call, or
<code class="constant">0</code> if the inputs exceed the maximum supported size
class and/or alignment. Behavior is undefined if
<em class="parameter"><code>size</code></em> is <code class="constant">0</code>.</p><p>The <code class="function">mallctl()</code> function provides a
general interface for introspecting the memory allocator, as well as
setting modifiable parameters and triggering actions. The
period-separated <em class="parameter"><code>name</code></em> argument specifies a
location in a tree-structured namespace; see the <a class="xref" href="#mallctl_namespace" title="MALLCTL NAMESPACE">MALLCTL NAMESPACE</a> section for
documentation on the tree contents. To read a value, pass a pointer via
<em class="parameter"><code>oldp</code></em> to adequate space to contain the value, and a
pointer to its length via <em class="parameter"><code>oldlenp</code></em>; otherwise pass
<code class="constant">NULL</code> and <code class="constant">NULL</code>. Similarly, to
write a value, pass a pointer to the value via
<em class="parameter"><code>newp</code></em>, and its length via
<em class="parameter"><code>newlen</code></em>; otherwise pass <code class="constant">NULL</code>
and <code class="constant">0</code>.</p><p>The <code class="function">mallctlnametomib()</code> function
provides a way to avoid repeated name lookups for applications that
repeatedly query the same portion of the namespace, by translating a name
to a <span class="quote">“<span class="quote">Management Information Base</span>”</span> (MIB) that can be passed
repeatedly to <code class="function">mallctlbymib()</code>. Upon
successful return from <code class="function">mallctlnametomib()</code>,
<em class="parameter"><code>mibp</code></em> contains an array of
<em class="parameter"><code>*miblenp</code></em> integers, where
<em class="parameter"><code>*miblenp</code></em> is the lesser of the number of components
in <em class="parameter"><code>name</code></em> and the input value of
<em class="parameter"><code>*miblenp</code></em>. Thus it is possible to pass a
<em class="parameter"><code>*miblenp</code></em> that is smaller than the number of
period-separated name components, which results in a partial MIB that can
be used as the basis for constructing a complete MIB. For name
components that are integers (e.g. the 2 in
<a class="link" href="#arenas.bin.i.size"><quote><code class="mallctl">arenas.bin.2.size</code></quote></a>),
the corresponding MIB component will always be that integer. Therefore,
it is legitimate to construct code like the following: </p><pre class="programlisting">
unsigned nbins, i;
size_t mib[4];
size_t len, miblen;
len = sizeof(nbins);
mallctl("arenas.nbins", &nbins, &len, NULL, 0);
miblen = 4;
mallctlnametomib("arenas.bin.0.size", mib, &miblen);
for (i = 0; i < nbins; i++) {
size_t bin_size;
mib[2] = i;
len = sizeof(bin_size);
mallctlbymib(mib, miblen, (void *)&bin_size, &len, NULL, 0);
/* Do something with bin_size... */
}</pre><dt><a name="malloc_stats_print_opts"/></dt><dd/><p>The <code class="function">malloc_stats_print()</code> function writes
summary statistics via the <em class="parameter"><code>write_cb</code></em> callback
function pointer and <em class="parameter"><code>cbopaque</code></em> data passed to
<em class="parameter"><code>write_cb</code></em>, or <code class="function">malloc_message()</code>
if <em class="parameter"><code>write_cb</code></em> is <code class="constant">NULL</code>. The
statistics are presented in human-readable form unless <span class="quote">“<span class="quote">J</span>”</span> is
specified as a character within the <em class="parameter"><code>opts</code></em> string, in
which case the statistics are presented in <a class="ulink" href="http://www.json.org/" target="_top">JSON format</a>. This function can be
called repeatedly. General information that never changes during
execution can be omitted by specifying <span class="quote">“<span class="quote">g</span>”</span> as a character
within the <em class="parameter"><code>opts</code></em> string. Note that
<code class="function">malloc_message()</code> uses the
<code class="function">mallctl*()</code> functions internally, so inconsistent
statistics can be reported if multiple threads use these functions
simultaneously. If <code class="option">--enable-stats</code> is specified during
configuration, <span class="quote">“<span class="quote">m</span>”</span>, <span class="quote">“<span class="quote">d</span>”</span>, and <span class="quote">“<span class="quote">a</span>”</span>
can be specified to omit merged arena, destroyed merged arena, and per
arena statistics, respectively; <span class="quote">“<span class="quote">b</span>”</span> and <span class="quote">“<span class="quote">l</span>”</span> can
be specified to omit per size class statistics for bins and large objects,
respectively; <span class="quote">“<span class="quote">x</span>”</span> can be specified to omit all mutex
statistics. Unrecognized characters are silently ignored. Note that
thread caching may prevent some statistics from being completely up to
date, since extra locking would be required to merge counters that track
thread cache operations.</p><p>The <code class="function">malloc_usable_size()</code> function
returns the usable size of the allocation pointed to by
<em class="parameter"><code>ptr</code></em>. The return value may be larger than the size
that was requested during allocation. The
<code class="function">malloc_usable_size()</code> function is not a
mechanism for in-place <code class="function">realloc()</code>; rather
it is provided solely as a tool for introspection purposes. Any
discrepancy between the requested allocation size and the size reported
by <code class="function">malloc_usable_size()</code> should not be
depended on, since such behavior is entirely implementation-dependent.
</p></div></div><div class="refsect1"><a name="tuning"/><h2>TUNING</h2><p>Once, when the first call is made to one of the memory allocation
routines, the allocator initializes its internals based in part on various
options that can be specified at compile- or run-time.</p><p>The string specified via <code class="option">--with-malloc-conf</code>, the
string pointed to by the global variable <code class="varname">malloc_conf</code>, the
<span class="quote">“<span class="quote">name</span>”</span> of the file referenced by the symbolic link named
<code class="filename">/etc/malloc.conf</code>, and the value of the
environment variable <code class="envar">MALLOC_CONF</code>, will be interpreted, in
that order, from left to right as options. Note that
<code class="varname">malloc_conf</code> may be read before
<code class="function">main()</code> is entered, so the declaration of
<code class="varname">malloc_conf</code> should specify an initializer that contains
the final value to be read by jemalloc. <code class="option">--with-malloc-conf</code>
and <code class="varname">malloc_conf</code> are compile-time mechanisms, whereas
<code class="filename">/etc/malloc.conf</code> and
<code class="envar">MALLOC_CONF</code> can be safely set any time prior to program
invocation.</p><p>An options string is a comma-separated list of option:value pairs.
There is one key corresponding to each <a class="link" href="#opt.abort"><quote><code class="mallctl">opt.*</code></quote></a> mallctl (see the <a class="xref" href="#mallctl_namespace" title="MALLCTL NAMESPACE">MALLCTL NAMESPACE</a> section for options
documentation). For example, <code class="literal">abort:true,narenas:1</code> sets
the <a class="link" href="#opt.abort"><quote><code class="mallctl">opt.abort</code></quote></a> and <a class="link" href="#opt.narenas"><quote><code class="mallctl">opt.narenas</code></quote></a> options. Some
options have boolean values (true/false), others have integer values (base
8, 10, or 16, depending on prefix), and yet others have raw string
values.</p></div><div class="refsect1"><a name="implementation_notes"/><h2>IMPLEMENTATION NOTES</h2><p>Traditionally, allocators have used
<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span> to obtain memory, which is
suboptimal for several reasons, including race conditions, increased
fragmentation, and artificial limitations on maximum usable memory. If
<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span> is supported by the operating
system, this allocator uses both
<span class="citerefentry"><span class="refentrytitle">mmap</span>(2)</span> and
<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span>, in that order of preference;
otherwise only <span class="citerefentry"><span class="refentrytitle">mmap</span>(2)</span> is used.</p><p>This allocator uses multiple arenas in order to reduce lock
contention for threaded programs on multi-processor systems. This works
well with regard to threading scalability, but incurs some costs. There is
a small fixed per-arena overhead, and additionally, arenas manage memory
completely independently of each other, which means a small fixed increase
in overall memory fragmentation. These overheads are not generally an
issue, given the number of arenas normally used. Note that using
substantially more arenas than the default is not likely to improve
performance, mainly due to reduced cache performance. However, it may make
sense to reduce the number of arenas if an application does not make much
use of the allocation functions.</p><p>In addition to multiple arenas, this allocator supports
thread-specific caching, in order to make it possible to completely avoid
synchronization for most allocation requests. Such caching allows very fast
allocation in the common case, but it increases memory usage and
fragmentation, since a bounded number of objects can remain allocated in
each thread cache.</p><p>Memory is conceptually broken into extents. Extents are always
aligned to multiples of the page size. This alignment makes it possible to
find metadata for user objects quickly. User objects are broken into two
categories according to size: small and large. Contiguous small objects
comprise a slab, which resides within a single extent, whereas large objects
each have their own extents backing them.</p><p>Small objects are managed in groups by slabs. Each slab maintains
a bitmap to track which regions are in use. Allocation requests that are no
more than half the quantum (8 or 16, depending on architecture) are rounded
up to the nearest power of two that is at least <code class="code">sizeof(<span class="type">double</span>)</code>. All other object size
classes are multiples of the quantum, spaced such that there are four size
classes for each doubling in size, which limits internal fragmentation to
approximately 20% for all but the smallest size classes. Small size classes
are smaller than four times the page size, and large size classes extend
from four times the page size up to the largest size class that does not
exceed <code class="constant">PTRDIFF_MAX</code>.</p><p>Allocations are packed tightly together, which can be an issue for
multi-threaded applications. If you need to assure that allocations do not
suffer from cacheline sharing, round your allocation requests up to the
nearest multiple of the cacheline size, or specify cacheline alignment when
allocating.</p><p>The <code class="function">realloc()</code>,
<code class="function">rallocx()</code>, and
<code class="function">xallocx()</code> functions may resize allocations
without moving them under limited circumstances. Unlike the
<code class="function">*allocx()</code> API, the standard API does not
officially round up the usable size of an allocation to the nearest size
class, so technically it is necessary to call
<code class="function">realloc()</code> to grow e.g. a 9-byte allocation to
16 bytes, or shrink a 16-byte allocation to 9 bytes. Growth and shrinkage
trivially succeeds in place as long as the pre-size and post-size both round
up to the same size class. No other API guarantees are made regarding
in-place resizing, but the current implementation also tries to resize large
allocations in place, as long as the pre-size and post-size are both large.
For shrinkage to succeed, the extent allocator must support splitting (see
<a class="link" href="#arena.i.extent_hooks"><quote><code class="mallctl">arena.<i>.extent_hooks</code></quote></a>).
