project-base/server-skynet-source-3rd-jemalloc
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
30d74db08e
server-skynet-source-3rd-je.../include/jemalloc/internal/jemalloc_internal.h.in

1229 lines
34 KiB
C
Raw Normal View History

Replace fprintf with malloc_printf in tests.
2012-04-16 22:30:26 +08:00
#ifndef JEMALLOC_INTERNAL_H
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define JEMALLOC_INTERNAL_H
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
jemalloc cpp new/delete bindings Adds cpp bindings for jemalloc, along with necessary autoconf settings. This is mostly to add sized deallocation support, which can't be added from C directly. Sized deallocation is ~10% microbench improvement. * Import ax_cxx_compile_stdcxx.m4 from the autoconf repo, seems like the easiest way to get c++14 detection. * Adds various other changes, like CXXFLAGS, to configure.ac. * Adds new rules to Makefile.in for src/jemalloc-cpp.cpp, and a basic unittest. * Both new and delete are overridden, to ensure jemalloc is used for both. * TODO future enhancement of avoiding extra PLT thunks for new and delete - sdallocx and malloc are publicly exported jemalloc symbols, using an alias would link them directly. Unfortunately, was having trouble getting it to play nice with jemalloc's namespace support. Testing: Tested gcc 4.8, gcc 5, gcc 5.2, clang 4.0. Only gcc >= 5 has sized deallocation support, verified that the rest build correctly. Tested mac osx and Centos. Tested --with-jemalloc-prefix and --without-export. This resolves #202.
2016-10-24 06:56:30 +08:00
#ifdef __cplusplus
extern "C" {
#endif
Refactor to support more varied testing. Refactor the test harness to support three types of tests: - unit: White box unit tests. These tests have full access to all internal jemalloc library symbols. Though in actuality all symbols are prefixed by jet_, macro-based name mangling abstracts this away from test code. - integration: Black box integration tests. These tests link with the installable shared jemalloc library, and with the exception of some utility code and configure-generated macro definitions, they have no access to jemalloc internals. - stress: Black box stress tests. These tests link with the installable shared jemalloc library, as well as with an internal allocator with symbols prefixed by jet_ (same as for unit tests) that can be used to allocate data structures that are internal to the test code. Move existing tests into test/{unit,integration}/ as appropriate. Split out internal parts of jemalloc_defs.h.in and put them in jemalloc_internal_defs.h.in. This reduces internals exposure to applications that #include <jemalloc/jemalloc.h>. Refactor jemalloc.h header generation so that a single header file results, and the prototypes can be used to generate jet_ prototypes for tests. Split jemalloc.h.in into multiple parts (jemalloc_defs.h.in, jemalloc_macros.h.in, jemalloc_protos.h.in, jemalloc_mangle.h.in) and use a shell script to generate a unified jemalloc.h at configure time. Change the default private namespace prefix from "" to "je_". Add missing private namespace mangling. Remove hard-coded private_namespace.h. Instead generate it and private_unnamespace.h from private_symbols.txt. Use similar logic for public symbols, which aids in name mangling for jet_ symbols. Add test_warn() and test_fail(). Replace existing exit(1) calls with test_fail() calls.
2013-12-01 07:25:42 +08:00
#include "jemalloc_internal_defs.h"
Move platform headers and tricks from jemalloc_internal.h.in to a new jemalloc_internal_decls.h header
2014-05-28 11:39:13 +08:00
#include "jemalloc/internal/jemalloc_internal_decls.h"
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 utrace(2)-based tracing (--enable-utrace).
2012-04-06 04:36:17 +08:00
#ifdef JEMALLOC_UTRACE
#include <sys/ktrace.h>
#endif
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define JEMALLOC_NO_DEMANGLE
Refactor to support more varied testing. Refactor the test harness to support three types of tests: - unit: White box unit tests. These tests have full access to all internal jemalloc library symbols. Though in actuality all symbols are prefixed by jet_, macro-based name mangling abstracts this away from test code. - integration: Black box integration tests. These tests link with the installable shared jemalloc library, and with the exception of some utility code and configure-generated macro definitions, they have no access to jemalloc internals. - stress: Black box stress tests. These tests link with the installable shared jemalloc library, as well as with an internal allocator with symbols prefixed by jet_ (same as for unit tests) that can be used to allocate data structures that are internal to the test code. Move existing tests into test/{unit,integration}/ as appropriate. Split out internal parts of jemalloc_defs.h.in and put them in jemalloc_internal_defs.h.in. This reduces internals exposure to applications that #include <jemalloc/jemalloc.h>. Refactor jemalloc.h header generation so that a single header file results, and the prototypes can be used to generate jet_ prototypes for tests. Split jemalloc.h.in into multiple parts (jemalloc_defs.h.in, jemalloc_macros.h.in, jemalloc_protos.h.in, jemalloc_mangle.h.in) and use a shell script to generate a unified jemalloc.h at configure time. Change the default private namespace prefix from "" to "je_". Add missing private namespace mangling. Remove hard-coded private_namespace.h. Instead generate it and private_unnamespace.h from private_symbols.txt. Use similar logic for public symbols, which aids in name mangling for jet_ symbols. Add test_warn() and test_fail(). Replace existing exit(1) calls with test_fail() calls.
2013-12-01 07:25:42 +08:00
#ifdef JEMALLOC_JET
# define JEMALLOC_N(n) jet_##n
# include "jemalloc/internal/public_namespace.h"
# define JEMALLOC_NO_RENAME
# include "../jemalloc@install_suffix@.h"
Fix name mangling for stress tests. Fix stress tests such that testlib code uses the jet_ allocator, but test code uses libjemalloc. Generate jemalloc_{rename,mangle}.h, the former because it's needed for the stress test name mangling fix, and the latter for consistency. As an artifact of this change, some (but not all) definitions related to the experimental API are absent from the headers unless the feature is enabled at configure time.
2014-01-17 09:38:01 +08:00
# undef JEMALLOC_NO_RENAME
Refactor to support more varied testing. Refactor the test harness to support three types of tests: - unit: White box unit tests. These tests have full access to all internal jemalloc library symbols. Though in actuality all symbols are prefixed by jet_, macro-based name mangling abstracts this away from test code. - integration: Black box integration tests. These tests link with the installable shared jemalloc library, and with the exception of some utility code and configure-generated macro definitions, they have no access to jemalloc internals. - stress: Black box stress tests. These tests link with the installable shared jemalloc library, as well as with an internal allocator with symbols prefixed by jet_ (same as for unit tests) that can be used to allocate data structures that are internal to the test code. Move existing tests into test/{unit,integration}/ as appropriate. Split out internal parts of jemalloc_defs.h.in and put them in jemalloc_internal_defs.h.in. This reduces internals exposure to applications that #include <jemalloc/jemalloc.h>. Refactor jemalloc.h header generation so that a single header file results, and the prototypes can be used to generate jet_ prototypes for tests. Split jemalloc.h.in into multiple parts (jemalloc_defs.h.in, jemalloc_macros.h.in, jemalloc_protos.h.in, jemalloc_mangle.h.in) and use a shell script to generate a unified jemalloc.h at configure time. Change the default private namespace prefix from "" to "je_". Add missing private namespace mangling. Remove hard-coded private_namespace.h. Instead generate it and private_unnamespace.h from private_symbols.txt. Use similar logic for public symbols, which aids in name mangling for jet_ symbols. Add test_warn() and test_fail(). Replace existing exit(1) calls with test_fail() calls.
2013-12-01 07:25:42 +08:00
#else
# define JEMALLOC_N(n) @private_namespace@##n
# include "../jemalloc@install_suffix@.h"
#endif
Adjust relative #include for private_namespace.h.
2011-07-31 08:58:07 +08:00
#include "jemalloc/internal/private_namespace.h"
Add the --with-private-namespace option. Add the --with-private-namespace option to make it possible to work around library-private symbols being exposed in static libraries.
2011-07-31 07:40:52 +08:00
Reduce cpp conditional logic complexity. Convert configuration-related cpp conditional logic to use static constant variables, e.g.: #ifdef JEMALLOC_DEBUG [...] #endif becomes: if (config_debug) { [...] } The advantage is clearer, more concise code. The main disadvantage is that data structures no longer have conditionally defined fields, so they pay the cost of all fields regardless of whether they are used. In practice, this is only a minor concern; config_stats will go away in an upcoming change, and config_prof is the only other major feature that depends on more than a few special-purpose fields.
2012-02-11 12:22:09 +08:00
static const bool config_debug =
#ifdef JEMALLOC_DEBUG
true
#else
false
#endif
;
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
static const bool have_dss =
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
#ifdef JEMALLOC_DSS
true
#else
false
#endif
;
static const bool config_fill =
#ifdef JEMALLOC_FILL
true
#else
false
#endif
;
static const bool config_lazy_lock =
#ifdef JEMALLOC_LAZY_LOCK
true
#else
false
#endif
;
Add --with-malloc-conf. Add --with-malloc-conf, which makes it possible to embed a default options string during configuration.
2016-02-08 06:23:22 +08:00
static const char * const config_malloc_conf = JEMALLOC_CONFIG_MALLOC_CONF;
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
static const bool config_prof =
#ifdef JEMALLOC_PROF
true
#else
false
#endif
;
static const bool config_prof_libgcc =
#ifdef JEMALLOC_PROF_LIBGCC
true
#else
false
#endif
;
static const bool config_prof_libunwind =
#ifdef JEMALLOC_PROF_LIBUNWIND
true
#else
false
#endif
;
Implement support for non-coalescing maps on MinGW. - Do not reallocate huge objects in place if the number of backing chunks would change. - Do not cache multi-chunk mappings. This resolves #213.
2015-07-25 09:21:42 +08:00
static const bool maps_coalesce =
#ifdef JEMALLOC_MAPS_COALESCE
true
#else
false
#endif
;
Disable munmap() if it causes VM map holes. Add a configure test to determine whether common mmap()/munmap() patterns cause VM map holes, and only use munmap() to discard unused chunks if the problem does not exist. Unify the chunk caching for mmap and dss. Fix options processing to limit lg_chunk to be large enough that redzones will always fit.
2012-04-13 11:20:58 +08:00
static const bool config_munmap =
#ifdef JEMALLOC_MUNMAP
true
#else
false
#endif
;
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
static const bool config_stats =
#ifdef JEMALLOC_STATS
true
#else
false
#endif
;
static const bool config_tcache =
#ifdef JEMALLOC_TCACHE
true
#else
false
#endif
;
static const bool config_tls =
#ifdef JEMALLOC_TLS
true
#else
false
#endif
;
Add utrace(2)-based tracing (--enable-utrace).
2012-04-06 04:36:17 +08:00
static const bool config_utrace =
#ifdef JEMALLOC_UTRACE
true
#else
false
#endif
;
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
static const bool config_xmalloc =
#ifdef JEMALLOC_XMALLOC
true
#else
false
#endif
;
Restore --enable-ivsalloc. However, unlike before it was removed do not force --enable-ivsalloc when Darwin zone allocator integration is enabled, since the zone allocator code uses ivsalloc() regardless of whether malloc_usable_size() and sallocx() do. This resolves #211.
2015-03-19 12:06:58 +08:00
static const bool config_ivsalloc =
#ifdef JEMALLOC_IVSALLOC
true
#else
false
#endif
;
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 bool config_cache_oblivious =
#ifdef JEMALLOC_CACHE_OBLIVIOUS
true
#else
false
#endif
;
Add huge page configuration and pages_[no}huge(). Add the --with-lg-hugepage configure option, but automatically configure LG_HUGEPAGE even if it isn't specified. Add the pages_[no]huge() functions, which toggle huge page state via madvise(..., MADV_[NO]HUGEPAGE) calls.
2016-11-18 05:36:17 +08:00
static const bool have_thp =
#ifdef JEMALLOC_THP
true
#else
false
#endif
;
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
#ifdef JEMALLOC_HAVE_SCHED_GETCPU
/* Currently percpu_arena depends on sched_getcpu. */
#define JEMALLOC_PERCPU_ARENA
#endif
static const bool have_percpu_arena =
#ifdef JEMALLOC_PERCPU_ARENA
true
#else
false
#endif
;
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
Use OSSpinLock*() for locking on OS X. pthread_mutex_lock() can call malloc() on OS X (!!!), which causes deadlock. Work around this by using spinlocks that are built of more primitive stuff.
