project-base/server-skynet-source-3rd-jemalloc
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
32dca1191c
server-skynet-source-3rd-je.../include/jemalloc/internal/arena.h

1252 lines
38 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
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
#define LARGE_MINCLASS (ZU(1) << LG_LARGE_MINCLASS)
Avoid overflow in arena_run_regind(). Fix a regression due to: Remove an arena_bin_run_size_calc() constraint. 2a6f2af6e446a98a635caadd281a23ca09a491cb The removed constraint required that small run headers fit in one page, which indirectly limited runs such that they would not cause overflow in arena_run_regind(). Add an explicit constraint to arena_bin_run_size_calc() based on the largest number of regions that arena_run_regind() can handle (2^11 as currently configured).
2011-03-23 00:00:56 +08:00
/* Maximum number of regions in one run. */
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
#define LG_RUN_MAXREGS (LG_PAGE - LG_TINY_MIN)
Avoid overflow in arena_run_regind(). Fix a regression due to: Remove an arena_bin_run_size_calc() constraint. 2a6f2af6e446a98a635caadd281a23ca09a491cb The removed constraint required that small run headers fit in one page, which indirectly limited runs such that they would not cause overflow in arena_run_regind(). Add an explicit constraint to arena_bin_run_size_calc() based on the largest number of regions that arena_run_regind() can handle (2^11 as currently configured).
2011-03-23 00:00:56 +08:00
#define RUN_MAXREGS (1U << LG_RUN_MAXREGS)
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
/*
* Minimum redzone size. Redzones may be larger than this if necessary to
* preserve region alignment.
*/
#define REDZONE_MINSIZE 16
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
/*
* The minimum ratio of active:dirty pages per arena is computed as:
*
Implement dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
* (nactive >> lg_dirty_mult) >= ndirty
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 dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
* So, supposing that lg_dirty_mult is 3, there can be no less than 8 times as
* many active pages as dirty pages.
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
*/
Purge unused dirty pages in a fragmentation-reducing order. Purge unused dirty pages in an order that first performs clean/dirty run defragmentation, in order to mitigate available run fragmentation. Remove the limitation that prevented purging unless at least one chunk worth of dirty pages had accumulated in an arena. This limitation was intended to avoid excessive purging for small applications, but the threshold was arbitrary, and the effect of questionable utility. Relax opt_lg_dirty_mult from 5 to 3. This compensates for increased likelihood of allocating clean runs, given the same ratio of clean:dirty runs, and reduces the potential for repeated purging in pathological large malloc/free loops that push the active:dirty page ratio just over the purge threshold.
2012-10-31 06:42:37 +08:00
#define LG_DIRTY_MULT_DEFAULT 3
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
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
typedef struct arena_runs_dirty_link_s arena_runs_dirty_link_t;
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
typedef struct arena_run_s arena_run_t;
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
typedef struct arena_chunk_map_bits_s arena_chunk_map_bits_t;
typedef struct arena_chunk_map_misc_s arena_chunk_map_misc_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 arena_chunk_s arena_chunk_t;
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
typedef struct arena_bin_info_s arena_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 arena_bin_s arena_bin_t;
typedef struct arena_s arena_t;
#endif /* JEMALLOC_H_TYPES */
/******************************************************************************/
#ifdef JEMALLOC_H_STRUCTS
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
#ifdef JEMALLOC_ARENA_STRUCTS_A
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
struct arena_run_s {
Remove arena_dalloc_bin_run() clean page preservation. Remove code in arena_dalloc_bin_run() that preserved the "clean" state of trailing clean pages by splitting them into a separate run during deallocation. This was a useful mechanism for reducing dirty page churn when bin runs comprised many pages, but bin runs are now quite small. Remove the nextind field from arena_run_t now that it is no longer needed, and change arena_run_t's bin field (arena_bin_t *) to binind (index_t). These two changes remove 8 bytes of chunk header overhead per page, which saves 1/512 of all arena chunk memory.
2014-10-11 14:01:03 +08:00
/* Index of bin this run is associated with. */
index_t binind;
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
/* Number of free regions in run. */
unsigned nfree;
/* Per region allocated/deallocated bitmap. */
bitmap_t bitmap[BITMAP_GROUPS_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
/* Each element of the chunk map corresponds to one page within the chunk. */
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
struct arena_chunk_map_bits_s {
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
/*
* Run address (or size) and various flags are stored together. The bit
* layout looks like (assuming 32-bit system):
*
Update a comment.
2012-05-10 15:18:46 +08:00
* ???????? ???????? ????nnnn nnnndula
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
*
* ? : Unallocated: Run address for first/last pages, unset for internal
* pages.
Track dirty and clean runs separately. Split arena->runs_avail into arena->runs_avail_{clean,dirty}, and preferentially allocate dirty runs.
2010-03-19 11:36:40 +08:00
* Small: Run page offset.
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
* Large: Run size for first page, unset for trailing pages.
Update a comment.
2012-05-10 15:18:46 +08:00
* n : binind for small size class, BININD_INVALID for large size class.
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
* d : dirty?
Update various comments.
2010-12-18 10:07:53 +08:00
* u : unzeroed?
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
* l : large?
* a : allocated?
*
* Following are example bit patterns for the three types of runs.
*
* p : run page offset
* s : run size
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
* n : binind for size class; large objects set these to BININD_INVALID
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
* x : don't care
* - : 0
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
* + : 1
Change CHUNK_MAP_ZEROED to CHUNK_MAP_UNZEROED. Invert the chunk map bit that tracks whether a page is zeroed, so that for zeroed arena chunks, the interior of the page map does not need to be initialized (as it consists entirely of zero bytes).
2010-10-02 08:53:37 +08:00
* [DULA] : bit set
* [dula] : bit unset
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
*
Track dirty and clean runs separately. Split arena->runs_avail into arena->runs_avail_{clean,dirty}, and preferentially allocate dirty runs.
2010-03-19 11:36:40 +08:00
* Unallocated (clean):
Update a comment.
2012-05-10 15:18:46 +08:00
* ssssssss ssssssss ssss++++ ++++du-a
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
* xxxxxxxx xxxxxxxx xxxxxxxx xxxx-Uxx
Update a comment.
2012-05-10 15:18:46 +08:00
* ssssssss ssssssss ssss++++ ++++dU-a
Track dirty and clean runs separately. Split arena->runs_avail into arena->runs_avail_{clean,dirty}, and preferentially allocate dirty runs.
2010-03-19 11:36:40 +08:00
*
* Unallocated (dirty):
Update a comment.
2012-05-10 15:18:46 +08:00
* ssssssss ssssssss ssss++++ ++++D--a
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
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
Update a comment.
2012-05-10 15:18:46 +08:00
* ssssssss ssssssss ssss++++ ++++D--a
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
* Small:
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
* pppppppp pppppppp ppppnnnn nnnnd--A
* pppppppp pppppppp ppppnnnn nnnn---A
* pppppppp pppppppp ppppnnnn nnnnd--A
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
*
* Large:
Update a comment.
2012-05-10 15:18:46 +08:00
* ssssssss ssssssss ssss++++ ++++D-LA
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
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
Update a comment.
2012-05-10 15:18:46 +08:00
* -------- -------- ----++++ ++++D-LA
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
*
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
* Large (sampled, size <= LARGE_MINCLASS):
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
* ssssssss ssssssss ssssnnnn nnnnD-LA
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
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
* -------- -------- ----++++ ++++D-LA
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
*
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
* Large (not sampled, size == LARGE_MINCLASS):
Update a comment.
2012-05-10 15:18:46 +08:00
* ssssssss ssssssss ssss++++ ++++D-LA
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
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
* -------- -------- ----++++ ++++D-LA
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
*/
size_t bits;
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
#define CHUNK_MAP_BININD_SHIFT 4
#define BININD_INVALID ((size_t)0xffU)
/* CHUNK_MAP_BININD_MASK == (BININD_INVALID << CHUNK_MAP_BININD_SHIFT) */
#define CHUNK_MAP_BININD_MASK ((size_t)0xff0U)
#define CHUNK_MAP_BININD_INVALID CHUNK_MAP_BININD_MASK
#define CHUNK_MAP_FLAGS_MASK ((size_t)0xcU)
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
#define CHUNK_MAP_DIRTY ((size_t)0x8U)
Change CHUNK_MAP_ZEROED to CHUNK_MAP_UNZEROED. Invert the chunk map bit that tracks whether a page is zeroed, so that for zeroed arena chunks, the interior of the page map does not need to be initialized (as it consists entirely of zero bytes).
2010-10-02 08:53:37 +08:00
#define CHUNK_MAP_UNZEROED ((size_t)0x4U)
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
#define CHUNK_MAP_LARGE ((size_t)0x2U)
#define CHUNK_MAP_ALLOCATED ((size_t)0x1U)
#define CHUNK_MAP_KEY CHUNK_MAP_ALLOCATED
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
};
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
struct arena_runs_dirty_link_s {
qr(arena_runs_dirty_link_t) rd_link;
};
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
/*
* Each arena_chunk_map_misc_t corresponds to one page within the chunk, just
* like arena_chunk_map_bits_t. Two separate arrays are stored within each
* chunk header in order to improve cache locality.
*/
struct arena_chunk_map_misc_s {
/*
* Linkage for run trees. There are two disjoint uses:
*
* 1) arena_t's runs_avail tree.
* 2) arena_run_t conceptually uses this linkage for in-use non-full
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
* runs, rather than directly embedding linkage.
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
*/
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
rb_node(arena_chunk_map_misc_t) rb_link;
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
union {
/* Linkage for list of dirty runs. */
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
arena_runs_dirty_link_t rd;
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
/* Profile counters, used for large object runs. */
Fix type punning in calls to atomic operation functions.
2015-05-08 13:35:40 +08:00
union {
void *prof_tctx_pun;
prof_tctx_t *prof_tctx;
};
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
/* Small region run metadata. */
arena_run_t run;
};
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
};
typedef rb_tree(arena_chunk_map_misc_t) arena_avail_tree_t;
typedef rb_tree(arena_chunk_map_misc_t) arena_run_tree_t;
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
#endif /* JEMALLOC_ARENA_STRUCTS_A */
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
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
#ifdef JEMALLOC_ARENA_STRUCTS_B
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
/* Arena chunk header. */
struct arena_chunk_s {
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
/*
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
* A pointer to the arena that owns the chunk is stored within the node.
* This field as a whole is used by chunks_rtree to support both
* ivsalloc() and core-based debugging.
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
*/
extent_node_t node;
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
Omit chunk header in arena chunk map. Omit the first map_bias elements of the map in arena_chunk_t. This avoids barely spilling over into an extra chunk header page for common chunk sizes.
2010-10-02 08:35:43 +08:00
/*
* Map of pages within chunk that keeps track of free/large/small. The
* first map_bias entries are omitted, since the chunk header does not
* need to be tracked in the map. This omission saves a header page
* for common chunk sizes (e.g. 4 MiB).