Growth only succeeds if the trailing memory is currently available, and the
extent allocator supports merging.</p><p>Assuming 4 KiB pages and a 16-byte quantum on a 64-bit system, the
size classes in each category are as shown in <a class="xref" href="#size_classes" title="Table 1. Size classes">Table 1</a>.</p><div class="table"><a name="size_classes"/><p class="title"><b>Table 1. Size classes</b></p><div class="table-contents"><table summary="Size classes" border="1"><colgroup><col align="left" class="c1"/><col align="right" class="c2"/><col align="left" class="c3"/></colgroup><thead><tr><th align="left">Category</th><th align="right">Spacing</th><th align="left">Size</th></tr></thead><tbody><tr><td rowspan="9" align="left">Small</td><td align="right">lg</td><td align="left">[8]</td></tr><tr><td align="right">16</td><td align="left">[16, 32, 48, 64, 80, 96, 112, 128]</td></tr><tr><td align="right">32</td><td align="left">[160, 192, 224, 256]</td></tr><tr><td align="right">64</td><td align="left">[320, 384, 448, 512]</td></tr><tr><td align="right">128</td><td align="left">[640, 768, 896, 1024]</td></tr><tr><td align="right">256</td><td align="left">[1280, 1536, 1792, 2048]</td></tr><tr><td align="right">512</td><td align="left">[2560, 3072, 3584, 4096]</td></tr><tr><td align="right">1 KiB</td><td align="left">[5 KiB, 6 KiB, 7 KiB, 8 KiB]</td></tr><tr><td align="right">2 KiB</td><td align="left">[10 KiB, 12 KiB, 14 KiB]</td></tr><tr><td rowspan="16" align="left">Large</td><td align="right">2 KiB</td><td align="left">[16 KiB]</td></tr><tr><td align="right">4 KiB</td><td align="left">[20 KiB, 24 KiB, 28 KiB, 32 KiB]</td></tr><tr><td align="right">8 KiB</td><td align="left">[40 KiB, 48 KiB, 54 KiB, 64 KiB]</td></tr><tr><td align="right">16 KiB</td><td align="left">[80 KiB, 96 KiB, 112 KiB, 128 KiB]</td></tr><tr><td align="right">32 KiB</td><td align="left">[160 KiB, 192 KiB, 224 KiB, 256 KiB]</td></tr><tr><td align="right">64 KiB</td><td align="left">[320 KiB, 384 KiB, 448 KiB, 512 KiB]</td></tr><tr><td align="right">128 KiB</td><td align="left">[640 KiB, 768 KiB, 896 KiB, 1 MiB]</td></tr><tr><td align="right">256 KiB</td><td align="left">[1280 KiB, 1536 KiB, 1792 KiB, 2 MiB]</td></tr><tr><td align="right">512 KiB</td><td align="left">[2560 KiB, 3 MiB, 3584 KiB, 4 MiB]</td></tr><tr><td align="right">1 MiB</td><td align="left">[5 MiB, 6 MiB, 7 MiB, 8 MiB]</td></tr><tr><td align="right">2 MiB</td><td align="left">[10 MiB, 12 MiB, 14 MiB, 16 MiB]</td></tr><tr><td align="right">4 MiB</td><td align="left">[20 MiB, 24 MiB, 28 MiB, 32 MiB]</td></tr><tr><td align="right">8 MiB</td><td align="left">[40 MiB, 48 MiB, 56 MiB, 64 MiB]</td></tr><tr><td align="right">...</td><td align="left">...</td></tr><tr><td align="right">512 PiB</td><td align="left">[2560 PiB, 3 EiB, 3584 PiB, 4 EiB]</td></tr><tr><td align="right">1 EiB</td><td align="left">[5 EiB, 6 EiB, 7 EiB]</td></tr></tbody></table></div></div><br class="table-break"/></div><div class="refsect1"><a name="mallctl_namespace"/><h2>MALLCTL NAMESPACE</h2><p>The following names are defined in the namespace accessible via the
<code class="function">mallctl*()</code> functions. Value types are specified in
parentheses, their readable/writable statuses are encoded as
<code class="literal">rw</code>, <code class="literal">r-</code>, <code class="literal">-w</code>, or
<code class="literal">--</code>, and required build configuration flags follow, if
any. A name element encoded as <code class="literal"><i></code> or
<code class="literal"><j></code> indicates an integer component, where the
integer varies from 0 to some upper value that must be determined via
introspection. In the case of <quote><code class="mallctl">stats.arenas.<i>.*</code></quote>
and <quote><code class="mallctl">arena.<i>.{initialized,purge,decay,dss}</code></quote>,
<code class="literal"><i></code> equal to
<code class="constant">MALLCTL_ARENAS_ALL</code> can be used to operate on all arenas
or access the summation of statistics from all arenas; similarly
<code class="literal"><i></code> equal to
<code class="constant">MALLCTL_ARENAS_DESTROYED</code> can be used to access the
summation of statistics from all destroyed arenas. These constants can be
utilized either via <code class="function">mallctlnametomib()</code> followed by
<code class="function">mallctlbymib()</code>, or via code such as the following:
</p><pre class="programlisting">
#define STRINGIFY_HELPER(x) #x
#define STRINGIFY(x) STRINGIFY_HELPER(x)
mallctl("arena." STRINGIFY(MALLCTL_ARENAS_ALL) ".decay",
NULL, NULL, NULL, 0);</pre><p>
Take special note of the <a class="link" href="#epoch"><quote><code class="mallctl">epoch</code></quote></a> mallctl, which controls
refreshing of cached dynamic statistics.</p><div class="variablelist"><dl class="variablelist"><dt><a name="version"/><span class="term">
<quote><code class="mallctl">version</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Return the jemalloc version string.</p></dd><dt><a name="epoch"/><span class="term">
<quote><code class="mallctl">epoch</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>If a value is passed in, refresh the data from which
the <code class="function">mallctl*()</code> functions report values,
and increment the epoch. Return the current epoch. This is useful for
detecting whether another thread caused a refresh.</p></dd><dt><a name="background_thread"/><span class="term">
<quote><code class="mallctl">background_thread</code></quote>
(<span class="type">bool</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Enable/disable internal background worker threads. When
set to true, background threads are created on demand (the number of
background threads will be no more than the number of CPUs or active
arenas). Threads run periodically, and handle <a class="link" href="#arena.i.decay">purging</a> asynchronously. When switching
off, background threads are terminated synchronously. Note that after
<span class="citerefentry"><span class="refentrytitle">fork</span>(2)</span>
function, the state in the child process will be disabled regardless
the state in parent process. See <a class="link" href="#stats.background_thread.num_threads"><quote><code class="mallctl">stats.background_thread</code></quote></a>
for related stats. <a class="link" href="#opt.background_thread"><quote><code class="mallctl">opt.background_thread</code></quote></a>
can be used to set the default option. This option is only available on
selected pthread-based platforms.</p></dd><dt><a name="max_background_threads"/><span class="term">
<quote><code class="mallctl">max_background_threads</code></quote>
(<span class="type">size_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Maximum number of background worker threads that will
be created. This value is capped at <a class="link" href="#opt.max_background_threads"><quote><code class="mallctl">opt.max_background_threads</code></quote></a> at
startup.</p></dd><dt><a name="config.cache_oblivious"/><span class="term">
<quote><code class="mallctl">config.cache_oblivious</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-cache-oblivious</code> was specified
during build configuration.</p></dd><dt><a name="config.debug"/><span class="term">
<quote><code class="mallctl">config.debug</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-debug</code> was specified during
build configuration.</p></dd><dt><a name="config.fill"/><span class="term">
<quote><code class="mallctl">config.fill</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-fill</code> was specified during
build configuration.</p></dd><dt><a name="config.lazy_lock"/><span class="term">
<quote><code class="mallctl">config.lazy_lock</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-lazy-lock</code> was specified
during build configuration.</p></dd><dt><a name="config.malloc_conf"/><span class="term">
<quote><code class="mallctl">config.malloc_conf</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Embedded configure-time-specified run-time options
string, empty unless <code class="option">--with-malloc-conf</code> was specified
during build configuration.</p></dd><dt><a name="config.prof"/><span class="term">
<quote><code class="mallctl">config.prof</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-prof</code> was specified during
build configuration.</p></dd><dt><a name="config.prof_libgcc"/><span class="term">
<quote><code class="mallctl">config.prof_libgcc</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--disable-prof-libgcc</code> was not
specified during build configuration.</p></dd><dt><a name="config.prof_libunwind"/><span class="term">
<quote><code class="mallctl">config.prof_libunwind</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-prof-libunwind</code> was specified
during build configuration.</p></dd><dt><a name="config.stats"/><span class="term">
<quote><code class="mallctl">config.stats</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-stats</code> was specified during
build configuration.</p></dd><dt><a name="config.utrace"/><span class="term">
<quote><code class="mallctl">config.utrace</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-utrace</code> was specified during
build configuration.</p></dd><dt><a name="config.xmalloc"/><span class="term">
<quote><code class="mallctl">config.xmalloc</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p><code class="option">--enable-xmalloc</code> was specified during
build configuration.</p></dd><dt><a name="opt.abort"/><span class="term">
<quote><code class="mallctl">opt.abort</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Abort-on-warning enabled/disabled. If true, most
warnings are fatal. Note that runtime option warnings are not included
(see <a class="link" href="#opt.abort_conf"><quote><code class="mallctl">opt.abort_conf</code></quote></a> for
that). The process will call
<span class="citerefentry"><span class="refentrytitle">abort</span>(3)</span> in these cases. This option is
disabled by default unless <code class="option">--enable-debug</code> is
specified during configuration, in which case it is enabled by default.
</p></dd><dt><a name="opt.abort_conf"/><span class="term">
<quote><code class="mallctl">opt.abort_conf</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Abort-on-invalid-configuration enabled/disabled. If
true, invalid runtime options are fatal. The process will call
<span class="citerefentry"><span class="refentrytitle">abort</span>(3)</span> in these cases. This option is
disabled by default unless <code class="option">--enable-debug</code> is
specified during configuration, in which case it is enabled by default.
</p></dd><dt><a name="opt.metadata_thp"/><span class="term">
<quote><code class="mallctl">opt.metadata_thp</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Controls whether to allow jemalloc to use transparent
huge page (THP) for internal metadata (see <a class="link" href="#stats.metadata">stats.metadata</a>). <span class="quote">“<span class="quote">always</span>”</span>
allows such usage. <span class="quote">“<span class="quote">auto</span>”</span> uses no THP initially, but may
begin to do so when metadata usage reaches certain level. The default
is <span class="quote">“<span class="quote">disabled</span>”</span>.</p></dd><dt><a name="opt.retain"/><span class="term">
<quote><code class="mallctl">opt.retain</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>If true, retain unused virtual memory for later reuse
rather than discarding it by calling
<span class="citerefentry"><span class="refentrytitle">munmap</span>(2)</span> or equivalent (see <a class="link" href="#stats.retained">stats.retained</a> for related details).
This option is disabled by default unless discarding virtual memory is
known to trigger
platform-specific performance problems, e.g. for [64-bit] Linux, which
has a quirk in its virtual memory allocation algorithm that causes
semi-permanent VM map holes under normal jemalloc operation. Although
<span class="citerefentry"><span class="refentrytitle">munmap</span>(2)</span> causes issues on 32-bit Linux as
well, retaining virtual memory for 32-bit Linux is disabled by default
due to the practical possibility of address space exhaustion.
</p></dd><dt><a name="opt.dss"/><span class="term">
<quote><code class="mallctl">opt.dss</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>dss (<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span>) allocation precedence as
related to <span class="citerefentry"><span class="refentrytitle">mmap</span>(2)</span> allocation. The following
settings are supported if
<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span> is supported by the operating
system: <span class="quote">“<span class="quote">disabled</span>”</span>, <span class="quote">“<span class="quote">primary</span>”</span>, and
<span class="quote">“<span class="quote">secondary</span>”</span>; otherwise only <span class="quote">“<span class="quote">disabled</span>”</span> is
supported. The default is <span class="quote">“<span class="quote">secondary</span>”</span> if
<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span> is supported by the operating
system; <span class="quote">“<span class="quote">disabled</span>”</span> otherwise.
</p></dd><dt><a name="opt.narenas"/><span class="term">
<quote><code class="mallctl">opt.narenas</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Maximum number of arenas to use for automatic
multiplexing of threads and arenas. The default is four times the
number of CPUs, or one if there is a single CPU.</p></dd><dt><a name="opt.percpu_arena"/><span class="term">
<quote><code class="mallctl">opt.percpu_arena</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Per CPU arena mode. Use the <span class="quote">“<span class="quote">percpu</span>”</span>
setting to enable this feature, which uses number of CPUs to determine
number of arenas, and bind threads to arenas dynamically based on the
CPU the thread runs on currently. <span class="quote">“<span class="quote">phycpu</span>”</span> setting uses
one arena per physical CPU, which means the two hyper threads on the
same CPU share one arena. Note that no runtime checking regarding the
availability of hyper threading is done at the moment. When set to
<span class="quote">“<span class="quote">disabled</span>”</span>, narenas and thread to arena association will
not be impacted by this option. The default is <span class="quote">“<span class="quote">disabled</span>”</span>.
