jemalloc — general purpose memory allocation functions
This manual describes jemalloc 5.3.0-0-g54eaed1d8b56b1aa528be3bdd1877e59c56fa90c. More information can be found at the jemalloc website.
#include <jemalloc/jemalloc.h>void *malloc( | size_t size); |
void *calloc( | size_t number, |
size_t size); |
int posix_memalign( | void **ptr, |
| size_t alignment, | |
size_t size); |
void *aligned_alloc( | size_t alignment, |
size_t size); |
void *realloc( | void *ptr, |
size_t size); |
void free( | void *ptr); |
void *mallocx( | size_t size, |
int flags); |
void *rallocx( | void *ptr, |
| size_t size, | |
int flags); |
size_t xallocx( | void *ptr, |
| size_t size, | |
| size_t extra, | |
int flags); |
size_t sallocx( | void *ptr, |
int flags); |
void dallocx( | void *ptr, |
int flags); |
void sdallocx( | void *ptr, |
| size_t size, | |
int flags); |
size_t nallocx( | size_t size, |
int flags); |
int mallctl( | const char *name, |
| void *oldp, | |
| size_t *oldlenp, | |
| void *newp, | |
size_t newlen); |
int mallctlnametomib( | const char *name, |
| size_t *mibp, | |
size_t *miblenp); |
int mallctlbymib( | const size_t *mib, |
| size_t miblen, | |
| void *oldp, | |
| size_t *oldlenp, | |
| void *newp, | |
size_t newlen); |
void malloc_stats_print( | void (*write_cb)
(void *, const char *)
, |
| void *cbopaque, | |
const char *opts); |
size_t malloc_usable_size( | const void *ptr); |
void (*malloc_message)( | void *cbopaque, |
const char *s); |
const char *malloc_conf;
The malloc() function allocates
size bytes of uninitialized memory. The allocated
space is suitably aligned (after possible pointer coercion) for storage
of any type of object.
The calloc() function allocates
space for number objects, each
size bytes in length. The result is identical to
calling malloc() with an argument of
number * size, with the
exception that the allocated memory is explicitly initialized to zero
bytes.
The posix_memalign() function
allocates size bytes of memory such that the
allocation's base address is a multiple of
alignment, and returns the allocation in the value
pointed to by ptr. The requested
alignment must be a power of 2 at least as large as
sizeof(void *).
The aligned_alloc() function
allocates size bytes of memory such that the
allocation's base address is a multiple of
alignment. The requested
alignment must be a power of 2. Behavior is
undefined if size is not an integral multiple of
alignment.
The realloc() function changes the
size of the previously allocated memory referenced by
ptr to size 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 ptr is freed and a pointer to the newly
allocated memory is returned. Note that
realloc() may move the memory allocation,
resulting in a different return value than ptr.
If ptr is NULL, the
realloc() function behaves identically to
malloc() for the specified size.
The free() function causes the
allocated memory referenced by ptr to be made
available for future allocations. If ptr is
NULL, no action occurs.
The mallocx(),
rallocx(),
xallocx(),
sallocx(),
dallocx(),
sdallocx(), and
nallocx() functions all have a
flags argument that can be used to specify
options. The functions only check the options that are contextually
relevant. Use bitwise or (|) operations to
specify one or more of the following:
MALLOCX_LG_ALIGN(la)
Align the memory allocation to start at an address
that is a multiple of (1 <<
. This macro does not validate
that la)la is within the valid
range.
MALLOCX_ALIGN(a)
Align the memory allocation to start at an address
that is a multiple of a, where
a is a power of two. This macro does not
validate that a is a power of 2.
MALLOCX_ZEROInitialize 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.
MALLOCX_TCACHE(tc)
Use the thread-specific cache (tcache) specified by
the identifier tc, which must have been
acquired via the
mallctl. This macro does not validate that
tcache.createtc specifies a valid
identifier.
MALLOCX_TCACHE_NONEDo not use a thread-specific cache (tcache). Unless
MALLOCX_TCACHE( or
tc)MALLOCX_TCACHE_NONE is specified, an
automatically managed tcache will be used under many circumstances.
This macro cannot be used in the same flags
argument as
MALLOCX_TCACHE(.tc)
MALLOCX_ARENA(a)
Use the arena specified by the index
a. This macro has no effect for regions that
were allocated via an arena other than the one specified. This
macro does not validate that a specifies an
arena index in the valid range.
The mallocx() function allocates at
least size bytes of memory, and returns a pointer
to the base address of the allocation. Behavior is undefined if
size is 0.
The rallocx() function resizes the
allocation at ptr to be at least
size 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
size is 0.
The xallocx() function resizes the
allocation at ptr in place to be at least
size bytes, and returns the real size of the
allocation. If extra is non-zero, an attempt is
made to resize the allocation to be at least ( bytes, though inability to allocate
the extra byte(s) will not by itself result in failure to resize.
Behavior is undefined if size +
extra)size is
0, or if (.size + extra
> SIZE_T_MAX)
The sallocx() function returns the
real size of the allocation at ptr.
The dallocx() function causes the
memory referenced by ptr to be made available for
future allocations.
The sdallocx() function is an
extension of dallocx() with a
size 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
nallocx() or
sallocx().
The nallocx() function allocates no
memory, but it performs the same size computation as the
mallocx() function, and returns the real
size of the allocation that would result from the equivalent
mallocx() function call, or
0 if the inputs exceed the maximum supported size
class and/or alignment. Behavior is undefined if
size is 0.
The mallctl() function provides a
general interface for introspecting the memory allocator, as well as
setting modifiable parameters and triggering actions. The
period-separated name argument specifies a
location in a tree-structured namespace; see the MALLCTL NAMESPACE section for
documentation on the tree contents. To read a value, pass a pointer via
oldp to adequate space to contain the value, and a
pointer to its length via oldlenp; otherwise pass
NULL and NULL. Similarly, to
write a value, pass a pointer to the value via
newp, and its length via
newlen; otherwise pass NULL
and 0.
The mallctlnametomib() 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 “Management Information Base” (MIB) that can be passed
repeatedly to mallctlbymib(). Upon
successful return from mallctlnametomib(),
mibp contains an array of
*miblenp integers, where
*miblenp is the lesser of the number of components
in name and the input value of
*miblenp. Thus it is possible to pass a
*miblenp 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
),
the corresponding MIB component will always be that integer. Therefore,
it is legitimate to construct code like the following: arenas.bin.2.size
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... */
}The malloc_stats_print() function writes
summary statistics via the write_cb callback
function pointer and cbopaque data passed to
write_cb, or malloc_message()
if write_cb is NULL. The
statistics are presented in human-readable form unless “J” is
specified as a character within the opts string, in
which case the statistics are presented in JSON format. This function can be
called repeatedly. General information that never changes during
execution can be omitted by specifying “g” as a character
within the opts string. Note that
malloc_stats_print() uses the
mallctl*() functions internally, so inconsistent
statistics can be reported if multiple threads use these functions
simultaneously. If --enable-stats is specified during
configuration, “m”, “d”, and “a”
can be specified to omit merged arena, destroyed merged arena, and per
arena statistics, respectively; “b” and “l” can
be specified to omit per size class statistics for bins and large objects,
respectively; “x” can be specified to omit all mutex
statistics; “e” can be used to omit extent 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.