*/
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
arena_chunk_map_bits_t map_bits[1]; /* Dynamically sized. */
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
};
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +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
* Read-only information associated with each element of arena_t's bins array
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
* is stored separately, partly to reduce memory usage (only one copy, rather
* than one per arena), but mainly to avoid false cacheline sharing.
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
*
* Each run has the following layout:
*
* /--------------------\
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
* | pad? |
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
* |--------------------|
* | redzone |
* reg0_offset | region 0 |
* | redzone |
* |--------------------| \
* | redzone | |
* | region 1 | > reg_interval
* | redzone | /
* |--------------------|
* | ... |
* | ... |
* | ... |
* |--------------------|
* | redzone |
* | region nregs-1 |
* | redzone |
* |--------------------|
* | alignment pad? |
* \--------------------/
*
* reg_interval has at least the same minimum alignment as reg_size; this
* preserves the alignment constraint that sa2u() depends on. Alignment pad is
* either 0 or redzone_size; it is present only if needed to align reg0_offset.
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
*/
struct arena_bin_info_s {
/* Size of regions in a run for this bin's size class. */
size_t reg_size;
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
/* Redzone size. */
size_t redzone_size;
/* Interval between regions (reg_size + (redzone_size << 1)). */
size_t reg_interval;
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
/* Total size of a run for this bin's size class. */
size_t run_size;
/* Total number of regions in a run for this bin's size class. */
uint32_t nregs;
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
/*
* Metadata used to manipulate bitmaps for runs associated with this
* bin.
*/
bitmap_info_t bitmap_info;
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
/* Offset of first region in a run for this bin's size class. */
uint32_t reg0_offset;
};
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 arena_bin_s {
Push locks into arena bins. For bin-related allocation, protect data structures with bin locks rather than arena locks. Arena locks remain for run allocation/deallocation and other miscellaneous operations. Restructure statistics counters to maintain per bin allocated/nmalloc/ndalloc, but continue to provide arena-wide statistics via aggregation in the ctl code.
2010-03-14 12:32:56 +08:00
/*
* All operations on runcur, runs, and stats require that lock be
* locked. Run allocation/deallocation are protected by the arena lock,
* which may be acquired while holding one or more bin locks, but not
* vise versa.
*/
malloc_mutex_t lock;
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
/*
* Current run being used to service allocations of this bin's size
* class.
*/
arena_run_t *runcur;
/*
* Tree of non-full runs. This tree is used when looking for an
* existing run when runcur is no longer usable. We choose the
* non-full run that is lowest in memory; this policy tends to keep
* objects packed well, and it can also help reduce the number of
* almost-empty chunks.
*/
arena_run_tree_t runs;
/* Bin statistics. */
malloc_bin_stats_t stats;
};
struct arena_s {
Implement allocation profiling and leack checking. Add the --enable-prof and --enable-prof-libunwind configure options. Add the B/b, F/f, I/i, L/l, and U/u JEMALLOC_OPTIONS. Interval-based profile dump triggering is not yet implemented. Add supporting generic code: * Add memory barriers. * Add prn (LCG PRNG). * Add hash (Murmur hash function). * Add ckh (cuckoo hash tables).
2010-02-11 02:37:56 +08:00
/* This arena's index within the arenas array. */
unsigned ind;
Push locks into arena bins. For bin-related allocation, protect data structures with bin locks rather than arena locks. Arena locks remain for run allocation/deallocation and other miscellaneous operations. Restructure statistics counters to maintain per bin allocated/nmalloc/ndalloc, but continue to provide arena-wide statistics via aggregation in the ctl code.
2010-03-14 12:32:56 +08:00
/*
Improve thread-->arena assignment. Rather than blindly assigning threads to arenas in round-robin fashion, choose the lowest-numbered arena that currently has the smallest number of threads assigned to it. Add the "stats.arenas.<i>.nthreads" mallctl.
2011-03-19 04:41:33 +08:00
* Number of threads currently assigned to this arena. This field is
* protected by arenas_lock.
*/
unsigned nthreads;
/*
* There are three classes of arena operations from a locking
* perspective:
Style and spelling fixes.
2014-12-09 06:40:14 +08:00
* 1) Thread assignment (modifies nthreads) is protected by arenas_lock.
Improve thread-->arena assignment. Rather than blindly assigning threads to arenas in round-robin fashion, choose the lowest-numbered arena that currently has the smallest number of threads assigned to it. Add the "stats.arenas.<i>.nthreads" mallctl.
2011-03-19 04:41:33 +08:00
* 2) Bin-related operations are protected by bin locks.
* 3) Chunk- and run-related operations are protected by this mutex.
Push locks into arena bins. For bin-related allocation, protect data structures with bin locks rather than arena locks. Arena locks remain for run allocation/deallocation and other miscellaneous operations. Restructure statistics counters to maintain per bin allocated/nmalloc/ndalloc, but continue to provide arena-wide statistics via aggregation in the ctl code.
2010-03-14 12:32:56 +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
malloc_mutex_t lock;
arena_stats_t stats;
/*
* List of tcaches for extant threads associated with this 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
* Stats from these are merged incrementally, and at exit if
* opt_stats_print is enabled.
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
*/
ql_head(tcache_t) tcache_ql;
Implement interval-based heap profile dumping. Add mallctl interfaces for profiling parameters. Fix a file descriptor leak in heap profile dumping.
2010-02-12 05:19:21 +08:00
uint64_t prof_accumbytes;
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
/*
* PRNG state for cache index randomization of large allocation base
* pointers.
*/
uint64_t offset_state;
Add arena-specific and selective dss allocation. Add the "arenas.extend" mallctl, so that it is possible to create new arenas that are outside the set that jemalloc automatically multiplexes threads onto. Add the ALLOCM_ARENA() flag for {,r,d}allocm(), so that it is possible to explicitly allocate from a particular arena. Add the "opt.dss" mallctl, which controls the default precedence of dss allocation relative to mmap allocation. Add the "arena.<i>.dss" mallctl, which makes it possible to set the default dss precedence on a per arena or global basis. Add the "arena.<i>.purge" mallctl, which obsoletes "arenas.purge". Add the "stats.arenas.<i>.dss" mallctl.
2012-10-12 04:53:15 +08:00
dss_prec_t dss_prec;
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
/*
* In order to avoid rapid chunk allocation/deallocation when an arena
* oscillates right on the cusp of needing a new chunk, cache the most
* recently freed chunk. The spare is left in the arena's chunk trees
* until it is deleted.
*
* There is one spare chunk per arena, rather than one spare total, in
* order to avoid interactions between multiple threads that could make
* a single spare inadequate.
*/
arena_chunk_t *spare;
Implement dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
/* Minimum ratio (log base 2) of nactive:ndirty. */
ssize_t lg_dirty_mult;
Convert arena_maybe_purge() recursion to iteration. This resolves #235.
2015-06-23 09:50:32 +08:00
/* True if a thread is currently executing arena_purge(). */
bool purging;
Refactor huge allocation to be managed by arenas. Refactor huge allocation to be managed by arenas (though the global red-black tree of huge allocations remains for lookup during deallocation). This is the logical conclusion of recent changes that 1) made per arena dss precedence apply to huge allocation, and 2) made it possible to replace the per arena chunk allocation/deallocation functions. Remove the top level huge stats, and replace them with per arena huge stats. Normalize function names and types to *dalloc* (some were *dealloc*). Remove the --enable-mremap option. As jemalloc currently operates, this is a performace regression for some applications, but planned work to logarithmically space huge size classes should provide similar amortized performance. The motivation for this change was that mremap-based huge reallocation forced leaky abstractions that prevented refactoring.
2014-05-16 13:22:27 +08:00
/* Number of pages in active runs and huge regions. */
Various minor cleanups. Clean up whitespace. Lock access of swap_avail when printing stats. Use inttypes.h for portable printf() format specifiers, specifically for uint64_t (PRIu64). Change highchunks and curchunks stats from (unsigned long) to (size_t).
2010-01-25 08:41:01 +08:00
size_t nactive;
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
/*
* Current count of pages within unused runs that are potentially
* dirty, and for which madvise(... MADV_DONTNEED) has not been called.
* By tracking this, we can institute a limit on how much dirty unused
* memory is mapped for each arena.
*/
size_t ndirty;
/*
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
* Size/address-ordered tree of this arena's available runs. The tree
Use first-fit rather than first-best-fit run/chunk allocation. This tends to more effectively pack active memory toward low addresses. However, additional tree searches are required in many cases, so whether this change stands the test of time will depend on real-world benchmarks.
2015-03-07 11:57:36 +08:00
* is used for first-fit run allocation.
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
*/
Purge unused dirty pages in a fragmentation-reducing order. Purge unused dirty pages in an order that first performs clean/dirty run defragmentation, in order to mitigate available run fragmentation. Remove the limitation that prevented purging unless at least one chunk worth of dirty pages had accumulated in an arena. This limitation was intended to avoid excessive purging for small applications, but the threshold was arbitrary, and the effect of questionable utility. Relax opt_lg_dirty_mult from 5 to 3. This compensates for increased likelihood of allocating clean runs, given the same ratio of clean:dirty runs, and reduces the potential for repeated purging in pathological large malloc/free loops that push the active:dirty page ratio just over the purge threshold.
2012-10-31 06:42:37 +08:00
arena_avail_tree_t runs_avail;
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
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
/*
* Unused dirty memory this arena manages. Dirty memory is conceptually
Rename "dirty chunks" to "cached chunks". Rename "dirty chunks" to "cached chunks", in order to avoid overloading the term "dirty". Fix the regression caused by 339c2b23b2d61993ac768afcc72af135662c6771 (Fix chunk_unmap() to propagate dirty state.), and actually address what that change attempted, which is to only purge chunks once, and propagate whether zeroed pages resulted into chunk_record().
2015-02-18 17:15:50 +08:00
* tracked as an arbitrarily interleaved LRU of dirty runs and cached
* chunks, but the list linkage is actually semi-duplicated in order to
* avoid extra arena_chunk_map_misc_t space overhead.
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
*
* LRU-----------------------------------------------------------MRU
*
Normalize rdelm/rd structure field naming.
2015-03-11 09:29:49 +08:00
* /-- arena ---\
* | |
* | |
* |------------| /- chunk -\
* ...->|chunks_cache|<--------------------------->| /----\ |<--...
* |------------| | |node| |
* | | | | | |
* | | /- run -\ /- run -\ | | | |
* | | | | | | | | | |
* | | | | | | | | | |
* |------------| |-------| |-------| | |----| |
* ...->|runs_dirty |<-->|rd |<-->|rd |<---->|rd |<----...
* |------------| |-------| |-------| | |----| |
* | | | | | | | | | |
* | | | | | | | \----/ |
* | | \-------/ \-------/ | |
* | | | |
* | | | |
* \------------/ \---------/
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
*/
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
arena_runs_dirty_link_t runs_dirty;
Rename "dirty chunks" to "cached chunks". Rename "dirty chunks" to "cached chunks", in order to avoid overloading the term "dirty". Fix the regression caused by 339c2b23b2d61993ac768afcc72af135662c6771 (Fix chunk_unmap() to propagate dirty state.), and actually address what that change attempted, which is to only purge chunks once, and propagate whether zeroed pages resulted into chunk_record().