</p></dd><dt><a name="opt.background_thread"/><span class="term">
<quote><code class="mallctl">opt.background_thread</code></quote>
(<span class="type">const bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Internal background worker threads enabled/disabled.
Because of potential circular dependencies, enabling background thread
using this option may cause crash or deadlock during initialization. For
a reliable way to use this feature, see <a class="link" href="#background_thread">background_thread</a> for dynamic control
options and details. This option is disabled by
default.</p></dd><dt><a name="opt.max_background_threads"/><span class="term">
<quote><code class="mallctl">opt.max_background_threads</code></quote>
(<span class="type">const size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Maximum number of background threads that will be created
if <a class="link" href="#background_thread">background_thread</a> is set.
Defaults to number of cpus.</p></dd><dt><a name="opt.dirty_decay_ms"/><span class="term">
<quote><code class="mallctl">opt.dirty_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Approximate time in milliseconds from the creation of a
set of unused dirty pages until an equivalent set of unused dirty pages
is purged (i.e. converted to muzzy via e.g.
<code class="function">madvise(<em class="parameter"><code>...</code></em><em class="parameter"><code><code class="constant">MADV_FREE</code></code></em>)</code>
if supported by the operating system, or converted to clean otherwise)
and/or reused. Dirty pages are defined as previously having been
potentially written to by the application, and therefore consuming
physical memory, yet having no current use. The pages are incrementally
purged according to a sigmoidal decay curve that starts and ends with
zero purge rate. A decay time of 0 causes all unused dirty pages to be
purged immediately upon creation. A decay time of -1 disables purging.
The default decay time is 10 seconds. See <a class="link" href="#arenas.dirty_decay_ms"><quote><code class="mallctl">arenas.dirty_decay_ms</code></quote></a>
and <a class="link" href="#arena.i.dirty_decay_ms"><quote><code class="mallctl">arena.<i>.dirty_decay_ms</code></quote></a>
for related dynamic control options. See <a class="link" href="#opt.muzzy_decay_ms"><quote><code class="mallctl">opt.muzzy_decay_ms</code></quote></a>
for a description of muzzy pages.</p></dd><dt><a name="opt.muzzy_decay_ms"/><span class="term">
<quote><code class="mallctl">opt.muzzy_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Approximate time in milliseconds from the creation of a
set of unused muzzy pages until an equivalent set of unused muzzy pages
is purged (i.e. converted to clean) and/or reused. Muzzy pages are
defined as previously having been unused dirty pages that were
subsequently purged in a manner that left them subject to the
reclamation whims of the operating system (e.g.
<code class="function">madvise(<em class="parameter"><code>...</code></em><em class="parameter"><code><code class="constant">MADV_FREE</code></code></em>)</code>),
and therefore in an indeterminate state. The pages are incrementally
purged according to a sigmoidal decay curve that starts and ends with
zero purge rate. A decay time of 0 causes all unused muzzy pages to be
purged immediately upon creation. A decay time of -1 disables purging.
The default decay time is 10 seconds. See <a class="link" href="#arenas.muzzy_decay_ms"><quote><code class="mallctl">arenas.muzzy_decay_ms</code></quote></a>
and <a class="link" href="#arena.i.muzzy_decay_ms"><quote><code class="mallctl">arena.<i>.muzzy_decay_ms</code></quote></a>
for related dynamic control options.</p></dd><dt><a name="opt.lg_extent_max_active_fit"/><span class="term">
<quote><code class="mallctl">opt.lg_extent_max_active_fit</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>When reusing dirty extents, this determines the (log
base 2 of the) maximum ratio between the size of the active extent
selected (to split off from) and the size of the requested allocation.
This prevents the splitting of large active extents for smaller
allocations, which can reduce fragmentation over the long run
(especially for non-active extents). Lower value may reduce
fragmentation, at the cost of extra active extents. The default value
is 6, which gives a maximum ratio of 64 (2^6).</p></dd><dt><a name="opt.stats_print"/><span class="term">
<quote><code class="mallctl">opt.stats_print</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Enable/disable statistics printing at exit. If
enabled, the <code class="function">malloc_stats_print()</code>
function is called at program exit via an
<span class="citerefentry"><span class="refentrytitle">atexit</span>(3)</span> function. <a class="link" href="#opt.stats_print_opts"><quote><code class="mallctl">opt.stats_print_opts</code></quote></a>
can be combined to specify output options. If
<code class="option">--enable-stats</code> is specified during configuration, this
has the potential to cause deadlock for a multi-threaded process that
exits while one or more threads are executing in the memory allocation
functions. Furthermore, <code class="function">atexit()</code> may
allocate memory during application initialization and then deadlock
internally when jemalloc in turn calls
<code class="function">atexit()</code>, so this option is not
universally usable (though the application can register its own
<code class="function">atexit()</code> function with equivalent
functionality). Therefore, this option should only be used with care;
it is primarily intended as a performance tuning aid during application
development. This option is disabled by default.</p></dd><dt><a name="opt.stats_print_opts"/><span class="term">
<quote><code class="mallctl">opt.stats_print_opts</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Options (the <em class="parameter"><code>opts</code></em> string) to pass
to the <code class="function">malloc_stats_print()</code> at exit (enabled
through <a class="link" href="#opt.stats_print"><quote><code class="mallctl">opt.stats_print</code></quote></a>). See
available options in <a class="link" href="#malloc_stats_print_opts"><code class="function">malloc_stats_print()</code></a>.
Has no effect unless <a class="link" href="#opt.stats_print"><quote><code class="mallctl">opt.stats_print</code></quote></a> is
enabled. The default is <span class="quote">“<span class="quote"/>”</span>.</p></dd><dt><a name="opt.junk"/><span class="term">
<quote><code class="mallctl">opt.junk</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
[<code class="option">--enable-fill</code>]
</span></dt><dd><p>Junk filling. If set to <span class="quote">“<span class="quote">alloc</span>”</span>, each byte
of uninitialized allocated memory will be initialized to
<code class="literal">0xa5</code>. If set to <span class="quote">“<span class="quote">free</span>”</span>, all deallocated
memory will be initialized to <code class="literal">0x5a</code>. If set to
<span class="quote">“<span class="quote">true</span>”</span>, both allocated and deallocated memory will be
initialized, and if set to <span class="quote">“<span class="quote">false</span>”</span>, junk filling be
disabled entirely. This is intended for debugging and will impact
performance negatively. This option is <span class="quote">“<span class="quote">false</span>”</span> by default
unless <code class="option">--enable-debug</code> is specified during
configuration, in which case it is <span class="quote">“<span class="quote">true</span>”</span> by
default.</p></dd><dt><a name="opt.zero"/><span class="term">
<quote><code class="mallctl">opt.zero</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-fill</code>]
</span></dt><dd><p>Zero filling enabled/disabled. If enabled, each byte
of uninitialized allocated memory will be initialized to 0. Note that
this initialization only happens once for each byte, so
<code class="function">realloc()</code> and
<code class="function">rallocx()</code> calls do not zero memory that
was previously allocated. This is intended for debugging and will
impact performance negatively. This option is disabled by default.
</p></dd><dt><a name="opt.utrace"/><span class="term">
<quote><code class="mallctl">opt.utrace</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-utrace</code>]
</span></dt><dd><p>Allocation tracing based on
<span class="citerefentry"><span class="refentrytitle">utrace</span>(2)</span> enabled/disabled. This option
is disabled by default.</p></dd><dt><a name="opt.xmalloc"/><span class="term">
<quote><code class="mallctl">opt.xmalloc</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-xmalloc</code>]
</span></dt><dd><p>Abort-on-out-of-memory enabled/disabled. If enabled,
rather than returning failure for any allocation function, display a
diagnostic message on <code class="constant">STDERR_FILENO</code> and cause the
program to drop core (using
<span class="citerefentry"><span class="refentrytitle">abort</span>(3)</span>). If an application is
designed to depend on this behavior, set the option at compile time by
including the following in the source code:
</p><pre class="programlisting">
malloc_conf = "xmalloc:true";</pre><p>
This option is disabled by default.</p></dd><dt><a name="opt.tcache"/><span class="term">
<quote><code class="mallctl">opt.tcache</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Thread-specific caching (tcache) enabled/disabled. When
there are multiple threads, each thread uses a tcache for objects up to
a certain size. Thread-specific caching allows many allocations to be
satisfied without performing any thread synchronization, at the cost of
increased memory use. See the <a class="link" href="#opt.lg_tcache_max"><quote><code class="mallctl">opt.lg_tcache_max</code></quote></a>
option for related tuning information. This option is enabled by
default.</p></dd><dt><a name="opt.lg_tcache_max"/><span class="term">
<quote><code class="mallctl">opt.lg_tcache_max</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Maximum size class (log base 2) to cache in the
thread-specific cache (tcache). At a minimum, all small size classes
are cached, and at a maximum all large size classes are cached. The
default maximum is 32 KiB (2^15).</p></dd><dt><a name="opt.thp"/><span class="term">
<quote><code class="mallctl">opt.thp</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Transparent hugepage (THP) mode. Settings "always",
"never" and "default" are available if THP is supported by the operating
system. The "always" setting enables transparent hugepage for all user
memory mappings with
<em class="parameter"><code><code class="constant">MADV_HUGEPAGE</code></code></em>; "never"
ensures no transparent hugepage with
<em class="parameter"><code><code class="constant">MADV_NOHUGEPAGE</code></code></em>; the default
setting "default" makes no changes. Note that: this option does not
affect THP for jemalloc internal metadata (see <a class="link" href="#opt.metadata_thp"><quote><code class="mallctl">opt.metadata_thp</code></quote></a>);
in addition, for arenas with customized <a class="link" href="#arena.i.extent_hooks"><quote><code class="mallctl">extent_hooks</code></quote></a>,
this option is bypassed as it is implemented as part of the default
extent hooks.</p></dd><dt><a name="opt.prof"/><span class="term">
<quote><code class="mallctl">opt.prof</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Memory profiling enabled/disabled. If enabled, profile
memory allocation activity. See the <a class="link" href="#opt.prof_active"><quote><code class="mallctl">opt.prof_active</code></quote></a>
option for on-the-fly activation/deactivation. See the <a class="link" href="#opt.lg_prof_sample"><quote><code class="mallctl">opt.lg_prof_sample</code></quote></a>
option for probabilistic sampling control. See the <a class="link" href="#opt.prof_accum"><quote><code class="mallctl">opt.prof_accum</code></quote></a>
option for control of cumulative sample reporting. See the <a class="link" href="#opt.lg_prof_interval"><quote><code class="mallctl">opt.lg_prof_interval</code></quote></a>
option for information on interval-triggered profile dumping, the <a class="link" href="#opt.prof_gdump"><quote><code class="mallctl">opt.prof_gdump</code></quote></a>
option for information on high-water-triggered profile dumping, and the
<a class="link" href="#opt.prof_final"><quote><code class="mallctl">opt.prof_final</code></quote></a>
option for final profile dumping. Profile output is compatible with
the <span class="command"><strong>jeprof</strong></span> command, which is based on the
<span class="command"><strong>pprof</strong></span> that is developed as part of the <a class="ulink" href="http://code.google.com/p/gperftools/" target="_top">gperftools
package</a>. See <a class="link" href="#heap_profile_format" title="HEAP PROFILE FORMAT">HEAP PROFILE
FORMAT</a> for heap profile format documentation.</p></dd><dt><a name="opt.prof_prefix"/><span class="term">
<quote><code class="mallctl">opt.prof_prefix</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Filename prefix for profile dumps. If the prefix is
set to the empty string, no automatic dumps will occur; this is
primarily useful for disabling the automatic final heap dump (which
also disables leak reporting, if enabled). The default prefix is
<code class="filename">jeprof</code>.</p></dd><dt><a name="opt.prof_active"/><span class="term">
<quote><code class="mallctl">opt.prof_active</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Profiling activated/deactivated. This is a secondary
control mechanism that makes it possible to start the application with
profiling enabled (see the <a class="link" href="#opt.prof"><quote><code class="mallctl">opt.prof</code></quote></a> option) but
inactive, then toggle profiling at any time during program execution
with the <a class="link" href="#prof.active"><quote><code class="mallctl">prof.active</code></quote></a> mallctl.