The malloc_usable_size() function
returns the usable size of the allocation pointed to by
ptr. The return value may be larger than the size
that was requested during allocation. The
malloc_usable_size() function is not a
mechanism for in-place realloc(); rather
it is provided solely as a tool for introspection purposes. Any
discrepancy between the requested allocation size and the size reported
by malloc_usable_size() should not be
depended on, since such behavior is entirely implementation-dependent.
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.
The string specified via --with-malloc-conf, the
string pointed to by the global variable malloc_conf, the
“name” of the file referenced by the symbolic link named
/etc/malloc.conf, and the value of the
environment variable MALLOC_CONF, will be interpreted, in
that order, from left to right as options. Note that
malloc_conf may be read before
main() is entered, so the declaration of
malloc_conf should specify an initializer that contains
the final value to be read by jemalloc. --with-malloc-conf
and malloc_conf are compile-time mechanisms, whereas
/etc/malloc.conf and
MALLOC_CONF can be safely set any time prior to program
invocation.
An options string is a comma-separated list of option:value pairs.
There is one key corresponding to each
mallctl (see the MALLCTL NAMESPACE section for options
documentation). For example, opt.*abort:true,narenas:1 sets
the
and opt.abort
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.opt.narenas
Traditionally, allocators have used sbrk(2) to obtain memory, which is suboptimal for several reasons, including race conditions, increased fragmentation, and artificial limitations on maximum usable memory. If sbrk(2) is supported by the operating system, this allocator uses both mmap(2) and sbrk(2), in that order of preference; otherwise only mmap(2) is used.
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.
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.
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.
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 sizeof(double). 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 PTRDIFF_MAX.
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.
The realloc(),
rallocx(), and
xallocx() functions may resize allocations
without moving them under limited circumstances. Unlike the
*allocx() 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
realloc() 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
).
Growth only succeeds if the trailing memory is currently available, and the
extent allocator supports merging.arena.<i>.extent_hooks
Assuming 4 KiB pages and a 16-byte quantum on a 64-bit system, the size classes in each category are as shown in Table 1.
Table 1. Size classes
| Category | Spacing | Size |
|---|---|---|
| Small | lg | [8] |
| 16 | [16, 32, 48, 64, 80, 96, 112, 128] | |
| 32 | [160, 192, 224, 256] | |
| 64 | [320, 384, 448, 512] | |
| 128 | [640, 768, 896, 1024] | |
| 256 | [1280, 1536, 1792, 2048] | |
| 512 | [2560, 3072, 3584, 4096] | |
| 1 KiB | [5 KiB, 6 KiB, 7 KiB, 8 KiB] | |
| 2 KiB | [10 KiB, 12 KiB, 14 KiB] | |
| Large | 2 KiB | [16 KiB] |
| 4 KiB | [20 KiB, 24 KiB, 28 KiB, 32 KiB] | |
| 8 KiB | [40 KiB, 48 KiB, 56 KiB, 64 KiB] | |
| 16 KiB | [80 KiB, 96 KiB, 112 KiB, 128 KiB] | |
| 32 KiB | [160 KiB, 192 KiB, 224 KiB, 256 KiB] | |
| 64 KiB | [320 KiB, 384 KiB, 448 KiB, 512 KiB] | |
| 128 KiB | [640 KiB, 768 KiB, 896 KiB, 1 MiB] | |
| 256 KiB | [1280 KiB, 1536 KiB, 1792 KiB, 2 MiB] | |
| 512 KiB | [2560 KiB, 3 MiB, 3584 KiB, 4 MiB] | |
| 1 MiB | [5 MiB, 6 MiB, 7 MiB, 8 MiB] | |
| 2 MiB | [10 MiB, 12 MiB, 14 MiB, 16 MiB] | |
| 4 MiB | [20 MiB, 24 MiB, 28 MiB, 32 MiB] | |
| 8 MiB | [40 MiB, 48 MiB, 56 MiB, 64 MiB] | |
| ... | ... | |
| 512 PiB | [2560 PiB, 3 EiB, 3584 PiB, 4 EiB] | |
| 1 EiB | [5 EiB, 6 EiB, 7 EiB] |
The following names are defined in the namespace accessible via the
mallctl*() functions. Value types are specified in
parentheses, their readable/writable statuses are encoded as
rw, r-, -w, or
--, and required build configuration flags follow, if
any. A name element encoded as <i> or
<j> indicates an integer component, where the
integer varies from 0 to some upper value that must be determined via
introspection. In the case of
and stats.arenas.<i>.*
,
arena.<i>.{initialized,purge,decay,dss}<i> equal to
MALLCTL_ARENAS_ALL can be used to operate on all arenas
or access the summation of statistics from all arenas; similarly
<i> equal to
MALLCTL_ARENAS_DESTROYED can be used to access the
summation of statistics from all destroyed arenas. These constants can be
utilized either via mallctlnametomib() followed by
mallctlbymib(), or via code such as the following:
#define STRINGIFY_HELPER(x) #x
#define STRINGIFY(x) STRINGIFY_HELPER(x)
mallctl("arena." STRINGIFY(MALLCTL_ARENAS_ALL) ".decay",
NULL, NULL, NULL, 0);
Take special note of the
mallctl, which controls
refreshing of cached dynamic statistics.epoch
versionr-
Return the jemalloc version string.
epochrw
If a value is passed in, refresh the data from which
the mallctl*() functions report values,
and increment the epoch. Return the current epoch. This is useful for
detecting whether another thread caused a refresh.
background_threadrw
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 purging asynchronously. When switching
off, background threads are terminated synchronously. Note that after
fork(2)
function, the state in the child process will be disabled regardless
the state in parent process. See
for related stats. stats.background_thread
can be used to set the default option. This option is only available on
selected pthread-based platforms.opt.background_thread
max_background_threadsrw
Maximum number of background worker threads that will
be created. This value is capped at
at
startup.opt.max_background_threads
config.cache_obliviousr-
--enable-cache-oblivious was specified
during build configuration.
config.debugr-
--enable-debug was specified during
build configuration.
config.fillr-
--enable-fill was specified during
build configuration.
config.lazy_lockr-
--enable-lazy-lock was specified
during build configuration.
