project-base/server-skynet-source-3rd-jemalloc
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
667eca2ac2
server-skynet-source-3rd-je.../include/jemalloc/internal/tcache.h

470 lines
13 KiB
C
Raw Normal View History

Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
/******************************************************************************/
#ifdef JEMALLOC_H_TYPES
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
typedef struct tcache_bin_info_s tcache_bin_info_t;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
typedef struct tcache_bin_s tcache_bin_t;
typedef struct tcache_s tcache_t;
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
typedef struct tcaches_s tcaches_t;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
/*
* tcache pointers close to NULL are used to encode state information that is
* used for two purposes: preventing thread caching on a per thread basis and
* cleaning up during thread shutdown.
*/
#define TCACHE_STATE_DISABLED ((tcache_t *)(uintptr_t)1)
#define TCACHE_STATE_REINCARNATED ((tcache_t *)(uintptr_t)2)
#define TCACHE_STATE_PURGATORY ((tcache_t *)(uintptr_t)3)
#define TCACHE_STATE_MAX TCACHE_STATE_PURGATORY
Impose a minimum tcache count for small size classes. Now that small allocation runs have fewer regions due to run metadata residing in chunk headers, an explicit minimum tcache count is needed to make sure that tcache adequately amortizes synchronization overhead.
2015-05-20 08:47:16 +08:00
/*
* Absolute minimum number of cache slots for each small bin.
*/
#define TCACHE_NSLOTS_SMALL_MIN 20
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
/*
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
* Absolute maximum number of cache slots for each small bin in the thread
* cache. This is an additional constraint beyond that imposed as: twice the
* number of regions per run for this size class.
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
*
* This constant must be an even number.
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
*/
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
#define TCACHE_NSLOTS_SMALL_MAX 200
/* Number of cache slots for large size classes. */
#define TCACHE_NSLOTS_LARGE 20
Replace JEMALLOC_OPTIONS with MALLOC_CONF. Replace the single-character run-time flags with key/value pairs, which can be set via the malloc_conf global, /etc/malloc.conf, and the MALLOC_CONF environment variable. Replace the JEMALLOC_PROF_PREFIX environment variable with the "opt.prof_prefix" option. Replace umax2s() with u2s().
2010-10-24 09:37:06 +08:00
/* (1U << opt_lg_tcache_max) is used to compute tcache_maxclass. */
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
#define LG_TCACHE_MAXCLASS_DEFAULT 15
/*
Remove the lg_tcache_gc_sweep option. Remove the lg_tcache_gc_sweep option, because it is no longer very useful. Prior to the addition of dynamic adjustment of tcache fill count, it was possible for fill/flush overhead to be a problem, but this problem no longer occurs.
2012-03-06 06:34:37 +08:00
* TCACHE_GC_SWEEP is the approximate number of allocation events between
* full GC sweeps. Integer rounding may cause the actual number to be
* slightly higher, since GC is performed incrementally.
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
*/
Remove the lg_tcache_gc_sweep option. Remove the lg_tcache_gc_sweep option, because it is no longer very useful. Prior to the addition of dynamic adjustment of tcache fill count, it was possible for fill/flush overhead to be a problem, but this problem no longer occurs.
2012-03-06 06:34:37 +08:00
#define TCACHE_GC_SWEEP 8192
/* Number of tcache allocation/deallocation events between incremental GCs. */
#define TCACHE_GC_INCR \
((TCACHE_GC_SWEEP / NBINS) + ((TCACHE_GC_SWEEP / NBINS == 0) ? 0 : 1))
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
#endif /* JEMALLOC_H_TYPES */
/******************************************************************************/
#ifdef JEMALLOC_H_STRUCTS
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
typedef enum {
tcache_enabled_false = 0, /* Enable cast to/from bool. */
tcache_enabled_true = 1,
tcache_enabled_default = 2
} tcache_enabled_t;
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
/*
* Read-only information associated with each element of tcache_t's tbins array
* is stored separately, mainly to reduce memory usage.