This option is enabled by default.</p></dd><dt><a name="opt.prof_thread_active_init"/><span class="term">
<quote><code class="mallctl">opt.prof_thread_active_init</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Initial setting for <a class="link" href="#thread.prof.active"><quote><code class="mallctl">thread.prof.active</code></quote></a>
in newly created threads. The initial setting for newly created threads
can also be changed during execution via the <a class="link" href="#prof.thread_active_init"><quote><code class="mallctl">prof.thread_active_init</code></quote></a>
mallctl. This option is enabled by default.</p></dd><dt><a name="opt.lg_prof_sample"/><span class="term">
<quote><code class="mallctl">opt.lg_prof_sample</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Average interval (log base 2) between allocation
samples, as measured in bytes of allocation activity. Increasing the
sampling interval decreases profile fidelity, but also decreases the
computational overhead. The default sample interval is 512 KiB (2^19
B).</p></dd><dt><a name="opt.prof_accum"/><span class="term">
<quote><code class="mallctl">opt.prof_accum</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Reporting of cumulative object/byte counts in profile
dumps enabled/disabled. If this option is enabled, every unique
backtrace must be stored for the duration of execution. Depending on
the application, this can impose a large memory overhead, and the
cumulative counts are not always of interest. This option is disabled
by default.</p></dd><dt><a name="opt.lg_prof_interval"/><span class="term">
<quote><code class="mallctl">opt.lg_prof_interval</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Average interval (log base 2) between memory profile
dumps, as measured in bytes of allocation activity. The actual
interval between dumps may be sporadic because decentralized allocation
counters are used to avoid synchronization bottlenecks. Profiles are
dumped to files named according to the pattern
<code class="filename"><prefix>.<pid>.<seq>.i<iseq>.heap</code>,
where <code class="literal"><prefix></code> is controlled by the
<a class="link" href="#opt.prof_prefix"><quote><code class="mallctl">opt.prof_prefix</code></quote></a>
option. By default, interval-triggered profile dumping is disabled
(encoded as -1).
</p></dd><dt><a name="opt.prof_gdump"/><span class="term">
<quote><code class="mallctl">opt.prof_gdump</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Set the initial state of <a class="link" href="#prof.gdump"><quote><code class="mallctl">prof.gdump</code></quote></a>, which when
enabled triggers a memory profile dump every time the total virtual
memory exceeds the previous maximum. This option is disabled by
default.</p></dd><dt><a name="opt.prof_final"/><span class="term">
<quote><code class="mallctl">opt.prof_final</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Use an
<span class="citerefentry"><span class="refentrytitle">atexit</span>(3)</span> function to dump final memory
usage to a file named according to the pattern
<code class="filename"><prefix>.<pid>.<seq>.f.heap</code>,
where <code class="literal"><prefix></code> is controlled by the <a class="link" href="#opt.prof_prefix"><quote><code class="mallctl">opt.prof_prefix</code></quote></a>
option. Note that <code class="function">atexit()</code> may allocate
memory during application initialization and then deadlock internally
when jemalloc in turn calls <code class="function">atexit()</code>, so
this option is not universally usable (though the application can
register its own <code class="function">atexit()</code> function with
equivalent functionality). This option is disabled by
default.</p></dd><dt><a name="opt.prof_leak"/><span class="term">
<quote><code class="mallctl">opt.prof_leak</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Leak reporting enabled/disabled. If enabled, use an
<span class="citerefentry"><span class="refentrytitle">atexit</span>(3)</span> function to report memory leaks
detected by allocation sampling. See the
<a class="link" href="#opt.prof"><quote><code class="mallctl">opt.prof</code></quote></a> option for
information on analyzing heap profile output. This option is disabled
by default.</p></dd><dt><a name="thread.arena"/><span class="term">
<quote><code class="mallctl">thread.arena</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Get or set the arena associated with the calling
thread. If the specified arena was not initialized beforehand (see the
<a class="link" href="#arena.i.initialized"><quote><code class="mallctl">arena.i.initialized</code></quote></a>
mallctl), it will be automatically initialized as a side effect of
calling this interface.</p></dd><dt><a name="thread.allocated"/><span class="term">
<quote><code class="mallctl">thread.allocated</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Get the total number of bytes ever allocated by the
calling thread. This counter has the potential to wrap around; it is
up to the application to appropriately interpret the counter in such
cases.</p></dd><dt><a name="thread.allocatedp"/><span class="term">
<quote><code class="mallctl">thread.allocatedp</code></quote>
(<span class="type">uint64_t *</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Get a pointer to the the value that is returned by the
<a class="link" href="#thread.allocated"><quote><code class="mallctl">thread.allocated</code></quote></a>
mallctl. This is useful for avoiding the overhead of repeated
<code class="function">mallctl*()</code> calls.</p></dd><dt><a name="thread.deallocated"/><span class="term">
<quote><code class="mallctl">thread.deallocated</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Get the total number of bytes ever deallocated by the
calling thread. This counter has the potential to wrap around; it is
up to the application to appropriately interpret the counter in such
cases.</p></dd><dt><a name="thread.deallocatedp"/><span class="term">
<quote><code class="mallctl">thread.deallocatedp</code></quote>
(<span class="type">uint64_t *</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Get a pointer to the the value that is returned by the
<a class="link" href="#thread.deallocated"><quote><code class="mallctl">thread.deallocated</code></quote></a>
mallctl. This is useful for avoiding the overhead of repeated
<code class="function">mallctl*()</code> calls.</p></dd><dt><a name="thread.tcache.enabled"/><span class="term">
<quote><code class="mallctl">thread.tcache.enabled</code></quote>
(<span class="type">bool</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Enable/disable calling thread's tcache. The tcache is
implicitly flushed as a side effect of becoming
disabled (see <a class="link" href="#thread.tcache.flush"><quote><code class="mallctl">thread.tcache.flush</code></quote></a>).
</p></dd><dt><a name="thread.tcache.flush"/><span class="term">
<quote><code class="mallctl">thread.tcache.flush</code></quote>
(<span class="type">void</span>)
<code class="literal">--</code>
</span></dt><dd><p>Flush calling thread's thread-specific cache (tcache).
This interface releases all cached objects and internal data structures
associated with the calling thread's tcache. Ordinarily, this interface
need not be called, since automatic periodic incremental garbage
collection occurs, and the thread cache is automatically discarded when
a thread exits. However, garbage collection is triggered by allocation
activity, so it is possible for a thread that stops
allocating/deallocating to retain its cache indefinitely, in which case
the developer may find manual flushing useful.</p></dd><dt><a name="thread.prof.name"/><span class="term">
<quote><code class="mallctl">thread.prof.name</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code> or
<code class="literal">-w</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Get/set the descriptive name associated with the calling
thread in memory profile dumps. An internal copy of the name string is
created, so the input string need not be maintained after this interface
completes execution. The output string of this interface should be
copied for non-ephemeral uses, because multiple implementation details
can cause asynchronous string deallocation. Furthermore, each
invocation of this interface can only read or write; simultaneous
read/write is not supported due to string lifetime limitations. The
name string must be nil-terminated and comprised only of characters in
the sets recognized
by <span class="citerefentry"><span class="refentrytitle">isgraph</span>(3)</span> and
<span class="citerefentry"><span class="refentrytitle">isblank</span>(3)</span>.</p></dd><dt><a name="thread.prof.active"/><span class="term">
<quote><code class="mallctl">thread.prof.active</code></quote>
(<span class="type">bool</span>)
<code class="literal">rw</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Control whether sampling is currently active for the
calling thread. This is an activation mechanism in addition to <a class="link" href="#prof.active"><quote><code class="mallctl">prof.active</code></quote></a>; both must
be active for the calling thread to sample. This flag is enabled by
default.</p></dd><dt><a name="tcache.create"/><span class="term">
<quote><code class="mallctl">tcache.create</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Create an explicit thread-specific cache (tcache) and
return an identifier that can be passed to the <a class="link" href="#MALLOCX_TCACHE"><code class="constant">MALLOCX_TCACHE(<em class="parameter"><code>tc</code></em>)</code></a>
macro to explicitly use the specified cache rather than the
automatically managed one that is used by default. Each explicit cache
can be used by only one thread at a time; the application must assure
that this constraint holds.
</p></dd><dt><a name="tcache.flush"/><span class="term">
<quote><code class="mallctl">tcache.flush</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">-w</code>
</span></dt><dd><p>Flush the specified thread-specific cache (tcache). The
same considerations apply to this interface as to <a class="link" href="#thread.tcache.flush"><quote><code class="mallctl">thread.tcache.flush</code></quote></a>,
except that the tcache will never be automatically discarded.
</p></dd><dt><a name="tcache.destroy"/><span class="term">
<quote><code class="mallctl">tcache.destroy</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">-w</code>
</span></dt><dd><p>Flush the specified thread-specific cache (tcache) and
make the identifier available for use during a future tcache creation.
</p></dd><dt><a name="arena.i.initialized"/><span class="term">
<quote><code class="mallctl">arena.<i>.initialized</code></quote>
(<span class="type">bool</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Get whether the specified arena's statistics are
initialized (i.e. the arena was initialized prior to the current epoch).
This interface can also be nominally used to query whether the merged
statistics corresponding to <code class="constant">MALLCTL_ARENAS_ALL</code> are
initialized (always true).</p></dd><dt><a name="arena.i.decay"/><span class="term">
<quote><code class="mallctl">arena.<i>.decay</code></quote>
(<span class="type">void</span>)
<code class="literal">--</code>
</span></dt><dd><p>Trigger decay-based purging of unused dirty/muzzy pages
for arena <i>, or for all arenas if <i> equals
<code class="constant">MALLCTL_ARENAS_ALL</code>. The proportion of unused
dirty/muzzy pages to be purged depends on the current time; see <a class="link" href="#opt.dirty_decay_ms"><quote><code class="mallctl">opt.dirty_decay_ms</code></quote></a>
and <a class="link" href="#opt.muzzy_decay_ms"><quote><code class="mallctl">opt.muzy_decay_ms</code></quote></a>
for details.</p></dd><dt><a name="arena.i.purge"/><span class="term">
<quote><code class="mallctl">arena.<i>.purge</code></quote>
(<span class="type">void</span>)
<code class="literal">--</code>
</span></dt><dd><p>Purge all unused dirty pages for arena <i>, or for
all arenas if <i> equals <code class="constant">MALLCTL_ARENAS_ALL</code>.