2011-03-19 10:30:18 +08:00
#if (defined(JEMALLOC_OSATOMIC) || defined(JEMALLOC_OSSPIN))
Add atomic operation support for OS X.
2011-03-19 10:10:31 +08:00
#include <libkern/OSAtomic.h>
#endif
Port to Mac OS X. Add Mac OS X support, based in large part on the OS X support in Mozilla's version of jemalloc.
2010-09-06 01:35:13 +08:00
#ifdef JEMALLOC_ZONE
#include <mach/mach_error.h>
#include <mach/mach_init.h>
#include <mach/vm_map.h>
#endif
Refactor/fix ph. Refactor ph to support configurable comparison functions. Use a cpp macro code generation form equivalent to the rb macros so that pairing heaps can be used for both run heaps and chunk heaps. Remove per node parent pointers, and instead use leftmost siblings' prev pointers to track parents. Fix multi-pass sibling merging to iterate over intermediate results using a FIFO, rather than a LIFO. Use this fixed sibling merging implementation for both merge phases of the auxiliary twopass algorithm (first merging the aux list, then replacing the root with its merged children). This fixes both degenerate merge behavior and the potential for deep recursion. This regression was introduced by 6bafa6678fc36483e638f1c3a0a9bf79fb89bfc9 (Pairing heap). This resolves #371.
2016-03-27 08:30:37 +08:00
#include "jemalloc/internal/ph.h"
Don't use compact red-black trees with the pgi compiler Some bug (either in the red-black tree code, or in the pgi compiler) seems to cause red-black trees to become unbalanced. This issue seems to go away if we don't use compact red-black trees. Since red-black trees don't seem to be used much anymore, I opted for what seems to be an easy fix here instead of digging in and trying to find the root cause of the bug. Some context in case it's helpful: I experienced a ton of segfaults while using pgi as Chapel's target compiler with jemalloc 4.0.4. The little bit of debugging I did pointed me somewhere deep in red-black tree manipulation, but I didn't get a chance to investigate further. It looks like 4.2.0 replaced most uses of red-black trees with pairing-heaps, which seems to avoid whatever bug I was hitting. However, `make check_unit` was still failing on the rb test, so I figured the core issue was just being masked. Here's the `make check_unit` failure: ```sh === test/unit/rb === test_rb_empty: pass tree_recurse:test/unit/rb.c:90: Failed assertion: (((_Bool) (((uintptr_t) (left_node)->link.rbn_right_red) & ((size_t)1)))) == (false) --> true != false: Node should be black test_rb_random:test/unit/rb.c:274: Failed assertion: (imbalances) == (0) --> 1 != 0: Tree is unbalanced tree_recurse:test/unit/rb.c:90: Failed assertion: (((_Bool) (((uintptr_t) (left_node)->link.rbn_right_red) & ((size_t)1)))) == (false) --> true != false: Node should be black test_rb_random:test/unit/rb.c:274: Failed assertion: (imbalances) == (0) --> 1 != 0: Tree is unbalanced node_remove:test/unit/rb.c:190: Failed assertion: (imbalances) == (0) --> 2 != 0: Tree is unbalanced <jemalloc>: test/unit/rb.c:43: Failed assertion: "pathp[-1].cmp < 0" test/test.sh: line 22: 12926 Aborted Test harness error ``` While starting to debug I saw the RB_COMPACT option and decided to check if turning that off resolved the bug. It seems to have fixed it (`make check_unit` passes and the segfaults under Chapel are gone) so it seems like on okay work-around. I'd imagine this has performance implications for red-black trees under pgi, but if they're not going to be used much anymore it's probably not a big deal.
2016-06-09 05:48:55 +08:00
#ifndef __PGI
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define RB_COMPACT
Don't use compact red-black trees with the pgi compiler Some bug (either in the red-black tree code, or in the pgi compiler) seems to cause red-black trees to become unbalanced. This issue seems to go away if we don't use compact red-black trees. Since red-black trees don't seem to be used much anymore, I opted for what seems to be an easy fix here instead of digging in and trying to find the root cause of the bug. Some context in case it's helpful: I experienced a ton of segfaults while using pgi as Chapel's target compiler with jemalloc 4.0.4. The little bit of debugging I did pointed me somewhere deep in red-black tree manipulation, but I didn't get a chance to investigate further. It looks like 4.2.0 replaced most uses of red-black trees with pairing-heaps, which seems to avoid whatever bug I was hitting. However, `make check_unit` was still failing on the rb test, so I figured the core issue was just being masked. Here's the `make check_unit` failure: ```sh === test/unit/rb === test_rb_empty: pass tree_recurse:test/unit/rb.c:90: Failed assertion: (((_Bool) (((uintptr_t) (left_node)->link.rbn_right_red) & ((size_t)1)))) == (false) --> true != false: Node should be black test_rb_random:test/unit/rb.c:274: Failed assertion: (imbalances) == (0) --> 1 != 0: Tree is unbalanced tree_recurse:test/unit/rb.c:90: Failed assertion: (((_Bool) (((uintptr_t) (left_node)->link.rbn_right_red) & ((size_t)1)))) == (false) --> true != false: Node should be black test_rb_random:test/unit/rb.c:274: Failed assertion: (imbalances) == (0) --> 1 != 0: Tree is unbalanced node_remove:test/unit/rb.c:190: Failed assertion: (imbalances) == (0) --> 2 != 0: Tree is unbalanced <jemalloc>: test/unit/rb.c:43: Failed assertion: "pathp[-1].cmp < 0" test/test.sh: line 22: 12926 Aborted Test harness error ``` While starting to debug I saw the RB_COMPACT option and decided to check if turning that off resolved the bug. It seems to have fixed it (`make check_unit` passes and the segfaults under Chapel are gone) so it seems like on okay work-around. I'd imagine this has performance implications for red-black trees under pgi, but if they're not going to be used much anymore it's probably not a big deal.
2016-06-09 05:48:55 +08:00
#endif
Restructure source tree.
2010-02-12 06:45:59 +08:00
#include "jemalloc/internal/rb.h"
#include "jemalloc/internal/qr.h"
#include "jemalloc/internal/ql.h"
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 tracing (--enable-trace). Remove all functionality related to tracing. This functionality was useful for understanding memory fragmentation during early algorithmic design of jemalloc, but it had little utility for non-trivial applications, due to the sheer volume of data written to disk.
2010-02-12 05:38:12 +08:00
* jemalloc can conceptually be broken into components (arena, tcache, etc.),
* but there are circular dependencies that cannot be broken without
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
* substantial performance degradation.
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
*
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
* Historically, we dealt with this by each header into four sections (types,
* structs, externs, and inlines), and included each header file multiple times
* in this file, picking out the portion we want on each pass using the
* following #defines:
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
* JEMALLOC_H_TYPES : Preprocessor-defined constants and psuedo-opaque data
* types.
* JEMALLOC_H_STRUCTS : Data structures.
* JEMALLOC_H_EXTERNS : Extern data declarations and function prototypes.
* JEMALLOC_H_INLINES : Inline functions.
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
*
* We're moving toward a world in which the dependencies are explicit; each file
* will #include the headers it depends on (rather than relying on them being
* implicitly available via this file including every header file in the
* project).
*
* We're now in an intermediate state: we've broken up the header files to avoid
* having to include each one multiple times, but have not yet moved the
* dependency information into the header files (i.e. we still rely on the
* ordering in this file to ensure all a header's dependencies are available in
* its translation unit). Each component is now broken up into multiple header
* files, corresponding to the sections above (e.g. instead of "tsd.h", we now
* have "tsd_types.h", "tsd_structs.h", "tsd_externs.h", "tsd_inlines.h").
Introduce a backport of C11 atomics This introduces a backport of C11 atomics. It has four implementations; ranked in order of preference, they are: - GCC/Clang __atomic builtins - GCC/Clang __sync builtins - MSVC _Interlocked builtins - C11 atomics, from <stdatomic.h> The primary advantages are: - Close adherence to the standard API gives us a defined memory model. - Type safety: atomic objects are now separate types from non-atomic ones, so that it's impossible to mix up atomic and non-atomic updates (which is undefined behavior that compilers are starting to take advantage of). - Efficiency: we can specify ordering for operations, avoiding fences and atomic operations on strongly ordered architectures (example: `atomic_write_u32(ptr, val);` involves a CAS loop, whereas `atomic_store(ptr, val, ATOMIC_RELEASE);` is a plain store. This diff leaves in the current atomics API (implementing them in terms of the backport). This lets us transition uses over piecemeal. Testing: This is by nature hard to test. I've manually tested the first three options on Linux on gcc by futzing with the #defines manually, on freebsd with gcc and clang, on MSVC, and on OS X with clang. All of these were x86 machines though, and we don't have any test infrastructure set up for non-x86 platforms.
2017-01-26 01:54:27 +08:00
*
* Those files which have been converted to explicitly include their
* inter-component dependencies are now in the initial HERMETIC HEADERS
* section. These headers may still rely on this file for system headers and
* global jemalloc headers, however.
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
*/
Fix more test refactoring issues.
2013-12-06 13:43:46 +08:00
#include "jemalloc/internal/jemalloc_internal_macros.h"
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
Introduce a backport of C11 atomics This introduces a backport of C11 atomics. It has four implementations; ranked in order of preference, they are: - GCC/Clang __atomic builtins - GCC/Clang __sync builtins - MSVC _Interlocked builtins - C11 atomics, from <stdatomic.h> The primary advantages are: - Close adherence to the standard API gives us a defined memory model. - Type safety: atomic objects are now separate types from non-atomic ones, so that it's impossible to mix up atomic and non-atomic updates (which is undefined behavior that compilers are starting to take advantage of). - Efficiency: we can specify ordering for operations, avoiding fences and atomic operations on strongly ordered architectures (example: `atomic_write_u32(ptr, val);` involves a CAS loop, whereas `atomic_store(ptr, val, ATOMIC_RELEASE);` is a plain store. This diff leaves in the current atomics API (implementing them in terms of the backport). This lets us transition uses over piecemeal. Testing: This is by nature hard to test. I've manually tested the first three options on Linux on gcc by futzing with the #defines manually, on freebsd with gcc and clang, on MSVC, and on OS X with clang. All of these were x86 machines though, and we don't have any test infrastructure set up for non-x86 platforms.
2017-01-26 01:54:27 +08:00
/******************************************************************************/
/* HERMETIC HEADERS */
/******************************************************************************/
Disentangle assert and util This is the first header refactoring diff, #533. It splits the assert and util components into separate, hermetic, header files. In the process, it splits out two of the large sub-components of util (the stdio.h replacement, and bit manipulation routines) into their own components (malloc_io.h and bit_util.h). This is mostly to break up cyclic dependencies, but it also breaks off a good chunk of the catch-all-ness of util, which is nice.
2017-03-04 02:10:08 +08:00
#include "jemalloc/internal/assert.h"
Introduce a backport of C11 atomics This introduces a backport of C11 atomics. It has four implementations; ranked in order of preference, they are: - GCC/Clang __atomic builtins - GCC/Clang __sync builtins - MSVC _Interlocked builtins - C11 atomics, from <stdatomic.h> The primary advantages are: - Close adherence to the standard API gives us a defined memory model. - Type safety: atomic objects are now separate types from non-atomic ones, so that it's impossible to mix up atomic and non-atomic updates (which is undefined behavior that compilers are starting to take advantage of). - Efficiency: we can specify ordering for operations, avoiding fences and atomic operations on strongly ordered architectures (example: `atomic_write_u32(ptr, val);` involves a CAS loop, whereas `atomic_store(ptr, val, ATOMIC_RELEASE);` is a plain store. This diff leaves in the current atomics API (implementing them in terms of the backport). This lets us transition uses over piecemeal. Testing: This is by nature hard to test. I've manually tested the first three options on Linux on gcc by futzing with the #defines manually, on freebsd with gcc and clang, on MSVC, and on OS X with clang. All of these were x86 machines though, and we don't have any test infrastructure set up for non-x86 platforms.