*/
struct tcache_bin_info_s {
unsigned ncached_max; /* Upper limit on ncached. */
};
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
struct tcache_bin_s {
tcache_bin_stats_t tstats;
Dynamically adjust tcache fill count. Dynamically adjust tcache fill count (number of objects allocated per tcache refill) such that if GC has to flush inactive objects, the fill count gradually decreases. Conversely, if refills occur while the fill count is depressed, the fill count gradually increases back to its maximum value.
2011-03-21 15:18:17 +08:00
int low_water; /* Min # cached since last GC. */
unsigned lg_fill_div; /* Fill (ncached_max >> lg_fill_div). */
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
unsigned ncached; /* # of cached objects. */
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
/*
* To make use of adjacent cacheline prefetch, the items in the avail
* stack goes to higher address for newer allocations. avail points
* just above the available space, which means that
Fix a comment.
2015-11-13 02:51:32 +08:00
* avail[-ncached, ... -1] are available items and the lowest item will
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
* be allocated first.
*/
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
void **avail; /* Stack of available objects. */
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
};
struct tcache_s {
ql_elm(tcache_t) link; /* Used for aggregating stats. */
Style and spelling fixes.
2014-12-09 06:40:14 +08:00
uint64_t prof_accumbytes;/* Cleared after arena_prof_accum(). */
Use ticker for incremental tcache GC.
2016-02-03 12:37:24 +08:00
ticker_t gc_ticker; /* Drives incremental GC. */
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
szind_t next_gc_bin; /* Next bin to GC. */
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
tcache_bin_t tbins[1]; /* Dynamically sized. */
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
/*
* The pointer stacks associated with tbins follow as a contiguous
* array. During tcache initialization, the avail pointer in each
* element of tbins is initialized to point to the proper offset within
* this array.
*/
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
};
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
/* Linkage for list of available (previously used) explicit tcache IDs. */
struct tcaches_s {
union {
tcache_t *tcache;
tcaches_t *next;
};
};
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
#endif /* JEMALLOC_H_STRUCTS */
/******************************************************************************/
#ifdef JEMALLOC_H_EXTERNS
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
extern bool opt_tcache;
Replace JEMALLOC_OPTIONS with MALLOC_CONF. Replace the single-character run-time flags with key/value pairs, which can be set via the malloc_conf global, /etc/malloc.conf, and the MALLOC_CONF environment variable. Replace the JEMALLOC_PROF_PREFIX environment variable with the "opt.prof_prefix" option. Replace umax2s() with u2s().
2010-10-24 09:37:06 +08:00
extern ssize_t opt_lg_tcache_max;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
extern tcache_bin_info_t *tcache_bin_info;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
/*
Simplify small size class infrastructure. Program-generate small size class tables for all valid combinations of LG_TINY_MIN, LG_QUANTUM, and PAGE_SHIFT. Use the appropriate table to generate all relevant data structures, and remove the distinction between tiny/quantum/cacheline/subpage bins. Remove --enable-dynamic-page-shift. This option didn't prove useful in practice, and it prevented optimizations. Add Tilera architecture support.
2012-02-29 08:50:47 +08:00
* Number of tcache bins. There are NBINS small-object bins, plus 0 or more
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
* large-object bins.
*/
Make nhbins unsigned rather than size_t.
2016-02-25 03:02:14 +08:00
extern unsigned nhbins;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
/* Maximum cached size class. */
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
extern size_t tcache_maxclass;
/*
* Explicit tcaches, managed via the tcache.{create,flush,destroy} mallctls and
* usable via the MALLOCX_TCACHE() flag. The automatic per thread tcaches are
* completely disjoint from this data structure. tcaches starts off as a sparse
* array, so it has no physical memory footprint until individual pages are
* touched. This allows the entire array to be allocated the first time an
* explicit tcache is created without a disproportionate impact on memory usage.
*/
extern tcaches_t *tcaches;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
size_t tcache_salloc(const void *ptr);
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
void tcache_event_hard(tsd_t *tsd, tcache_t *tcache);
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
void *tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
tcache_bin_t *tbin, szind_t binind, bool *tcache_success);
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
void tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
szind_t binind, unsigned rem);
void tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
unsigned rem, tcache_t *tcache);
Implement tsd. Implement tsd, which is a TLS/TSD abstraction that uses one or both internally. Modify bootstrapping such that no tsd's are utilized until allocation is safe. Remove malloc_[v]tprintf(), and use malloc_snprintf() instead. Fix %p argument size handling in malloc_vsnprintf(). Fix a long-standing statistics-related bug in the "thread.arena" mallctl that could cause crashes due to linked list corruption.