</p></dd><dt><a name="arena.i.reset"/><span class="term">
<quote><code class="mallctl">arena.<i>.reset</code></quote>
(<span class="type">void</span>)
<code class="literal">--</code>
</span></dt><dd><p>Discard all of the arena's extant allocations. This
interface can only be used with arenas explicitly created via <a class="link" href="#arenas.create"><quote><code class="mallctl">arenas.create</code></quote></a>. None
of the arena's discarded/cached allocations may accessed afterward. As
part of this requirement, all thread caches which were used to
allocate/deallocate in conjunction with the arena must be flushed
beforehand.</p></dd><dt><a name="arena.i.destroy"/><span class="term">
<quote><code class="mallctl">arena.<i>.destroy</code></quote>
(<span class="type">void</span>)
<code class="literal">--</code>
</span></dt><dd><p>Destroy the arena. Discard all of the arena's extant
allocations using the same mechanism as for <a class="link" href="#arena.i.reset"><quote><code class="mallctl">arena.<i>.reset</code></quote></a>
(with all the same constraints and side effects), merge the arena stats
into those accessible at arena index
<code class="constant">MALLCTL_ARENAS_DESTROYED</code>, and then completely
discard all metadata associated with the arena. Future calls to <a class="link" href="#arenas.create"><quote><code class="mallctl">arenas.create</code></quote></a> may
recycle the arena index. Destruction will fail if any threads are
currently associated with the arena as a result of calls to <a class="link" href="#thread.arena"><quote><code class="mallctl">thread.arena</code></quote></a>.</p></dd><dt><a name="arena.i.dss"/><span class="term">
<quote><code class="mallctl">arena.<i>.dss</code></quote>
(<span class="type">const char *</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Set the precedence of dss allocation as related to mmap
allocation for arena <i>, or for all arenas if <i> equals
<code class="constant">MALLCTL_ARENAS_ALL</code>. See <a class="link" href="#opt.dss"><quote><code class="mallctl">opt.dss</code></quote></a> for supported
settings.</p></dd><dt><a name="arena.i.dirty_decay_ms"/><span class="term">
<quote><code class="mallctl">arena.<i>.dirty_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Current per-arena approximate time in milliseconds from
the creation of a set of unused dirty pages until an equivalent set of
unused dirty pages is purged and/or reused. Each time this interface is
set, all currently unused dirty pages are considered to have fully
decayed, which causes immediate purging of all unused dirty pages unless
the decay time is set to -1 (i.e. purging disabled). See <a class="link" href="#opt.dirty_decay_ms"><quote><code class="mallctl">opt.dirty_decay_ms</code></quote></a>
for additional information.</p></dd><dt><a name="arena.i.muzzy_decay_ms"/><span class="term">
<quote><code class="mallctl">arena.<i>.muzzy_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Current per-arena approximate time in milliseconds from
the creation of a set of unused muzzy pages until an equivalent set of
unused muzzy pages is purged and/or reused. Each time this interface is
set, all currently unused muzzy pages are considered to have fully
decayed, which causes immediate purging of all unused muzzy pages unless
the decay time is set to -1 (i.e. purging disabled). See <a class="link" href="#opt.muzzy_decay_ms"><quote><code class="mallctl">opt.muzzy_decay_ms</code></quote></a>
for additional information.</p></dd><dt><a name="arena.i.retain_grow_limit"/><span class="term">
<quote><code class="mallctl">arena.<i>.retain_grow_limit</code></quote>
(<span class="type">size_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Maximum size to grow retained region (only relevant when
<a class="link" href="#opt.retain"><quote><code class="mallctl">opt.retain</code></quote></a> is
enabled). This controls the maximum increment to expand virtual memory,
or allocation through <a class="link" href="#arena.i.extent_hooks"><quote><code class="mallctl">arena.<i>extent_hooks</code></quote></a>.
In particular, if customized extent hooks reserve physical memory
(e.g. 1G huge pages), this is useful to control the allocation hook's
input size. The default is no limit.</p></dd><dt><a name="arena.i.extent_hooks"/><span class="term">
<quote><code class="mallctl">arena.<i>.extent_hooks</code></quote>
(<span class="type">extent_hooks_t *</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Get or set the extent management hook functions for
arena <i>. The functions must be capable of operating on all
extant extents associated with arena <i>, usually by passing
unknown extents to the replaced functions. In practice, it is feasible
to control allocation for arenas explicitly created via <a class="link" href="#arenas.create"><quote><code class="mallctl">arenas.create</code></quote></a> such
that all extents originate from an application-supplied extent allocator
(by specifying the custom extent hook functions during arena creation),
but the automatically created arenas will have already created extents
prior to the application having an opportunity to take over extent
allocation.</p><pre class="programlisting">
typedef extent_hooks_s extent_hooks_t;
struct extent_hooks_s {
extent_alloc_t *alloc;
extent_dalloc_t *dalloc;
extent_destroy_t *destroy;
extent_commit_t *commit;
extent_decommit_t *decommit;
extent_purge_t *purge_lazy;
extent_purge_t *purge_forced;
extent_split_t *split;
extent_merge_t *merge;
};</pre><p>The <span class="type">extent_hooks_t</span> structure comprises function
pointers which are described individually below. jemalloc uses these
functions to manage extent lifetime, which starts off with allocation of
mapped committed memory, in the simplest case followed by deallocation.
However, there are performance and platform reasons to retain extents
for later reuse. Cleanup attempts cascade from deallocation to decommit
to forced purging to lazy purging, which gives the extent management
functions opportunities to reject the most permanent cleanup operations
in favor of less permanent (and often less costly) operations. All
operations except allocation can be universally opted out of by setting
the hook pointers to <code class="constant">NULL</code>, or selectively opted out
of by returning failure. Note that once the extent hook is set, the
structure is accessed directly by the associated arenas, so it must
remain valid for the entire lifetime of the arenas.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef void *<b class="fsfunc">(extent_alloc_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">new_addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">alignment</var>, </td></tr><tr><td> </td><td>bool *<var class="pdparam">zero</var>, </td></tr><tr><td> </td><td>bool *<var class="pdparam">commit</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>An extent allocation function conforms to the
<span class="type">extent_alloc_t</span> type and upon success returns a pointer to
<em class="parameter"><code>size</code></em> bytes of mapped memory on behalf of arena
<em class="parameter"><code>arena_ind</code></em> such that the extent's base address is
a multiple of <em class="parameter"><code>alignment</code></em>, as well as setting
<em class="parameter"><code>*zero</code></em> to indicate whether the extent is zeroed
and <em class="parameter"><code>*commit</code></em> to indicate whether the extent is
committed. Upon error the function returns <code class="constant">NULL</code>
and leaves <em class="parameter"><code>*zero</code></em> and
<em class="parameter"><code>*commit</code></em> unmodified. The
<em class="parameter"><code>size</code></em> parameter is always a multiple of the page
size. The <em class="parameter"><code>alignment</code></em> parameter is always a power
of two at least as large as the page size. Zeroing is mandatory if
<em class="parameter"><code>*zero</code></em> is true upon function entry. Committing is
mandatory if <em class="parameter"><code>*commit</code></em> is true upon function entry.
If <em class="parameter"><code>new_addr</code></em> is not <code class="constant">NULL</code>, the
returned pointer must be <em class="parameter"><code>new_addr</code></em> on success or
<code class="constant">NULL</code> on error. Committed memory may be committed
in absolute terms as on a system that does not overcommit, or in
implicit terms as on a system that overcommits and satisfies physical
memory needs on demand via soft page faults. Note that replacing the
default extent allocation function makes the arena's <a class="link" href="#arena.i.dss"><quote><code class="mallctl">arena.<i>.dss</code></quote></a>
setting irrelevant.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef bool <b class="fsfunc">(extent_dalloc_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>bool <var class="pdparam">committed</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>
An extent deallocation function conforms to the
<span class="type">extent_dalloc_t</span> type and deallocates an extent at given
<em class="parameter"><code>addr</code></em> and <em class="parameter"><code>size</code></em> with
<em class="parameter"><code>committed</code></em>/decommited memory as indicated, on
behalf of arena <em class="parameter"><code>arena_ind</code></em>, returning false upon
success. If the function returns true, this indicates opt-out from
deallocation; the virtual memory mapping associated with the extent
remains mapped, in the same commit state, and available for future use,
in which case it will be automatically retained for later reuse.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef void <b class="fsfunc">(extent_destroy_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>bool <var class="pdparam">committed</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>
An extent destruction function conforms to the
<span class="type">extent_destroy_t</span> type and unconditionally destroys an
extent at given <em class="parameter"><code>addr</code></em> and
<em class="parameter"><code>size</code></em> with
<em class="parameter"><code>committed</code></em>/decommited memory as indicated, on
behalf of arena <em class="parameter"><code>arena_ind</code></em>. This function may be
called to destroy retained extents during arena destruction (see <a class="link" href="#arena.i.destroy"><quote><code class="mallctl">arena.<i>.destroy</code></quote></a>).</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef bool <b class="fsfunc">(extent_commit_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">offset</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">length</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>An extent commit function conforms to the
<span class="type">extent_commit_t</span> type and commits zeroed physical memory to
back pages within an extent at given <em class="parameter"><code>addr</code></em> and
<em class="parameter"><code>size</code></em> at <em class="parameter"><code>offset</code></em> bytes,
extending for <em class="parameter"><code>length</code></em> on behalf of arena
<em class="parameter"><code>arena_ind</code></em>, returning false upon success.
Committed memory may be committed in absolute terms as on a system that
does not overcommit, or in implicit terms as on a system that
overcommits and satisfies physical memory needs on demand via soft page
faults. If the function returns true, this indicates insufficient
physical memory to satisfy the request.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef bool <b class="fsfunc">(extent_decommit_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">offset</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">length</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>An extent decommit function conforms to the
<span class="type">extent_decommit_t</span> type and decommits any physical memory
that is backing pages within an extent at given
<em class="parameter"><code>addr</code></em> and <em class="parameter"><code>size</code></em> at
<em class="parameter"><code>offset</code></em> bytes, extending for
<em class="parameter"><code>length</code></em> on behalf of arena
<em class="parameter"><code>arena_ind</code></em>, returning false upon success, in which
case the pages will be committed via the extent commit function before
being reused. If the function returns true, this indicates opt-out from
decommit; the memory remains committed and available for future use, in
which case it will be automatically retained for later reuse.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef bool <b class="fsfunc">(extent_purge_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">offset</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">length</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>An extent purge function conforms to the
<span class="type">extent_purge_t</span> type and discards physical pages
within the virtual memory mapping associated with an extent at given
<em class="parameter"><code>addr</code></em> and <em class="parameter"><code>size</code></em> at
<em class="parameter"><code>offset</code></em> bytes, extending for
<em class="parameter"><code>length</code></em> on behalf of arena
<em class="parameter"><code>arena_ind</code></em>. A lazy extent purge function (e.g.