2017-01-26 01:54:27 +08:00
#include "jemalloc/internal/atomic.h"
Disentangle assert and util This is the first header refactoring diff, #533. It splits the assert and util components into separate, hermetic, header files. In the process, it splits out two of the large sub-components of util (the stdio.h replacement, and bit manipulation routines) into their own components (malloc_io.h and bit_util.h). This is mostly to break up cyclic dependencies, but it also breaks off a good chunk of the catch-all-ness of util, which is nice.
2017-03-04 02:10:08 +08:00
#include "jemalloc/internal/bit_util.h"
#include "jemalloc/internal/malloc_io.h"
#include "jemalloc/internal/util.h"
Introduce a backport of C11 atomics This introduces a backport of C11 atomics. It has four implementations; ranked in order of preference, they are: - GCC/Clang __atomic builtins - GCC/Clang __sync builtins - MSVC _Interlocked builtins - C11 atomics, from <stdatomic.h> The primary advantages are: - Close adherence to the standard API gives us a defined memory model. - Type safety: atomic objects are now separate types from non-atomic ones, so that it's impossible to mix up atomic and non-atomic updates (which is undefined behavior that compilers are starting to take advantage of). - Efficiency: we can specify ordering for operations, avoiding fences and atomic operations on strongly ordered architectures (example: `atomic_write_u32(ptr, val);` involves a CAS loop, whereas `atomic_store(ptr, val, ATOMIC_RELEASE);` is a plain store. This diff leaves in the current atomics API (implementing them in terms of the backport). This lets us transition uses over piecemeal. Testing: This is by nature hard to test. I've manually tested the first three options on Linux on gcc by futzing with the #defines manually, on freebsd with gcc and clang, on MSVC, and on OS X with clang. All of these were x86 machines though, and we don't have any test infrastructure set up for non-x86 platforms.
2017-01-26 01:54:27 +08:00
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
/******************************************************************************/
/* TYPES */
/******************************************************************************/
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
/* Page size index type. */
typedef unsigned pszind_t;
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 class index type. */
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
typedef unsigned szind_t;
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
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
/* Processor / core id type. */
typedef int malloc_cpuid_t;
Implement explicit tcache support. Add the MALLOCX_TCACHE() and MALLOCX_TCACHE_NONE macros, which can be used in conjunction with the *allocx() API. Add the tcache.create, tcache.flush, and tcache.destroy mallctls. This resolves #145.
2015-01-30 07:30:47 +08:00
/*
* Flags bits:
*
* a: arena
* t: tcache
* 0: unused
* z: zero
* n: alignment
*
* aaaaaaaa aaaatttt tttttttt 0znnnnnn
*/
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_ARENA_BITS 12
#define MALLOCX_TCACHE_BITS 12
#define MALLOCX_LG_ALIGN_BITS 6
#define MALLOCX_ARENA_SHIFT 20
#define MALLOCX_TCACHE_SHIFT 8
#define MALLOCX_ARENA_MASK \
Refactor ctl_stats_t. Refactor ctl_stats_t to be a demand-zeroed non-growing data structure. To keep the size from being onerous (~60 MiB) on 32-bit systems, convert the arenas field to contain pointers rather than directly embedded ctl_arena_stats_t elements.
2017-01-04 04:40:54 +08:00
(((1 << MALLOCX_ARENA_BITS) - 1) << MALLOCX_ARENA_SHIFT)
Implement arena.<i>.destroy . Add MALLCTL_ARENAS_DESTROYED for accessing destroyed arena stats as an analogue to MALLCTL_ARENAS_ALL. This resolves #382.
2017-01-04 09:21:59 +08:00
/* NB: Arena index bias decreases the maximum number of arenas by 1. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_ARENA_MAX ((1 << MALLOCX_ARENA_BITS) - 2)
#define MALLOCX_TCACHE_MASK \
Refactor ctl_stats_t. Refactor ctl_stats_t to be a demand-zeroed non-growing data structure. To keep the size from being onerous (~60 MiB) on 32-bit systems, convert the arenas field to contain pointers rather than directly embedded ctl_arena_stats_t elements.
2017-01-04 04:40:54 +08:00
(((1 << MALLOCX_TCACHE_BITS) - 1) << MALLOCX_TCACHE_SHIFT)
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_TCACHE_MAX ((1 << MALLOCX_TCACHE_BITS) - 3)
#define MALLOCX_LG_ALIGN_MASK ((1 << MALLOCX_LG_ALIGN_BITS) - 1)
Optimize [nmd]alloc() fast paths. Optimize [nmd]alloc() fast paths such that the (flags == 0) case is streamlined, flags decoding only happens to the minimum degree necessary, and no conditionals are repeated.
2014-09-08 05:40:19 +08:00
/* Use MALLOCX_ALIGN_GET() if alignment may not be specified in flags. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_ALIGN_GET_SPECIFIED(flags) \
Optimize [nmd]alloc() fast paths. Optimize [nmd]alloc() fast paths such that the (flags == 0) case is streamlined, flags decoding only happens to the minimum degree necessary, and no conditionals are repeated.
2014-09-08 05:40:19 +08:00
(ZU(1) << (flags & MALLOCX_LG_ALIGN_MASK))
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_ALIGN_GET(flags) \
Optimize [nmd]alloc() fast paths. Optimize [nmd]alloc() fast paths such that the (flags == 0) case is streamlined, flags decoding only happens to the minimum degree necessary, and no conditionals are repeated.
2014-09-08 05:40:19 +08:00
(MALLOCX_ALIGN_GET_SPECIFIED(flags) & (SIZE_T_MAX-1))
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_ZERO_GET(flags) \
Optimize [nmd]alloc() fast paths. Optimize [nmd]alloc() fast paths such that the (flags == 0) case is streamlined, flags decoding only happens to the minimum degree necessary, and no conditionals are repeated.
2014-09-08 05:40:19 +08:00
((bool)(flags & MALLOCX_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
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_TCACHE_GET(flags) \
Refactor ctl_stats_t. Refactor ctl_stats_t to be a demand-zeroed non-growing data structure. To keep the size from being onerous (~60 MiB) on 32-bit systems, convert the arenas field to contain pointers rather than directly embedded ctl_arena_stats_t elements.
2017-01-04 04:40:54 +08:00
(((unsigned)((flags & MALLOCX_TCACHE_MASK) >> MALLOCX_TCACHE_SHIFT)) - 2)
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define MALLOCX_ARENA_GET(flags) \
Refactor ctl_stats_t. Refactor ctl_stats_t to be a demand-zeroed non-growing data structure. To keep the size from being onerous (~60 MiB) on 32-bit systems, convert the arenas field to contain pointers rather than directly embedded ctl_arena_stats_t elements.
2017-01-04 04:40:54 +08:00
(((unsigned)(((unsigned)flags) >> MALLOCX_ARENA_SHIFT)) - 1)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
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
/* Smallest size class to support. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define TINY_MIN (1U << LG_TINY_MIN)
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
/*
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
* Minimum allocation alignment is 2^LG_QUANTUM bytes (ignoring tiny size
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
* classes).
*/
#ifndef LG_QUANTUM
Add support for MSVC Tested with MSVC 8 32 and 64 bits.
2012-04-30 18:38:31 +08:00
# if (defined(__i386__) || defined(_M_IX86))
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
# define LG_QUANTUM 4
# endif
# ifdef __ia64__
# define LG_QUANTUM 4
# endif
# ifdef __alpha__
# define LG_QUANTUM 4
# endif
Fix LG_QUANTUM definition for sparc64 GCC 4.9.3 cross-compiled for sparc64 defines __sparc_v9__, not __sparc64__ nor __sparcv9. This prevents LG_QUANTUM from being defined properly. Adding this new value to the check solves the issue.
2016-07-15 04:44:01 +08:00
# if (defined(__sparc64__) || defined(__sparcv9) || defined(__sparc_v9__))
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
# define LG_QUANTUM 4
# endif
Add support for MSVC Tested with MSVC 8 32 and 64 bits.
2012-04-30 18:38:31 +08:00
# if (defined(__amd64__) || defined(__x86_64__) || defined(_M_X64))
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
# define LG_QUANTUM 4
# endif
# ifdef __arm__
# define LG_QUANTUM 3
# endif
Add aarch64 LG_QUANTUM size definition Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
2013-03-18 22:40:20 +08:00
# ifdef __aarch64__
# define LG_QUANTUM 4
# endif
Define LG_QUANTUM for hppa. Submitted by Jory Pratt.
2012-10-09 06:41:06 +08:00
# ifdef __hppa__
# define LG_QUANTUM 4
# endif
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
# ifdef __mips__
# define LG_QUANTUM 3
# endif
Add OpenRISC/or1k LG_QUANTUM size definition
2014-07-30 06:11:26 +08:00
# ifdef __or1k__
# define LG_QUANTUM 3
# endif
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
# ifdef __powerpc__
# define LG_QUANTUM 4
# endif
Add LG_QUANTUM definition for the RISC-V architecture.
2016-05-07 08:15:32 +08:00
# ifdef __riscv__
# define LG_QUANTUM 4
# endif
fix building for s390 systems Checking for __s390x__ means you work on s390x, but not s390 (32bit) systems. So use __s390__ which works for both. With this, `make check` passes on s390. Signed-off-by: Mike Frysinger <vapier@gentoo.org>
2013-01-29 04:19:34 +08:00
# ifdef __s390__
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
# define LG_QUANTUM 4
# endif
Add SH4 and mips architecture support. Submitted by Andreas Vinsander.
2012-03-06 04:16:57 +08:00
# ifdef __SH4__
# define LG_QUANTUM 4
# endif
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
# ifdef __tile__
# define LG_QUANTUM 4
# endif
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
# ifdef __le32__
# define LG_QUANTUM 4
# endif
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
# ifndef LG_QUANTUM
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
# error "Unknown minimum alignment for architecture; specify via "
"--with-lg-quantum"
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
# endif
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
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define QUANTUM ((size_t)(1U << LG_QUANTUM))
#define QUANTUM_MASK (QUANTUM - 1)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
/* Return the smallest quantum multiple that is >= a. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define QUANTUM_CEILING(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
(((a) + QUANTUM_MASK) & ~QUANTUM_MASK)
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define LONG ((size_t)(1U << LG_SIZEOF_LONG))
#define LONG_MASK (LONG - 1)
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
/* Return the smallest long multiple that is >= a. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define LONG_CEILING(a) \
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
(((a) + LONG_MASK) & ~LONG_MASK)
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define SIZEOF_PTR (1U << LG_SIZEOF_PTR)
#define PTR_MASK (SIZEOF_PTR - 1)
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
/* Return the smallest (void *) multiple that is >= a. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define PTR_CEILING(a) \
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
(((a) + PTR_MASK) & ~PTR_MASK)
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
/*
* Maximum size of L1 cache line. This is used to avoid cache line aliasing.
* In addition, this controls the spacing of cacheline-spaced size classes.
Make CACHELINE a raw constant. Make CACHELINE a raw constant in order to work around a __declspec(align()) limitation. Submitted by Mike Hommey.
2012-05-02 16:22:16 +08:00
*
* CACHELINE cannot be based on LG_CACHELINE because __declspec(align()) can
* only handle raw constants.
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
*/
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define LG_CACHELINE 6
#define CACHELINE 64
#define CACHELINE_MASK (CACHELINE - 1)
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
/* Return the smallest cacheline multiple that is >= s. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define CACHELINE_CEILING(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
(((s) + CACHELINE_MASK) & ~CACHELINE_MASK)
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
/* Return the nearest aligned address at or below a. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define ALIGNMENT_ADDR2BASE(a, alignment) \
Avoid negation of unsigned numbers. Rather than relying on two's complement negation for alignment mask generation, use bitwise not and addition. This dodges warnings from MSVC, and should be strength-reduced by compiler optimization anyway.