2015-02-18 17:15:50 +08:00
extent_node_t chunks_cache;
Fix and refactor runs_dirty-based purging. Fix runs_dirty-based purging to also purge dirty pages in the spare chunk. Refactor runs_dirty manipulation into arena_dirty_{insert,remove}(), and move the arena->ndirty accounting into those functions. Remove the u.ql_link field from arena_chunk_map_t, and get rid of the enclosing union for u.rb_link, since only rb_link remains. Remove the ndirty field from arena_chunk_t.
2014-08-15 05:45:58 +08:00
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
/* Extant huge allocations. */
ql_head(extent_node_t) huge;
/* Synchronizes all huge allocation/update/deallocation. */
malloc_mutex_t huge_mtx;
/*
* Trees of chunks that were previously allocated (trees differ only in
* node ordering). These are used when allocating chunks, in an attempt
* to re-use address space. Depending on function, different tree
* orderings are needed, which is why there are two trees with the same
* contents.
*/
Rename "dirty chunks" to "cached chunks". Rename "dirty chunks" to "cached chunks", in order to avoid overloading the term "dirty". Fix the regression caused by 339c2b23b2d61993ac768afcc72af135662c6771 (Fix chunk_unmap() to propagate dirty state.), and actually address what that change attempted, which is to only purge chunks once, and propagate whether zeroed pages resulted into chunk_record().
2015-02-18 17:15:50 +08:00
extent_tree_t chunks_szad_cache;
extent_tree_t chunks_ad_cache;
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
extent_tree_t chunks_szad_mmap;
extent_tree_t chunks_ad_mmap;
extent_tree_t chunks_szad_dss;
extent_tree_t chunks_ad_dss;
malloc_mutex_t chunks_mtx;
/* Cache of nodes that were allocated via base_alloc(). */
ql_head(extent_node_t) node_cache;
malloc_mutex_t node_cache_mtx;
Add support for user-specified chunk allocators/deallocators. Add new mallctl endpoints "arena<i>.chunk.alloc" and "arena<i>.chunk.dealloc" to allow userspace to configure jemalloc's chunk allocator and deallocator on a per-arena basis.
2014-05-06 06:16:56 +08:00
/*
Implement dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
* User-configurable chunk allocation/deallocation/purge functions.
Add support for user-specified chunk allocators/deallocators. Add new mallctl endpoints "arena<i>.chunk.alloc" and "arena<i>.chunk.dealloc" to allow userspace to configure jemalloc's chunk allocator and deallocator on a per-arena basis.
2014-05-06 06:16:56 +08:00
*/
chunk_alloc_t *chunk_alloc;
Refactor huge allocation to be managed by arenas. Refactor huge allocation to be managed by arenas (though the global red-black tree of huge allocations remains for lookup during deallocation). This is the logical conclusion of recent changes that 1) made per arena dss precedence apply to huge allocation, and 2) made it possible to replace the per arena chunk allocation/deallocation functions. Remove the top level huge stats, and replace them with per arena huge stats. Normalize function names and types to *dalloc* (some were *dealloc*). Remove the --enable-mremap option. As jemalloc currently operates, this is a performace regression for some applications, but planned work to logarithmically space huge size classes should provide similar amortized performance. The motivation for this change was that mremap-based huge reallocation forced leaky abstractions that prevented refactoring.
2014-05-16 13:22:27 +08:00
chunk_dalloc_t *chunk_dalloc;
Implement dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
chunk_purge_t *chunk_purge;
Add support for user-specified chunk allocators/deallocators. Add new mallctl endpoints "arena<i>.chunk.alloc" and "arena<i>.chunk.dealloc" to allow userspace to configure jemalloc's chunk allocator and deallocator on a per-arena basis.
2014-05-06 06:16:56 +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
/* bins is used to store trees of free regions. */
arena_bin_t bins[NBINS];
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
};
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
#endif /* JEMALLOC_ARENA_STRUCTS_B */
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
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
static const size_t large_pad =
#ifdef JEMALLOC_CACHE_OBLIVIOUS
PAGE
#else
0
#endif
;
extern ssize_t opt_lg_dirty_mult;
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
extern arena_bin_info_t arena_bin_info[NBINS];
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
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
extern size_t map_bias; /* Number of arena chunk header pages. */
extern size_t map_misc_offset;
extern size_t arena_maxrun; /* Max run size for arenas. */
extern size_t arena_maxclass; /* Max size class for arenas. */
Add per size class huge allocation statistics. Add per size class huge allocation statistics, and normalize various stats: - Change the arenas.nlruns type from size_t to unsigned. - Add the arenas.nhchunks and arenas.hchunks.<i>.size mallctl's. - Replace the stats.arenas.<i>.bins.<j>.allocated mallctl with stats.arenas.<i>.bins.<j>.curregs . - Add the stats.arenas.<i>.hchunks.<j>.nmalloc, stats.arenas.<i>.hchunks.<j>.ndalloc, stats.arenas.<i>.hchunks.<j>.nrequests, and stats.arenas.<i>.hchunks.<j>.curhchunks mallctl's.
2014-10-13 13:53:59 +08:00
extern unsigned nlclasses; /* Number of large size classes. */
extern unsigned nhclasses; /* Number of huge size classes. */
Implement mallctl{nametomib,bymib}(). Replace chunk stats code that was missing locking; this fixes a race condition that could corrupt chunk statistics. Converting malloc_stats_print() to use mallctl*(). Add a missing semicolon in th DSS code. Convert malloc_tcache_flush() to a mallctl. Convert malloc_swap_enable() to a set of mallctl's.
2010-01-28 05:10:55 +08:00
Rename "dirty chunks" to "cached chunks". Rename "dirty chunks" to "cached chunks", in order to avoid overloading the term "dirty". Fix the regression caused by 339c2b23b2d61993ac768afcc72af135662c6771 (Fix chunk_unmap() to propagate dirty state.), and actually address what that change attempted, which is to only purge chunks once, and propagate whether zeroed pages resulted into chunk_record().
2015-02-18 17:15:50 +08:00
void arena_chunk_cache_maybe_insert(arena_t *arena, extent_node_t *node,
bool cache);
void arena_chunk_cache_maybe_remove(arena_t *arena, extent_node_t *node,
bool cache);
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
extent_node_t *arena_node_alloc(arena_t *arena);
void arena_node_dalloc(arena_t *arena, extent_node_t *node);
Fix huge allocation statistics.
2014-10-15 13:20:00 +08:00
void *arena_chunk_alloc_huge(arena_t *arena, size_t usize, size_t alignment,
bool *zero);
Add per size class huge allocation statistics. Add per size class huge allocation statistics, and normalize various stats: - Change the arenas.nlruns type from size_t to unsigned. - Add the arenas.nhchunks and arenas.hchunks.<i>.size mallctl's. - Replace the stats.arenas.<i>.bins.<j>.allocated mallctl with stats.arenas.<i>.bins.<j>.curregs . - Add the stats.arenas.<i>.hchunks.<j>.nmalloc, stats.arenas.<i>.hchunks.<j>.ndalloc, stats.arenas.<i>.hchunks.<j>.nrequests, and stats.arenas.<i>.hchunks.<j>.curhchunks mallctl's.
2014-10-13 13:53:59 +08:00
void arena_chunk_dalloc_huge(arena_t *arena, void *chunk, size_t usize);
Fix huge allocation statistics.
2014-10-15 13:20:00 +08:00
void arena_chunk_ralloc_huge_similar(arena_t *arena, void *chunk,
size_t oldsize, size_t usize);
void arena_chunk_ralloc_huge_shrink(arena_t *arena, void *chunk,
size_t oldsize, size_t usize);
bool arena_chunk_ralloc_huge_expand(arena_t *arena, void *chunk,
size_t oldsize, size_t usize, bool *zero);
Implement dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
ssize_t arena_lg_dirty_mult_get(arena_t *arena);
bool arena_lg_dirty_mult_set(arena_t *arena, ssize_t lg_dirty_mult);
Fix chunk cache races. These regressions were introduced by ee41ad409a43d12900a5a3108f6c14f84e4eb0eb (Integrate whole chunks into unused dirty page purging machinery.).
2015-02-19 08:40:53 +08:00
void arena_maybe_purge(arena_t *arena);
Add the "arenas.purge" mallctl.
2010-10-01 07:55:08 +08:00
void arena_purge_all(arena_t *arena);
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 arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin,
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
index_t binind, uint64_t prof_accumbytes);
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
void arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info,
bool zero);
Add quarantine unit tests. Verify that freed regions are quarantined, and that redzone corruption is detected. Introduce a testing idiom for intercepting/replacing internal functions. In this case the replaced function is ordinarily a static function, but the idiom should work similarly for library-private functions.
2013-12-18 07:14:36 +08:00
#ifdef JEMALLOC_JET
typedef void (arena_redzone_corruption_t)(void *, size_t, bool, size_t,
uint8_t);
Add junk/zero filling unit tests, and fix discovered bugs. Fix growing large reallocation to junk fill new space. Fix huge deallocation to junk fill when munmap is disabled.
2014-01-08 08:47:56 +08:00
extern arena_redzone_corruption_t *arena_redzone_corruption;
typedef void (arena_dalloc_junk_small_t)(void *, arena_bin_info_t *);
extern arena_dalloc_junk_small_t *arena_dalloc_junk_small;
#else
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
void arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info);
Add junk/zero filling unit tests, and fix discovered bugs. Fix growing large reallocation to junk fill new space. Fix huge deallocation to junk fill when munmap is disabled.
2014-01-08 08:47:56 +08:00
#endif
Add quarantine unit tests. Verify that freed regions are quarantined, and that redzone corruption is detected. Introduce a testing idiom for intercepting/replacing internal functions. In this case the replaced function is ordinarily a static function, but the idiom should work similarly for library-private functions.
2013-12-18 07:14:36 +08:00
void arena_quarantine_junk_small(void *ptr, size_t 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
void *arena_malloc_small(arena_t *arena, size_t size, bool 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
void *arena_malloc_large(arena_t *arena, size_t size, bool zero);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
void *arena_palloc(tsd_t *tsd, arena_t *arena, size_t usize,
size_t alignment, bool zero, tcache_t *tcache);
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
void arena_prof_promoted(const void *ptr, size_t size);
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
void arena_dalloc_bin_junked_locked(arena_t *arena, arena_chunk_t *chunk,
void *ptr, arena_chunk_map_bits_t *bitselm);
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
void arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr,
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
size_t pageind, arena_chunk_map_bits_t *bitselm);
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
void arena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr,
size_t pageind);
Add junk/zero filling unit tests, and fix discovered bugs. Fix growing large reallocation to junk fill new space. Fix huge deallocation to junk fill when munmap is disabled.