config.malloc_confr-
Embedded configure-time-specified run-time options
string, empty unless --with-malloc-conf was specified
during build configuration.
config.profr-
--enable-prof was specified during
build configuration.
config.prof_libgccr-
--disable-prof-libgcc was not
specified during build configuration.
config.prof_libunwindr-
--enable-prof-libunwind was specified
during build configuration.
config.statsr-
--enable-stats was specified during
build configuration.
config.utracer-
--enable-utrace was specified during
build configuration.
config.xmallocr-
--enable-xmalloc was specified during
build configuration.
opt.abortr-
Abort-on-warning enabled/disabled. If true, most
warnings are fatal. Note that runtime option warnings are not included
(see
for
that). The process will call
abort(3) in these cases. This option is
disabled by default unless opt.abort_conf--enable-debug is
specified during configuration, in which case it is enabled by default.
opt.confirm_confr-
Confirm-runtime-options-when-program-starts
enabled/disabled. If true, the string specified via
--with-malloc-conf, the string pointed to by the
global variable malloc_conf, the “name”
of the file referenced by the symbolic link named
/etc/malloc.conf, and the value of
the environment variable MALLOC_CONF, will be printed in
order. Then, each option being set will be individually printed. This
option is disabled by default.
opt.abort_confr-
Abort-on-invalid-configuration enabled/disabled. If
true, invalid runtime options are fatal. The process will call
abort(3) in these cases. This option is
disabled by default unless --enable-debug is
specified during configuration, in which case it is enabled by default.
opt.cache_obliviousr-
Enable / Disable cache-oblivious large allocation alignment, for large requests with no alignment constraints. If this feature is disabled, all large allocations are page-aligned as an implementation artifact, which can severely harm CPU cache utilization. However, the cache-oblivious layout comes at the cost of one extra page per large allocation, which in the most extreme case increases physical memory usage for the 16 KiB size class to 20 KiB. This option is enabled by default.
opt.metadata_thpr-
Controls whether to allow jemalloc to use transparent huge page (THP) for internal metadata (see stats.metadata). “always” allows such usage. “auto” uses no THP initially, but may begin to do so when metadata usage reaches certain level. The default is “disabled”.
opt.trust_madviser-
If true, do not perform runtime check for MADV_DONTNEED, to check that it actually zeros pages. The default is disabled on Linux and enabled elsewhere.
opt.retainr-
If true, retain unused virtual memory for later reuse
rather than discarding it by calling
munmap(2) or equivalent (see stats.retained for related details).
It also makes jemalloc use mmap(2) or equivalent in a more greedy way, mapping larger
chunks in one go. This option is disabled by default unless discarding
virtual memory is known to trigger platform-specific performance
problems, namely 1) 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; and 2) for [64-bit] Windows, which
disallows split / merged regions with
MEM_RELEASE
opt.dssr-
dss (sbrk(2)) allocation precedence as related to mmap(2) allocation. The following settings are supported if sbrk(2) is supported by the operating system: “disabled”, “primary”, and “secondary”; otherwise only “disabled” is supported. The default is “secondary” if sbrk(2) is supported by the operating system; “disabled” otherwise.
opt.narenasr-
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.
opt.oversize_thresholdr-
The threshold in bytes of which requests are considered
oversize. Allocation requests with greater sizes are fulfilled from a
dedicated arena (automatically managed, however not within
narenas), in order to reduce fragmentation by not
mixing huge allocations with small ones. In addition, the decay API
guarantees on the extents greater than the specified threshold may be
overridden. Note that requests with arena index specified via
MALLOCX_ARENA, or threads associated with explicit
arenas will not be considered. The default threshold is 8MiB. Values
not within large size classes disables this feature.
opt.percpu_arenar-
Per CPU arena mode. Use the “percpu” 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. “phycpu” 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 “disabled”, narenas and thread to arena association will not be impacted by this option. The default is “disabled”.
opt.background_threadr-
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 background_thread for dynamic control options and details. This option is disabled by default.
opt.max_background_threadsr-
Maximum number of background threads that will be created if background_thread is set. Defaults to number of cpus.
opt.dirty_decay_msr-
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.
madvise(
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 ...MADV_FREE
and arenas.dirty_decay_ms
for related dynamic control options. See arena.<i>.dirty_decay_ms
for a description of muzzy pages.for a description of muzzy pages. Note
that when the opt.muzzy_decay_ms
feature is enabled, the arenas reserved for oversize requests may have
its own default decay settings.oversize_threshold
opt.muzzy_decay_msr-
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.
madvise(),
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 ...MADV_FREE
and arenas.muzzy_decay_ms
for related dynamic control options.arena.<i>.muzzy_decay_ms
opt.lg_extent_max_active_fitr-
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).
opt.stats_printr-
Enable/disable statistics printing at exit. If
enabled, the malloc_stats_print()
function is called at program exit via an
atexit(3) function.
can be combined to specify output options. If
opt.stats_print_opts--enable-stats 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, atexit() may
allocate memory during application initialization and then deadlock
internally when jemalloc in turn calls
atexit(), so this option is not
universally usable (though the application can register its own
atexit() 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.
opt.stats_print_optsr-
Options (the opts string) to pass
to the malloc_stats_print() at exit (enabled
through
). See
available options in opt.stats_printmalloc_stats_print().
Has no effect unless
is
enabled. The default is “”.opt.stats_print
opt.stats_intervalr-
Average interval between statistics outputs, as measured
in bytes of allocation activity. The actual interval may be sporadic
because decentralized event counters are used to avoid synchronization
bottlenecks. The output may be triggered on any thread, which then
calls malloc_stats_print().
can be combined to specify output options. By default,
interval-triggered stats output is disabled (encoded as
-1).opt.stats_interval_opts
opt.stats_interval_optsr-
Options (the opts string) to pass
to the malloc_stats_print() for interval based
statistics printing (enabled
through
). See
available options in opt.stats_intervalmalloc_stats_print().
