| 1.1.1.1 |
| 19-Apr-2026 |
christos | Import jemalloc-5.3.1 (previous was 5.3.0)
This release includes over 390 commits spanning bug fixes, new features,
performance optimizations, and portability improvements. Multiple percent
of system-level metric improvements were measured in tested production
workloads. The release has gone through large-scale production testing
at Meta.
New features:
Support pvalloc. (@Lapenkov: 5b1f2cc)
Add double free detection for the debug build. (@izaitsevfb:
36366f3, @guangli-dai: 42daa1a, @divanorama: 1897f18)
Add compile-time option --enable-pageid to enable memory mapping
annotation. (@devnexen: 4fc5c4f)
Add runtime option prof_bt_max to control the max stack depth for
profiling. (@guangli-dai: a0734fd)
Add compile-time option --enable-force-getenv to use getenv instead
of secure_getenv. (@interwq: 481bbfc)
Add compile-time option --disable-dss to disable the usage of
sbrk(2). (@Svetlitski: ea5b7be)
Add runtime option tcache_ncached_max to control the number of items
in each size bin in the thread cache. (@guangli-dai: 8a22d10)
Add runtime option calloc_madvise_threshold to determine if kernel or
memset is used to zero the allocations for calloc. (@nullptr0-0:
5081c16)
Add compile-time option --disable-user-config to disable reading the
runtime configurations from /etc/malloc.conf or environment variable
MALLOC_CONF. (@roblabla: c17bf8b)
Add runtime option disable_large_size_classes to guard the new usable
size calculation, which minimizes the memory overhead for large
allocations, i.e., >= 4 * PAGE. (@guangli-dai: c067a55, 8347f10)
Enable process_madvise usage, add runtime option
process_madvise_max_batch to control the max # of regions in each
madvise batch. (@interwq: 22440a0, @spredolac: 4246475)
Add mallctl interfaces:
opt.prof_bt_max (@guangli-dai: a0734fd)
arena.<i>.name to set and get arena names. (@guangli-dai: ba19d2c)
thread.tcache.max to set and get the tcache_max of the current
thread. (@guangli-dai: a442d9b)
thread.tcache.ncached_max.write and
thread.tcache.ncached_max.read_sizeclass to set and get the
ncached_max setup of the current thread. (@guangli-dai: 630f7de,
6b197fd)
arenas.hugepage to return the hugepage size used, also exported to
malloc stats. (@ilvokhin: 90c627e)
approximate_stats.active to return an estimate of the current active
bytes, which should not be compared with other stats retrieved.
(@guangli-dai: 0988583)
Bug fixes:
Prevent potential deadlocks in decaying during reentrancy. (@interwq:
434a68e)
Fix segfault in extent coalescing. (@Svetlitski: 12311fe)
Add null pointer detections in mallctl calls. (@Svetlitski: dc0a184,
0288126)
Make mallctl arenas.lookup triable without crashing on invalid
pointers. (@auxten: 019cccc, 5bac384)
Demote sampled allocations for proper deallocations during
arena_reset. (@Svetlitski: 62648c8)
Fix jemalloc's read(2) and write(2). (@Svetlitski: d2c9ed3, @lexprfuncall:
9fdc116)
Fix the pkg-config metadata file. (@BtbN: ed7e6fe, ce8ce99)
Fix the autogen.sh so that it accepts quoted extra options.
(@honggyukim: f6fe6ab)
Fix rallocx() to set errno to ENOMEM upon OOMing. (@arter97: 38056fe,
@interwq: 83b0757)
Avoid stack overflow for internal variable array usage. (@nullptr0-0:
47c9bcd, 48f66cf, @xinydev: 9169e92)
Fix background thread initialization race. (@puzpuzpuz: 4d0ffa0)
Guard os_page_id against a NULL address. (@lexprfuncall: 79cc7dc)
Handle tcache init failures gracefully. (@lexprfuncall: a056c20)
Fix missing release of acquired neighbor edata in
extent_try_coalesce_impl. (@spredolac: 675ab07)
Fix memory leak of old curr_reg on san_bump_grow_locked failure.