2012-03-22 09:33:03 +08:00
void tcache_arena_associate(tcache_t *tcache, arena_t *arena);
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
void tcache_arena_reassociate(tcache_t *tcache, arena_t *oldarena,
arena_t *newarena);
void tcache_arena_dissociate(tcache_t *tcache, arena_t *arena);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tcache_t *tcache_get_hard(tsd_t *tsd);
Add configure options. Add: --with-lg-page --with-lg-page-sizes --with-lg-size-class-group --with-lg-quantum Get rid of STATIC_PAGE_SHIFT, in favor of directly setting LG_PAGE. Fix various edge conditions exposed by the configure options.
2014-10-10 08:54:06 +08:00
tcache_t *tcache_create(tsd_t *tsd, arena_t *arena);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
void tcache_cleanup(tsd_t *tsd);
void tcache_enabled_cleanup(tsd_t *tsd);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
void tcache_stats_merge(tcache_t *tcache, arena_t *arena);
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
bool tcaches_create(tsd_t *tsd, unsigned *r_ind);
void tcaches_flush(tsd_t *tsd, unsigned ind);
void tcaches_destroy(tsd_t *tsd, unsigned ind);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
bool tcache_boot(void);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
#endif /* JEMALLOC_H_EXTERNS */
/******************************************************************************/
#ifdef JEMALLOC_H_INLINES
#ifndef JEMALLOC_ENABLE_INLINE
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
void tcache_event(tsd_t *tsd, tcache_t *tcache);
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
void tcache_flush(void);
bool tcache_enabled_get(void);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tcache_t *tcache_get(tsd_t *tsd, bool create);
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
void tcache_enabled_set(bool enabled);
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
void *tcache_alloc_easy(tcache_bin_t *tbin, bool *tcache_success);
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
void *tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
size_t size, szind_t ind, bool zero, bool slow_path);
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
void *tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
size_t size, szind_t ind, bool zero, bool slow_path);
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
void tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
szind_t binind, bool slow_path);
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
void tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
size_t size, bool slow_path);
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_t *tcaches_get(tsd_t *tsd, unsigned ind);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
#endif
#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TCACHE_C_))
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
JEMALLOC_INLINE void
tcache_flush(void)
{
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tsd_t *tsd;
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
cassert(config_tcache);
Fix tsd cleanup regressions. Fix tsd cleanup regressions that were introduced in 5460aa6f6676c7f253bfcb75c028dfd38cae8aaf (Convert all tsd variables to reside in a single tsd structure.). These regressions were twofold: 1) tsd_tryget() should never (and need never) return NULL. Rename it to tsd_fetch() and simplify all callers. 2) tsd_*_set() must only be called when tsd is in the nominal state, because cleanup happens during the nominal-->purgatory transition, and re-initialization must not happen while in the purgatory state. Add tsd_nominal() and use it as needed. Note that tsd_*{p,}_get() can still be used as long as no re-initialization that would require cleanup occurs. This means that e.g. the thread_allocated counter can be updated unconditionally.
2014-10-05 02:12:53 +08:00
tsd = tsd_fetch();
tcache_cleanup(tsd);
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
}
JEMALLOC_INLINE bool
tcache_enabled_get(void)
{
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tsd_t *tsd;
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
tcache_enabled_t tcache_enabled;
cassert(config_tcache);
Fix tsd cleanup regressions. Fix tsd cleanup regressions that were introduced in 5460aa6f6676c7f253bfcb75c028dfd38cae8aaf (Convert all tsd variables to reside in a single tsd structure.). These regressions were twofold: 1) tsd_tryget() should never (and need never) return NULL. Rename it to tsd_fetch() and simplify all callers. 2) tsd_*_set() must only be called when tsd is in the nominal state, because cleanup happens during the nominal-->purgatory transition, and re-initialization must not happen while in the purgatory state. Add tsd_nominal() and use it as needed. Note that tsd_*{p,}_get() can still be used as long as no re-initialization that would require cleanup occurs. This means that e.g. the thread_allocated counter can be updated unconditionally.