Has no effect unless
is
enabled. The default is “”.opt.stats_interval
opt.junkr-
[--enable-fill]
Junk filling. If set to “alloc”, each byte
of uninitialized allocated memory will be initialized to
0xa5. If set to “free”, all deallocated
memory will be initialized to 0x5a. If set to
“true”, both allocated and deallocated memory will be
initialized, and if set to “false”, junk filling be
disabled entirely. This is intended for debugging and will impact
performance negatively. This option is “false” by default
unless --enable-debug is specified during
configuration, in which case it is “true” by
default.
opt.zeror-
[--enable-fill]
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
realloc() and
rallocx() 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.
opt.utracer-
[--enable-utrace]
Allocation tracing based on utrace(2) enabled/disabled. This option is disabled by default.
opt.xmallocr-
[--enable-xmalloc]
Abort-on-out-of-memory enabled/disabled. If enabled,
rather than returning failure for any allocation function, display a
diagnostic message on STDERR_FILENO and cause the
program to drop core (using
abort(3)). If an application is
designed to depend on this behavior, set the option at compile time by
including the following in the source code:
malloc_conf = "xmalloc:true";
This option is disabled by default.
opt.tcacher-
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
option for related tuning information. This option is enabled by
default.opt.tcache_max
opt.tcache_maxr-
Maximum size class to cache in the thread-specific cache (tcache). At a minimum, the first size class is cached; and at a maximum, size classes up to 8 MiB can be cached. The default maximum is 32 KiB (2^15). As a convenience, this may also be set by specifying lg_tcache_max, which will be taken to be the base-2 logarithm of the setting of tcache_max.
opt.thpr-
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
MADV_HUGEPAGEMADV_NOHUGEPAGE
);
in addition, for arenas with customized opt.metadata_thp
,
this option is bypassed as it is implemented as part of the default
extent hooks.extent_hooks
opt.profr-
[--enable-prof]
Memory profiling enabled/disabled. If enabled, profile
memory allocation activity. See the
option for on-the-fly activation/deactivation. See the opt.prof_active
option for probabilistic sampling control. See the opt.lg_prof_sample
option for control of cumulative sample reporting. See the opt.prof_accum
option for information on interval-triggered profile dumping, the opt.lg_prof_interval
option for information on high-water-triggered profile dumping, and the
opt.prof_gdump
option for final profile dumping. Profile output is compatible with
the jeprof command, which is based on the
pprof that is developed as part of the gperftools
package. See HEAP PROFILE
FORMAT for heap profile format documentation.opt.prof_final
opt.prof_prefixr-
[--enable-prof]
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
jeprof. This prefix value can be overridden by
.
prof.prefix
opt.prof_activer-
[--enable-prof]
Profiling activated/deactivated. This is a secondary
control mechanism that makes it possible to start the application with
profiling enabled (see the
option) but
inactive, then toggle profiling at any time during program execution
with the opt.prof
mallctl.
This option is enabled by default.prof.active
opt.prof_thread_active_initr-
[--enable-prof]
Initial setting for
in newly created threads. The initial setting for newly created threads
can also be changed during execution via the thread.prof.active
mallctl. This option is enabled by default.prof.thread_active_init
opt.lg_prof_sampler-
[--enable-prof]
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).
opt.prof_accumr-
[--enable-prof]
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.
opt.lg_prof_intervalr-
[--enable-prof]
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
<prefix>.<pid>.<seq>.i<iseq>.heap,
where <prefix> is controlled by the
and
opt.prof_prefix
options. By default, interval-triggered profile dumping is disabled
(encoded as -1).
prof.prefix
opt.prof_gdumpr-
[--enable-prof]
Set the initial state of
, which when
enabled triggers a memory profile dump every time the total virtual
memory exceeds the previous maximum. This option is disabled by
default.prof.gdump
opt.prof_finalr-
[--enable-prof]
Use an
atexit(3) function to dump final memory
usage to a file named according to the pattern
<prefix>.<pid>.<seq>.f.heap,
where <prefix> is controlled by the
and
opt.prof_prefix
options. Note that prof.prefixatexit() may allocate
memory during application initialization and then deadlock internally
when jemalloc in turn calls atexit(), so
this option is not universally usable (though the application can
register its own atexit() function with
equivalent functionality). This option is disabled by
default.
opt.prof_leakr-
[--enable-prof]
Leak reporting enabled/disabled. If enabled, use an
atexit(3) function to report memory leaks
detected by allocation sampling. See the
option for
information on analyzing heap profile output. Works only when combined
with opt.prof
, otherwise does nothing. This option is disabled by default.
opt.prof_final
opt.prof_leak_errorr-
[--enable-prof]
Similar to
, but makes the process exit with error
code 1 if a memory leak is detected. This option supersedes
opt.prof_leak
,
meaning that if both are specified, this option takes precedence. When
enabled, also enables opt.prof_leak
. Works only when combined with
opt.prof_leak
,
otherwise does nothing. This option is disabled by default.
opt.prof_final
opt.zero_reallocr-
Determines the behavior of
realloc() when passed a value of zero for the new
size. “alloc” treats this as an allocation of size zero
(and returns a non-null result except in case of resource exhaustion).
“free” treats this as a deallocation of the pointer, and
returns NULL without setting
errno. “abort” aborts the process if
zero is passed. The default is “free” on Linux and
Windows, and “alloc” elsewhere.
There is considerable divergence of behaviors across
implementations in handling this case. Many have the behavior of
“free”. This can introduce security vulnerabilities, since
a NULL return value indicates failure, and the
continued validity of the passed-in pointer (per POSIX and C11).
“alloc” is safe, but can cause leaks in programs that
expect the common behavior. Programs intended to be portable and
leak-free cannot assume either behavior, and must therefore never call
realloc with a size of 0. The “abort” option enables these
testing this behavior.
thread.arenarw
Get or set the arena associated with the calling
thread. If the specified arena was not initialized beforehand (see the
mallctl), it will be automatically initialized as a side effect of
calling this interface.arena.i.initialized
thread.allocatedr-
[--enable-stats]
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.
thread.allocatedpr-
[--enable-stats]
Get a pointer to the the value that is returned by the
mallctl. This is useful for avoiding the overhead of repeated
thread.allocatedmallctl*() calls. Note that the underlying counter
should not be modified by the application.
thread.deallocatedr-
[--enable-stats]
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.
thread.deallocatedpr-
[--enable-stats]
Get a pointer to the the value that is returned by the
mallctl. This is useful for avoiding the overhead of repeated
thread.deallocatedmallctl*() calls. Note that the underlying counter
should not be modified by the application.
thread.peak.readr-
[--enable-stats]
Get an approximation of the maximum value of the
difference between the number of bytes allocated and the number of bytes
deallocated by the calling thread since the last call to
,
or since the thread's creation if it has not called thread.peak.reset
.
No guarantees are made about the quality of the approximation, but
jemalloc currently endeavors to maintain accuracy to within one hundred
kilobytes.
thread.peak.reset
thread.peak.reset--
[--enable-stats]
Resets the counter for net bytes allocated in the calling
thread to zero. This affects subsequent calls to
,
but not the values returned by thread.peak.read
or thread.allocated
.
thread.deallocated
thread.tcache.enabledrw
Enable/disable calling thread's tcache. The tcache is
implicitly flushed as a side effect of becoming
disabled (see
).
thread.tcache.flush
thread.tcache.flush--
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.
thread.prof.namer- or
-w
[--enable-prof]
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 isgraph(3) and isblank(3).
thread.prof.activerw
[--enable-prof]
Control whether sampling is currently active for the
calling thread. This is an activation mechanism in addition to
; both must
be active for the calling thread to sample. This flag is enabled by
default.prof.active
thread.idle--
Hints to jemalloc that the calling thread will be idle for some nontrivial period of time (say, on the order of seconds), and that doing some cleanup operations may be beneficial. There are no guarantees as to what specific operations will be performed; currently this flushes the caller's tcache and may (according to some heuristic) purge its associated arena.