2016-10-28 12:26:33 +08:00
((void *)((uintptr_t)(a) & ((~(alignment)) + 1)))
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
/* Return the offset between a and the nearest aligned address at or below a. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define ALIGNMENT_ADDR2OFFSET(a, alignment) \
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
((size_t)((uintptr_t)(a) & (alignment - 1)))
/* Return the smallest alignment multiple that is >= s. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define ALIGNMENT_CEILING(s, alignment) \
Avoid negation of unsigned numbers. Rather than relying on two's complement negation for alignment mask generation, use bitwise not and addition. This dodges warnings from MSVC, and should be strength-reduced by compiler optimization anyway.
2016-10-28 12:26:33 +08:00
(((s) + (alignment - 1)) & ((~(alignment)) + 1))
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
Style and spelling fixes.
2014-12-09 06:40:14 +08:00
/* Declare a variable-length array. */
Avoid variable length arrays and remove declarations within code MSVC doesn't support C99, and building as C++ to be able to use them is dangerous, as C++ and C99 are incompatible. Introduce a VARIABLE_ARRAY macro that either uses VLA when supported, or alloca() otherwise. Note that using alloca() inside loops doesn't quite work like VLAs, thus the use of VARIABLE_ARRAY there is discouraged. It might be worth investigating ways to check whether VARIABLE_ARRAY is used in such context at runtime in debug builds and bail out if that happens.
2012-04-25 05:22:02 +08:00
#if __STDC_VERSION__ < 199901L
# ifdef _MSC_VER
# include <malloc.h>
# define alloca _alloca
# else
Fix build break on *BSD Linux uses alloca.h; many other operating systems define alloca(3) in stdlib.h. Signed-off-by: Garrett Cooper <yanegomi@gmail.com>
2012-12-03 09:58:40 +08:00
# ifdef JEMALLOC_HAS_ALLOCA_H
# include <alloca.h>
# else
# include <stdlib.h>
# endif
Avoid variable length arrays and remove declarations within code MSVC doesn't support C99, and building as C++ to be able to use them is dangerous, as C++ and C99 are incompatible. Introduce a VARIABLE_ARRAY macro that either uses VLA when supported, or alloca() otherwise. Note that using alloca() inside loops doesn't quite work like VLAs, thus the use of VARIABLE_ARRAY there is discouraged. It might be worth investigating ways to check whether VARIABLE_ARRAY is used in such context at runtime in debug builds and bail out if that happens.
2012-04-25 05:22:02 +08:00
# endif
# define VARIABLE_ARRAY(type, name, count) \
prof_backtrace: use unw_backtrace unw_backtrace: - does internal per-thread caching - doesn't acquire an internal lock
2014-04-22 11:52:35 +08:00
type *name = alloca(sizeof(type) * (count))
Avoid variable length arrays and remove declarations within code MSVC doesn't support C99, and building as C++ to be able to use them is dangerous, as C++ and C99 are incompatible. Introduce a VARIABLE_ARRAY macro that either uses VLA when supported, or alloca() otherwise. Note that using alloca() inside loops doesn't quite work like VLAs, thus the use of VARIABLE_ARRAY there is discouraged. It might be worth investigating ways to check whether VARIABLE_ARRAY is used in such context at runtime in debug builds and bail out if that happens.
2012-04-25 05:22:02 +08:00
#else
prof_backtrace: use unw_backtrace unw_backtrace: - does internal per-thread caching - doesn't acquire an internal lock
2014-04-22 11:52:35 +08:00
# define VARIABLE_ARRAY(type, name, count) type name[(count)]
Avoid variable length arrays and remove declarations within code MSVC doesn't support C99, and building as C++ to be able to use them is dangerous, as C++ and C99 are incompatible. Introduce a VARIABLE_ARRAY macro that either uses VLA when supported, or alloca() otherwise. Note that using alloca() inside loops doesn't quite work like VLAs, thus the use of VARIABLE_ARRAY there is discouraged. It might be worth investigating ways to check whether VARIABLE_ARRAY is used in such context at runtime in debug builds and bail out if that happens.
2012-04-25 05:22:02 +08:00
#endif
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/nstime_types.h"
#include "jemalloc/internal/spin_types.h"
#include "jemalloc/internal/prng_types.h"
#include "jemalloc/internal/ticker_types.h"
#include "jemalloc/internal/ckh_types.h"
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
#include "jemalloc/internal/size_classes.h"
Implement smoothstep table generation. Check in a generated smootherstep table as smoothstep.h rather than generating it at configure time, since not all systems (e.g. Windows) have dc.
2016-02-06 16:46:19 +08:00
#include "jemalloc/internal/smoothstep.h"
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/stats_types.h"
#include "jemalloc/internal/ctl_types.h"
#include "jemalloc/internal/witness_types.h"
#include "jemalloc/internal/mutex_types.h"
#include "jemalloc/internal/tsd_types.h"
#include "jemalloc/internal/extent_types.h"
#include "jemalloc/internal/extent_dss_types.h"
#include "jemalloc/internal/base_types.h"
#include "jemalloc/internal/arena_types.h"
#include "jemalloc/internal/bitmap_types.h"
#include "jemalloc/internal/rtree_types.h"
#include "jemalloc/internal/pages_types.h"
#include "jemalloc/internal/tcache_types.h"
#include "jemalloc/internal/prof_types.h"
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
/******************************************************************************/
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
/* STRUCTS */
/******************************************************************************/
#include "jemalloc/internal/nstime_structs.h"
#include "jemalloc/internal/spin_structs.h"
#include "jemalloc/internal/ticker_structs.h"
#include "jemalloc/internal/ckh_structs.h"
#include "jemalloc/internal/witness_structs.h"
#include "jemalloc/internal/mutex_structs.h"
Convert arena->stats synchronization to atomics.
2017-02-13 09:43:33 +08:00
#include "jemalloc/internal/stats_structs.h"
#include "jemalloc/internal/ctl_structs.h"
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/bitmap_structs.h"
#include "jemalloc/internal/arena_structs_a.h"
#include "jemalloc/internal/extent_structs.h"
#include "jemalloc/internal/extent_dss_structs.h"
#include "jemalloc/internal/base_structs.h"
Convert arena->prof_accumbytes synchronization to atomics.
2017-02-13 09:03:46 +08:00
#include "jemalloc/internal/prof_structs.h"
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/arena_structs_b.h"
#include "jemalloc/internal/rtree_structs.h"
#include "jemalloc/internal/tcache_structs.h"
#include "jemalloc/internal/tsd_structs.h"
/******************************************************************************/
/* EXTERNS */
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
/******************************************************************************/
extern bool opt_abort;
Introduce two new modes of junk filling: "alloc" and "free". In addition to true/false, opt.junk can now be either "alloc" or "free", giving applications the possibility of junking memory only on allocation or deallocation. This resolves #172.
2014-12-09 05:12:41 +08:00
extern const char *opt_junk;
extern bool opt_junk_alloc;
extern bool opt_junk_free;
Add utrace(2)-based tracing (--enable-utrace).
2012-04-06 04:36:17 +08:00
extern bool opt_utrace;
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
extern bool opt_xmalloc;
extern bool opt_zero;
Make opt_narenas unsigned rather than size_t.
2016-02-25 03:03:40 +08:00
extern unsigned opt_narenas;
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
/* Number of CPUs. */
Refactor arenas array (fixes deadlock). Refactor the arenas array, which contains pointers to all extant arenas, such that it starts out as a sparse array of maximum size, and use double-checked atomics-based reads as the basis for fast and simple arena_get(). Additionally, reduce arenas_lock's role such that it only protects against arena initalization races. These changes remove the possibility for arena lookups to trigger locking, which resolves at least one known (fork-related) deadlock. This resolves #315.
2016-02-25 15:58:10 +08:00
extern unsigned ncpus;
Do not allocate metadata via non-auto arenas, nor tcaches. This assures that all internally allocated metadata come from the first opt_narenas arenas, i.e. the automatically multiplexed arenas.
2016-04-23 05:34:14 +08:00
/* Number of arenas used for automatic multiplexing of threads and arenas. */
extern unsigned narenas_auto;
Refactor arenas array (fixes deadlock). Refactor the arenas array, which contains pointers to all extant arenas, such that it starts out as a sparse array of maximum size, and use double-checked atomics-based reads as the basis for fast and simple arena_get(). Additionally, reduce arenas_lock's role such that it only protects against arena initalization races. These changes remove the possibility for arena lookups to trigger locking, which resolves at least one known (fork-related) deadlock. This resolves #315.
2016-02-25 15:58:10 +08:00
/*
* Arenas that are used to service external requests. Not all elements of the
* arenas array are necessarily used; arenas are created lazily as needed.
*/
Move arena-tracking atomics in jemalloc.c to C11-style
2017-04-05 06:12:24 +08:00
extern atomic_p_t arenas[];
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 runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
/*
* pind2sz_tab encodes the same information as could be computed by
* pind2sz_compute().
*/
Fix psz/pind edge cases. Add an "over-size" extent heap in which to store extents which exceed the maximum size class (plus cache-oblivious padding, if enabled). Remove psz2ind_clamp() and use psz2ind() instead so that trying to allocate the maximum size class can in principle succeed. In practice, this allows assertions to hold so that OOM errors can be successfully generated.
2016-11-04 12:18:50 +08:00
extern size_t const pind2sz_tab[NPSIZES+1];
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
/*
* index2size_tab encodes the same information as could be computed (at
* unacceptable cost in some code paths) by index2size_compute().
*/
Refactor runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
extern size_t const index2size_tab[NSIZES];
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
/*
* size2index_tab is a compact lookup table that rounds request sizes up to
* size classes. In order to reduce cache footprint, the table is compressed,
* and all accesses are via size2index().
*/
extern uint8_t const size2index_tab[];
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 *a0malloc(size_t size);
Refactor bootstrapping to delay tsd initialization. Refactor bootstrapping to delay tsd initialization, primarily to support integration with FreeBSD's libc. Refactor a0*() for internal-only use, and add the bootstrap_{malloc,calloc,free}() API for use by FreeBSD's libc. This separation limits use of the a0*() functions to metadata allocation, which doesn't require malloc/calloc/free API compatibility. This resolves #170.
2015-01-21 07:37:51 +08:00
void a0dalloc(void *ptr);
Add missing prototypes for bootstrap_{malloc,calloc,free}().
2015-02-04 04:39:55 +08:00
void *bootstrap_malloc(size_t size);
void *bootstrap_calloc(size_t num, size_t size);
void bootstrap_free(void *ptr);
Implement arena.<i>.destroy . Add MALLCTL_ARENAS_DESTROYED for accessing destroyed arena stats as an analogue to MALLCTL_ARENAS_ALL. This resolves #382.
2017-01-04 09:21:59 +08:00
void arena_set(unsigned ind, arena_t *arena);
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
unsigned narenas_total_get(void);
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
arena_t *arena_init(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks);
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
arena_tdata_t *arena_tdata_get_hard(tsd_t *tsd, unsigned ind);
Do not allocate metadata via non-auto arenas, nor tcaches. This assures that all internally allocated metadata come from the first opt_narenas arenas, i.e. the automatically multiplexed arenas.
2016-04-23 05:34:14 +08:00
arena_t *arena_choose_hard(tsd_t *tsd, bool internal);
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
void arena_migrate(tsd_t *tsd, unsigned oldind, unsigned newind);
Do not allocate metadata via non-auto arenas, nor tcaches. This assures that all internally allocated metadata come from the first opt_narenas arenas, i.e. the automatically multiplexed arenas.
2016-04-23 05:34:14 +08:00
void iarena_cleanup(tsd_t *tsd);
Convert all tsd variables to reside in a single tsd structure.
2014-09-23 12:09:23 +08:00
void arena_cleanup(tsd_t *tsd);
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
void arenas_tdata_cleanup(tsd_t *tsd);
Port to Mac OS X. Add Mac OS X support, based in large part on the OS X support in Mozilla's version of jemalloc.