2014-10-05 02:12:53 +08:00
tsd = tsd_fetch();
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tcache_enabled = tsd_tcache_enabled_get(tsd);
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
if (tcache_enabled == tcache_enabled_default) {
tcache_enabled = (tcache_enabled_t)opt_tcache;
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tsd_tcache_enabled_set(tsd, tcache_enabled);
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
}
return ((bool)tcache_enabled);
}
JEMALLOC_INLINE void
tcache_enabled_set(bool enabled)
{
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tsd_t *tsd;
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
tcache_enabled_t tcache_enabled;
cassert(config_tcache);
Fix tsd cleanup regressions. Fix tsd cleanup regressions that were introduced in 5460aa6f6676c7f253bfcb75c028dfd38cae8aaf (Convert all tsd variables to reside in a single tsd structure.). These regressions were twofold: 1) tsd_tryget() should never (and need never) return NULL. Rename it to tsd_fetch() and simplify all callers. 2) tsd_*_set() must only be called when tsd is in the nominal state, because cleanup happens during the nominal-->purgatory transition, and re-initialization must not happen while in the purgatory state. Add tsd_nominal() and use it as needed. Note that tsd_*{p,}_get() can still be used as long as no re-initialization that would require cleanup occurs. This means that e.g. the thread_allocated counter can be updated unconditionally.
2014-10-05 02:12:53 +08:00
tsd = tsd_fetch();
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
tcache_enabled = (tcache_enabled_t)enabled;
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tsd_tcache_enabled_set(tsd, tcache_enabled);
if (!enabled)
tcache_cleanup(tsd);
Add the "thread.tcache.enabled" mallctl.
2012-03-27 09:54:44 +08:00
}
Implement tsd. Implement tsd, which is a TLS/TSD abstraction that uses one or both internally. Modify bootstrapping such that no tsd's are utilized until allocation is safe. Remove malloc_[v]tprintf(), and use malloc_snprintf() instead. Fix %p argument size handling in malloc_vsnprintf(). Fix a long-standing statistics-related bug in the "thread.arena" mallctl that could cause crashes due to linked list corruption.
2012-03-22 09:33:03 +08:00
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE tcache_t *
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tcache_get(tsd_t *tsd, bool create)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
{
tcache_t *tcache;
Convert to uniform style: cond == false --> !cond
2014-10-04 01:16:09 +08:00
if (!config_tcache)
Streamline tcache-related malloc/free fast paths. tcache_get() is inlined, so do the config_tcache check inside tcache_get() and simplify its callers. Make arena_malloc() an inline function, since it is part of the malloc() fast path. Remove conditional logic that cause build issues if --disable-tcache was specified.
2012-02-14 04:29:49 +08:00
return (NULL);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tcache = tsd_tcache_get(tsd);
if (!create)
return (tcache);
Fix tsd cleanup regressions. Fix tsd cleanup regressions that were introduced in 5460aa6f6676c7f253bfcb75c028dfd38cae8aaf (Convert all tsd variables to reside in a single tsd structure.). These regressions were twofold: 1) tsd_tryget() should never (and need never) return NULL. Rename it to tsd_fetch() and simplify all callers. 2) tsd_*_set() must only be called when tsd is in the nominal state, because cleanup happens during the nominal-->purgatory transition, and re-initialization must not happen while in the purgatory state. Add tsd_nominal() and use it as needed. Note that tsd_*{p,}_get() can still be used as long as no re-initialization that would require cleanup occurs. This means that e.g. the thread_allocated counter can be updated unconditionally.
2014-10-05 02:12:53 +08:00
if (unlikely(tcache == NULL) && tsd_nominal(tsd)) {
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
tcache = tcache_get_hard(tsd);
tsd_tcache_set(tsd, tcache);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
return (tcache);
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_event(tsd_t *tsd, tcache_t *tcache)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
{
Remove the lg_tcache_gc_sweep option. Remove the lg_tcache_gc_sweep option, because it is no longer very useful. Prior to the addition of dynamic adjustment of tcache fill count, it was possible for fill/flush overhead to be a problem, but this problem no longer occurs.