This is not intended to be a general-purpose background activity
mechanism, and threads should not wake up multiple times solely to call
it. Rather, a thread waiting for a task should do a timed wait first,
call
if no task appears in the timeout interval, and then do an untimed wait.
For such a background activity mechanism, see
thread.idle
.
background_thread
tcache.creater-
Create an explicit thread-specific cache (tcache) and
return an identifier that can be passed to the MALLOCX_TCACHE(
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.
tc)
If the amount of space supplied for storing the thread-specific
cache identifier does not equal
sizeof(unsigned), no
thread-specific cache will be created, no data will be written to the
space pointed by oldp, and
*oldlenp will be set to 0.
tcache.flush-w
Flush the specified thread-specific cache (tcache). The
same considerations apply to this interface as to
,
except that the tcache will never be automatically discarded.
thread.tcache.flush
tcache.destroy-w
Flush the specified thread-specific cache (tcache) and make the identifier available for use during a future tcache creation.
arena.<i>.initializedr-
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 MALLCTL_ARENAS_ALL are
initialized (always true).
arena.<i>.decay--
Trigger decay-based purging of unused dirty/muzzy pages
for arena <i>, or for all arenas if <i> equals
MALLCTL_ARENAS_ALL. The proportion of unused
dirty/muzzy pages to be purged depends on the current time; see
and opt.dirty_decay_ms
for details.opt.muzy_decay_ms
arena.<i>.purge--
Purge all unused dirty pages for arena <i>, or for
all arenas if <i> equals MALLCTL_ARENAS_ALL.
arena.<i>.reset--
Discard all of the arena's extant allocations. This
interface can only be used with arenas explicitly created via
. 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.arenas.create
arena.<i>.destroy--
Destroy the arena. Discard all of the arena's extant
allocations using the same mechanism as for
(with all the same constraints and side effects), merge the arena stats
into those accessible at arena index
arena.<i>.resetMALLCTL_ARENAS_DESTROYED, and then completely
discard all metadata associated with the arena. Future calls to
may
recycle the arena index. Destruction will fail if any threads are
currently associated with the arena as a result of calls to arenas.create
.thread.arena
arena.<i>.dssrw
Set the precedence of dss allocation as related to mmap
allocation for arena <i>, or for all arenas if <i> equals
MALLCTL_ARENAS_ALL. See
for supported
settings.opt.dss
arena.<i>.dirty_decay_msrw
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
for additional information.opt.dirty_decay_ms
arena.<i>.muzzy_decay_msrw
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
for additional information.opt.muzzy_decay_ms
arena.<i>.retain_grow_limitrw
Maximum size to grow retained region (only relevant when
is
enabled). This controls the maximum increment to expand virtual memory,
or allocation through opt.retain
.
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.arena.<i>extent_hooks
arena.<i>.extent_hooksrw
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
such
that all extents originate from an application-supplied extent allocator
(by specifying the custom extent hook functions during arena creation).
However, the API guarantees for the automatically created arenas may be
relaxed -- hooks set there may be called in a "best effort" fashion; in
addition there may be extents created prior to the application having an
opportunity to take over extent allocation.arenas.create
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;
};The extent_hooks_t 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 NULL, 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.
typedef void *(extent_alloc_t)( | extent_hooks_t *extent_hooks, |
| void *new_addr, | |
| size_t size, | |
| size_t alignment, | |
| bool *zero, | |
| bool *commit, | |
unsigned arena_ind); |
An extent allocation function conforms to the
extent_alloc_t type and upon success returns a pointer to
size bytes of mapped memory on behalf of arena
arena_ind such that the extent's base address is
a multiple of alignment, as well as setting
*zero to indicate whether the extent is zeroed
and *commit to indicate whether the extent is
committed. Upon error the function returns NULL
and leaves *zero and
*commit unmodified. The
size parameter is always a multiple of the page
size. The alignment parameter is always a power
of two at least as large as the page size. Zeroing is mandatory if
*zero is true upon function entry. Committing is
mandatory if *commit is true upon function entry.
If new_addr is not NULL, the
returned pointer must be new_addr on success or
NULL 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
setting irrelevant.arena.<i>.dss
typedef bool (extent_dalloc_t)( | extent_hooks_t *extent_hooks, |
| void *addr, | |
| size_t size, | |
| bool committed, | |
unsigned arena_ind); |
An extent deallocation function conforms to the
extent_dalloc_t type and deallocates an extent at given
addr and size with
committed/decommited memory as indicated, on
behalf of arena arena_ind, 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.
typedef void (extent_destroy_t)( | extent_hooks_t *extent_hooks, |
| void *addr, | |
| size_t size, | |
| bool committed, | |
unsigned arena_ind); |
An extent destruction function conforms to the
extent_destroy_t type and unconditionally destroys an
extent at given addr and
size with
committed/decommited memory as indicated, on
behalf of arena arena_ind. This function may be
called to destroy retained extents during arena destruction (see
).arena.<i>.destroy
typedef bool (extent_commit_t)( | extent_hooks_t *extent_hooks, |
| void *addr, | |
| size_t size, | |
| size_t offset, | |
| size_t length, | |
unsigned arena_ind); |
An extent commit function conforms to the
extent_commit_t type and commits zeroed physical memory to
back pages within an extent at given addr and
size at offset bytes,
extending for length on behalf of arena
arena_ind, 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.
typedef bool (extent_decommit_t)( | extent_hooks_t *extent_hooks, |
| void *addr, | |
| size_t size, | |
| size_t offset, | |
| size_t length, | |
unsigned arena_ind); |
An extent decommit function conforms to the
extent_decommit_t type and decommits any physical memory
that is backing pages within an extent at given
addr and size at
offset bytes, extending for
length on behalf of arena
arena_ind, 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.
typedef bool (extent_purge_t)( | extent_hooks_t *extent_hooks, |
| void *addr, | |
| size_t size, | |
| size_t offset, | |
| size_t length, | |
unsigned arena_ind); |
An extent purge function conforms to the
extent_purge_t type and discards physical pages
within the virtual memory mapping associated with an extent at given
addr and size at
offset bytes, extending for
length on behalf of arena
arena_ind. A lazy extent purge function (e.g.