2010-09-06 01:35:13 +08:00
void jemalloc_prefork(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 jemalloc_postfork_parent(void);
void jemalloc_postfork_child(void);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/nstime_externs.h"
#include "jemalloc/internal/ckh_externs.h"
#include "jemalloc/internal/stats_externs.h"
#include "jemalloc/internal/ctl_externs.h"
#include "jemalloc/internal/witness_externs.h"
#include "jemalloc/internal/mutex_externs.h"
#include "jemalloc/internal/bitmap_externs.h"
#include "jemalloc/internal/extent_externs.h"
#include "jemalloc/internal/extent_dss_externs.h"
#include "jemalloc/internal/extent_mmap_externs.h"
#include "jemalloc/internal/base_externs.h"
#include "jemalloc/internal/arena_externs.h"
#include "jemalloc/internal/rtree_externs.h"
#include "jemalloc/internal/pages_externs.h"
#include "jemalloc/internal/large_externs.h"
#include "jemalloc/internal/tcache_externs.h"
#include "jemalloc/internal/prof_externs.h"
#include "jemalloc/internal/tsd_externs.h"
/******************************************************************************/
/* INLINES */
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
/******************************************************************************/
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/spin_inlines.h"
#include "jemalloc/internal/prng_inlines.h"
#include "jemalloc/internal/ticker_inlines.h"
#include "jemalloc/internal/tsd_inlines.h"
#include "jemalloc/internal/witness_inlines.h"
#include "jemalloc/internal/mutex_inlines.h"
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
#ifndef JEMALLOC_ENABLE_INLINE
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
pszind_t psz2ind(size_t psz);
Refactor runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
size_t pind2sz_compute(pszind_t pind);
size_t pind2sz_lookup(pszind_t pind);
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
size_t pind2sz(pszind_t pind);
size_t psz2u(size_t psz);
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
szind_t size2index_compute(size_t size);
szind_t size2index_lookup(size_t size);
szind_t size2index(size_t size);
size_t index2size_compute(szind_t index);
size_t index2size_lookup(szind_t index);
size_t index2size(szind_t index);
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 s2u_compute(size_t size);
size_t s2u_lookup(size_t size);
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
size_t s2u(size_t size);
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
size_t sa2u(size_t size, size_t alignment);
Fix huge_palloc() regression. Split arena_choose() into arena_[i]choose() and use arena_ichoose() for arena lookup during internal allocation. This fixes huge_palloc() so that it always succeeds during extent node allocation. This regression was introduced by 66cd953514a18477eb49732e40d5c2ab5f1b12c5 (Do not allocate metadata via non-auto arenas, nor tcaches.).
2016-05-04 06:00:42 +08:00
arena_t *arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal);
arena_t *arena_choose(tsd_t *tsd, arena_t *arena);
Do not (recursively) allocate within tsd_fetch(). Refactor tsd so that tsdn_fetch() does not trigger allocation, since allocation could cause infinite recursion. This resolves #458.
2016-10-21 14:59:12 +08:00
arena_t *arena_ichoose(tsd_t *tsd, arena_t *arena);
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
arena_tdata_t *arena_tdata_get(tsd_t *tsd, unsigned ind,
bool refresh_if_missing);
Resolve bootstrapping issues when embedded in FreeBSD libc. b2c0d6322d2307458ae2b28545f8a5c9903d7ef5 (Add witness, a simple online locking validator.) caused a broad propagation of tsd throughout the internal API, but tsd_fetch() was designed to fail prior to tsd bootstrapping. Fix this by splitting tsd_t into non-nullable tsd_t and nullable tsdn_t, and modifying all internal APIs that do not critically rely on tsd to take nullable pointers. Furthermore, add the tsd_booted_get() function so that tsdn_fetch() can probe whether tsd bootstrapping is complete and return NULL if not. All dangerous conversions of nullable pointers are tsdn_tsd() calls that assert-fail on invalid conversion.
2016-05-11 13:21:10 +08:00
arena_t *arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing);
Implement decay-based unused dirty page purging. This is an alternative to the existing ratio-based unused dirty page purging, and is intended to eventually become the sole purging mechanism. Add mallctls: - opt.purge - opt.decay_time - arena.<i>.decay - arena.<i>.decay_time - arenas.decay_time - stats.arenas.<i>.decay_time This resolves #325.
2016-02-20 12:09:31 +08:00
ticker_t *decay_ticker_get(tsd_t *tsd, unsigned ind);
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
malloc_cpuid_t malloc_getcpu(void);
unsigned percpu_arena_choose(void);
unsigned percpu_arena_ind_limit(void);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
#endif
#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))
Fix size class overflow bugs. Avoid calling s2u() on raw extent sizes in extent_recycle(). Clamp psz2ind() (implemented as psz2ind_clamp()) when inserting/removing into/from size-segregated extent heaps.
2016-10-04 05:18:55 +08:00
JEMALLOC_ALWAYS_INLINE pszind_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
psz2ind(size_t psz) {
if (unlikely(psz > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NPSIZES;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
{
pszind_t x = lg_floor((psz<<1)-1);
pszind_t shift = (x < LG_SIZE_CLASS_GROUP + LG_PAGE) ? 0 : x -
(LG_SIZE_CLASS_GROUP + LG_PAGE);
pszind_t grp = shift << LG_SIZE_CLASS_GROUP;
pszind_t lg_delta = (x < LG_SIZE_CLASS_GROUP + LG_PAGE + 1) ?
LG_PAGE : x - LG_SIZE_CLASS_GROUP - 1;
Make type abbreviations consistent: ssize_t is zd everywhere
2017-03-07 03:40:29 +08:00
size_t delta_inverse_mask = ZD(-1) << lg_delta;
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
pszind_t mod = ((((psz-1) & delta_inverse_mask) >> lg_delta)) &
((ZU(1) << LG_SIZE_CLASS_GROUP) - 1);
pszind_t ind = grp + mod;
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ind;
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
}
}
JEMALLOC_INLINE size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
pind2sz_compute(pszind_t pind) {
if (unlikely(pind == NPSIZES)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return LARGE_MAXCLASS + PAGE;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
{
size_t grp = pind >> LG_SIZE_CLASS_GROUP;
size_t mod = pind & ((ZU(1) << LG_SIZE_CLASS_GROUP) - 1);
size_t grp_size_mask = ~((!!grp)-1);
size_t grp_size = ((ZU(1) << (LG_PAGE +
(LG_SIZE_CLASS_GROUP-1))) << grp) & grp_size_mask;
size_t shift = (grp == 0) ? 1 : grp;
size_t lg_delta = shift + (LG_PAGE-1);
size_t mod_size = (mod+1) << lg_delta;
size_t sz = grp_size + mod_size;
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return sz;
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
}
}
Refactor runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
JEMALLOC_INLINE size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
pind2sz_lookup(pszind_t pind) {
Refactor runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
size_t ret = (size_t)pind2sz_tab[pind];
assert(ret == pind2sz_compute(pind));
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
Refactor runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
}
JEMALLOC_INLINE size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
pind2sz(pszind_t pind) {
Fix psz/pind edge cases. Add an "over-size" extent heap in which to store extents which exceed the maximum size class (plus cache-oblivious padding, if enabled). Remove psz2ind_clamp() and use psz2ind() instead so that trying to allocate the maximum size class can in principle succeed. In practice, this allows assertions to hold so that OOM errors can be successfully generated.
2016-11-04 12:18:50 +08:00
assert(pind < NPSIZES+1);
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return pind2sz_lookup(pind);
Refactor runs_avail. Use pszind_t size classes rather than szind_t size classes, and always reserve space for NPSIZES elements. This removes unused heaps that are not multiples of the page size, and adds (currently) unused heaps for all huge size classes, with the immediate benefit that the size of arena_t allocations is constant (no longer dependent on chunk size).
2016-04-09 05:17:57 +08:00
}
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
JEMALLOC_INLINE size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
psz2u(size_t psz) {
if (unlikely(psz > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return LARGE_MAXCLASS + PAGE;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
{
size_t x = lg_floor((psz<<1)-1);
size_t lg_delta = (x < LG_SIZE_CLASS_GROUP + LG_PAGE + 1) ?
LG_PAGE : x - LG_SIZE_CLASS_GROUP - 1;
size_t delta = ZU(1) << lg_delta;
size_t delta_mask = delta - 1;
size_t usize = (psz + delta_mask) & ~delta_mask;
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return usize;
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
}
}
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
JEMALLOC_INLINE szind_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
size2index_compute(size_t size) {
if (unlikely(size > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NSIZES;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +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
#if (NTBINS != 0)
if (size <= (ZU(1) << LG_TINY_MAXCLASS)) {
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t lg_tmin = LG_TINY_MAXCLASS - NTBINS + 1;
szind_t lg_ceil = lg_floor(pow2_ceil_zu(size));
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
return (lg_ceil < lg_tmin ? 0 : lg_ceil - lg_tmin);
Fix size class overflow handling when profiling is enabled. Fix size class overflow handling for malloc(), posix_memalign(), memalign(), calloc(), and realloc() when profiling is enabled. Remove an assertion that erroneously caused arena_sdalloc() to fail when profiling was enabled. This resolves #232.
2015-06-24 09:47:07 +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
{
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
szind_t x = lg_floor((size<<1)-1);
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t shift = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM) ? 0 :
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
x - (LG_SIZE_CLASS_GROUP + LG_QUANTUM);
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t grp = shift << LG_SIZE_CLASS_GROUP;
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
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t lg_delta = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM + 1)
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
? LG_QUANTUM : x - LG_SIZE_CLASS_GROUP - 1;
Make type abbreviations consistent: ssize_t is zd everywhere
2017-03-07 03:40:29 +08:00
size_t delta_inverse_mask = ZD(-1) << lg_delta;
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t mod = ((((size-1) & delta_inverse_mask) >> lg_delta)) &
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
((ZU(1) << LG_SIZE_CLASS_GROUP) - 1);
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t index = NTBINS + grp + mod;
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return index;
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
}
}
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
JEMALLOC_ALWAYS_INLINE szind_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
size2index_lookup(size_t size) {
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
assert(size <= LOOKUP_MAXCLASS);
{
Change lg_floor() return type from size_t to unsigned.
2016-02-25 03:04:51 +08:00
szind_t ret = (size2index_tab[(size-1) >> LG_TINY_MIN]);
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
assert(ret == size2index_compute(size));
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
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
}
}
Rename index_t to szind_t to avoid an existing type on Solaris. This resolves #256.
2015-08-20 06:21:32 +08:00
JEMALLOC_ALWAYS_INLINE szind_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
size2index(size_t size) {
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
assert(size > 0);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (likely(size <= LOOKUP_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return size2index_lookup(size);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return size2index_compute(size);
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_INLINE size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
index2size_compute(szind_t index) {
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
#if (NTBINS > 0)
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (index < NTBINS) {
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
return (ZU(1) << (LG_TINY_MAXCLASS - NTBINS + 1 + index));
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +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
{
size_t reduced_index = index - NTBINS;
size_t grp = reduced_index >> LG_SIZE_CLASS_GROUP;
size_t mod = reduced_index & ((ZU(1) << LG_SIZE_CLASS_GROUP) -
1);
size_t grp_size_mask = ~((!!grp)-1);
size_t grp_size = ((ZU(1) << (LG_QUANTUM +
(LG_SIZE_CLASS_GROUP-1))) << grp) & grp_size_mask;
size_t shift = (grp == 0) ? 1 : grp;
size_t lg_delta = shift + (LG_QUANTUM-1);
size_t mod_size = (mod+1) << lg_delta;
size_t usize = grp_size + mod_size;
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return usize;
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 size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
index2size_lookup(szind_t index) {
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 ret = (size_t)index2size_tab[index];
assert(ret == index2size_compute(index));
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
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 size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
index2size(szind_t index) {
Make *allocx() size class overflow behavior defined. Limit supported size and alignment to HUGE_MAXCLASS, which in turn is now limited to be less than PTRDIFF_MAX. This resolves #278 and #295.
2016-02-26 07:29:49 +08:00
assert(index < NSIZES);
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return index2size_lookup(index);
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 size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
s2u_compute(size_t size) {
if (unlikely(size > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return 0;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +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
#if (NTBINS > 0)
if (size <= (ZU(1) << LG_TINY_MAXCLASS)) {
size_t lg_tmin = LG_TINY_MAXCLASS - NTBINS + 1;
Refactor prng* from cpp macros into inline functions. Remove 32-bit variant, convert prng64() to prng_lg_range(), and add prng_range().