implemented via
<code class="function">madvise(<em class="parameter"><code>...</code></em><em class="parameter"><code><code class="constant">MADV_FREE</code></code></em>)</code>)
can delay purging indefinitely and leave the pages within the purged
virtual memory range in an indeterminite state, whereas a forced extent
purge function immediately purges, and the pages within the virtual
memory range will be zero-filled the next time they are accessed. If
the function returns true, this indicates failure to purge.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef bool <b class="fsfunc">(extent_split_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size_a</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size_b</var>, </td></tr><tr><td> </td><td>bool <var class="pdparam">committed</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>An extent split function conforms to the
<span class="type">extent_split_t</span> type and optionally splits an extent at
given <em class="parameter"><code>addr</code></em> and <em class="parameter"><code>size</code></em> into
two adjacent extents, the first of <em class="parameter"><code>size_a</code></em> bytes,
and the second of <em class="parameter"><code>size_b</code></em> bytes, operating on
<em class="parameter"><code>committed</code></em>/decommitted memory as indicated, on
behalf of arena <em class="parameter"><code>arena_ind</code></em>, returning false upon
success. If the function returns true, this indicates that the extent
remains unsplit and therefore should continue to be operated on as a
whole.</p><div class="funcsynopsis"><table border="0" class="funcprototype-table" summary="Function synopsis" style="cellspacing: 0; cellpadding: 0;"><tr><td><code class="funcdef">typedef bool <b class="fsfunc">(extent_merge_t)</b>(</code></td><td>extent_hooks_t *<var class="pdparam">extent_hooks</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr_a</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size_a</var>, </td></tr><tr><td> </td><td>void *<var class="pdparam">addr_b</var>, </td></tr><tr><td> </td><td>size_t <var class="pdparam">size_b</var>, </td></tr><tr><td> </td><td>bool <var class="pdparam">committed</var>, </td></tr><tr><td> </td><td>unsigned <var class="pdparam">arena_ind</var><code>)</code>;</td></tr></table><div class="funcprototype-spacer"> </div></div><div class="literallayout"><p/></div><p>An extent merge function conforms to the
<span class="type">extent_merge_t</span> type and optionally merges adjacent extents,
at given <em class="parameter"><code>addr_a</code></em> and <em class="parameter"><code>size_a</code></em>
with given <em class="parameter"><code>addr_b</code></em> and
<em class="parameter"><code>size_b</code></em> into one contiguous extent, operating on
<em class="parameter"><code>committed</code></em>/decommitted memory as indicated, on
behalf of arena <em class="parameter"><code>arena_ind</code></em>, returning false upon
success. If the function returns true, this indicates that the extents
remain distinct mappings and therefore should continue to be operated on
independently.</p></dd><dt><a name="arenas.narenas"/><span class="term">
<quote><code class="mallctl">arenas.narenas</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Current limit on number of arenas.</p></dd><dt><a name="arenas.dirty_decay_ms"/><span class="term">
<quote><code class="mallctl">arenas.dirty_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Current default per-arena approximate time in
milliseconds from the creation of a set of unused dirty pages until an
equivalent set of unused dirty pages is purged and/or reused, used to
initialize <a class="link" href="#arena.i.dirty_decay_ms"><quote><code class="mallctl">arena.<i>.dirty_decay_ms</code></quote></a>
during arena creation. See <a class="link" href="#opt.dirty_decay_ms"><quote><code class="mallctl">opt.dirty_decay_ms</code></quote></a>
for additional information.</p></dd><dt><a name="arenas.muzzy_decay_ms"/><span class="term">
<quote><code class="mallctl">arenas.muzzy_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Current default per-arena approximate time in
milliseconds from the creation of a set of unused muzzy pages until an
equivalent set of unused muzzy pages is purged and/or reused, used to
initialize <a class="link" href="#arena.i.muzzy_decay_ms"><quote><code class="mallctl">arena.<i>.muzzy_decay_ms</code></quote></a>
during arena creation. See <a class="link" href="#opt.muzzy_decay_ms"><quote><code class="mallctl">opt.muzzy_decay_ms</code></quote></a>
for additional information.</p></dd><dt><a name="arenas.quantum"/><span class="term">
<quote><code class="mallctl">arenas.quantum</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Quantum size.</p></dd><dt><a name="arenas.page"/><span class="term">
<quote><code class="mallctl">arenas.page</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Page size.</p></dd><dt><a name="arenas.tcache_max"/><span class="term">
<quote><code class="mallctl">arenas.tcache_max</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Maximum thread-cached size class.</p></dd><dt><a name="arenas.nbins"/><span class="term">
<quote><code class="mallctl">arenas.nbins</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of bin size classes.</p></dd><dt><a name="arenas.nhbins"/><span class="term">
<quote><code class="mallctl">arenas.nhbins</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Total number of thread cache bin size
classes.</p></dd><dt><a name="arenas.bin.i.size"/><span class="term">
<quote><code class="mallctl">arenas.bin.<i>.size</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Maximum size supported by size class.</p></dd><dt><a name="arenas.bin.i.nregs"/><span class="term">
<quote><code class="mallctl">arenas.bin.<i>.nregs</code></quote>
(<span class="type">uint32_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of regions per slab.</p></dd><dt><a name="arenas.bin.i.slab_size"/><span class="term">
<quote><code class="mallctl">arenas.bin.<i>.slab_size</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of bytes per slab.</p></dd><dt><a name="arenas.nlextents"/><span class="term">
<quote><code class="mallctl">arenas.nlextents</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Total number of large size classes.</p></dd><dt><a name="arenas.lextent.i.size"/><span class="term">
<quote><code class="mallctl">arenas.lextent.<i>.size</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Maximum size supported by this large size
class.</p></dd><dt><a name="arenas.create"/><span class="term">
<quote><code class="mallctl">arenas.create</code></quote>
(<span class="type">unsigned</span>, <span class="type">extent_hooks_t *</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Explicitly create a new arena outside the range of
automatically managed arenas, with optionally specified extent hooks,
and return the new arena index.</p></dd><dt><a name="arenas.lookup"/><span class="term">
<quote><code class="mallctl">arenas.lookup</code></quote>
(<span class="type">unsigned</span>, <span class="type">void*</span>)
<code class="literal">rw</code>
</span></dt><dd><p>Index of the arena to which an allocation belongs to.</p></dd><dt><a name="prof.thread_active_init"/><span class="term">
<quote><code class="mallctl">prof.thread_active_init</code></quote>
(<span class="type">bool</span>)
<code class="literal">rw</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Control the initial setting for <a class="link" href="#thread.prof.active"><quote><code class="mallctl">thread.prof.active</code></quote></a>
in newly created threads. See the <a class="link" href="#opt.prof_thread_active_init"><quote><code class="mallctl">opt.prof_thread_active_init</code></quote></a>
option for additional information.</p></dd><dt><a name="prof.active"/><span class="term">
<quote><code class="mallctl">prof.active</code></quote>
(<span class="type">bool</span>)
<code class="literal">rw</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Control whether sampling is currently active. See the
<a class="link" href="#opt.prof_active"><quote><code class="mallctl">opt.prof_active</code></quote></a>
option for additional information, as well as the interrelated <a class="link" href="#thread.prof.active"><quote><code class="mallctl">thread.prof.active</code></quote></a>
mallctl.</p></dd><dt><a name="prof.dump"/><span class="term">
<quote><code class="mallctl">prof.dump</code></quote>
(<span class="type">const char *</span>)
<code class="literal">-w</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Dump a memory profile to the specified file, or if NULL
is specified, to a file according to the pattern
<code class="filename"><prefix>.<pid>.<seq>.m<mseq>.heap</code>,
where <code class="literal"><prefix></code> is controlled by the
<a class="link" href="#opt.prof_prefix"><quote><code class="mallctl">opt.prof_prefix</code></quote></a>
option.</p></dd><dt><a name="prof.gdump"/><span class="term">
<quote><code class="mallctl">prof.gdump</code></quote>
(<span class="type">bool</span>)
<code class="literal">rw</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>When enabled, trigger a memory profile dump every time
the total virtual memory exceeds the previous maximum. Profiles are
dumped to files named according to the pattern
<code class="filename"><prefix>.<pid>.<seq>.u<useq>.heap</code>,
where <code class="literal"><prefix></code> is controlled by the <a class="link" href="#opt.prof_prefix"><quote><code class="mallctl">opt.prof_prefix</code></quote></a>
option.</p></dd><dt><a name="prof.reset"/><span class="term">
<quote><code class="mallctl">prof.reset</code></quote>
(<span class="type">size_t</span>)
<code class="literal">-w</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Reset all memory profile statistics, and optionally
update the sample rate (see <a class="link" href="#opt.lg_prof_sample"><quote><code class="mallctl">opt.lg_prof_sample</code></quote></a>
and <a class="link" href="#prof.lg_sample"><quote><code class="mallctl">prof.lg_sample</code></quote></a>).
</p></dd><dt><a name="prof.lg_sample"/><span class="term">
<quote><code class="mallctl">prof.lg_sample</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Get the current sample rate (see <a class="link" href="#opt.lg_prof_sample"><quote><code class="mallctl">opt.lg_prof_sample</code></quote></a>).
</p></dd><dt><a name="prof.interval"/><span class="term">
<quote><code class="mallctl">prof.interval</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-prof</code>]
</span></dt><dd><p>Average number of bytes allocated between
interval-based profile dumps. See the
<a class="link" href="#opt.lg_prof_interval"><quote><code class="mallctl">opt.lg_prof_interval</code></quote></a>
option for additional information.</p></dd><dt><a name="stats.allocated"/><span class="term">
<quote><code class="mallctl">stats.allocated</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Total number of bytes allocated by the
application.</p></dd><dt><a name="stats.active"/><span class="term">
<quote><code class="mallctl">stats.active</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Total number of bytes in active pages allocated by the
application. This is a multiple of the page size, and greater than or
equal to <a class="link" href="#stats.allocated"><quote><code class="mallctl">stats.allocated</code></quote></a>.
This does not include <a class="link" href="#stats.arenas.i.pdirty">
<quote><code class="mallctl">stats.arenas.<i>.pdirty</code></quote></a>,
<a class="link" href="#stats.arenas.i.pmuzzy">
<quote><code class="mallctl">stats.arenas.<i>.pmuzzy</code></quote></a>, nor pages
entirely devoted to allocator metadata.</p></dd><dt><a name="stats.metadata"/><span class="term">
<quote><code class="mallctl">stats.metadata</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Total number of bytes dedicated to metadata, which
comprise base allocations used for bootstrap-sensitive allocator
metadata structures (see <a class="link" href="#stats.arenas.i.base"><quote><code class="mallctl">stats.arenas.<i>.base</code></quote></a>)
and internal allocations (see <a class="link" href="#stats.arenas.i.internal"><quote><code class="mallctl">stats.arenas.<i>.internal</code></quote></a>).
Transparent huge page (enabled with <a class="link" href="#opt.metadata_thp">opt.metadata_thp</a>) usage is not
considered.</p></dd><dt><a name="stats.metadata_thp"/><span class="term">
<quote><code class="mallctl">stats.metadata_thp</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of transparent huge pages (THP) used for
metadata. See <a class="link" href="#stats.metadata"><quote><code class="mallctl">stats.metadata</code></quote></a> and
<a class="link" href="#opt.metadata_thp">opt.metadata_thp</a>) for
details.</p></dd><dt><a name="stats.resident"/><span class="term">
<quote><code class="mallctl">stats.resident</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Maximum number of bytes in physically resident data
pages mapped by the allocator, comprising all pages dedicated to
allocator metadata, pages backing active allocations, and unused dirty
pages. This is a maximum rather than precise because pages may not
actually be physically resident if they correspond to demand-zeroed
virtual memory that has not yet been touched. This is a multiple of the
page size, and is larger than <a class="link" href="#stats.active"><quote><code class="mallctl">stats.active</code></quote></a>.</p></dd><dt><a name="stats.mapped"/><span class="term">
<quote><code class="mallctl">stats.mapped</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Total number of bytes in active extents mapped by the
allocator. This is larger than <a class="link" href="#stats.active"><quote><code class="mallctl">stats.active</code></quote></a>. This
does not include inactive extents, even those that contain unused dirty
pages, which means that there is no strict ordering between this and
<a class="link" href="#stats.resident"><quote><code class="mallctl">stats.resident</code></quote></a>.</p></dd><dt><a name="stats.retained"/><span class="term">
<quote><code class="mallctl">stats.retained</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Total number of bytes in virtual memory mappings that
were retained rather than being returned to the operating system via
e.g. <span class="citerefentry"><span class="refentrytitle">munmap</span>(2)</span> or similar. Retained virtual
memory is typically untouched, decommitted, or purged, so it has no
strongly associated physical memory (see <a class="link" href="#arena.i.extent_hooks">extent hooks</a> for details).
Retained memory is excluded from mapped memory statistics, e.g. <a class="link" href="#stats.mapped"><quote><code class="mallctl">stats.mapped</code></quote></a>.