implemented via
madvise()
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....MADV_FREE
typedef bool (extent_split_t)( | extent_hooks_t *extent_hooks, |
| void *addr, | |
| size_t size, | |
| size_t size_a, | |
| size_t size_b, | |
| bool committed, | |
unsigned arena_ind); |
An extent split function conforms to the
extent_split_t type and optionally splits an extent at
given addr and size into
two adjacent extents, the first of size_a bytes,
and the second of size_b bytes, operating on
committed/decommitted memory as indicated, on
behalf of arena arena_ind, 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.
typedef bool (extent_merge_t)( | extent_hooks_t *extent_hooks, |
| void *addr_a, | |
| size_t size_a, | |
| void *addr_b, | |
| size_t size_b, | |
| bool committed, | |
unsigned arena_ind); |
An extent merge function conforms to the
extent_merge_t type and optionally merges adjacent extents,
at given addr_a and size_a
with given addr_b and
size_b into one contiguous extent, operating on
committed/decommitted memory as indicated, on
behalf of arena arena_ind, 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.
arenas.narenasr-
Current limit on number of arenas.
arenas.dirty_decay_msrw
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
during arena creation. See arena.<i>.dirty_decay_ms
for additional information.opt.dirty_decay_ms
arenas.muzzy_decay_msrw
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
during arena creation. See arena.<i>.muzzy_decay_ms
for additional information.opt.muzzy_decay_ms
arenas.quantumr-
Quantum size.
arenas.pager-
Page size.
arenas.tcache_maxr-
Maximum thread-cached size class.
arenas.nbinsr-
Number of bin size classes.
arenas.nhbinsr-
Total number of thread cache bin size classes.
arenas.bin.<i>.sizer-
Maximum size supported by size class.
arenas.bin.<i>.nregsr-
Number of regions per slab.
arenas.bin.<i>.slab_sizer-
Number of bytes per slab.
arenas.nlextentsr-
Total number of large size classes.
arenas.lextent.<i>.sizer-
Maximum size supported by this large size class.
arenas.createrw
Explicitly create a new arena outside the range of automatically managed arenas, with optionally specified extent hooks, and return the new arena index.
If the amount of space supplied for storing the arena index does
not equal sizeof(unsigned), no
arena will be created, no data will be written to the space pointed by
oldp, and *oldlenp will
be set to 0.
arenas.lookuprw
Index of the arena to which an allocation belongs to.
prof.thread_active_initrw
[--enable-prof]
Control the initial setting for
in newly created threads. See the thread.prof.active
option for additional information.opt.prof_thread_active_init
prof.activerw
[--enable-prof]
Control whether sampling is currently active. See the
option for additional information, as well as the interrelated opt.prof_active
mallctl.thread.prof.active
prof.dump-w
[--enable-prof]
Dump a memory profile to the specified file, or if NULL
is specified, to a file according to the pattern
<prefix>.<pid>.<seq>.m<mseq>.heap,
where <prefix> is controlled by the
and opt.prof_prefix
options.prof.prefix
prof.prefix-w
[--enable-prof]
Set the filename prefix for profile dumps. See
for the default setting. This can be useful to differentiate profile
dumps such as from forked processes.
opt.prof_prefix
prof.gdumprw
[--enable-prof]
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
<prefix>.<pid>.<seq>.u<useq>.heap,
where <prefix> is controlled by the
and
opt.prof_prefix
options.prof.prefix
prof.reset-w
[--enable-prof]
Reset all memory profile statistics, and optionally
update the sample rate (see
and opt.lg_prof_sample
).
prof.lg_sample
prof.lg_sampler-
[--enable-prof]
Get the current sample rate (see
).
opt.lg_prof_sample
prof.intervalr-
[--enable-prof]
Average number of bytes allocated between
interval-based profile dumps. See the
option for additional information.opt.lg_prof_interval
stats.allocatedr-
[--enable-stats]
Total number of bytes allocated by the application.
stats.activer-
[--enable-stats]
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
.
This does not include
stats.allocated
,
stats.arenas.<i>.pdirty
, nor pages
entirely devoted to allocator metadata.stats.arenas.<i>.pmuzzy
stats.metadatar-
[--enable-stats]
Total number of bytes dedicated to metadata, which
comprise base allocations used for bootstrap-sensitive allocator
metadata structures (see
)
and internal allocations (see stats.arenas.<i>.base
).
Transparent huge page (enabled with opt.metadata_thp) usage is not
considered.stats.arenas.<i>.internal
stats.metadata_thpr-
[--enable-stats]
Number of transparent huge pages (THP) used for
metadata. See
and
opt.metadata_thp) for
details.stats.metadata
stats.residentr-
[--enable-stats]
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
.stats.active
stats.mappedr-
[--enable-stats]
Total number of bytes in active extents mapped by the
allocator. This is larger than
. This
does not include inactive extents, even those that contain unused dirty
pages, which means that there is no strict ordering between this and
stats.active
.stats.resident
stats.retainedr-
[--enable-stats]
Total number of bytes in virtual memory mappings that
were retained rather than being returned to the operating system via
e.g. munmap(2) or similar. Retained virtual
memory is typically untouched, decommitted, or purged, so it has no
strongly associated physical memory (see extent hooks for details).
Retained memory is excluded from mapped memory statistics, e.g.
.
stats.mapped
stats.zero_reallocsr-
[--enable-stats]
Number of times that the realloc()
was called with a non-NULL pointer argument and a
0 size argument. This is a fundamentally unsafe
pattern in portable programs; see
for details.
opt.zero_realloc
stats.background_thread.num_threadsr-
[--enable-stats]
Number of background threads running currently.
stats.background_thread.num_runsr-
[--enable-stats]
Total number of runs from all background threads.
stats.background_thread.run_intervalr-
[--enable-stats]
Average run interval in nanoseconds of background threads.
stats.mutexes.ctl.{counter};r-
[--enable-stats]
Statistics on ctl mutex (global
scope; mallctl related).
is one of the
counters below:{counter}
num_ops (uint64_t):
Total number of lock acquisition operations on this mutex.
num_spin_acq (uint64_t): 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.
num_wait (uint64_t): 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.
max_wait_time (uint64_t):
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.
total_wait_time (uint64_t):
Cumulative time in nanoseconds spent on wait-acquired lock operations.