2016-02-10 08:28:40 +08:00
size_t lg_ceil = lg_floor(pow2_ceil_zu(size));
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
return (lg_ceil < lg_tmin ? (ZU(1) << lg_tmin) :
(ZU(1) << lg_ceil));
Fix size class overflow handling when profiling is enabled. Fix size class overflow handling for malloc(), posix_memalign(), memalign(), calloc(), and realloc() when profiling is enabled. Remove an assertion that erroneously caused arena_sdalloc() to fail when profiling was enabled. This resolves #232.
2015-06-24 09:47:07 +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
{
Implement pz2ind(), pind2sz(), and psz2u(). These compute size classes and indices similarly to size2index(), index2size() and s2u(), respectively, but using the subset of size classes that are multiples of the page size. Note that pszind_t and szind_t are not interchangeable.
2016-04-18 07:16:11 +08:00
size_t x = lg_floor((size<<1)-1);
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 lg_delta = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM + 1)
? LG_QUANTUM : x - LG_SIZE_CLASS_GROUP - 1;
size_t delta = ZU(1) << lg_delta;
size_t delta_mask = delta - 1;
size_t usize = (size + delta_mask) & ~delta_mask;
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return usize;
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 size_t
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
s2u_lookup(size_t size) {
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 ret = index2size_lookup(size2index_lookup(size));
assert(ret == s2u_compute(size));
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
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
}
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
/*
* Compute usable size that would result from allocating an object with the
* specified size.
*/
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
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
s2u(size_t size) {
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
assert(size > 0);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (likely(size <= LOOKUP_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return s2u_lookup(size);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return s2u_compute(size);
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
}
/*
* Compute usable size that would result from allocating an object with the
* specified size and alignment.
*/
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
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
sa2u(size_t size, size_t alignment) {
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
size_t usize;
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
assert(alignment != 0 && ((alignment - 1) & alignment) == 0);
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
/* Try for a small size class. */
if (size <= SMALL_MAXCLASS && alignment < PAGE) {
/*
* Round size up to the nearest multiple of alignment.
*
* This done, we can take advantage of the fact that for each
* small size class, every object is aligned at the smallest
* power of two that is non-zero in the base two representation
* of the size. For example:
*
* Size | Base 2 | Minimum alignment
* -----+----------+------------------
* 96 | 1100000 | 32
* 144 | 10100000 | 32
* 192 | 11000000 | 64
*/
usize = s2u(ALIGNMENT_CEILING(size, alignment));
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (usize < LARGE_MINCLASS) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return usize;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
}
Rename huge to large.
2016-06-01 05:50:21 +08:00
/* Large size class. Beware of overflow. */
Make *allocx() size class overflow behavior defined. Limit supported size and alignment to HUGE_MAXCLASS, which in turn is now limited to be less than PTRDIFF_MAX. This resolves #278 and #295.
2016-02-26 07:29:49 +08:00
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (unlikely(alignment > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return 0;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +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
Use huge size class infrastructure for large size classes.
2016-05-28 15:17:28 +08:00
/* Make sure result is a large size class. */
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (size <= LARGE_MINCLASS) {
Use huge size class infrastructure for large size classes.
2016-05-28 15:17:28 +08:00
usize = LARGE_MINCLASS;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
} else {
Normalize size classes. Normalize size classes to use the same number of size classes per size doubling (currently hard coded to 4), across the intire range of size classes. Small size classes already used this spacing, but in order to support this change, additional small size classes now fill [4 KiB .. 16 KiB). Large size classes range from [16 KiB .. 4 MiB). Huge size classes now support non-multiples of the chunk size in order to fill (4 MiB .. 16 MiB).
2014-10-06 08:54:10 +08:00
usize = s2u(size);
if (usize < size) {
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
/* size_t overflow. */
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return 0;
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +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
}
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +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
/*
Rename huge to large.
2016-06-01 05:50:21 +08:00
* Calculate the multi-page mapping that large_palloc() would need in
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
* order to guarantee the alignment.
*/
Fix potential VM map fragmentation regression. Revert 245ae6036c09cc11a72fab4335495d95cddd5beb (Support --with-lg-page values larger than actual page size.), because it could cause VM map fragmentation if the kernel grows mmap()ed memory downward. This resolves #391.
2016-06-08 05:15:49 +08:00
if (usize + large_pad + PAGE_CEILING(alignment) - PAGE < usize) {
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 overflow. */
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return 0;
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return usize;
Add per thread allocation counters, and enhance heap sampling. Add the "thread.allocated" and "thread.deallocated" mallctls, which can be used to query the total number of bytes ever allocated/deallocated by the calling thread. Add s2u() and sa2u(), which can be used to compute the usable size that will result from an allocation request of a particular size/alignment. Re-factor ipalloc() to use sa2u(). Enhance the heap profiler to trigger samples based on usable size, rather than request size. This has a subtle, but important, impact on the accuracy of heap sampling. For example, previous to this change, 16- and 17-byte objects were sampled at nearly the same rate, but 17-byte objects actually consume 32 bytes each. Therefore it was possible for the sample to be somewhat skewed compared to actual memory usage of the allocated objects.
2010-10-21 08:39:18 +08:00
}
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
JEMALLOC_ALWAYS_INLINE malloc_cpuid_t
malloc_getcpu(void) {
assert(have_percpu_arena);
#if defined(JEMALLOC_HAVE_SCHED_GETCPU)
return (malloc_cpuid_t)sched_getcpu();
#else
not_reached();
return -1;
#endif
}
Add a0malloc(), a0calloc(), and a0free(). Add a0malloc(), a0calloc(), and a0free(), which are used by FreeBSD's libc to allocate/deallocate TLS in static binaries.
2012-04-04 00:28:00 +08:00
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
/* Return the chosen arena index based on current cpu. */
JEMALLOC_ALWAYS_INLINE unsigned
percpu_arena_choose(void) {
unsigned arena_ind;
assert(have_percpu_arena && (percpu_arena_mode != percpu_arena_disabled));
malloc_cpuid_t cpuid = malloc_getcpu();
assert(cpuid >= 0);
if ((percpu_arena_mode == percpu_arena) ||
((unsigned)cpuid < ncpus / 2)) {
arena_ind = cpuid;
} else {
assert(percpu_arena_mode == per_phycpu_arena);
/* Hyper threads on the same physical CPU share arena. */
arena_ind = cpuid - ncpus / 2;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +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
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
return arena_ind;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
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
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
/* Return the limit of percpu auto arena range, i.e. arenas[0...ind_limit). */
JEMALLOC_ALWAYS_INLINE unsigned
percpu_arena_ind_limit(void) {
assert(have_percpu_arena && (percpu_arena_mode != percpu_arena_disabled));
if (percpu_arena_mode == per_phycpu_arena && ncpus > 1) {
if (ncpus % 2) {
/* This likely means a misconfig. */
return ncpus / 2 + 1;
}
return ncpus / 2;
} else {
return ncpus;
}
Fix huge_palloc() regression. Split arena_choose() into arena_[i]choose() and use arena_ichoose() for arena lookup during internal allocation. This fixes huge_palloc() so that it always succeeds during extent node allocation. This regression was introduced by 66cd953514a18477eb49732e40d5c2ab5f1b12c5 (Do not allocate metadata via non-auto arenas, nor tcaches.).
2016-05-04 06:00:42 +08:00
}
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
JEMALLOC_INLINE arena_tdata_t *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing) {
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
arena_tdata_t *tdata;
arena_tdata_t *arenas_tdata = tsd_arenas_tdata_get(tsd);
if (unlikely(arenas_tdata == NULL)) {
/* arenas_tdata hasn't been initialized yet. */
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return arena_tdata_get_hard(tsd, ind);
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
}
if (unlikely(ind >= tsd_narenas_tdata_get(tsd))) {
/*
* ind is invalid, cache is old (too small), or tdata to be
* initialized.
*/
return (refresh_if_missing ? arena_tdata_get_hard(tsd, ind) :
NULL);
}
tdata = &arenas_tdata[ind];
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (likely(tdata != NULL) || !refresh_if_missing) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return tdata;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return arena_tdata_get_hard(tsd, ind);
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
}
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
JEMALLOC_INLINE arena_t *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing) {
Refactor arenas array (fixes deadlock). Refactor the arenas array, which contains pointers to all extant arenas, such that it starts out as a sparse array of maximum size, and use double-checked atomics-based reads as the basis for fast and simple arena_get(). Additionally, reduce arenas_lock's role such that it only protects against arena initalization races. These changes remove the possibility for arena lookups to trigger locking, which resolves at least one known (fork-related) deadlock. This resolves #315.
2016-02-25 15:58:10 +08:00
arena_t *ret;
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
Refactor arenas array (fixes deadlock). Refactor the arenas array, which contains pointers to all extant arenas, such that it starts out as a sparse array of maximum size, and use double-checked atomics-based reads as the basis for fast and simple arena_get(). Additionally, reduce arenas_lock's role such that it only protects against arena initalization races. These changes remove the possibility for arena lookups to trigger locking, which resolves at least one known (fork-related) deadlock. This resolves #315.
2016-02-25 15:58:10 +08:00
assert(ind <= MALLOCX_ARENA_MAX);
Refactor arenas_cache tsd. Refactor arenas_cache tsd into arenas_tdata, which is a structure of type arena_tdata_t.
2016-02-20 11:37:10 +08:00
Move arena-tracking atomics in jemalloc.c to C11-style
2017-04-05 06:12:24 +08:00
ret = (arena_t *)atomic_load_p(&arenas[ind], ATOMIC_ACQUIRE);
Refactor arenas array (fixes deadlock). Refactor the arenas array, which contains pointers to all extant arenas, such that it starts out as a sparse array of maximum size, and use double-checked atomics-based reads as the basis for fast and simple arena_get(). Additionally, reduce arenas_lock's role such that it only protects against arena initalization races. These changes remove the possibility for arena lookups to trigger locking, which resolves at least one known (fork-related) deadlock. This resolves #315.
2016-02-25 15:58:10 +08:00
if (unlikely(ret == NULL)) {
Move arena-tracking atomics in jemalloc.c to C11-style
2017-04-05 06:12:24 +08:00
if (init_if_missing) {
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
ret = arena_init(tsdn, ind,
(extent_hooks_t *)&extent_hooks_default);
}
Refactor arenas array (fixes deadlock). Refactor the arenas array, which contains pointers to all extant arenas, such that it starts out as a sparse array of maximum size, and use double-checked atomics-based reads as the basis for fast and simple arena_get(). Additionally, reduce arenas_lock's role such that it only protects against arena initalization races. These changes remove the possibility for arena lookups to trigger locking, which resolves at least one known (fork-related) deadlock. This resolves #315.
2016-02-25 15:58:10 +08:00
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
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
}
Implement decay-based unused dirty page purging. This is an alternative to the existing ratio-based unused dirty page purging, and is intended to eventually become the sole purging mechanism. Add mallctls: - opt.purge - opt.decay_time - arena.<i>.decay - arena.<i>.decay_time - arenas.decay_time - stats.arenas.<i>.decay_time This resolves #325.
2016-02-20 12:09:31 +08:00
JEMALLOC_INLINE ticker_t *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
decay_ticker_get(tsd_t *tsd, unsigned ind) {
Implement decay-based unused dirty page purging. This is an alternative to the existing ratio-based unused dirty page purging, and is intended to eventually become the sole purging mechanism. Add mallctls: - opt.purge - opt.decay_time - arena.<i>.decay - arena.<i>.decay_time - arenas.decay_time - stats.arenas.<i>.decay_time This resolves #325.
2016-02-20 12:09:31 +08:00
arena_tdata_t *tdata;
tdata = arena_tdata_get(tsd, ind, true);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (unlikely(tdata == NULL)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NULL;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return &tdata->decay_ticker;
Implement decay-based unused dirty page purging. This is an alternative to the existing ratio-based unused dirty page purging, and is intended to eventually become the sole purging mechanism. Add mallctls: - opt.purge - opt.decay_time - arena.<i>.decay - arena.<i>.decay_time - arenas.decay_time - stats.arenas.<i>.decay_time This resolves #325.