2012-03-06 06:34:37 +08:00
if (TCACHE_GC_INCR == 0)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
return;
Use ticker for incremental tcache GC.
2016-02-03 12:37:24 +08:00
if (unlikely(ticker_tick(&tcache->gc_ticker)))
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_event_hard(tsd, tcache);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void *
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
tcache_alloc_easy(tcache_bin_t *tbin, bool *tcache_success)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
{
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
void *ret;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (unlikely(tbin->ncached == 0)) {
Dynamically adjust tcache fill count. Dynamically adjust tcache fill count (number of objects allocated per tcache refill) such that if GC has to flush inactive objects, the fill count gradually decreases. Conversely, if refills occur while the fill count is depressed, the fill count gradually increases back to its maximum value.
2011-03-21 15:18:17 +08:00
tbin->low_water = -1;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
*tcache_success = false;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
return (NULL);
Dynamically adjust tcache fill count. Dynamically adjust tcache fill count (number of objects allocated per tcache refill) such that if GC has to flush inactive objects, the fill count gradually decreases. Conversely, if refills occur while the fill count is depressed, the fill count gradually increases back to its maximum value.
2011-03-21 15:18:17 +08:00
}
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
/*
* tcache_success (instead of ret) should be checked upon the return of
* this function. We avoid checking (ret == NULL) because there is
* never a null stored on the avail stack (which is unknown to the
* compiler), and eagerly checking ret would cause pipeline stall
* (waiting for the cacheline).
*/
*tcache_success = true;
ret = *(tbin->avail - tbin->ncached);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
tbin->ncached--;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (unlikely((int)tbin->ncached < tbin->low_water))
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
tbin->low_water = tbin->ncached;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
return (ret);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void *
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
szind_t binind, bool zero, bool slow_path)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
{
void *ret;
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
tcache_bin_t *tbin;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
bool tcache_success;
size_t usize JEMALLOC_CC_SILENCE_INIT(0);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Simplify small size class infrastructure. Program-generate small size class tables for all valid combinations of LG_TINY_MIN, LG_QUANTUM, and PAGE_SHIFT. Use the appropriate table to generate all relevant data structures, and remove the distinction between tiny/quantum/cacheline/subpage bins. Remove --enable-dynamic-page-shift. This option didn't prove useful in practice, and it prevented optimizations. Add Tilera architecture support.
2012-02-29 08:50:47 +08:00
assert(binind < NBINS);
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
tbin = &tcache->tbins[binind];
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
ret = tcache_alloc_easy(tbin, &tcache_success);
assert(tcache_success == (ret != NULL));
if (unlikely(!tcache_success)) {
bool tcache_hard_success;
arena = arena_choose(tsd, arena);
if (unlikely(arena == NULL))
return (NULL);
ret = tcache_alloc_small_hard(tsd, arena, tcache, tbin, binind,
&tcache_hard_success);
if (tcache_hard_success == false)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
return (NULL);
}
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
assert(ret);
/*
* Only compute usize if required. The checks in the following if
* statement are all static.
*/
if (config_prof || (slow_path && config_fill) || unlikely(zero)) {
usize = index2size(binind);
assert(tcache_salloc(ret) == usize);
}
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Convert to uniform style: cond == false --> !cond
2014-10-04 01:16:09 +08:00
if (likely(!zero)) {
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
if (slow_path && config_fill) {
Introduce two new modes of junk filling: "alloc" and "free". In addition to true/false, opt.junk can now be either "alloc" or "free", giving applications the possibility of junking memory only on allocation or deallocation. This resolves #172.
2014-12-09 05:12:41 +08:00
if (unlikely(opt_junk_alloc)) {
Implement Valgrind support, redzones, and quarantine. Implement Valgrind support, as well as the redzone and quarantine features, which help Valgrind detect memory errors. Redzones are only implemented for small objects because the changes necessary to support redzones around large and huge objects are complicated by in-place reallocation, to the point that it isn't clear that the maintenance burden is worth the incremental improvement to Valgrind support. Merge arena_salloc() and arena_salloc_demote(). Refactor i[v]salloc() to expose the 'demote' option.