Similarly, spin-acquired cases are not considered.
max_num_thds (uint32_t): Maximum
number of threads waiting on this mutex simultaneously. Similarly,
spin-acquired cases are not considered.
num_owner_switch (uint64_t):
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.
stats.mutexes.background_thread.{counter}r-
[--enable-stats]
Statistics on background_thread mutex
(global scope;
related). background_thread
is one of the counters in mutex profiling
counters.{counter}
stats.mutexes.prof.{counter}r-
[--enable-stats]
Statistics on prof mutex (global
scope; profiling related).
is one of the
counters in mutex profiling
counters.{counter}
stats.mutexes.prof_thds_data.{counter}r-
[--enable-stats]
Statistics on prof threads data mutex
(global scope; profiling related).
is one
of the counters in mutex profiling
counters.{counter}
stats.mutexes.prof_dump.{counter}r-
[--enable-stats]
Statistics on prof dumping mutex
(global scope; profiling related).
is one
of the counters in mutex profiling
counters.{counter}
stats.mutexes.reset--
[--enable-stats]
Reset all mutex profile statistics, including global mutexes, arena mutexes and bin mutexes.
stats.arenas.<i>.dssr-
dss (sbrk(2)) allocation precedence as
related to mmap(2) allocation. See
for details.
opt.dss
stats.arenas.<i>.dirty_decay_msr-
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
for details.opt.dirty_decay_ms
stats.arenas.<i>.muzzy_decay_msr-
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
for details.opt.muzzy_decay_ms
stats.arenas.<i>.nthreadsr-
Number of threads currently assigned to arena.
stats.arenas.<i>.uptimer-
Time elapsed (in nanoseconds) since the arena was
created. If <i> equals 0 or
MALLCTL_ARENAS_ALL, this is the uptime since malloc
initialization.
stats.arenas.<i>.pactiver-
Number of pages in active extents.
stats.arenas.<i>.pdirtyr-
Number of pages within unused extents that are
potentially dirty, and for which madvise() or
similar has not been called. See
for a description of dirty pages.opt.dirty_decay_ms
stats.arenas.<i>.pmuzzyr-
Number of pages within unused extents that are muzzy.
See
for a description of muzzy pages.opt.muzzy_decay_ms
stats.arenas.<i>.mappedr-
[--enable-stats]
Number of mapped bytes.
stats.arenas.<i>.retainedr-
[--enable-stats]
Number of retained bytes. See
for
details.stats.retained
stats.arenas.<i>.extent_availr-
[--enable-stats]
Number of allocated (but unused) extent structs in this arena.
stats.arenas.<i>.baser-
[--enable-stats]
Number of bytes dedicated to bootstrap-sensitive allocator metadata structures.
stats.arenas.<i>.internalr-
[--enable-stats]
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.
stats.arenas.<i>.metadata_thpr-
[--enable-stats]
Number of transparent huge pages (THP) used for metadata. See opt.metadata_thp for details.
stats.arenas.<i>.residentr-
[--enable-stats]
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.
stats.arenas.<i>.dirty_npurger-
[--enable-stats]
Number of dirty page purge sweeps performed.
stats.arenas.<i>.dirty_nmadviser-
[--enable-stats]
Number of madvise() or similar
calls made to purge dirty pages.
stats.arenas.<i>.dirty_purgedr-
[--enable-stats]
Number of dirty pages purged.
stats.arenas.<i>.muzzy_npurger-
[--enable-stats]
Number of muzzy page purge sweeps performed.
stats.arenas.<i>.muzzy_nmadviser-
[--enable-stats]
Number of madvise() or similar
calls made to purge muzzy pages.
stats.arenas.<i>.muzzy_purgedr-
[--enable-stats]
Number of muzzy pages purged.
stats.arenas.<i>.small.allocatedr-
[--enable-stats]
Number of bytes currently allocated by small objects.
stats.arenas.<i>.small.nmallocr-
[--enable-stats]
Cumulative number of times a small allocation was
requested from the arena's bins, whether to fill the relevant tcache if
is
enabled, or to directly satisfy an allocation request
otherwise.opt.tcache
stats.arenas.<i>.small.ndallocr-
[--enable-stats]
Cumulative number of times a small allocation was
returned to the arena's bins, whether to flush the relevant tcache if
is
enabled, or to directly deallocate an allocation
otherwise.opt.tcache
stats.arenas.<i>.small.nrequestsr-
[--enable-stats]
Cumulative number of allocation requests satisfied by all bin size classes.
stats.arenas.<i>.small.nfillsr-
[--enable-stats]
Cumulative number of tcache fills by all small size classes.
stats.arenas.<i>.small.nflushesr-
[--enable-stats]
Cumulative number of tcache flushes by all small size classes.
stats.arenas.<i>.large.allocatedr-
[--enable-stats]
Number of bytes currently allocated by large objects.
stats.arenas.<i>.large.nmallocr-
[--enable-stats]
Cumulative number of times a large extent was allocated
from the arena, whether to fill the relevant tcache if
is enabled and
the size class is within the range being cached, or to directly satisfy
an allocation request otherwise.opt.tcache
stats.arenas.<i>.large.ndallocr-
[--enable-stats]
Cumulative number of times a large extent was returned
to the arena, whether to flush the relevant tcache if
is enabled and
the size class is within the range being cached, or to directly
deallocate an allocation otherwise.opt.tcache
stats.arenas.<i>.large.nrequestsr-
[--enable-stats]
Cumulative number of allocation requests satisfied by all large size classes.
stats.arenas.<i>.large.nfillsr-
[--enable-stats]
Cumulative number of tcache fills by all large size classes.
stats.arenas.<i>.large.nflushesr-
[--enable-stats]
Cumulative number of tcache flushes by all large size classes.
stats.arenas.<i>.bins.<j>.nmallocr-
[--enable-stats]
Cumulative number of times a bin region of the
corresponding size class was allocated from the arena, whether to fill
the relevant tcache if
is enabled, or
to directly satisfy an allocation request otherwise.opt.tcache
stats.arenas.<i>.bins.<j>.ndallocr-
[--enable-stats]
Cumulative number of times a bin region of the
corresponding size class was returned to the arena, whether to flush the
relevant tcache if
is enabled, or
to directly deallocate an allocation otherwise.opt.tcache
stats.arenas.<i>.bins.<j>.nrequestsr-
[--enable-stats]
Cumulative number of allocation requests satisfied by bin regions of the corresponding size class.
stats.arenas.<i>.bins.<j>.curregsr-
[--enable-stats]
Current number of regions for this size class.
stats.arenas.<i>.bins.<j>.nfillsr-
Cumulative number of tcache fills.
stats.arenas.<i>.bins.<j>.nflushesr-
Cumulative number of tcache flushes.
stats.arenas.<i>.bins.<j>.nslabsr-
[--enable-stats]
Cumulative number of slabs created.
stats.arenas.<i>.bins.<j>.nreslabsr-
[--enable-stats]
Cumulative number of times the current slab from which to allocate changed.
stats.arenas.<i>.bins.<j>.curslabsr-
[--enable-stats]
Current number of slabs.
stats.arenas.<i>.bins.<j>.nonfull_slabsr-
[--enable-stats]
Current number of nonfull slabs.
stats.arenas.<i>.bins.<j>.mutex.{counter}r-
[--enable-stats]
Statistics on
arena.<i>.bins.<j> mutex (arena bin
scope; bin operation related).