2016-02-20 12:09:31 +08:00
}
Fix "thread.{de,}allocatedp" mallctl. For the non-TLS case (as on OS X), if the "thread.{de,}allocatedp" mallctl was called before any allocation occurred for that thread, the TSD was still NULL, thus putting the application at risk of dereferencing NULL. Fix this by refactoring the initialization code, and making it part of the conditional logic for all per thread allocation counter accesses.
2011-02-14 10:11:54 +08:00
#endif
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
Incorporate szind/slab into rtree leaves. Expand and restructure the rtree API such that all common operations can be achieved with minimal work, regardless of whether the rtree leaf fields are independent versus packed into a single atomic pointer.
2017-03-17 08:57:52 +08:00
#include "jemalloc/internal/rtree_inlines.h"
Convert extent_t's usize to szind. Rather than storing usize only for large (and prof-promoted) allocations, store the size class index for allocations that reside within the extent, such that the size class index is valid for all extents that contain extant allocations, and invalid otherwise (mainly to make debugging simpler).
2017-03-14 08:36:57 +08:00
#include "jemalloc/internal/base_inlines.h"
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/bitmap_inlines.h"
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
/*
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
* Include portions of arena code interleaved with tcache code in order to
* resolve circular dependencies.
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
*/
Convert arena->prof_accumbytes synchronization to atomics.
2017-02-13 09:03:46 +08:00
#include "jemalloc/internal/prof_inlines_a.h"
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/arena_inlines_a.h"
Store arena index rather than (arena_t *) in extent_t.
2017-03-25 02:25:43 +08:00
#include "jemalloc/internal/extent_inlines.h"
Use rtree-based chunk lookups rather than pointer bit twiddling. Look up chunk metadata via the radix tree, rather than using CHUNK_ADDR2BASE(). Propagate pointer's containing extent. Minimize extent lookups by doing a single lookup (e.g. in free()) and propagating the pointer's extent into nearly all the functions that may need it.
2016-03-24 11:29:33 +08:00
#ifndef JEMALLOC_ENABLE_INLINE
Refactor rtree to always use base_alloc() for node allocation.
2016-04-16 15:36:11 +08:00
extent_t *iealloc(tsdn_t *tsdn, const void *ptr);
Use rtree-based chunk lookups rather than pointer bit twiddling. Look up chunk metadata via the radix tree, rather than using CHUNK_ADDR2BASE(). Propagate pointer's containing extent. Minimize extent lookups by doing a single lookup (e.g. in free()) and propagating the pointer's extent into nearly all the functions that may need it.
2016-03-24 11:29:33 +08:00
#endif
#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))
Implement per-CPU arena. The new feature, opt.percpu_arena, determines thread-arena association dynamically based CPU id. Three modes are supported: "percpu", "phycpu" and disabled. "percpu" uses the current core id (with help from sched_getcpu()) directly as the arena index, while "phycpu" will assign threads on the same physical CPU to the same arena. In other words, "percpu" means # of arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of CPUs). Note that no runtime check on whether hyper threading is enabled is added yet. When enabled, threads will be migrated between arenas when a CPU change is detected. In the current design, to reduce overhead from reading CPU id, each arena tracks the thread accessed most recently. When a new thread comes in, we will read CPU id and update arena if necessary.
2017-02-03 09:02:05 +08:00
/* Choose an arena based on a per-thread value. */
JEMALLOC_INLINE arena_t *
arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal) {
arena_t *ret;
if (arena != NULL) {
return arena;
}
ret = internal ? tsd_iarena_get(tsd) : tsd_arena_get(tsd);
if (unlikely(ret == NULL)) {
ret = arena_choose_hard(tsd, internal);
}
assert(ret != NULL);
/*
* Note that for percpu arena, if the current arena is outside of the
* auto percpu arena range, (i.e. thread is assigned to a manually
* managed arena), then percpu arena is skipped.
*/
if (have_percpu_arena && (percpu_arena_mode != percpu_arena_disabled) &&
(arena_ind_get(ret) < percpu_arena_ind_limit()) &&
(ret->last_thd != tsd_tsdn(tsd))) {
unsigned ind = percpu_arena_choose();
if (arena_ind_get(ret) != ind) {
percpu_arena_update(tsd, ind);
ret = tsd_arena_get(tsd);
}
ret->last_thd = tsd_tsdn(tsd);
}
return ret;
}
JEMALLOC_INLINE arena_t *
arena_choose(tsd_t *tsd, arena_t *arena) {
return arena_choose_impl(tsd, arena, false);
}
JEMALLOC_INLINE arena_t *
arena_ichoose(tsd_t *tsd, arena_t *arena) {
return arena_choose_impl(tsd, arena, true);
}
Use rtree-based chunk lookups rather than pointer bit twiddling. Look up chunk metadata via the radix tree, rather than using CHUNK_ADDR2BASE(). Propagate pointer's containing extent. Minimize extent lookups by doing a single lookup (e.g. in free()) and propagating the pointer's extent into nearly all the functions that may need it.
2016-03-24 11:29:33 +08:00
JEMALLOC_ALWAYS_INLINE extent_t *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
iealloc(tsdn_t *tsdn, const void *ptr) {
Incorporate szind/slab into rtree leaves. Expand and restructure the rtree API such that all common operations can be achieved with minimal work, regardless of whether the rtree leaf fields are independent versus packed into a single atomic pointer.
2017-03-17 08:57:52 +08:00
rtree_ctx_t rtree_ctx_fallback;
rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
return rtree_extent_read(tsdn, &extents_rtree, rtree_ctx,
(uintptr_t)ptr, true);
Use rtree-based chunk lookups rather than pointer bit twiddling. Look up chunk metadata via the radix tree, rather than using CHUNK_ADDR2BASE(). Propagate pointer's containing extent. Minimize extent lookups by doing a single lookup (e.g. in free()) and propagating the pointer's extent into nearly all the functions that may need it.
2016-03-24 11:29:33 +08:00
}
#endif
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#include "jemalloc/internal/tcache_inlines.h"
#include "jemalloc/internal/arena_inlines_b.h"
#include "jemalloc/internal/hash_inlines.h"
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
#ifndef JEMALLOC_ENABLE_INLINE
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
arena_t *iaalloc(tsdn_t *tsdn, const void *ptr);
size_t isalloc(tsdn_t *tsdn, const void *ptr);
void *iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero,
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
tcache_t *tcache, bool is_internal, arena_t *arena, bool slow_path);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void *ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero,
Optimize the fast paths of calloc() and [m,d,sd]allocx(). This is a broader application of optimizations to malloc() and free() in f4a0f32d340985de477bbe329ecdaecd69ed1055 (Fast-path improvement: reduce # of branches and unnecessary operations.). This resolves #321.
2016-05-07 03:16:00 +08:00
bool slow_path);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void *ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
tcache_t *tcache, bool is_internal, arena_t *arena);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void *ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, 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, arena_t *arena);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void *ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero);
size_t ivsalloc(tsdn_t *tsdn, const void *ptr);
void idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_internal,
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
bool slow_path);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void idalloc(tsd_t *tsd, void *ptr);
void isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
bool slow_path);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void *iralloct_realign(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
size_t extra, size_t alignment, bool zero, tcache_t *tcache,
arena_t *arena);
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
void *iralloct(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
size_t alignment, bool zero, tcache_t *tcache, arena_t *arena);
void *iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
size_t alignment, bool zero);
bool ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t extra,
size_t alignment, bool zero);
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_C_))
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 *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
iaalloc(tsdn_t *tsdn, const void *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
assert(ptr != NULL);
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return arena_aalloc(tsdn, 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
}
JEMALLOC_ALWAYS_INLINE size_t
Remove extent arg from isalloc() and arena_salloc().
2017-03-17 16:25:12 +08:00
isalloc(tsdn_t *tsdn, const void *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
assert(ptr != NULL);
Remove extent arg from isalloc() and arena_salloc().
2017-03-17 16:25:12 +08:00
return arena_salloc(tsdn, 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
}
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 *
Resolve bootstrapping issues when embedded in FreeBSD libc. b2c0d6322d2307458ae2b28545f8a5c9903d7ef5 (Add witness, a simple online locking validator.) caused a broad propagation of tsd throughout the internal API, but tsd_fetch() was designed to fail prior to tsd bootstrapping. Fix this by splitting tsd_t into non-nullable tsd_t and nullable tsdn_t, and modifying all internal APIs that do not critically rely on tsd to take nullable pointers. Furthermore, add the tsd_booted_get() function so that tsdn_fetch() can probe whether tsd bootstrapping is complete and return NULL if not. All dangerous conversions of nullable pointers are tsdn_tsd() calls that assert-fail on invalid conversion.
2016-05-11 13:21:10 +08:00
iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero, tcache_t *tcache,
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
bool is_internal, arena_t *arena, bool slow_path) {
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 *ret;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
assert(size != 0);
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
assert(!is_internal || tcache == NULL);
assert(!is_internal || arena == NULL || arena_ind_get(arena) <
narenas_auto);
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Resolve bootstrapping issues when embedded in FreeBSD libc. b2c0d6322d2307458ae2b28545f8a5c9903d7ef5 (Add witness, a simple online locking validator.) caused a broad propagation of tsd throughout the internal API, but tsd_fetch() was designed to fail prior to tsd bootstrapping. Fix this by splitting tsd_t into non-nullable tsd_t and nullable tsdn_t, and modifying all internal APIs that do not critically rely on tsd to take nullable pointers. Furthermore, add the tsd_booted_get() function so that tsdn_fetch() can probe whether tsd bootstrapping is complete and return NULL if not. All dangerous conversions of nullable pointers are tsdn_tsd() calls that assert-fail on invalid conversion.
2016-05-11 13:21:10 +08:00
ret = arena_malloc(tsdn, arena, size, ind, zero, tcache, slow_path);
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
if (config_stats && is_internal && likely(ret != NULL)) {
Remove extent arg from isalloc() and arena_salloc().
2017-03-17 16:25:12 +08:00
arena_internal_add(iaalloc(tsdn, ret), isalloc(tsdn, ret));
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
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
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 void *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero, bool slow_path) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return iallocztm(tsd_tsdn(tsd), size, ind, zero, tcache_get(tsd, true),
false, NULL, slow_path);
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
}
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 *
Resolve bootstrapping issues when embedded in FreeBSD libc. b2c0d6322d2307458ae2b28545f8a5c9903d7ef5 (Add witness, a simple online locking validator.) caused a broad propagation of tsd throughout the internal API, but tsd_fetch() was designed to fail prior to tsd bootstrapping. Fix this by splitting tsd_t into non-nullable tsd_t and nullable tsdn_t, and modifying all internal APIs that do not critically rely on tsd to take nullable pointers. Furthermore, add the tsd_booted_get() function so that tsdn_fetch() can probe whether tsd bootstrapping is complete and return NULL if not. All dangerous conversions of nullable pointers are tsdn_tsd() calls that assert-fail on invalid conversion.
2016-05-11 13:21:10 +08:00
ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
tcache_t *tcache, bool is_internal, arena_t *arena) {
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
void *ret;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
Fix error detection for ipalloc() when profiling. sa2u() returns 0 on overflow, but the profiling code was blindly calling sa2u() and allowing the error to silently propagate, ultimately ending in a later assertion failure. Refactor all ipalloc() callers to call sa2u(), check for overflow before calling ipalloc(), and pass usize rather than size. This allows ipalloc() to avoid calling sa2u() in the common case.
2011-03-23 15:37:29 +08:00
assert(usize != 0);
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
assert(usize == sa2u(usize, alignment));
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
assert(!is_internal || tcache == NULL);
assert(!is_internal || arena == NULL || arena_ind_get(arena) <
narenas_auto);
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
Fix error detection for ipalloc() when profiling. sa2u() returns 0 on overflow, but the profiling code was blindly calling sa2u() and allowing the error to silently propagate, ultimately ending in a later assertion failure. Refactor all ipalloc() callers to call sa2u(), check for overflow before calling ipalloc(), and pass usize rather than size. This allows ipalloc() to avoid calling sa2u() in the common case.