2012-04-06 15:35:09 +08:00
arena_alloc_junk_small(ret,
&arena_bin_info[binind], false);
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
} else if (unlikely(opt_zero))
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
memset(ret, 0, usize);
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
}
Implement Valgrind support, redzones, and quarantine. Implement Valgrind support, as well as the redzone and quarantine features, which help Valgrind detect memory errors. Redzones are only implemented for small objects because the changes necessary to support redzones around large and huge objects are complicated by in-place reallocation, to the point that it isn't clear that the maintenance burden is worth the incremental improvement to Valgrind support. Merge arena_salloc() and arena_salloc_demote(). Refactor i[v]salloc() to expose the 'demote' option.
2012-04-06 15:35:09 +08:00
} else {
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
if (slow_path && config_fill && unlikely(opt_junk_alloc)) {
Implement Valgrind support, redzones, and quarantine. Implement Valgrind support, as well as the redzone and quarantine features, which help Valgrind detect memory errors. Redzones are only implemented for small objects because the changes necessary to support redzones around large and huge objects are complicated by in-place reallocation, to the point that it isn't clear that the maintenance burden is worth the incremental improvement to Valgrind support. Merge arena_salloc() and arena_salloc_demote(). Refactor i[v]salloc() to expose the 'demote' option.
2012-04-06 15:35:09 +08:00
arena_alloc_junk_small(ret, &arena_bin_info[binind],
true);
}
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
memset(ret, 0, usize);
Implement Valgrind support, redzones, and quarantine. Implement Valgrind support, as well as the redzone and quarantine features, which help Valgrind detect memory errors. Redzones are only implemented for small objects because the changes necessary to support redzones around large and huge objects are complicated by in-place reallocation, to the point that it isn't clear that the maintenance burden is worth the incremental improvement to Valgrind support. Merge arena_salloc() and arena_salloc_demote(). Refactor i[v]salloc() to expose the 'demote' option.
2012-04-06 15:35:09 +08:00
}
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
if (config_stats)
tbin->tstats.nrequests++;
if (config_prof)
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
tcache->prof_accumbytes += usize;
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_event(tsd, tcache);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
return (ret);
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void *
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
szind_t binind, bool zero, bool slow_path)
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
{
void *ret;
tcache_bin_t *tbin;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
bool tcache_success;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
assert(binind < nhbins);
tbin = &tcache->tbins[binind];
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
ret = tcache_alloc_easy(tbin, &tcache_success);
assert(tcache_success == (ret != NULL));
if (unlikely(!tcache_success)) {
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
/*
* Only allocate one large object at a time, because it's quite
* expensive to create one and not use it.
*/
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
arena = arena_choose(tsd, arena);
if (unlikely(arena == NULL))
return (NULL);
Make *allocx() size class overflow behavior defined. Limit supported size and alignment to HUGE_MAXCLASS, which in turn is now limited to be less than PTRDIFF_MAX. This resolves #278 and #295.
2016-02-26 07:29:49 +08:00
ret = arena_malloc_large(tsd, arena, binind, zero);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
if (ret == NULL)
return (NULL);
} else {
Make *allocx() size class overflow behavior defined. Limit supported size and alignment to HUGE_MAXCLASS, which in turn is now limited to be less than PTRDIFF_MAX. This resolves #278 and #295.
2016-02-26 07:29:49 +08:00
size_t usize JEMALLOC_CC_SILENCE_INIT(0);
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
/* Only compute usize on demand */
Make *allocx() size class overflow behavior defined. Limit supported size and alignment to HUGE_MAXCLASS, which in turn is now limited to be less than PTRDIFF_MAX. This resolves #278 and #295.