2014-01-08 08:47:56 +08:00
#ifdef JEMALLOC_JET
typedef void (arena_dalloc_junk_large_t)(void *, size_t);
extern arena_dalloc_junk_large_t *arena_dalloc_junk_large;
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
#else
void arena_dalloc_junk_large(void *ptr, size_t usize);
Add junk/zero filling unit tests, and fix discovered bugs. Fix growing large reallocation to junk fill new space. Fix huge deallocation to junk fill when munmap is disabled.
2014-01-08 08:47:56 +08:00
#endif
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
void arena_dalloc_large_junked_locked(arena_t *arena, arena_chunk_t *chunk,
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
void *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
void arena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr);
Add junk/zero filling unit tests, and fix discovered bugs. Fix growing large reallocation to junk fill new space. Fix huge deallocation to junk fill when munmap is disabled.
2014-01-08 08:47:56 +08:00
#ifdef JEMALLOC_JET
typedef void (arena_ralloc_junk_large_t)(void *, size_t, size_t);
extern arena_ralloc_junk_large_t *arena_ralloc_junk_large;
#endif
Extract profiling code from [re]allocation functions. Extract profiling code from malloc(), imemalign(), calloc(), realloc(), mallocx(), rallocx(), and xallocx(). This slightly reduces the amount of code compiled into the fast paths, but the primary benefit is the combinatorial complexity reduction. Simplify iralloc[t]() by creating a separate ixalloc() that handles the no-move cases. Further simplify [mrxn]allocx() (and by implication [mrn]allocm()) to make request size overflows due to size class and/or alignment constraints trigger undefined behavior (detected by debug-only assertions). Report ENOMEM rather than EINVAL if an OOM occurs during heap profiling backtrace creation in imemalign(). This bug impacted posix_memalign() and aligned_alloc().
2014-01-13 07:05:44 +08:00
bool arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size,
Add {,r,s,d}allocm(). Add allocm(), rallocm(), sallocm(), and dallocm(), which are a functional superset of malloc(), calloc(), posix_memalign(), malloc_usable_size(), and free().
2010-09-18 06:46:18 +08:00
size_t extra, bool zero);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
void *arena_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
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
size_t size, size_t extra, size_t alignment, bool zero, tcache_t *tcache);
Add arena-specific and selective dss allocation. Add the "arenas.extend" mallctl, so that it is possible to create new arenas that are outside the set that jemalloc automatically multiplexes threads onto. Add the ALLOCM_ARENA() flag for {,r,d}allocm(), so that it is possible to explicitly allocate from a particular arena. Add the "opt.dss" mallctl, which controls the default precedence of dss allocation relative to mmap allocation. Add the "arena.<i>.dss" mallctl, which makes it possible to set the default dss precedence on a per arena or global basis. Add the "arena.<i>.purge" mallctl, which obsoletes "arenas.purge". Add the "stats.arenas.<i>.dss" mallctl.
2012-10-12 04:53:15 +08:00
dss_prec_t arena_dss_prec_get(arena_t *arena);
Make dss non-optional, and fix an "arena.<i>.dss" mallctl bug. Make dss non-optional on all platforms which support sbrk(2). Fix the "arena.<i>.dss" mallctl to return an error if "primary" or "secondary" precedence is specified, but sbrk(2) is not supported.
2014-04-16 03:09:48 +08:00
bool arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);
Implement dynamic per arena control over dirty page purging. Add mallctls: - arenas.lg_dirty_mult is initialized via opt.lg_dirty_mult, and can be modified to change the initial lg_dirty_mult setting for newly created arenas. - arena.<i>.lg_dirty_mult controls an individual arena's dirty page purging threshold, and synchronously triggers any purging that may be necessary to maintain the constraint. - arena.<i>.chunk.purge allows the per arena dirty page purging function to be replaced. This resolves #93.
2015-03-19 09:55:33 +08:00
ssize_t arena_lg_dirty_mult_default_get(void);
bool arena_lg_dirty_mult_default_set(ssize_t lg_dirty_mult);
Add the "stats.arenas.<i>.lg_dirty_mult" mallctl.
2015-03-25 07:36:12 +08:00
void arena_stats_merge(arena_t *arena, const char **dss,
ssize_t *lg_dirty_mult, size_t *nactive, size_t *ndirty,
arena_stats_t *astats, malloc_bin_stats_t *bstats,
Add per size class huge allocation statistics. Add per size class huge allocation statistics, and normalize various stats: - Change the arenas.nlruns type from size_t to unsigned. - Add the arenas.nhchunks and arenas.hchunks.<i>.size mallctl's. - Replace the stats.arenas.<i>.bins.<j>.allocated mallctl with stats.arenas.<i>.bins.<j>.curregs . - Add the stats.arenas.<i>.hchunks.<j>.nmalloc, stats.arenas.<i>.hchunks.<j>.ndalloc, stats.arenas.<i>.hchunks.<j>.nrequests, and stats.arenas.<i>.hchunks.<j>.curhchunks mallctl's.
2014-10-13 13:53:59 +08:00
malloc_large_stats_t *lstats, malloc_huge_stats_t *hstats);
Refactor/fix arenas manipulation. Abstract arenas access to use arena_get() (or a0get() where appropriate) rather than directly reading e.g. arenas[ind]. Prior to the addition of the arenas.extend mallctl, the worst possible outcome of directly accessing arenas was a stale read, but arenas.extend may allocate and assign a new array to arenas. Add a tsd-based arenas_cache, which amortizes arenas reads. This introduces some subtle bootstrapping issues, with tsd_boot() now being split into tsd_boot[01]() to support tsd wrapper allocation bootstrapping, as well as an arenas_cache_bypass tsd variable which dynamically terminates allocation of arenas_cache itself. Promote a0malloc(), a0calloc(), and a0free() to be generally useful for internal allocation, and use them in several places (more may be appropriate). Abstract arena->nthreads management and fix a missing decrement during thread destruction (recent tsd refactoring left arenas_cleanup() unused). Change arena_choose() to propagate OOM, and handle OOM in all callers. This is important for providing consistent allocation behavior when the MALLOCX_ARENA() flag is being used. Prior to this fix, it was possible for an OOM to result in allocation silently allocating from a different arena than the one specified.
2014-10-08 14:14:57 +08:00
arena_t *arena_new(unsigned ind);
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
bool arena_boot(void);
Fix fork-related bugs. Acquire/release arena bin locks as part of the prefork/postfork. This bug made deadlock in the child between fork and exec a possibility. Split jemalloc_postfork() into jemalloc_postfork_{parent,child}() so that the child can reinitialize mutexes rather than unlocking them. In practice, this bug tended not to cause problems.
2012-03-14 07:31:41 +08:00
void arena_prefork(arena_t *arena);
void arena_postfork_parent(arena_t *arena);
void arena_postfork_child(arena_t *arena);
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
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
arena_chunk_map_bits_t *arena_bitselm_get(arena_chunk_t *chunk,
size_t pageind);
arena_chunk_map_misc_t *arena_miscelm_get(arena_chunk_t *chunk,
size_t pageind);
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
size_t arena_miscelm_to_pageind(arena_chunk_map_misc_t *miscelm);
void *arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm);
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
arena_chunk_map_misc_t *arena_rd_to_miscelm(arena_runs_dirty_link_t *rd);
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
arena_chunk_map_misc_t *arena_run_to_miscelm(arena_run_t *run);
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
size_t *arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind);
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t arena_mapbitsp_read(size_t *mapbitsp);
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
size_t arena_mapbits_get(arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_unallocated_size_get(arena_chunk_t *chunk,
size_t pageind);
size_t arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind);
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
index_t arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind);
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
size_t arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind);
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
void arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits);
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
void arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind,
size_t size, size_t flags);
void arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
size_t size);
void arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind,
size_t size, size_t flags);
void arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
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
index_t binind);
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
void arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind,
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
size_t runind, index_t binind, size_t flags);
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
void arena_mapbits_unzeroed_set(arena_chunk_t *chunk, size_t pageind,
size_t unzeroed);
Implement metadata statistics. There are three categories of metadata: - Base allocations are used for bootstrap-sensitive internal allocator data structures. - Arena chunk headers comprise pages which track the states of the non-metadata pages. - Internal allocations differ from application-originated allocations in that they are for internal use, and that they are omitted from heap profiles. The metadata statistics comprise the metadata categories as follows: - stats.metadata: All metadata -- base + arena chunk headers + internal allocations. - stats.arenas.<i>.metadata.mapped: Arena chunk headers. - stats.arenas.<i>.metadata.allocated: Internal allocations. This is reported separately from the other metadata statistics because it overlaps with the allocated and active statistics, whereas the other metadata statistics do not. Base allocations are not reported separately, though their magnitude can be computed by subtracting the arena-specific metadata. This resolves #163.
2014-11-28 03:22:36 +08:00
void arena_metadata_allocated_add(arena_t *arena, size_t size);
void arena_metadata_allocated_sub(arena_t *arena, size_t size);
size_t arena_metadata_allocated_get(arena_t *arena);
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
bool arena_prof_accum_impl(arena_t *arena, uint64_t accumbytes);
bool arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes);
bool arena_prof_accum(arena_t *arena, uint64_t accumbytes);
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
index_t arena_ptr_small_binind_get(const void *ptr, size_t mapbits);
index_t arena_bin_index(arena_t *arena, arena_bin_t *bin);
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
unsigned arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info,
Clean up after arena_bin_info_t change. Fix a couple of problems related to the addition of arena_bin_info_t.
2011-03-16 13:19:45 +08:00
const void *ptr);
Implement per thread heap profiling. Rename data structures (prof_thr_cnt_t-->prof_tctx_t, prof_ctx_t-->prof_gctx_t), and convert to storing a prof_tctx_t for sampled objects. Convert PROF_ALLOC_PREP() to prof_alloc_prep(), since precise backtrace depth within jemalloc functions is no longer an issue (pprof prunes irrelevant frames). Implement mallctl's: - prof.reset implements full sample data reset, and optional change of sample interval. - prof.lg_sample reads the current sample interval (opt.lg_prof_sample was the permanent source of truth prior to prof.reset). - thread.prof.name provides naming capability for threads within heap profile dumps. - thread.prof.active makes it possible to activate/deactivate heap profiling for individual threads. Modify the heap dump files to contain per thread heap profile data. This change is incompatible with the existing pprof, which will require enhancements to read and process the enriched data.
2014-08-19 07:22:13 +08:00
prof_tctx_t *arena_prof_tctx_get(const void *ptr);
void arena_prof_tctx_set(const void *ptr, prof_tctx_t *tctx);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
void *arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
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 *tcache);
Implement metadata statistics. There are three categories of metadata: - Base allocations are used for bootstrap-sensitive internal allocator data structures. - Arena chunk headers comprise pages which track the states of the non-metadata pages. - Internal allocations differ from application-originated allocations in that they are for internal use, and that they are omitted from heap profiles. The metadata statistics comprise the metadata categories as follows: - stats.metadata: All metadata -- base + arena chunk headers + internal allocations. - stats.arenas.<i>.metadata.mapped: Arena chunk headers. - stats.arenas.<i>.metadata.allocated: Internal allocations. This is reported separately from the other metadata statistics because it overlaps with the allocated and active statistics, whereas the other metadata statistics do not. Base allocations are not reported separately, though their magnitude can be computed by subtracting the arena-specific metadata. This resolves #163.