</p></dd><dt><a name="stats.background_thread.num_threads"/><span class="term">
<quote><code class="mallctl">stats.background_thread.num_threads</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p> Number of <a class="link" href="#background_thread">background
threads</a> running currently.</p></dd><dt><a name="stats.background_thread.num_runs"/><span class="term">
<quote><code class="mallctl">stats.background_thread.num_runs</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p> Total number of runs from all <a class="link" href="#background_thread">background threads</a>.</p></dd><dt><a name="stats.background_thread.run_interval"/><span class="term">
<quote><code class="mallctl">stats.background_thread.run_interval</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p> Average run interval in nanoseconds of <a class="link" href="#background_thread">background threads</a>.</p></dd><dt><a name="stats.mutexes.ctl"/><span class="term">
<quote><code class="mallctl">stats.mutexes.ctl.{counter};</code></quote>
(<span class="type">counter specific type</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">ctl</code> mutex (global
scope; mallctl related). <quote><code class="mallctl">{counter}</code></quote> is one of the
counters below:</p><dt><a name="mutex_counters"/></dt><dd><p><code class="varname">num_ops</code> (<span class="type">uint64_t</span>):
Total number of lock acquisition operations on this mutex.</p><p><code class="varname">num_spin_acq</code> (<span class="type">uint64_t</span>): Number
of times the mutex was spin-acquired. When the mutex is currently
locked and cannot be acquired immediately, a short period of
spin-retry within jemalloc will be performed. Acquired through spin
generally means the contention was lightweight and not causing context
switches.</p><p><code class="varname">num_wait</code> (<span class="type">uint64_t</span>): Number of
times the mutex was wait-acquired, which means the mutex contention
was not solved by spin-retry, and blocking operation was likely
involved in order to acquire the mutex. This event generally implies
higher cost / longer delay, and should be investigated if it happens
often.</p><p><code class="varname">max_wait_time</code> (<span class="type">uint64_t</span>):
Maximum length of time in nanoseconds spent on a single wait-acquired
lock operation. Note that to avoid profiling overhead on the common
path, this does not consider spin-acquired cases.</p><p><code class="varname">total_wait_time</code> (<span class="type">uint64_t</span>):
Cumulative time in nanoseconds spent on wait-acquired lock operations.
Similarly, spin-acquired cases are not considered.</p><p><code class="varname">max_num_thds</code> (<span class="type">uint32_t</span>): Maximum
number of threads waiting on this mutex simultaneously. Similarly,
spin-acquired cases are not considered.</p><p><code class="varname">num_owner_switch</code> (<span class="type">uint64_t</span>):
Number of times the current mutex owner is different from the previous
one. This event does not generally imply an issue; rather it is an
indicator of how often the protected data are accessed by different
threads.
</p></dd></dd><dt><a name="stats.mutexes.background_thread"/><span class="term">
<quote><code class="mallctl">stats.mutexes.background_thread.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">background_thread</code> mutex
(global scope; <a class="link" href="#background_thread"><quote><code class="mallctl">background_thread</code></quote></a>
related). <quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.mutexes.prof"/><span class="term">
<quote><code class="mallctl">stats.mutexes.prof.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">prof</code> mutex (global
scope; profiling related). <quote><code class="mallctl">{counter}</code></quote> is one of the
counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.mutexes.reset"/><span class="term">
<quote><code class="mallctl">stats.mutexes.reset</code></quote>
(<span class="type">void</span>) <code class="literal">--</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Reset all mutex profile statistics, including global
mutexes, arena mutexes and bin mutexes.</p></dd><dt><a name="stats.arenas.i.dss"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.dss</code></quote>
(<span class="type">const char *</span>)
<code class="literal">r-</code>
</span></dt><dd><p>dss (<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span>) allocation precedence as
related to <span class="citerefentry"><span class="refentrytitle">mmap</span>(2)</span> allocation. See <a class="link" href="#opt.dss"><quote><code class="mallctl">opt.dss</code></quote></a> for details.
</p></dd><dt><a name="stats.arenas.i.dirty_decay_ms"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.dirty_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Approximate time in milliseconds from the creation of a
set of unused dirty pages until an equivalent set of unused dirty pages
is purged and/or reused. See <a class="link" href="#opt.dirty_decay_ms"><quote><code class="mallctl">opt.dirty_decay_ms</code></quote></a>
for details.</p></dd><dt><a name="stats.arenas.i.muzzy_decay_ms"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.muzzy_decay_ms</code></quote>
(<span class="type">ssize_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Approximate time in milliseconds from the creation of a
set of unused muzzy pages until an equivalent set of unused muzzy pages
is purged and/or reused. See <a class="link" href="#opt.muzzy_decay_ms"><quote><code class="mallctl">opt.muzzy_decay_ms</code></quote></a>
for details.</p></dd><dt><a name="stats.arenas.i.nthreads"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.nthreads</code></quote>
(<span class="type">unsigned</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of threads currently assigned to
arena.</p></dd><dt><a name="stats.arenas.i.uptime"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.uptime</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Time elapsed (in nanoseconds) since the arena was
created. If <i> equals <code class="constant">0</code> or
<code class="constant">MALLCTL_ARENAS_ALL</code>, this is the uptime since malloc
initialization.</p></dd><dt><a name="stats.arenas.i.pactive"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.pactive</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of pages in active extents.</p></dd><dt><a name="stats.arenas.i.pdirty"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.pdirty</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of pages within unused extents that are
potentially dirty, and for which <code class="function">madvise()</code> or
similar has not been called. See <a class="link" href="#opt.dirty_decay_ms"><quote><code class="mallctl">opt.dirty_decay_ms</code></quote></a>
for a description of dirty pages.</p></dd><dt><a name="stats.arenas.i.pmuzzy"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.pmuzzy</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Number of pages within unused extents that are muzzy.
See <a class="link" href="#opt.muzzy_decay_ms"><quote><code class="mallctl">opt.muzzy_decay_ms</code></quote></a>
for a description of muzzy pages.</p></dd><dt><a name="stats.arenas.i.mapped"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mapped</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of mapped bytes.</p></dd><dt><a name="stats.arenas.i.retained"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.retained</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of retained bytes. See <a class="link" href="#stats.retained"><quote><code class="mallctl">stats.retained</code></quote></a> for
details.</p></dd><dt><a name="stats.arenas.i.base"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.base</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>
Number of bytes dedicated to bootstrap-sensitive allocator metadata
structures.</p></dd><dt><a name="stats.arenas.i.internal"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.internal</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of bytes dedicated to internal allocations.
Internal allocations differ from application-originated allocations in
that they are for internal use, and that they are omitted from heap
profiles.</p></dd><dt><a name="stats.arenas.i.metadata_thp"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.metadata_thp</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of transparent huge pages (THP) used for
metadata. See <a class="link" href="#opt.metadata_thp">opt.metadata_thp</a>
for details.</p></dd><dt><a name="stats.arenas.i.resident"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.resident</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Maximum number of bytes in physically resident data
pages mapped by the arena, comprising all pages dedicated to allocator
metadata, pages backing active allocations, and unused dirty pages.
This is a maximum rather than precise because pages may not actually be
physically resident if they correspond to demand-zeroed virtual memory
that has not yet been touched. This is a multiple of the page
size.</p></dd><dt><a name="stats.arenas.i.dirty_npurge"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.dirty_npurge</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of dirty page purge sweeps performed.
</p></dd><dt><a name="stats.arenas.i.dirty_nmadvise"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.dirty_nmadvise</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of <code class="function">madvise()</code> or similar
calls made to purge dirty pages.</p></dd><dt><a name="stats.arenas.i.dirty_purged"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.dirty_purged</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of dirty pages purged.</p></dd><dt><a name="stats.arenas.i.muzzy_npurge"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.muzzy_npurge</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of muzzy page purge sweeps performed.
</p></dd><dt><a name="stats.arenas.i.muzzy_nmadvise"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.muzzy_nmadvise</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of <code class="function">madvise()</code> or similar
calls made to purge muzzy pages.</p></dd><dt><a name="stats.arenas.i.muzzy_purged"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.muzzy_purged</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of muzzy pages purged.</p></dd><dt><a name="stats.arenas.i.small.allocated"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.small.allocated</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of bytes currently allocated by small objects.
</p></dd><dt><a name="stats.arenas.i.small.nmalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.small.nmalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a small allocation was
requested from the arena's bins, whether to fill the relevant tcache if
<a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is
enabled, or to directly satisfy an allocation request
otherwise.</p></dd><dt><a name="stats.arenas.i.small.ndalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.small.ndalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a small allocation was
returned to the arena's bins, whether to flush the relevant tcache if
<a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is
enabled, or to directly deallocate an allocation
otherwise.</p></dd><dt><a name="stats.arenas.i.small.nrequests"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.small.nrequests</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of allocation requests satisfied by
all bin size classes.</p></dd><dt><a name="stats.arenas.i.large.allocated"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.large.allocated</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Number of bytes currently allocated by large objects.
</p></dd><dt><a name="stats.arenas.i.large.nmalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.large.nmalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a large extent was allocated
from the arena, whether to fill the relevant tcache if <a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is enabled and
the size class is within the range being cached, or to directly satisfy
an allocation request otherwise.</p></dd><dt><a name="stats.arenas.i.large.ndalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.large.ndalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a large extent was returned
to the arena, whether to flush the relevant tcache if <a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is enabled and
the size class is within the range being cached, or to directly
deallocate an allocation otherwise.</p></dd><dt><a name="stats.arenas.i.large.nrequests"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.large.nrequests</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of allocation requests satisfied by
all large size classes.</p></dd><dt><a name="stats.arenas.i.bins.j.nmalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.nmalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a bin region of the
corresponding size class was allocated from the arena, whether to fill
the relevant tcache if <a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is enabled, or
to directly satisfy an allocation request otherwise.</p></dd><dt><a name="stats.arenas.i.bins.j.ndalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.ndalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a bin region of the
corresponding size class was returned to the arena, whether to flush the
relevant tcache if <a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is enabled, or
to directly deallocate an allocation otherwise.</p></dd><dt><a name="stats.arenas.i.bins.j.nrequests"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.nrequests</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of allocation requests satisfied by
bin regions of the corresponding size class.</p></dd><dt><a name="stats.arenas.i.bins.j.curregs"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.curregs</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Current number of regions for this size
class.</p></dd><dt><a name="stats.arenas.i.bins.j.nfills"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.nfills</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Cumulative number of tcache fills.</p></dd><dt><a name="stats.arenas.i.bins.j.nflushes"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.nflushes</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
</span></dt><dd><p>Cumulative number of tcache flushes.</p></dd><dt><a name="stats.arenas.i.bins.j.nslabs"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.nslabs</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of slabs created.</p></dd><dt><a name="stats.arenas.i.bins.j.nreslabs"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.nreslabs</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times the current slab from which
to allocate changed.</p></dd><dt><a name="stats.arenas.i.bins.j.curslabs"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.curslabs</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Current number of slabs.</p></dd><dt><a name="stats.arenas.i.bins.mutex"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.bins.<j>.mutex.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on
<code class="varname">arena.<i>.bins.<j></code> mutex (arena bin
scope; bin operation related). <quote><code class="mallctl">{counter}</code></quote> is one of
the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.lextents.j.nmalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.lextents.<j>.nmalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a large extent of the
corresponding size class was allocated from the arena, whether to fill
the relevant tcache if <a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is enabled and
the size class is within the range being cached, or to directly satisfy
an allocation request otherwise.</p></dd><dt><a name="stats.arenas.i.lextents.j.ndalloc"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.lextents.<j>.ndalloc</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of times a large extent of the
corresponding size class was returned to the arena, whether to flush the
relevant tcache if <a class="link" href="#opt.tcache"><quote><code class="mallctl">opt.tcache</code></quote></a> is enabled and
the size class is within the range being cached, or to directly
deallocate an allocation otherwise.</p></dd><dt><a name="stats.arenas.i.lextents.j.nrequests"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.lextents.<j>.nrequests</code></quote>
(<span class="type">uint64_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Cumulative number of allocation requests satisfied by
large extents of the corresponding size class.</p></dd><dt><a name="stats.arenas.i.lextents.j.curlextents"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.lextents.<j>.curlextents</code></quote>
(<span class="type">size_t</span>)
<code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Current number of large allocations for this size class.
</p></dd><dt><a name="stats.arenas.i.mutexes.large"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.large.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.large</code>
mutex (arena scope; large allocation related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.extent_avail"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.extent_avail.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.extent_avail
</code> mutex (arena scope; extent avail related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.extents_dirty"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.extents_dirty.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.extents_dirty
</code> mutex (arena scope; dirty extents related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.extents_muzzy"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.extents_muzzy.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.extents_muzzy
</code> mutex (arena scope; muzzy extents related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.extents_retained"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.extents_retained.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.extents_retained
</code> mutex (arena scope; retained extents related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.decay_dirty"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.decay_dirty.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.decay_dirty
</code> mutex (arena scope; decay for dirty pages related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.decay_muzzy"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.decay_muzzy.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.decay_muzzy
</code> mutex (arena scope; decay for muzzy pages related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.base"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.base.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on <code class="varname">arena.<i>.base</code>
mutex (arena scope; base allocator related).