2011-03-23 15:37:29 +08:00
Resolve bootstrapping issues when embedded in FreeBSD libc. b2c0d6322d2307458ae2b28545f8a5c9903d7ef5 (Add witness, a simple online locking validator.) caused a broad propagation of tsd throughout the internal API, but tsd_fetch() was designed to fail prior to tsd bootstrapping. Fix this by splitting tsd_t into non-nullable tsd_t and nullable tsdn_t, and modifying all internal APIs that do not critically rely on tsd to take nullable pointers. Furthermore, add the tsd_booted_get() function so that tsdn_fetch() can probe whether tsd bootstrapping is complete and return NULL if not. All dangerous conversions of nullable pointers are tsdn_tsd() calls that assert-fail on invalid conversion.
2016-05-11 13:21:10 +08:00
ret = arena_palloc(tsdn, arena, usize, alignment, zero, tcache);
Normalize aligned allocation algorithms. Normalize arena_palloc(), chunk_alloc_mmap_slow(), and chunk_recycle_dss() to use the same algorithm for trimming over-allocation. Add the ALIGNMENT_ADDR2BASE(), ALIGNMENT_ADDR2OFFSET(), and ALIGNMENT_CEILING() macros, and use them where appropriate. Remove the run_size_p parameter from sa2u(). Fix a potential deadlock in chunk_recycle_dss() that was introduced by eae269036c9f702d9fa9be497a1a2aa1be13a29e (Add alignment support to chunk_alloc()).
2012-04-12 09:13:45 +08:00
assert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret);
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
if (config_stats && is_internal && likely(ret != NULL)) {
Remove extent arg from isalloc() and arena_salloc().
2017-03-17 16:25:12 +08:00
arena_internal_add(iaalloc(tsdn, ret), isalloc(tsdn, ret));
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
}
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ret;
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void *
Resolve bootstrapping issues when embedded in FreeBSD libc. b2c0d6322d2307458ae2b28545f8a5c9903d7ef5 (Add witness, a simple online locking validator.) caused a broad propagation of tsd throughout the internal API, but tsd_fetch() was designed to fail prior to tsd bootstrapping. Fix this by splitting tsd_t into non-nullable tsd_t and nullable tsdn_t, and modifying all internal APIs that do not critically rely on tsd to take nullable pointers. Furthermore, add the tsd_booted_get() function so that tsdn_fetch() can probe whether tsd bootstrapping is complete and return NULL if not. All dangerous conversions of nullable pointers are tsdn_tsd() calls that assert-fail on invalid conversion.
2016-05-11 13:21:10 +08:00
ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
tcache_t *tcache, arena_t *arena) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ipallocztm(tsdn, usize, alignment, zero, tcache, false, arena);
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
}
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 void *
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return ipallocztm(tsd_tsdn(tsd), usize, alignment, zero,
tcache_get(tsd, true), false, NULL);
Refactor jemalloc.c into multiple source files. Fix a stats bug in large object curruns accounting. Replace tcache_bin_fill() with arena_tcache_fill(), and fix a bug in an OOM error path. Fix API name mangling to coexist with __attribute__((malloc)).
2010-01-17 01:53:50 +08:00
}
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
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
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
ivsalloc(tsdn_t *tsdn, const void *ptr) {
Incorporate szind/slab into rtree leaves. Expand and restructure the rtree API such that all common operations can be achieved with minimal work, regardless of whether the rtree leaf fields are independent versus packed into a single atomic pointer.
2017-03-17 08:57:52 +08:00
return arena_vsalloc(tsdn, ptr);
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
}
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
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_internal,
bool slow_path) {
Revert "Avoid NULL check in free() and malloc_usable_size()." This reverts commit 96d4120ac08db3f2d566e8e5c3bc134a24aa0afc. ivsalloc() depends on chunks_rtree being initialized. This can be worked around via a NULL pointer check. However, thread_allocated_tsd_get() also depends on initialization having occurred, and there is no way to guard its call in free() that is cheaper than checking whether ptr is NULL.
2012-04-03 06:18:24 +08:00
assert(ptr != NULL);
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
assert(!is_internal || tcache == NULL);
assert(!is_internal || arena_ind_get(iaalloc(tsdn, ptr)) <
narenas_auto);
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
Implement per arena base allocators. Add/rename related mallctls: - Add stats.arenas.<i>.base . - Rename stats.arenas.<i>.metadata to stats.arenas.<i>.internal . - Add stats.arenas.<i>.resident . Modify the arenas.extend mallctl to take an optional (extent_hooks_t *) argument so that it is possible for all base allocations to be serviced by the specified extent hooks. This resolves #463.
2016-12-23 06:39:10 +08:00
if (config_stats && is_internal) {
Remove extent arg from isalloc() and arena_salloc().
2017-03-17 16:25:12 +08:00
arena_internal_sub(iaalloc(tsdn, ptr), isalloc(tsdn, 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
}
Revert "Avoid NULL check in free() and malloc_usable_size()." This reverts commit 96d4120ac08db3f2d566e8e5c3bc134a24aa0afc. ivsalloc() depends on chunks_rtree being initialized. This can be worked around via a NULL pointer check. However, thread_allocated_tsd_get() also depends on initialization having occurred, and there is no way to guard its call in free() that is cheaper than checking whether ptr is NULL.
2012-04-03 06:18:24 +08:00
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
arena_dalloc(tsdn, ptr, tcache, slow_path);
Fix compiler warnings. Add --enable-cc-silence, which can be used to silence harmless warnings. Fix an aliasing bug in ckh_pointer_hash().
2010-09-21 10:20:48 +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
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
idalloc(tsd_t *tsd, void *ptr) {
idalloctm(tsd_tsdn(tsd), ptr, tcache_get(tsd, false), false, true);
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
}
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 void
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
isdalloct(tsdn_t *tsdn, void *ptr, size_t size,
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
tcache_t *tcache, bool slow_path) {
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
arena_sdalloc(tsdn, ptr, size, tcache, slow_path);
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
}
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
JEMALLOC_ALWAYS_INLINE void *
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
iralloct_realign(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
size_t extra, size_t alignment, bool zero, tcache_t *tcache,
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
arena_t *arena) {
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
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
void *p;
size_t usize, copysize;
usize = sa2u(size + extra, alignment);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NULL;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Remove quarantine support.
2016-04-06 07:52:36 +08:00
p = ipalloct(tsdn, usize, alignment, zero, tcache, arena);
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
if (p == NULL) {
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (extra == 0) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NULL;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
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
/* Try again, without extra this time. */
usize = sa2u(size, alignment);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NULL;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
Remove quarantine support.
2016-04-06 07:52:36 +08:00
p = ipalloct(tsdn, usize, alignment, zero, tcache, arena);
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
if (p == NULL) {
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return NULL;
Update brace style. Add braces around single-line blocks, and remove line breaks before function-opening braces. This resolves #537.
2017-01-16 08:56:30 +08:00
}
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
}
/*
* Copy at most size bytes (not size+extra), since the caller has no
* expectation that the extra bytes will be reliably preserved.
*/
copysize = (size < oldsize) ? size : oldsize;
memcpy(p, ptr, copysize);
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
isdalloct(tsdn, ptr, oldsize, tcache, true);
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return p;
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
}
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 *
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
iralloct(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t alignment,
bool zero, tcache_t *tcache, arena_t *arena) {
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
assert(ptr != NULL);
assert(size != 0);
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
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
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
if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))
!= 0) {
/*
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
* Existing object alignment is inadequate; allocate new space
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
* and copy.
*/
Remove extent dereferences from the deallocation fast paths.
2017-03-17 17:45:12 +08:00
return iralloct_realign(tsdn, ptr, oldsize, size, 0, alignment,
zero, tcache, arena);
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
}
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
return arena_ralloc(tsdn, arena, ptr, oldsize, size, alignment, zero,
tcache);
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
}
Implement tsd. Implement tsd, which is a TLS/TSD abstraction that uses one or both internally. Modify bootstrapping such that no tsd's are utilized until allocation is safe. Remove malloc_[v]tprintf(), and use malloc_snprintf() instead. Fix %p argument size handling in malloc_vsnprintf(). Fix a long-standing statistics-related bug in the "thread.arena" mallctl that could cause crashes due to linked list corruption.
2012-03-22 09:33:03 +08:00
Add and use JEMALLOC_ALWAYS_INLINE. Add JEMALLOC_ALWAYS_INLINE and use it to guarantee that the entire fast paths of the primary allocation/deallocation functions are inlined.
2013-01-23 00:45:43 +08:00
JEMALLOC_ALWAYS_INLINE void *
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
bool zero) {
return iralloct(tsd_tsdn(tsd), ptr, oldsize, size, alignment, zero,
tcache_get(tsd, true), NULL);
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
}
JEMALLOC_ALWAYS_INLINE bool
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t extra,
size_t alignment, bool zero) {
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
assert(ptr != NULL);
assert(size != 0);
Disentangle arena and extent locking. Refactor arena and extent locking protocols such that arena and extent locks are never held when calling into the extent_*_wrapper() API. This requires extra care during purging since the arena lock no longer protects the inner purging logic. It also requires extra care to protect extents from being merged with adjacent extents. Convert extent_t's 'active' flag to an enumerated 'state', so that retained extents are explicitly marked as such, rather than depending on ring linkage state. Refactor the extent collections (and their synchronization) for cached and retained extents into extents_t. Incorporate LRU functionality to support purging. Incorporate page count accounting, which replaces arena->ndirty and arena->stats.retained. Assert that no core locks are held when entering any internal [de]allocation functions. This is in addition to existing assertions that no locks are held when entering external [de]allocation functions. Audit and document synchronization protocols for all arena_t fields. This fixes a potential deadlock due to recursive allocation during gdump, in a similar fashion to b49c649bc18fff4bd10a1c8adbaf1f25f6453cb6 (Fix lock order reversal during gdump.), but with a necessarily much broader code impact.
2017-01-30 13:57:14 +08:00
witness_assert_depth_to_rank(tsdn, WITNESS_RANK_CORE, 0);
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
if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))
!= 0) {
/* Existing object alignment is inadequate. */
Remove extraneous parens around return arguments. This resolves #540.
2017-01-20 10:15:45 +08:00
return true;
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
}
Push down iealloc() calls. Call iealloc() as deep into call chains as possible without causing redundant calls.
2017-03-21 02:00:07 +08:00
return arena_ralloc_no_move(tsdn, ptr, oldsize, size, extra, zero);
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
}
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
Convert arena->prof_accumbytes synchronization to atomics.
2017-02-13 09:03:46 +08:00
#include "jemalloc/internal/prof_inlines_b.h"
jemalloc cpp new/delete bindings Adds cpp bindings for jemalloc, along with necessary autoconf settings. This is mostly to add sized deallocation support, which can't be added from C directly. Sized deallocation is ~10% microbench improvement. * Import ax_cxx_compile_stdcxx.m4 from the autoconf repo, seems like the easiest way to get c++14 detection. * Adds various other changes, like CXXFLAGS, to configure.ac. * Adds new rules to Makefile.in for src/jemalloc-cpp.cpp, and a basic unittest. * Both new and delete are overridden, to ensure jemalloc is used for both. * TODO future enhancement of avoiding extra PLT thunks for new and delete - sdallocx and malloc are publicly exported jemalloc symbols, using an alias would link them directly. Unfortunately, was having trouble getting it to play nice with jemalloc's namespace support. Testing: Tested gcc 4.8, gcc 5, gcc 5.2, clang 4.0. Only gcc >= 5 has sized deallocation support, verified that the rest build correctly. Tested mac osx and Centos. Tested --with-jemalloc-prefix and --without-export. This resolves #202.
2016-10-24 06:56:30 +08:00
#ifdef __cplusplus
}
#endif
Replace fprintf with malloc_printf in tests.
2012-04-16 22:30:26 +08:00
#endif /* JEMALLOC_INTERNAL_H */
Reference in New Issue Copy Permalink