2016-02-26 07:29:49 +08:00
if (config_prof || (slow_path && config_fill) ||
unlikely(zero)) {
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
usize = index2size(binind);
assert(usize <= tcache_maxclass);
}
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
if (config_prof && usize == LARGE_MINCLASS) {
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
arena_chunk_t *chunk =
(arena_chunk_t *)CHUNK_ADDR2BASE(ret);
size_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >>
Clean up *PAGE* macros. s/PAGE_SHIFT/LG_PAGE/g and s/PAGE_SIZE/PAGE/g. Remove remnants of the dynamic-page-shift code. Rename the "arenas.pagesize" mallctl to "arenas.page". Remove the "arenas.chunksize" mallctl, which is redundant with "opt.lg_chunk".
2012-04-02 22:04:34 +08:00
LG_PAGE);
Optimize malloc() and free() fast paths. Embed the bin index for small page runs into the chunk page map, in order to omit [...] in the following dependent load sequence: ptr-->mapelm-->[run-->bin-->]bin_info Move various non-critcal code out of the inlined function chain into helper functions (tcache_event_hard(), arena_dalloc_small(), and locking).
2012-05-02 15:30:36 +08:00
arena_mapbits_large_binind_set(chunk, pageind,
BININD_INVALID);
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
}
Convert to uniform style: cond == false --> !cond
2014-10-04 01:16:09 +08:00
if (likely(!zero)) {
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
if (slow_path && config_fill) {
Add JEMALLOC_ALLOC_JUNK and JEMALLOC_FREE_JUNK macros Replace hardcoded 0xa5 and 0x5a junk values with JEMALLOC_ALLOC_JUNK and JEMALLOC_FREE_JUNK macros, respectively.
2016-03-28 14:28:39 +08:00
if (unlikely(opt_junk_alloc)) {
memset(ret, JEMALLOC_ALLOC_JUNK,
usize);
} else if (unlikely(opt_zero))
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
memset(ret, 0, usize);
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
}
Optimize Valgrind integration. Forcefully disable tcache if running inside Valgrind, and remove Valgrind calls in tcache-specific code. Restructure Valgrind-related code to move most Valgrind calls out of the fast path functions. Take advantage of static knowledge to elide some branches in JEMALLOC_VALGRIND_REALLOC().
2014-04-16 07:35:08 +08:00
} else
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
memset(ret, 0, usize);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
if (config_stats)
tbin->tstats.nrequests++;
if (config_prof)
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
tcache->prof_accumbytes += usize;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
}
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_event(tsd, tcache);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
return (ret);
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind,
bool slow_path)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
{
tcache_bin_t *tbin;
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
tcache_bin_info_t *tbin_info;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
assert(tcache_salloc(ptr) <= SMALL_MAXCLASS);
Reduce statistical heap sampling memory overhead. If the mean heap sampling interval is larger than one page, simulate sampled small objects with large objects. This allows profiling context pointers to be omitted for small objects. As a result, the memory overhead for sampling decreases as the sampling interval is increased. Fix a compilation error in the profiling code.
2010-04-01 07:45:04 +08:00
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
if (slow_path && config_fill && unlikely(opt_junk_free))
Implement Valgrind support, redzones, and quarantine. Implement Valgrind support, as well as the redzone and quarantine features, which help Valgrind detect memory errors. Redzones are only implemented for small objects because the changes necessary to support redzones around large and huge objects are complicated by in-place reallocation, to the point that it isn't clear that the maintenance burden is worth the incremental improvement to Valgrind support. Merge arena_salloc() and arena_salloc_demote(). Refactor i[v]salloc() to expose the 'demote' option.
2012-04-06 15:35:09 +08:00
arena_dalloc_junk_small(ptr, &arena_bin_info[binind]);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
tbin = &tcache->tbins[binind];
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
tbin_info = &tcache_bin_info[binind];
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (unlikely(tbin->ncached == tbin_info->ncached_max)) {
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
tcache_bin_flush_small(tsd, tcache, tbin, binind,
(tbin_info->ncached_max >> 1));
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
}
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
assert(tbin->ncached < tbin_info->ncached_max);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
tbin->ncached++;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
*(tbin->avail - tbin->ncached) = ptr;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_event(tsd, tcache);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size,
bool slow_path)
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
{
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
szind_t binind;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
tcache_bin_t *tbin;
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
tcache_bin_info_t *tbin_info;
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
assert((size & PAGE_MASK) == 0);
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
assert(tcache_salloc(ptr) > SMALL_MAXCLASS);
assert(tcache_salloc(ptr) <= tcache_maxclass);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
binind = size2index(size);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
if (slow_path && config_fill && unlikely(opt_junk_free))
Add configure options. Add: --with-lg-page --with-lg-page-sizes --with-lg-size-class-group --with-lg-quantum Get rid of STATIC_PAGE_SHIFT, in favor of directly setting LG_PAGE. Fix various edge conditions exposed by the configure options.