<quote><code class="mallctl">{counter}</code></quote> is one of the counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd><dt><a name="stats.arenas.i.mutexes.tcache_list"/><span class="term">
<quote><code class="mallctl">stats.arenas.<i>.mutexes.tcache_list.{counter}</code></quote>
(<span class="type">counter specific type</span>) <code class="literal">r-</code>
[<code class="option">--enable-stats</code>]
</span></dt><dd><p>Statistics on
<code class="varname">arena.<i>.tcache_list</code> mutex (arena scope;
tcache to arena association related). This mutex is expected to be
accessed less often. <quote><code class="mallctl">{counter}</code></quote> is one of the
counters in <a class="link" href="#mutex_counters">mutex profiling
counters</a>.</p></dd></dl></div></div><div class="refsect1"><a name="heap_profile_format"/><h2>HEAP PROFILE FORMAT</h2><p>Although the heap profiling functionality was originally designed to
be compatible with the
<span class="command"><strong>pprof</strong></span> command that is developed as part of the <a class="ulink" href="http://code.google.com/p/gperftools/" target="_top">gperftools
package</a>, the addition of per thread heap profiling functionality
required a different heap profile format. The <span class="command"><strong>jeprof</strong></span>
command is derived from <span class="command"><strong>pprof</strong></span>, with enhancements to
support the heap profile format described here.</p><p>In the following hypothetical heap profile, <code class="constant">[...]</code>
indicates elision for the sake of compactness. </p><pre class="programlisting">
heap_v2/524288
t*: 28106: 56637512 [0: 0]
[...]
t3: 352: 16777344 [0: 0]
[...]
t99: 17754: 29341640 [0: 0]
[...]
@ 0x5f86da8 0x5f5a1dc [...] 0x29e4d4e 0xa200316 0xabb2988 [...]
t*: 13: 6688 [0: 0]
t3: 12: 6496 [0: ]
t99: 1: 192 [0: 0]
[...]
MAPPED_LIBRARIES:
[...]</pre><p> The following matches the above heap profile, but most
tokens are replaced with <code class="constant"><description></code> to indicate
descriptions of the corresponding fields. </p><pre class="programlisting">
<heap_profile_format_version>/<mean_sample_interval>
<aggregate>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
[...]
<thread_3_aggregate>: <curobjs>: <curbytes>[<cumobjs>: <cumbytes>]
[...]
<thread_99_aggregate>: <curobjs>: <curbytes>[<cumobjs>: <cumbytes>]
[...]
@ <top_frame> <frame> [...] <frame> <frame> <frame> [...]
<backtrace_aggregate>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
<backtrace_thread_3>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
<backtrace_thread_99>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
[...]
MAPPED_LIBRARIES:
</proc/<pid>/maps></pre></div><div class="refsect1"><a name="debugging_malloc_problems"/><h2>DEBUGGING MALLOC PROBLEMS</h2><p>When debugging, it is a good idea to configure/build jemalloc with
the <code class="option">--enable-debug</code> and <code class="option">--enable-fill</code>
options, and recompile the program with suitable options and symbols for
debugger support. When so configured, jemalloc incorporates a wide variety
of run-time assertions that catch application errors such as double-free,
write-after-free, etc.</p><p>Programs often accidentally depend on <span class="quote">“<span class="quote">uninitialized</span>”</span>
memory actually being filled with zero bytes. Junk filling
(see the <a class="link" href="#opt.junk"><quote><code class="mallctl">opt.junk</code></quote></a>
option) tends to expose such bugs in the form of obviously incorrect
results and/or coredumps. Conversely, zero
filling (see the <a class="link" href="#opt.zero"><quote><code class="mallctl">opt.zero</code></quote></a> option) eliminates
the symptoms of such bugs. Between these two options, it is usually
possible to quickly detect, diagnose, and eliminate such bugs.</p><p>This implementation does not provide much detail about the problems
it detects, because the performance impact for storing such information
would be prohibitive.</p></div><div class="refsect1"><a name="diagnostic_messages"/><h2>DIAGNOSTIC MESSAGES</h2><p>If any of the memory allocation/deallocation functions detect an
error or warning condition, a message will be printed to file descriptor
<code class="constant">STDERR_FILENO</code>. Errors will result in the process
dumping core. If the <a class="link" href="#opt.abort"><quote><code class="mallctl">opt.abort</code></quote></a> option is set, most
warnings are treated as errors.</p><p>The <code class="varname">malloc_message</code> variable allows the programmer
to override the function which emits the text strings forming the errors
and warnings if for some reason the <code class="constant">STDERR_FILENO</code> file
descriptor is not suitable for this.
<code class="function">malloc_message()</code> takes the
<em class="parameter"><code>cbopaque</code></em> pointer argument that is
<code class="constant">NULL</code> unless overridden by the arguments in a call to
<code class="function">malloc_stats_print()</code>, followed by a string
pointer. Please note that doing anything which tries to allocate memory in
this function is likely to result in a crash or deadlock.</p><p>All messages are prefixed by
<span class="quote">“<span class="quote"><code class="computeroutput"><jemalloc>: </code></span>”</span>.</p></div><div class="refsect1"><a name="return_values"/><h2>RETURN VALUES</h2><div class="refsect2"><a name="idm140689764475728"/><h3>Standard API</h3><p>The <code class="function">malloc()</code> and
<code class="function">calloc()</code> functions return a pointer to the
allocated memory if successful; otherwise a <code class="constant">NULL</code>
pointer is returned and <code class="varname">errno</code> is set to
<span class="errorname">ENOMEM</span>.</p><p>The <code class="function">posix_memalign()</code> function
returns the value 0 if successful; otherwise it returns an error value.
The <code class="function">posix_memalign()</code> function will fail
if:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><span class="errorname">EINVAL</span></span></dt><dd><p>The <em class="parameter"><code>alignment</code></em> parameter is
not a power of 2 at least as large as
<code class="code">sizeof(<span class="type">void *</span>)</code>.
</p></dd><dt><span class="term"><span class="errorname">ENOMEM</span></span></dt><dd><p>Memory allocation error.</p></dd></dl></div><p>
</p><p>The <code class="function">aligned_alloc()</code> function returns
a pointer to the allocated memory if successful; otherwise a
<code class="constant">NULL</code> pointer is returned and
<code class="varname">errno</code> is set. The
<code class="function">aligned_alloc()</code> function will fail if:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><span class="errorname">EINVAL</span></span></dt><dd><p>The <em class="parameter"><code>alignment</code></em> parameter is
not a power of 2.
</p></dd><dt><span class="term"><span class="errorname">ENOMEM</span></span></dt><dd><p>Memory allocation error.</p></dd></dl></div><p>
</p><p>The <code class="function">realloc()</code> function returns a
pointer, possibly identical to <em class="parameter"><code>ptr</code></em>, to the
allocated memory if successful; otherwise a <code class="constant">NULL</code>
pointer is returned, and <code class="varname">errno</code> is set to
<span class="errorname">ENOMEM</span> if the error was the result of an
allocation failure. The <code class="function">realloc()</code>
function always leaves the original buffer intact when an error occurs.
</p><p>The <code class="function">free()</code> function returns no
value.</p></div><div class="refsect2"><a name="idm140689764454240"/><h3>Non-standard API</h3><p>The <code class="function">mallocx()</code> and
<code class="function">rallocx()</code> functions return a pointer to
the allocated memory if successful; otherwise a <code class="constant">NULL</code>
pointer is returned to indicate insufficient contiguous memory was
available to service the allocation request. </p><p>The <code class="function">xallocx()</code> function returns the
real size of the resulting resized allocation pointed to by
<em class="parameter"><code>ptr</code></em>, which is a value less than
<em class="parameter"><code>size</code></em> if the allocation could not be adequately
grown in place. </p><p>The <code class="function">sallocx()</code> function returns the
real size of the allocation pointed to by <em class="parameter"><code>ptr</code></em>.
</p><p>The <code class="function">nallocx()</code> returns the real size
that would result from a successful equivalent
<code class="function">mallocx()</code> function call, or zero if
insufficient memory is available to perform the size computation. </p><p>The <code class="function">mallctl()</code>,
<code class="function">mallctlnametomib()</code>, and
<code class="function">mallctlbymib()</code> functions return 0 on
success; otherwise they return an error value. The functions will fail
if:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><span class="errorname">EINVAL</span></span></dt><dd><p><em class="parameter"><code>newp</code></em> is not
<code class="constant">NULL</code>, and <em class="parameter"><code>newlen</code></em> is too
large or too small. Alternatively, <em class="parameter"><code>*oldlenp</code></em>
is too large or too small; in this case as much data as possible
are read despite the error.</p></dd><dt><span class="term"><span class="errorname">ENOENT</span></span></dt><dd><p><em class="parameter"><code>name</code></em> or
<em class="parameter"><code>mib</code></em> specifies an unknown/invalid
value.</p></dd><dt><span class="term"><span class="errorname">EPERM</span></span></dt><dd><p>Attempt to read or write void value, or attempt to
write read-only value.</p></dd><dt><span class="term"><span class="errorname">EAGAIN</span></span></dt><dd><p>A memory allocation failure
occurred.</p></dd><dt><span class="term"><span class="errorname">EFAULT</span></span></dt><dd><p>An interface with side effects failed in some way
not directly related to <code class="function">mallctl*()</code>
read/write processing.</p></dd></dl></div><p>
</p><p>The <code class="function">malloc_usable_size()</code> function
returns the usable size of the allocation pointed to by
<em class="parameter"><code>ptr</code></em>. </p></div></div><div class="refsect1"><a name="environment"/><h2>ENVIRONMENT</h2><p>The following environment variable affects the execution of the
allocation functions:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><code class="envar">MALLOC_CONF</code></span></dt><dd><p>If the environment variable
<code class="envar">MALLOC_CONF</code> is set, the characters it contains
will be interpreted as options.</p></dd></dl></div><p>
</p></div><div class="refsect1"><a name="examples"/><h2>EXAMPLES</h2><p>To dump core whenever a problem occurs:
</p><pre class="screen">ln -s 'abort:true' /etc/malloc.conf</pre><p>
</p><p>To specify in the source that only one arena should be automatically
created:
</p><pre class="programlisting">
malloc_conf = "narenas:1";</pre></div><div class="refsect1"><a name="see_also"/><h2>SEE ALSO</h2><p><span class="citerefentry"><span class="refentrytitle">madvise</span>(2)</span>,
<span class="citerefentry"><span class="refentrytitle">mmap</span>(2)</span>,
<span class="citerefentry"><span class="refentrytitle">sbrk</span>(2)</span>,
<span class="citerefentry"><span class="refentrytitle">utrace</span>(2)</span>,
<span class="citerefentry"><span class="refentrytitle">alloca</span>(3)</span>,
<span class="citerefentry"><span class="refentrytitle">atexit</span>(3)</span>,
<span class="citerefentry"><span class="refentrytitle">getpagesize</span>(3)</span></p></div><div class="refsect1"><a name="standards"/><h2>STANDARDS</h2><p>The <code class="function">malloc()</code>,
<code class="function">calloc()</code>,
<code class="function">realloc()</code>, and
<code class="function">free()</code> functions conform to ISO/IEC
9899:1990 (<span class="quote">“<span class="quote">ISO C90</span>”</span>).</p><p>The <code class="function">posix_memalign()</code> function conforms
to IEEE Std 1003.1-2001 (<span class="quote">“<span class="quote">POSIX.1</span>”</span>).</p></div></div></body></html>