2014-11-28 03:22:36 +08:00
arena_t *arena_aalloc(const void *ptr);
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
size_t arena_salloc(const void *ptr, bool demote);
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
void arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
void arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, 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
#endif
#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_))
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
# ifdef JEMALLOC_ARENA_INLINE_A
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
JEMALLOC_ALWAYS_INLINE arena_chunk_map_bits_t *
arena_bitselm_get(arena_chunk_t *chunk, size_t pageind)
{
assert(pageind >= map_bias);
assert(pageind < chunk_npages);
return (&chunk->map_bits[pageind-map_bias]);
}
JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
arena_miscelm_get(arena_chunk_t *chunk, size_t pageind)
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
{
assert(pageind >= map_bias);
assert(pageind < chunk_npages);
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
return ((arena_chunk_map_misc_t *)((uintptr_t)chunk +
(uintptr_t)map_misc_offset) + pageind-map_bias);
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
}
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
JEMALLOC_ALWAYS_INLINE size_t
arena_miscelm_to_pageind(arena_chunk_map_misc_t *miscelm)
{
arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(miscelm);
size_t pageind = ((uintptr_t)miscelm - ((uintptr_t)chunk +
map_misc_offset)) / sizeof(arena_chunk_map_misc_t) + map_bias;
assert(pageind >= map_bias);
assert(pageind < chunk_npages);
return (pageind);
}
JEMALLOC_ALWAYS_INLINE void *
arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm)
{
arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(miscelm);
size_t pageind = arena_miscelm_to_pageind(miscelm);
return ((void *)((uintptr_t)chunk + (pageind << LG_PAGE)));
}
Refactor dirty run linkage to reduce sizeof(extent_node_t).
2015-03-11 09:15:40 +08:00
JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
arena_rd_to_miscelm(arena_runs_dirty_link_t *rd)
{
arena_chunk_map_misc_t *miscelm = (arena_chunk_map_misc_t
*)((uintptr_t)rd - offsetof(arena_chunk_map_misc_t, rd));
assert(arena_miscelm_to_pageind(miscelm) >= map_bias);
assert(arena_miscelm_to_pageind(miscelm) < chunk_npages);
return (miscelm);
}
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
arena_run_to_miscelm(arena_run_t *run)
{
arena_chunk_map_misc_t *miscelm = (arena_chunk_map_misc_t
*)((uintptr_t)run - offsetof(arena_chunk_map_misc_t, run));
assert(arena_miscelm_to_pageind(miscelm) >= map_bias);
assert(arena_miscelm_to_pageind(miscelm) < chunk_npages);
return (miscelm);
}
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 size_t *
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_mapbitsp_get(arena_chunk_t *chunk, size_t pageind)
{
Refactor chunk map. Break the chunk map into two separate arrays, in order to improve cache locality. This is related to issue #23.
2014-08-30 04:34:40 +08:00
return (&arena_bitselm_get(chunk, pageind)->bits);
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
}
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbitsp_read(size_t *mapbitsp)
{
return (*mapbitsp);
}
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 size_t
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_get(arena_chunk_t *chunk, size_t pageind)
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
return (arena_mapbitsp_read(arena_mapbitsp_get(chunk, pageind)));
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
}
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 size_t
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_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);
return (mapbits & ~PAGE_MASK);
}
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 size_t
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_size_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) ==
(CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED));
return (mapbits & ~PAGE_MASK);
}
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 size_t
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_small_runind_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) ==
CHUNK_MAP_ALLOCATED);
return (mapbits >> LG_PAGE);
}
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
JEMALLOC_ALWAYS_INLINE index_t
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
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
index_t binind;
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
mapbits = arena_mapbits_get(chunk, pageind);
binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;
assert(binind < NBINS || binind == BININD_INVALID);
return (binind);
}
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 size_t
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_dirty_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
return (mapbits & CHUNK_MAP_DIRTY);
}
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 size_t
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_unzeroed_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
return (mapbits & CHUNK_MAP_UNZEROED);
}
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 size_t
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_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
return (mapbits & CHUNK_MAP_LARGE);
}
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 size_t
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_allocated_get(arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
return (mapbits & CHUNK_MAP_ALLOCATED);
}
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
JEMALLOC_ALWAYS_INLINE void
arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits)
{
*mapbitsp = mapbits;
}
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
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_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,
size_t flags)
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
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
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(size == PAGE_CEILING(size));
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
assert((flags & ~CHUNK_MAP_FLAGS_MASK) == 0);
Fix large calloc() zeroing bugs. Refactor code such that arena_mapbits_{large,small}_set() always preserves the unzeroed flag, and manually manipulate the unzeroed flag in the one case where it actually gets reset (in arena_chunk_purge()). This fixes unzeroed preservation bugs in arena_run_split() and arena_ralloc_large_grow(). These bugs caused large calloc() to return non-zeroed memory under some circumstances.
2012-05-11 11:59:39 +08:00
assert((flags & (CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == flags);
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
arena_mapbitsp_write(mapbitsp, size | CHUNK_MAP_BININD_INVALID | flags);
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
}
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
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_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
size_t size)
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
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
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(size == PAGE_CEILING(size));
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);
arena_mapbitsp_write(mapbitsp, size | (mapbits & PAGE_MASK));
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
}
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
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_set(arena_chunk_t *chunk, size_t pageind, size_t size,
size_t flags)
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
Fix large calloc() zeroing bugs. Refactor code such that arena_mapbits_{large,small}_set() always preserves the unzeroed flag, and manually manipulate the unzeroed flag in the one case where it actually gets reset (in arena_chunk_purge()). This fixes unzeroed preservation bugs in arena_run_split() and arena_ralloc_large_grow(). These bugs caused large calloc() to return non-zeroed memory under some circumstances.
2012-05-11 11:59:39 +08:00
size_t unzeroed;
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
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(size == PAGE_CEILING(size));
Fix large calloc() zeroing bugs. Refactor code such that arena_mapbits_{large,small}_set() always preserves the unzeroed flag, and manually manipulate the unzeroed flag in the one case where it actually gets reset (in arena_chunk_purge()). This fixes unzeroed preservation bugs in arena_run_split() and arena_ralloc_large_grow(). These bugs caused large calloc() to return non-zeroed memory under some circumstances.
2012-05-11 11:59:39 +08:00
assert((flags & CHUNK_MAP_DIRTY) == flags);
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
unzeroed = mapbits & CHUNK_MAP_UNZEROED; /* Preserve unzeroed. */
arena_mapbitsp_write(mapbitsp, size | CHUNK_MAP_BININD_INVALID | flags
| unzeroed | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED);
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
}
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
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(arena_chunk_t *chunk, size_t pageind,
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
index_t binind)
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
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
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
assert(binind <= BININD_INVALID);
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(arena_mapbits_large_size_get(chunk, pageind) == LARGE_MINCLASS +
large_pad);
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
arena_mapbitsp_write(mapbitsp, (mapbits & ~CHUNK_MAP_BININD_MASK) |
(binind << CHUNK_MAP_BININD_SHIFT));
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
}
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
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_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind,
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
index_t binind, size_t flags)
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
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
Fix large calloc() zeroing bugs. Refactor code such that arena_mapbits_{large,small}_set() always preserves the unzeroed flag, and manually manipulate the unzeroed flag in the one case where it actually gets reset (in arena_chunk_purge()). This fixes unzeroed preservation bugs in arena_run_split() and arena_ralloc_large_grow(). These bugs caused large calloc() to return non-zeroed memory under some circumstances.
2012-05-11 11:59:39 +08:00
size_t unzeroed;
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
assert(binind < BININD_INVALID);
assert(pageind - runind >= map_bias);
Fix large calloc() zeroing bugs. Refactor code such that arena_mapbits_{large,small}_set() always preserves the unzeroed flag, and manually manipulate the unzeroed flag in the one case where it actually gets reset (in arena_chunk_purge()). This fixes unzeroed preservation bugs in arena_run_split() and arena_ralloc_large_grow(). These bugs caused large calloc() to return non-zeroed memory under some circumstances.
2012-05-11 11:59:39 +08:00
assert((flags & CHUNK_MAP_DIRTY) == flags);
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
unzeroed = mapbits & CHUNK_MAP_UNZEROED; /* Preserve unzeroed. */
arena_mapbitsp_write(mapbitsp, (runind << LG_PAGE) | (binind <<
CHUNK_MAP_BININD_SHIFT) | flags | unzeroed | CHUNK_MAP_ALLOCATED);
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
}
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
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_unzeroed_set(arena_chunk_t *chunk, size_t pageind,
size_t unzeroed)
{
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
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
Fix a Valgrind integration flaw. Fix a Valgrind integration flaw that caused Valgrind warnings about reads of uninitialized memory in arena chunk headers.
2013-10-20 12:40:20 +08:00
arena_mapbitsp_write(mapbitsp, (mapbits & ~CHUNK_MAP_UNZEROED) |
unzeroed);
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
}
Implement metadata statistics. There are three categories of metadata: - Base allocations are used for bootstrap-sensitive internal allocator data structures. - Arena chunk headers comprise pages which track the states of the non-metadata pages. - Internal allocations differ from application-originated allocations in that they are for internal use, and that they are omitted from heap profiles. The metadata statistics comprise the metadata categories as follows: - stats.metadata: All metadata -- base + arena chunk headers + internal allocations. - stats.arenas.<i>.metadata.mapped: Arena chunk headers. - stats.arenas.<i>.metadata.allocated: Internal allocations. This is reported separately from the other metadata statistics because it overlaps with the allocated and active statistics, whereas the other metadata statistics do not. Base allocations are not reported separately, though their magnitude can be computed by subtracting the arena-specific metadata. This resolves #163.
2014-11-28 03:22:36 +08:00
JEMALLOC_INLINE void
arena_metadata_allocated_add(arena_t *arena, size_t size)
{
atomic_add_z(&arena->stats.metadata_allocated, size);
}
JEMALLOC_INLINE void
arena_metadata_allocated_sub(arena_t *arena, size_t size)
{
atomic_sub_z(&arena->stats.metadata_allocated, size);
}
JEMALLOC_INLINE size_t
arena_metadata_allocated_get(arena_t *arena)
{
return (atomic_read_z(&arena->stats.metadata_allocated));
}
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
JEMALLOC_INLINE bool
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +08:00
arena_prof_accum_impl(arena_t *arena, uint64_t accumbytes)
{
cassert(config_prof);
assert(prof_interval != 0);
arena->prof_accumbytes += accumbytes;
if (arena->prof_accumbytes >= prof_interval) {
arena->prof_accumbytes -= prof_interval;
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
return (true);
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +08:00
}
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
return (false);
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +08:00
}
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
JEMALLOC_INLINE bool
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +08:00
arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes)
{
cassert(config_prof);
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (likely(prof_interval == 0))
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
return (false);
return (arena_prof_accum_impl(arena, accumbytes));
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +08:00
}
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
JEMALLOC_INLINE bool
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +08:00
arena_prof_accum(arena_t *arena, uint64_t accumbytes)
{
cassert(config_prof);
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (likely(prof_interval == 0))
Fix a prof-related locking order bug. Fix a locking order bug that could cause deadlock during fork if heap profiling were enabled.