is one of
the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.extents.<j>.n{extent_type}r-
[--enable-stats]
Number of extents of the given type in this arena in the bucket corresponding to page size index <j>. The extent type is one of dirty, muzzy, or retained.
stats.arenas.<i>.extents.<j>.{extent_type}_bytesr-
[--enable-stats]
Sum of the bytes managed by extents of the given type in this arena in the bucket corresponding to page size index <j>. The extent type is one of dirty, muzzy, or retained.
stats.arenas.<i>.lextents.<j>.nmallocr-
[--enable-stats]
Cumulative number of times a large extent of the
corresponding size class was allocated from the arena, whether to fill
the relevant tcache if
is enabled and
the size class is within the range being cached, or to directly satisfy
an allocation request otherwise.opt.tcache
stats.arenas.<i>.lextents.<j>.ndallocr-
[--enable-stats]
Cumulative number of times a large extent of the
corresponding size class was returned to the arena, whether to flush the
relevant tcache if
is enabled and
the size class is within the range being cached, or to directly
deallocate an allocation otherwise.opt.tcache
stats.arenas.<i>.lextents.<j>.nrequestsr-
[--enable-stats]
Cumulative number of allocation requests satisfied by large extents of the corresponding size class.
stats.arenas.<i>.lextents.<j>.curlextentsr-
[--enable-stats]
Current number of large allocations for this size class.
stats.arenas.<i>.mutexes.large.{counter}r-
[--enable-stats]
Statistics on arena.<i>.large
mutex (arena scope; large allocation related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.extent_avail.{counter}r-
[--enable-stats]
Statistics on arena.<i>.extent_avail
mutex (arena scope; extent avail related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.extents_dirty.{counter}r-
[--enable-stats]
Statistics on arena.<i>.extents_dirty
mutex (arena scope; dirty extents related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.extents_muzzy.{counter}r-
[--enable-stats]
Statistics on arena.<i>.extents_muzzy
mutex (arena scope; muzzy extents related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.extents_retained.{counter}r-
[--enable-stats]
Statistics on arena.<i>.extents_retained
mutex (arena scope; retained extents related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.decay_dirty.{counter}r-
[--enable-stats]
Statistics on arena.<i>.decay_dirty
mutex (arena scope; decay for dirty pages related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.decay_muzzy.{counter}r-
[--enable-stats]
Statistics on arena.<i>.decay_muzzy
mutex (arena scope; decay for muzzy pages related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.base.{counter}r-
[--enable-stats]
Statistics on arena.<i>.base
mutex (arena scope; base allocator related).
is one of the counters in mutex profiling
counters.{counter}
stats.arenas.<i>.mutexes.tcache_list.{counter}r-
[--enable-stats]
Statistics on
arena.<i>.tcache_list mutex (arena scope;
tcache to arena association related). This mutex is expected to be
accessed less often.
is one of the
counters in mutex profiling
counters.{counter}
Although the heap profiling functionality was originally designed to be compatible with the pprof command that is developed as part of the gperftools package, the addition of per thread heap profiling functionality required a different heap profile format. The jeprof command is derived from pprof, with enhancements to support the heap profile format described here.
In the following hypothetical heap profile, [...]
indicates elision for the sake of compactness.
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: 0] t99: 1: 192 [0: 0] [...] MAPPED_LIBRARIES: [...]
The following matches the above heap profile, but most
tokens are replaced with <description> to indicate
descriptions of the corresponding fields.
<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>
When debugging, it is a good idea to configure/build jemalloc with
the --enable-debug and --enable-fill
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.
Programs often accidentally depend on “uninitialized”
memory actually being filled with zero bytes. Junk filling
(see the
option) tends to expose such bugs in the form of obviously incorrect
results and/or coredumps. Conversely, zero
filling (see the opt.junk
option) eliminates
the symptoms of such bugs. Between these two options, it is usually
possible to quickly detect, diagnose, and eliminate such bugs.opt.zero
This implementation does not provide much detail about the problems it detects, because the performance impact for storing such information would be prohibitive.
If any of the memory allocation/deallocation functions detect an
error or warning condition, a message will be printed to file descriptor
STDERR_FILENO. Errors will result in the process
dumping core. If the
option is set, most
warnings are treated as errors.opt.abort
The malloc_message variable allows the programmer
to override the function which emits the text strings forming the errors
and warnings if for some reason the STDERR_FILENO file
descriptor is not suitable for this.
malloc_message() takes the
cbopaque pointer argument that is
NULL unless overridden by the arguments in a call to
malloc_stats_print(), 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.
All messages are prefixed by
“<jemalloc>: ”.
The malloc() and
calloc() functions return a pointer to the
allocated memory if successful; otherwise a NULL
pointer is returned and errno is set to
ENOMEM.
The posix_memalign() function
returns the value 0 if successful; otherwise it returns an error value.
The posix_memalign() function will fail
if:
The alignment parameter is
not a power of 2 at least as large as
sizeof(void *).
Memory allocation error.
The aligned_alloc() function returns
a pointer to the allocated memory if successful; otherwise a
NULL pointer is returned and
errno is set. The
aligned_alloc() function will fail if:
The alignment parameter is
not a power of 2.
Memory allocation error.
The realloc() function returns a
pointer, possibly identical to ptr, to the
allocated memory if successful; otherwise a NULL
pointer is returned, and errno is set to
ENOMEM if the error was the result of an
allocation failure. The realloc()
function always leaves the original buffer intact when an error occurs.
The free() function returns no
value.
The mallocx() and
rallocx() functions return a pointer to
the allocated memory if successful; otherwise a NULL
pointer is returned to indicate insufficient contiguous memory was
available to service the allocation request.
The xallocx() function returns the
real size of the resulting resized allocation pointed to by
ptr, which is a value less than
size if the allocation could not be adequately
grown in place.
The sallocx() function returns the
real size of the allocation pointed to by ptr.
The nallocx() returns the real size
that would result from a successful equivalent
mallocx() function call, or zero if
insufficient memory is available to perform the size computation.
The mallctl(),
mallctlnametomib(), and
mallctlbymib() functions return 0 on
success; otherwise they return an error value. The functions will fail
if:
newp is not
NULL, and newlen is too
large or too small. Alternatively, *oldlenp
is too large or too small; when it happens, except for a very few
cases explicitly documented otherwise, as much data as possible
are read despite the error, with the amount of data read being
recorded in *oldlenp.
name or
mib specifies an unknown/invalid
value.
Attempt to read or write void value, or attempt to write read-only value.
A memory allocation failure occurred.
An interface with side effects failed in some way
not directly related to mallctl*()
read/write processing.
The malloc_usable_size() function
returns the usable size of the allocation pointed to by
ptr.
The following environment variable affects the execution of the allocation functions:
MALLOC_CONFIf the environment variable
MALLOC_CONF is set, the characters it contains
will be interpreted as options.