(@spredolac: 5904a42)
Fix large alloc nrequests under-counting on cache misses. (@spredolac:
3cc56d3)
Portability improvements:
Fix the build in C99. (@abaelhe: 56ddbea)
Add pthread_setaffinity_np detection for non Linux/BSD platforms.
(@devnexen: 4c95c95)
Make VARIABLE_ARRAY compatible with compilers not supporting VLA,
i.e., Visual Studio C compiler in C11 or C17 modes. (@madscientist:
be65438)
Fix the build on Linux using musl library. (@marv: aba1645, 45249cf)
Reduce the memory overhead in small allocation sampling for systems
with larger page sizes, e.g., ARM. (@Svetlitski: 5a858c6)
Add C23's free_sized and free_aligned_sized. (@Svetlitski:
cdb2c0e)
Enable heap profiling on MacOS. (@nullptr0-0: 4b555c1)
Fix incorrect printing on 32bit. (@sundb: 630434b)
Make JEMALLOC_CXX_THROW compatible with C++ versions newer than
C++17. (@r-barnes, @guangli-dai: 21bcc0a)
Fix mmap tag conflicts on MacOS. (@kdrag0n: c893fcd)
Fix monotonic timer assumption for win32. (@burtonli: 8dc97b1)
Fix VM over-reservation on systems with larger pages, e.g., aarch64.
(@interwq: cd05b19)
Remove unreachable() macro conditionally to prevent definition
conflicts for C23+. (@appujee: d8486b2, 4b88bdd)
Fix dlsym failure observed on FreeBSD. (@rhelmot: 86bbaba)
Change the default page size to 64KB on aarch64 Linux. (@lexprfuncall:
9442300)
Update config.guess and config.sub to the latest version.
(@lexprfuncall: c51949e)
Determine the page size on Android from NDK header files.
(@lexprfuncall: c51abba)
Improve the portability of grep patterns in configure.ac.
(@lexprfuncall: 365747b)
Add compile-time option --with-cxx-stdlib to specify the C++ standard
library. (@yuxuanchen1997: a10ef3e)
Optimizations and refactors:
Enable tcache for deallocation-only threads. (@interwq: 143e9c4)
Inline to accelerate operator delete. (@guangli-dai: e8f9f13)
Optimize pairing heap's performance. (@deadalnix: 5266152, be6da4f,
543e2d6, 10d7131, 92aa52c, @Svetlitski: 36ca0c1)
Inline the storage for thread name in the profiling data. (@interwq:
ce0b7ab, e62aa47)
Optimize a hot function edata_cmp_summary_comp to accelerate it.
(@Svetlitski: 6841110, @guangli-dai: 0181aaa)
Allocate thread cache using the base allocator, which enables thread
cache to use thp when metadata_thp is turned on. (@interwq:
72cfdce)
Allow oversize arena not to purge immediately when background threads
are enabled, although the default decay time is 0 to be back compatible.
(@interwq: d131331)
Optimize thread-local storage implementation on Windows. (@mcfi:
9e123a8, 3a0d9cd)
Optimize fast path to allow static size class computation. (@interwq:
323ed2e)
Redesign tcache GC to regulate the frequency and make it
locality-aware. The new design is default on, guarded by option
experimental_tcache_gc. (@nullptr0-0: 0c88be9, e2c9f3a,
14d5dc1, @deadalnix: 5afff2e)
Reduce the arena switching overhead by avoiding forced purging when
background thread is enabled. (@interwq: a3910b9)
Improve the reuse efficiency by limiting the maximum coalesced size for
large extents. (@jiebinn: 3c14707)
Refactor thread events to allow registration of users' thread events
and remove prof_threshold as the built-in event. (@spredolac: e6864c6,
015b017, 34ace91)
Documentation:
Update Windows building instructions. (@Lapenkov: 3713932)
Add vcpkg installation instructions. (@LilyWangLL: c0c9783)
Update profiling internals with an example. (@jordalgo: b04e766)
|