2013-02-07 03:59:30 +08:00
return (false);
{
bool ret;
malloc_mutex_lock(&arena->lock);
ret = arena_prof_accum_impl(arena, accumbytes);
malloc_mutex_unlock(&arena->lock);
return (ret);
}
Avoid arena_prof_accum()-related locking when possible. Refactor arena_prof_accum() and its callers to avoid arena locking when prof_interval is 0 (as when profiling is disabled). Reported by Ben Maurer.
2012-11-14 04:56:27 +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
JEMALLOC_ALWAYS_INLINE index_t
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
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
{
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
index_t binind;
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
binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;
if (config_debug) {
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
arena_chunk_t *chunk;
arena_t *arena;
size_t pageind;
size_t actual_mapbits;
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
size_t rpages_ind;
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
arena_run_t *run;
arena_bin_t *bin;
Remove arena_dalloc_bin_run() clean page preservation. Remove code in arena_dalloc_bin_run() that preserved the "clean" state of trailing clean pages by splitting them into a separate run during deallocation. This was a useful mechanism for reducing dirty page churn when bin runs comprised many pages, but bin runs are now quite small. Remove the nextind field from arena_run_t now that it is no longer needed, and change arena_run_t's bin field (arena_bin_t *) to binind (index_t). These two changes remove 8 bytes of chunk header overhead per page, which saves 1/512 of all arena chunk memory.
2014-10-11 14:01:03 +08:00
index_t run_binind, actual_binind;
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
arena_bin_info_t *bin_info;
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
arena_chunk_map_misc_t *miscelm;
void *rpages;
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
assert(binind != BININD_INVALID);
assert(binind < NBINS);
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
arena = extent_node_arena_get(&chunk->node);
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
actual_mapbits = arena_mapbits_get(chunk, pageind);
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
assert(mapbits == actual_mapbits);
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
assert(arena_mapbits_large_get(chunk, pageind) == 0);
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
rpages_ind = pageind - arena_mapbits_small_runind_get(chunk,
pageind);
miscelm = arena_miscelm_get(chunk, rpages_ind);
run = &miscelm->run;
Remove arena_dalloc_bin_run() clean page preservation. Remove code in arena_dalloc_bin_run() that preserved the "clean" state of trailing clean pages by splitting them into a separate run during deallocation. This was a useful mechanism for reducing dirty page churn when bin runs comprised many pages, but bin runs are now quite small. Remove the nextind field from arena_run_t now that it is no longer needed, and change arena_run_t's bin field (arena_bin_t *) to binind (index_t). These two changes remove 8 bytes of chunk header overhead per page, which saves 1/512 of all arena chunk memory.
2014-10-11 14:01:03 +08:00
run_binind = run->binind;
bin = &arena->bins[run_binind];
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
actual_binind = bin - arena->bins;
Remove arena_dalloc_bin_run() clean page preservation. Remove code in arena_dalloc_bin_run() that preserved the "clean" state of trailing clean pages by splitting them into a separate run during deallocation. This was a useful mechanism for reducing dirty page churn when bin runs comprised many pages, but bin runs are now quite small. Remove the nextind field from arena_run_t now that it is no longer needed, and change arena_run_t's bin field (arena_bin_t *) to binind (index_t). These two changes remove 8 bytes of chunk header overhead per page, which saves 1/512 of all arena chunk memory.
2014-10-11 14:01:03 +08:00
assert(run_binind == actual_binind);
Further optimize and harden arena_salloc(). Further optimize arena_salloc() to only look at the binind chunk map bits in the common case. Add more sanity checks to arena_salloc() that detect chunk map inconsistencies for large allocations (whether due to allocator bugs or application bugs).
2012-05-03 07:11:03 +08:00
bin_info = &arena_bin_info[actual_binind];
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
rpages = arena_miscelm_to_rpages(miscelm);
assert(((uintptr_t)ptr - ((uintptr_t)rpages +
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
(uintptr_t)bin_info->reg0_offset)) % bin_info->reg_interval
== 0);
}
return (binind);
}
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
# endif /* JEMALLOC_ARENA_INLINE_A */
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
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
# ifdef JEMALLOC_ARENA_INLINE_B
JEMALLOC_INLINE index_t
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
arena_bin_index(arena_t *arena, arena_bin_t *bin)
{
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
index_t binind = bin - arena->bins;
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);
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
return (binind);
}
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
JEMALLOC_INLINE unsigned
Clean up after arena_bin_info_t change. Fix a couple of problems related to the addition of arena_bin_info_t.
2011-03-16 13:19:45 +08:00
arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
{
unsigned shift, diff, regind;
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
size_t interval;
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
arena_chunk_map_misc_t *miscelm = arena_run_to_miscelm(run);
void *rpages = arena_miscelm_to_rpages(miscelm);
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +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
/*
* Freeing a pointer lower than region zero can cause assertion
* failure.
*/
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
assert((uintptr_t)ptr >= (uintptr_t)rpages +
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
(uintptr_t)bin_info->reg0_offset);
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
/*
* Avoid doing division with a variable divisor if possible. Using
* actual division here can reduce allocator throughput by over 20%!
*/
Move small run metadata into the arena chunk header. Move small run metadata into the arena chunk header, with multiple expected benefits: - Lower run fragmentation due to reduced run sizes; runs are more likely to completely drain when there are fewer total regions. - Improved cache behavior. Prior to this change, run headers were always page-aligned, which put extra pressure on some CPU cache sets. The degree to which this was a problem was hardware dependent, but it likely hurt some even for the most advanced modern hardware. - Buffer overruns/underruns are less likely to corrupt allocator metadata. - Size classes between 4 KiB and 16 KiB become reasonable to support without any special handling, and the runs are small enough that dirty unused pages aren't a significant concern.
2014-09-29 16:31:39 +08:00
diff = (unsigned)((uintptr_t)ptr - (uintptr_t)rpages -
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
bin_info->reg0_offset);
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
/* Rescale (factor powers of 2 out of the numerator and denominator). */
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
interval = bin_info->reg_interval;
Try to use __builtin_ffsl if ffsl is unavailable. Some platforms (like those using Newlib) don't have ffs/ffsl. This commit adds a check to configure.ac for __builtin_ffsl if ffsl isn't found. __builtin_ffsl performs the same function as ffsl, and has the added benefit of being available on any platform utilizing Gcc-compatible compiler. This change does not address the used of ffs in the MALLOCX_ARENA() macro.
2014-05-29 10:37:02 +08:00
shift = jemalloc_ffs(interval) - 1;
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
diff >>= shift;
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
interval >>= shift;
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +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
if (interval == 1) {
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
/* The divisor was a power of 2. */
regind = diff;
} else {
/*
* To divide by a number D that is not a power of two we
* multiply by (2^21 / D) and then right shift by 21 positions.
*
* X / D
*
* becomes
*
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
* (X * interval_invs[D - 3]) >> SIZE_INV_SHIFT
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
*
* We can omit the first three elements, because we never
* divide by 0, and 1 and 2 are both powers of two, which are
* handled above.
*/
Avoid overflow in arena_run_regind(). Fix a regression due to: Remove an arena_bin_run_size_calc() constraint. 2a6f2af6e446a98a635caadd281a23ca09a491cb The removed constraint required that small run headers fit in one page, which indirectly limited runs such that they would not cause overflow in arena_run_regind(). Add an explicit constraint to arena_bin_run_size_calc() based on the largest number of regions that arena_run_regind() can handle (2^11 as currently configured).
2011-03-23 00:00:56 +08:00
#define SIZE_INV_SHIFT ((sizeof(unsigned) << 3) - LG_RUN_MAXREGS)
#define SIZE_INV(s) (((1U << SIZE_INV_SHIFT) / (s)) + 1)
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
static const unsigned interval_invs[] = {
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
SIZE_INV(3),
SIZE_INV(4), SIZE_INV(5), SIZE_INV(6), SIZE_INV(7),
SIZE_INV(8), SIZE_INV(9), SIZE_INV(10), SIZE_INV(11),
SIZE_INV(12), SIZE_INV(13), SIZE_INV(14), SIZE_INV(15),
SIZE_INV(16), SIZE_INV(17), SIZE_INV(18), SIZE_INV(19),
SIZE_INV(20), SIZE_INV(21), SIZE_INV(22), SIZE_INV(23),
SIZE_INV(24), SIZE_INV(25), SIZE_INV(26), SIZE_INV(27),
SIZE_INV(28), SIZE_INV(29), SIZE_INV(30), SIZE_INV(31)
};
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (likely(interval <= ((sizeof(interval_invs) /
sizeof(unsigned)) + 2))) {
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
regind = (diff * interval_invs[interval - 3]) >>
SIZE_INV_SHIFT;
} else
regind = diff / interval;
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
#undef SIZE_INV
#undef SIZE_INV_SHIFT
}
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
assert(diff == regind * interval);
Create arena_bin_info_t. Move read-only fields from arena_bin_t into arena_bin_info_t, primarily in order to avoid false cacheline sharing.
2011-03-16 04:59:15 +08:00
assert(regind < bin_info->nregs);
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
return (regind);
}
Implement per thread heap profiling. Rename data structures (prof_thr_cnt_t-->prof_tctx_t, prof_ctx_t-->prof_gctx_t), and convert to storing a prof_tctx_t for sampled objects. Convert PROF_ALLOC_PREP() to prof_alloc_prep(), since precise backtrace depth within jemalloc functions is no longer an issue (pprof prunes irrelevant frames). Implement mallctl's: - prof.reset implements full sample data reset, and optional change of sample interval. - prof.lg_sample reads the current sample interval (opt.lg_prof_sample was the permanent source of truth prior to prof.reset). - thread.prof.name provides naming capability for threads within heap profile dumps. - thread.prof.active makes it possible to activate/deactivate heap profiling for individual threads. Modify the heap dump files to contain per thread heap profile data. This change is incompatible with the existing pprof, which will require enhancements to read and process the enriched data.
2014-08-19 07:22:13 +08:00
JEMALLOC_INLINE prof_tctx_t *
arena_prof_tctx_get(const void *ptr)
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
{
Implement per thread heap profiling. Rename data structures (prof_thr_cnt_t-->prof_tctx_t, prof_ctx_t-->prof_gctx_t), and convert to storing a prof_tctx_t for sampled objects. Convert PROF_ALLOC_PREP() to prof_alloc_prep(), since precise backtrace depth within jemalloc functions is no longer an issue (pprof prunes irrelevant frames). Implement mallctl's: - prof.reset implements full sample data reset, and optional change of sample interval. - prof.lg_sample reads the current sample interval (opt.lg_prof_sample was the permanent source of truth prior to prof.reset). - thread.prof.name provides naming capability for threads within heap profile dumps. - thread.prof.active makes it possible to activate/deactivate heap profiling for individual threads. Modify the heap dump files to contain per thread heap profile data. This change is incompatible with the existing pprof, which will require enhancements to read and process the enriched data.