2014-10-10 08:54:06 +08:00
arena_dalloc_junk_large(ptr, size);
Remove medium size classes. Remove medium size classes, because concurrent dirty page purging is no longer capable of purging inactive dirty pages inside active runs (due to recent arena/bin locking changes). Enhance tcache to support caching large objects, so that the same range of size classes is still cached, despite the removal of medium size class support.
2010-03-18 07:27:39 +08:00
tbin = &tcache->tbins[binind];
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
tbin_info = &tcache_bin_info[binind];
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (unlikely(tbin->ncached == tbin_info->ncached_max)) {
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_bin_flush_large(tsd, tbin, binind,
(tbin_info->ncached_max >> 1), tcache);
Simplify tcache object caching. Use chains of cached objects, rather than using arrays of pointers. Since tcache_bin_t is no longer dynamically sized, convert tcache_t's tbin to an array of structures, rather than an array of pointers. This implicitly removes tcache_bin_{create,destroy}(), which further simplifies the fast path for malloc/free. Use cacheline alignment for tcache_t allocations. Remove runtime configuration option for number of tcache bin slots, and replace it with a boolean option for enabling/disabling tcache. Limit the number of tcache objects to the lesser of TCACHE_NSLOTS_MAX and 2X the number of regions per run for the size class. For GC-triggered flush, discard 3/4 of the objects below the low water mark, rather than 1/2.
2010-03-08 07:34:14 +08:00
}
Use bitmaps to track small regions. The previous free list implementation, which embedded singly linked lists in available regions, had the unfortunate side effect of causing many cache misses during thread cache fills. Fix this in two places: - arena_run_t: Use a new bitmap implementation to track which regions are available. Furthermore, revert to preferring the lowest available region (as jemalloc did with its old bitmap-based approach). - tcache_t: Move read-only tcache_bin_t metadata into tcache_bin_info_t, and add a contiguous array of pointers to tcache_t in order to track cached objects. This substantially increases the size of tcache_t, but results in much higher data locality for common tcache operations. As a side benefit, it is again possible to efficiently flush the least recently used cached objects, so this change changes flushing from MRU to LRU. The new bitmap implementation uses a multi-level summary approach to make finding the lowest available region very fast. In practice, bitmaps only have one or two levels, though the implementation is general enough to handle extremely large bitmaps, mainly so that large page sizes can still be entertained. Fix tcache_bin_flush_large() to always flush statistics, in the same way that tcache_bin_flush_small() was recently fixed. Use JEMALLOC_DEBUG rather than NDEBUG. Add dassert(), and use it for debug-only asserts.
2011-03-17 01:30:13 +08:00
assert(tbin->ncached < tbin_info->ncached_max);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
tbin->ncached++;
Fast-path improvement: reduce # of branches and unnecessary operations. - Combine multiple runtime branches into a single malloc_slow check. - Avoid calling arena_choose / size2index / index2size on fast path. - A few micro optimizations.
2015-10-28 06:12:10 +08:00
*(tbin->avail - tbin->ncached) = ptr;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
tcache_event(tsd, tcache);
}
JEMALLOC_ALWAYS_INLINE tcache_t *
tcaches_get(tsd_t *tsd, unsigned ind)
{
tcaches_t *elm = &tcaches[ind];
if (unlikely(elm->tcache == NULL))
elm->tcache = tcache_create(tsd, arena_choose(tsd, NULL));
return (elm->tcache);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
#endif
#endif /* JEMALLOC_H_INLINES */
/******************************************************************************/
Reference in New Issue Copy Permalink