2014-08-19 07:22:13 +08:00
prof_tctx_t *ret;
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
arena_chunk_t *chunk;
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
cassert(config_prof);
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
assert(ptr != NULL);
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
if (likely(chunk != ptr)) {
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
size_t mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & CHUNK_MAP_ALLOCATED) != 0);
if (likely((mapbits & CHUNK_MAP_LARGE) == 0))
ret = (prof_tctx_t *)(uintptr_t)1U;
Make prof_tctx accesses atomic. Although exceedingly unlikely, it appears that writes to the prof_tctx field of arena_chunk_map_misc_t could be reordered such that a stale value could be read during deallocation, with profiler metadata corruption and invalid pointer dereferences being the most likely effects.
2015-02-13 07:54:53 +08:00
else {
arena_chunk_map_misc_t *elm = arena_miscelm_get(chunk,
pageind);
Fix type punning in calls to atomic operation functions.
2015-05-08 13:35:40 +08:00
ret = atomic_read_p(&elm->prof_tctx_pun);
Make prof_tctx accesses atomic. Although exceedingly unlikely, it appears that writes to the prof_tctx field of arena_chunk_map_misc_t could be reordered such that a stale value could be read during deallocation, with profiler metadata corruption and invalid pointer dereferences being the most likely effects.
2015-02-13 07:54:53 +08:00
}
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
} else
ret = huge_prof_tctx_get(ptr);
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +08:00
return (ret);
}
Fix heap profiling bugs. Fix a regression due to the recent heap profiling accuracy improvements: prof_{m,re}alloc() must set the object's profiling context regardless of whether it is sampled. Fix management of the CHUNK_MAP_CLASS chunk map bits, such that all large object (re-)allocation paths correctly initialize the bits. Prior to this fix, in-place realloc() cleared the bits, resulting in incorrect reported object size from arena_salloc_demote(). After this fix the non-demoted bit pattern is all zeros (instead of all ones), which makes it easier to assure that the bits are properly set.
2010-10-23 01:45:59 +08:00
JEMALLOC_INLINE void
Implement per thread heap profiling. Rename data structures (prof_thr_cnt_t-->prof_tctx_t, prof_ctx_t-->prof_gctx_t), and convert to storing a prof_tctx_t for sampled objects. Convert PROF_ALLOC_PREP() to prof_alloc_prep(), since precise backtrace depth within jemalloc functions is no longer an issue (pprof prunes irrelevant frames). Implement mallctl's: - prof.reset implements full sample data reset, and optional change of sample interval. - prof.lg_sample reads the current sample interval (opt.lg_prof_sample was the permanent source of truth prior to prof.reset). - thread.prof.name provides naming capability for threads within heap profile dumps. - thread.prof.active makes it possible to activate/deactivate heap profiling for individual threads. Modify the heap dump files to contain per thread heap profile data. This change is incompatible with the existing pprof, which will require enhancements to read and process the enriched data.
2014-08-19 07:22:13 +08:00
arena_prof_tctx_set(const void *ptr, prof_tctx_t *tctx)
Fix heap profiling bugs. Fix a regression due to the recent heap profiling accuracy improvements: prof_{m,re}alloc() must set the object's profiling context regardless of whether it is sampled. Fix management of the CHUNK_MAP_CLASS chunk map bits, such that all large object (re-)allocation paths correctly initialize the bits. Prior to this fix, in-place realloc() cleared the bits, resulting in incorrect reported object size from arena_salloc_demote(). After this fix the non-demoted bit pattern is all zeros (instead of all ones), which makes it easier to assure that the bits are properly set.
2010-10-23 01:45:59 +08:00
{
arena_chunk_t *chunk;
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
cassert(config_prof);
Fix heap profiling bugs. Fix a regression due to the recent heap profiling accuracy improvements: prof_{m,re}alloc() must set the object's profiling context regardless of whether it is sampled. Fix management of the CHUNK_MAP_CLASS chunk map bits, such that all large object (re-)allocation paths correctly initialize the bits. Prior to this fix, in-place realloc() cleared the bits, resulting in incorrect reported object size from arena_salloc_demote(). After this fix the non-demoted bit pattern is all zeros (instead of all ones), which makes it easier to assure that the bits are properly set.
2010-10-23 01:45:59 +08:00
assert(ptr != NULL);
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
if (likely(chunk != ptr)) {
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
Optimize arena_prof_ctx_set(). Refactor such that arena_prof_ctx_set() receives usize as an argument, and use it to determine whether to handle ptr as a small region, rather than reading the chunk page map.
2013-12-16 08:21:30 +08:00
Make prof_tctx accesses atomic. Although exceedingly unlikely, it appears that writes to the prof_tctx field of arena_chunk_map_misc_t could be reordered such that a stale value could be read during deallocation, with profiler metadata corruption and invalid pointer dereferences being the most likely effects.
2015-02-13 07:54:53 +08:00
if (unlikely(arena_mapbits_large_get(chunk, pageind) != 0)) {
arena_chunk_map_misc_t *elm = arena_miscelm_get(chunk,
pageind);
Fix type punning in calls to atomic operation functions.
2015-05-08 13:35:40 +08:00
atomic_write_p(&elm->prof_tctx_pun, tctx);
Make prof_tctx accesses atomic. Although exceedingly unlikely, it appears that writes to the prof_tctx field of arena_chunk_map_misc_t could be reordered such that a stale value could be read during deallocation, with profiler metadata corruption and invalid pointer dereferences being the most likely effects.
2015-02-13 07:54:53 +08:00
}
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
} else
huge_prof_tctx_set(ptr, tctx);
Fix heap profiling bugs. Fix a regression due to the recent heap profiling accuracy improvements: prof_{m,re}alloc() must set the object's profiling context regardless of whether it is sampled. Fix management of the CHUNK_MAP_CLASS chunk map bits, such that all large object (re-)allocation paths correctly initialize the bits. Prior to this fix, in-place realloc() cleared the bits, resulting in incorrect reported object size from arena_salloc_demote(). After this fix the non-demoted bit pattern is all zeros (instead of all ones), which makes it easier to assure that the bits are properly set.
2010-10-23 01:45:59 +08:00
}
Fix a heap profiling regression. Call prof_ctx_set() in all paths through prof_{m,re}alloc(). Inline arena_prof_ctx_get().
2010-10-21 11:52:00 +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 *
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
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 *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
{
assert(size != 0);
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
arena = arena_choose(tsd, arena);
if (unlikely(arena == NULL))
return (NULL);
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
if (likely(size <= SMALL_MAXCLASS)) {
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
if (likely(tcache != NULL)) {
return (tcache_alloc_small(tsd, arena, tcache, size,
zero));
} else
Refactor/fix arenas manipulation. Abstract arenas access to use arena_get() (or a0get() where appropriate) rather than directly reading e.g. arenas[ind]. Prior to the addition of the arenas.extend mallctl, the worst possible outcome of directly accessing arenas was a stale read, but arenas.extend may allocate and assign a new array to arenas. Add a tsd-based arenas_cache, which amortizes arenas reads. This introduces some subtle bootstrapping issues, with tsd_boot() now being split into tsd_boot[01]() to support tsd wrapper allocation bootstrapping, as well as an arenas_cache_bypass tsd variable which dynamically terminates allocation of arenas_cache itself. Promote a0malloc(), a0calloc(), and a0free() to be generally useful for internal allocation, and use them in several places (more may be appropriate). Abstract arena->nthreads management and fix a missing decrement during thread destruction (recent tsd refactoring left arenas_cleanup() unused). Change arena_choose() to propagate OOM, and handle OOM in all callers. This is important for providing consistent allocation behavior when the MALLOCX_ARENA() flag is being used. Prior to this fix, it was possible for an OOM to result in allocation silently allocating from a different arena than the one specified.
2014-10-08 14:14:57 +08:00
return (arena_malloc_small(arena, size, zero));
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
} else if (likely(size <= arena_maxclass)) {
Remove highruns statistics.
2012-02-14 07:18:19 +08:00
/*
* Initialize tcache after checking size in order to avoid
* infinite recursion during tcache initialization.
*/
If MALLOCX_ARENA(a) is specified, use it during tcache fill.
2015-02-14 07:28:56 +08:00
if (likely(tcache != NULL) && size <= tcache_maxclass) {
return (tcache_alloc_large(tsd, arena, tcache, size,
zero));
} else
Refactor/fix arenas manipulation. Abstract arenas access to use arena_get() (or a0get() where appropriate) rather than directly reading e.g. arenas[ind]. Prior to the addition of the arenas.extend mallctl, the worst possible outcome of directly accessing arenas was a stale read, but arenas.extend may allocate and assign a new array to arenas. Add a tsd-based arenas_cache, which amortizes arenas reads. This introduces some subtle bootstrapping issues, with tsd_boot() now being split into tsd_boot[01]() to support tsd wrapper allocation bootstrapping, as well as an arenas_cache_bypass tsd variable which dynamically terminates allocation of arenas_cache itself. Promote a0malloc(), a0calloc(), and a0free() to be generally useful for internal allocation, and use them in several places (more may be appropriate). Abstract arena->nthreads management and fix a missing decrement during thread destruction (recent tsd refactoring left arenas_cleanup() unused). Change arena_choose() to propagate OOM, and handle OOM in all callers. This is important for providing consistent allocation behavior when the MALLOCX_ARENA() flag is being used. Prior to this fix, it was possible for an OOM to result in allocation silently allocating from a different arena than the one specified.
2014-10-08 14:14:57 +08:00
return (arena_malloc_large(arena, size, zero));
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
} else
return (huge_malloc(tsd, arena, size, zero, 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
}
Implement metadata statistics. There are three categories of metadata: - Base allocations are used for bootstrap-sensitive internal allocator data structures. - Arena chunk headers comprise pages which track the states of the non-metadata pages. - Internal allocations differ from application-originated allocations in that they are for internal use, and that they are omitted from heap profiles. The metadata statistics comprise the metadata categories as follows: - stats.metadata: All metadata -- base + arena chunk headers + internal allocations. - stats.arenas.<i>.metadata.mapped: Arena chunk headers. - stats.arenas.<i>.metadata.allocated: Internal allocations. This is reported separately from the other metadata statistics because it overlaps with the allocated and active statistics, whereas the other metadata statistics do not. Base allocations are not reported separately, though their magnitude can be computed by subtracting the arena-specific metadata. This resolves #163.
2014-11-28 03:22:36 +08:00
JEMALLOC_ALWAYS_INLINE arena_t *
arena_aalloc(const void *ptr)
{
arena_chunk_t *chunk;
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
if (likely(chunk != ptr))
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
return (extent_node_arena_get(&chunk->node));
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
else
return (huge_aalloc(ptr));
Implement metadata statistics. There are three categories of metadata: - Base allocations are used for bootstrap-sensitive internal allocator data structures. - Arena chunk headers comprise pages which track the states of the non-metadata pages. - Internal allocations differ from application-originated allocations in that they are for internal use, and that they are omitted from heap profiles. The metadata statistics comprise the metadata categories as follows: - stats.metadata: All metadata -- base + arena chunk headers + internal allocations. - stats.arenas.<i>.metadata.mapped: Arena chunk headers. - stats.arenas.<i>.metadata.allocated: Internal allocations. This is reported separately from the other metadata statistics because it overlaps with the allocated and active statistics, whereas the other metadata statistics do not. Base allocations are not reported separately, though their magnitude can be computed by subtracting the arena-specific metadata. This resolves #163.
2014-11-28 03:22:36 +08:00
}
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
/* Return the size of the allocation pointed to by ptr. */
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 size_t
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
arena_salloc(const void *ptr, bool demote)
{
size_t ret;
arena_chunk_t *chunk;
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
size_t pageind;
index_t binind;
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +08:00
assert(ptr != NULL);
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
if (likely(chunk != ptr)) {
pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
binind = arena_mapbits_binind_get(chunk, pageind);
if (unlikely(binind == BININD_INVALID || (config_prof && !demote
&& arena_mapbits_large_get(chunk, pageind) != 0))) {
/*
* Large allocation. In the common case (demote), and
* as this is an inline function, most callers will only
* end up looking at binind to determine that ptr is a
* small allocation.
*/
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(config_cache_oblivious || ((uintptr_t)ptr &
PAGE_MASK) == 0);
ret = arena_mapbits_large_size_get(chunk, pageind) -
large_pad;
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
assert(ret != 0);
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(pageind + ((ret+large_pad)>>LG_PAGE) <=
chunk_npages);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
assert(arena_mapbits_dirty_get(chunk, pageind) ==
arena_mapbits_dirty_get(chunk,
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
pageind+((ret+large_pad)>>LG_PAGE)-1));
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
} else {
/*
* Small allocation (possibly promoted to a large
* object).
*/
assert(arena_mapbits_large_get(chunk, pageind) != 0 ||
arena_ptr_small_binind_get(ptr,
arena_mapbits_get(chunk, pageind)) == binind);
ret = index2size(binind);
}
} else
ret = huge_salloc(ptr);
Make arena_salloc() an inline function.
2012-04-20 09:28:03 +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
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
arena_dalloc(tsd_t *tsd, void *ptr, 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
{
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
arena_chunk_t *chunk;
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
size_t pageind, mapbits;
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
assert(ptr != NULL);
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
if (likely(chunk != ptr)) {
pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
mapbits = arena_mapbits_get(chunk, pageind);
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
if (likely((mapbits & CHUNK_MAP_LARGE) == 0)) {
/* Small allocation. */
if (likely(tcache != NULL)) {
index_t binind = arena_ptr_small_binind_get(ptr,
mapbits);
tcache_dalloc_small(tsd, tcache, ptr, binind);
} else {
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
arena_dalloc_small(extent_node_arena_get(
&chunk->node), chunk, ptr, pageind);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
}
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
} else {
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
size_t size = arena_mapbits_large_size_get(chunk,
Move centralized chunk management into arenas. Migrate all centralized data structures related to huge allocations and recyclable chunks into arena_t, so that each arena can manage huge allocations and recyclable virtual memory completely independently of other arenas. Add chunk node caching to arenas, in order to avoid contention on the base allocator. Use chunks_rtree to look up huge allocations rather than a red-black tree. Maintain a per arena unsorted list of huge allocations (which will be needed to enumerate huge allocations during arena reset). Remove the --enable-ivsalloc option, make ivsalloc() always available, and use it for size queries if --enable-debug is enabled. The only practical implications to this removal are that 1) ivsalloc() is now always available during live debugging (and the underlying radix tree is available during core-based debugging), and 2) size query validation can no longer be enabled independent of --enable-debug. Remove the stats.chunks.{current,total,high} mallctls, and replace their underlying statistics with simpler atomically updated counters used exclusively for gdump triggering. These statistics are no longer very useful because each arena manages chunks independently, and per arena statistics provide similar information. Simplify chunk synchronization code, now that base chunk allocation cannot cause recursive lock acquisition.
2015-02-12 04:24:27 +08:00
pageind);
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
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(config_cache_oblivious || ((uintptr_t)ptr &
PAGE_MASK) == 0);
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
Fix arena_dalloc() performance regression. Take into account large_pad when computing whether to pass the deallocation request to tcache_dalloc_large(), so that the largest cacheable size makes it back to tcache. This regression was introduced by 8a03cf039cd06f9fa6972711195055d865673966 (Implement cache index randomization for large allocations.).
2015-05-20 08:44:45 +08:00
if (likely(tcache != NULL) && size - large_pad <=
tcache_maxclass) {
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
tcache_dalloc_large(tsd, tcache, ptr, size -
large_pad);
} else {
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
arena_dalloc_large(extent_node_arena_get(
&chunk->node), chunk, ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
}
}
} else
huge_dalloc(tsd, ptr, 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 support for sized deallocation. This adds a new `sdallocx` function to the external API, allowing the size to be passed by the caller. It avoids some extra reads in the thread cache fast path. In the case where stats are enabled, this avoids the work of calculating the size from the pointer. An assertion validates the size that's passed in, so enabling debugging will allow users of the API to debug cases where an incorrect size is passed in. The performance win for a contrived microbenchmark doing an allocation and immediately freeing it is ~10%. It may have a different impact on a real workload. Closes #28
2014-08-29 03:41:48 +08:00
JEMALLOC_ALWAYS_INLINE void
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
Add support for sized deallocation. This adds a new `sdallocx` function to the external API, allowing the size to be passed by the caller. It avoids some extra reads in the thread cache fast path. In the case where stats are enabled, this avoids the work of calculating the size from the pointer. An assertion validates the size that's passed in, so enabling debugging will allow users of the API to debug cases where an incorrect size is passed in. The performance win for a contrived microbenchmark doing an allocation and immediately freeing it is ~10%. It may have a different impact on a real workload. Closes #28
2014-08-29 03:41:48 +08:00
{
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
arena_chunk_t *chunk;
Add support for sized deallocation. This adds a new `sdallocx` function to the external API, allowing the size to be passed by the caller. It avoids some extra reads in the thread cache fast path. In the case where stats are enabled, this avoids the work of calculating the size from the pointer. An assertion validates the size that's passed in, so enabling debugging will allow users of the API to debug cases where an incorrect size is passed in. The performance win for a contrived microbenchmark doing an allocation and immediately freeing it is ~10%. It may have a different impact on a real workload. Closes #28
2014-08-29 03:41:48 +08:00
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
if (likely(chunk != ptr)) {
if (config_prof && opt_prof) {
Apply likely()/unlikely() to allocation/deallocation fast paths.
2014-09-12 07:20:44 +08:00
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >>
LG_PAGE;
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
if (arena_mapbits_large_get(chunk, pageind) != 0) {
/*
* Make sure to use promoted size, not request
* size.
*/
size = arena_mapbits_large_size_get(chunk,
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
pageind) - large_pad;
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
}
Add support for sized deallocation. This adds a new `sdallocx` function to the external API, allowing the size to be passed by the caller. It avoids some extra reads in the thread cache fast path. In the case where stats are enabled, this avoids the work of calculating the size from the pointer. An assertion validates the size that's passed in, so enabling debugging will allow users of the API to debug cases where an incorrect size is passed in. The performance win for a contrived microbenchmark doing an allocation and immediately freeing it is ~10%. It may have a different impact on a real workload. Closes #28
2014-08-29 03:41:48 +08:00
}
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
assert(s2u(size) == s2u(arena_salloc(ptr, false)));
if (likely(size <= SMALL_MAXCLASS)) {
/* Small allocation. */
if (likely(tcache != NULL)) {
index_t binind = size2index(size);
tcache_dalloc_small(tsd, tcache, ptr, binind);
} else {
size_t pageind = ((uintptr_t)ptr -
(uintptr_t)chunk) >> LG_PAGE;
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
arena_dalloc_small(extent_node_arena_get(
&chunk->node), chunk, ptr, pageind);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
}
} else {
Implement cache index randomization for large allocations. Extract szad size quantization into {extent,run}_quantize(), and . quantize szad run sizes to the union of valid small region run sizes and large run sizes. Refactor iteration in arena_run_first_fit() to use run_quantize{,_first,_next(), and add support for padded large runs. For large allocations that have no specified alignment constraints, compute a pseudo-random offset from the beginning of the first backing page that is a multiple of the cache line size. Under typical configurations with 4-KiB pages and 64-byte cache lines this results in a uniform distribution among 64 page boundary offsets. Add the --disable-cache-oblivious option, primarily intended for performance testing. This resolves #13.
2015-05-05 00:58:36 +08:00
assert(config_cache_oblivious || ((uintptr_t)ptr &
PAGE_MASK) == 0);
Add support for sized deallocation. This adds a new `sdallocx` function to the external API, allowing the size to be passed by the caller. It avoids some extra reads in the thread cache fast path. In the case where stats are enabled, this avoids the work of calculating the size from the pointer. An assertion validates the size that's passed in, so enabling debugging will allow users of the API to debug cases where an incorrect size is passed in. The performance win for a contrived microbenchmark doing an allocation and immediately freeing it is ~10%. It may have a different impact on a real workload. Closes #28
2014-08-29 03:41:48 +08:00
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
if (likely(tcache != NULL) && size <= tcache_maxclass)
tcache_dalloc_large(tsd, tcache, ptr, size);
else {
Integrate whole chunks into unused dirty page purging machinery. Extend per arena unused dirty page purging to manage unused dirty chunks in aaddtion to unused dirty runs. Rather than immediately unmapping deallocated chunks (or purging them in the --disable-munmap case), store them in a separate set of trees, chunks_[sz]ad_dirty. Preferrentially allocate dirty chunks. When excessive unused dirty pages accumulate, purge runs and chunks in ingegrated LRU order (and unmap chunks in the --enable-munmap case). Refactor extent_node_t to provide accessor functions.
2015-02-16 10:04:46 +08:00
arena_dalloc_large(extent_node_arena_get(
&chunk->node), chunk, ptr);
Refactor huge_*() calls into arena internals. Make redirects to the huge_*() API the arena code's responsibility, since arenas now take responsibility for all allocation sizes.
2015-02-13 06:06:37 +08:00
}
}
} else
huge_dalloc(tsd, ptr, tcache);
Add support for sized deallocation. This adds a new `sdallocx` function to the external API, allowing the size to be passed by the caller. It avoids some extra reads in the thread cache fast path. In the case where stats are enabled, this avoids the work of calculating the size from the pointer. An assertion validates the size that's passed in, so enabling debugging will allow users of the API to debug cases where an incorrect size is passed in. The performance win for a contrived microbenchmark doing an allocation and immediately freeing it is ~10%. It may have a different impact on a real workload. Closes #28
2014-08-29 03:41:48 +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
# endif /* JEMALLOC_ARENA_INLINE_B */
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