Граф коммитов

76 Коммитов

Автор SHA1 Сообщение Дата
Leila Hadji b544725c2c Bug 498166 - malloc_init deadlock due to pthread_atfork. r=glandium 2015-05-22 06:01:00 -04:00
Paul Rouget d60a1bd75e Bug 1156628 - Don't limit the number of unused dirty pages kept by jemalloc on B2G desktop. r=gsvelto, r=glandium 2015-04-24 03:58:00 -04:00
Mike Hommey 548d773d4e Bug 1104634 - Force optimize mozjemalloc on --disable-optimize builds. r=gps
The reason for --disable-optimize is to make debugging easier, but not many
people actually need a high level of debuggability of the allocator itself.
This works around the issue that the Android NDK's definition of ffs is
broken when compiling without optimization, while avoiding to add yet another
configure test.
2014-12-10 10:50:44 +09:00
Emanuel Hoogeveen d573431119 Bug 1100485 - Lock chunks during double purging to avoid racing with allocation. r=glandium 2014-11-19 09:34:00 +01:00
Mike Hommey 6d15698158 Bug 818922 - Add bidirectional method calls with replace-malloc library. r=njn 2014-11-18 19:21:06 +09:00
Emanuel Hoogeveen f287680906 Bug 1073662 - Part 7: Enable limited chunk recycling on all platforms. r=glandium 2014-11-13 22:56:00 +01:00
Emanuel Hoogeveen 3768ec5def Bug 1073662 - Part 6: Add JEMALLOC_RECYCLE, a mode that keeps a limited amount of chunks alive. r=glandium 2014-11-13 22:55:00 +01:00
Emanuel Hoogeveen 92ebf2b732 Bug 1073662 - Part 5: Hook the chunk recycle code up to JEMALLOC_MUNMAP and make it safe for use with MALLOC_DECOMMIT and MALLOC_DOUBLE_PURGE. r=glandium 2014-11-13 22:52:00 +01:00
Emanuel Hoogeveen c6b56644de Bug 1073662 - Part 4: Import chunk recycling code from jemalloc3. r=glandium 2014-11-01 12:00:00 +01:00
Emanuel Hoogeveen 5d83aef48f Bug 1073662 - Part 3: Make all chunk-sized and larger allocations go through chunk_alloc. r=glandium 2014-09-26 13:29:00 +02:00
Emanuel Hoogeveen 8696f3a9ae Bug 1073662 - Part 2: Always use the jemalloc3 allocation logic and remove the old logic. r=glandium 2014-09-26 13:27:00 +02:00
Emanuel Hoogeveen db1349bf19 Bug 1073662 - Part 1: Remove support for the unused MALLOC_PAGEFILE. r=glandium 2014-09-26 13:27:00 +02:00
Mike Hommey ed70c5f377 Bug 1041941 - Use templates for programs, simple programs, libraries and C++ unit tests. r=gps 2014-09-03 14:10:54 +09:00
Nicholas Nethercote ee0e0c98ea Bug 1057754 (follow-up) - Allow Valgrind + jemalloc to be combined, because it makes sense when running Cachegrind. r=me. 2014-08-25 20:38:29 -07:00
Nicholas Nethercote 015fc70f98 Bug 1057754 - Remove the bogus Valgrind annotations from mozjemalloc. r=glandium.
--HG--
extra : rebase_source : c6b32f064181d28bf65d291189d87e516f5e0076
2014-08-25 16:59:43 -07:00
Mike Hommey 5fab42fcaf Bug 1036894 part 8 - Move most in-tree library linkage information to moz.build, as USE_LIBS. r=gps 2014-07-23 13:30:52 +09:00
Makoto Kato a0f8ec3066 Bug 1030899 - Remove issetugid usage on jemalloc. r=glandium 2014-07-08 17:52:09 +09:00
Patrick Wang (Chih-Kai Wang) 01cc0c8352 Bug 1026864: prevent malloc from calling wrapped pthread_mutex_lock and get rid of LibcAllocator. r=cyu,r=glandium,r=khuey 2014-06-25 10:44:58 +08:00
Ehsan Akhgari b7d21c0be3 Bug 1028684 - Fix the signature of _msize in jemalloc.c; r=glandium
This function accepts a void* argument, not a const void*.

--HG--
extra : rebase_source : 9c5cd5066b3730871e06465a84f9a27a6adc6204
2014-06-24 07:20:22 -07:00
Jed Davis 10b28157d6 Bug 1011350 - Use TaggedAnonymousMemory to distinguish our various mmap call sites. r=njn
We can probably get more granular information than this, given that we
can re-tag memory after the initial mapping, or tag subregions of an
individual mmap() call differently.  But this is a start.
2014-06-17 17:55:00 +02:00
Ehsan Akhgari de07b6927f Bug 1025906 - Include intrin.h in jemalloc.c; r=glandium
This fixes a linking error with clang-cl.
2014-06-16 22:34:26 -04:00
Eric Rahm a772941928 Bug 1006769 - Part 2: Backport jemalloc3 run_size_calc logic. r=glandium 2014-05-21 17:34:07 -07:00
Eric Rahm 33a51f0c36 Bug 1006769 - Part 1: Improve heap-overhead memory reporting. r=njn 2014-05-21 17:34:06 -07:00
Nicholas Nethercote b5aba0afc1 Bug 1013014 (attempt 2) - Disable Trace Malloc on TBPL debug builds. r=glandium.
--HG--
extra : rebase_source : 08fa200fd205225f85b9845bf8c18ac38fe26e0d
2014-05-19 18:16:42 -07:00
Mike Hommey 4915133ee5 Bug 984447 - Properly handle forks in mozjemalloc after bug 694896. r=smichaud 2014-03-29 16:53:49 +09:00
Thomas Zimmermann 724de1c4a9 Bug 979887: Fix GCC warnings about inline variable declarations, r=glandium
Some code is compiled in C90 mode, where inline declarations of
variables are not allowed.
2014-03-06 11:19:57 +01:00
Ms2ger 33da18c67e Bug 968856 - Move unconditional LOCAL_INCLUDES into moz.build; r=mshal 2014-02-15 21:24:59 +01:00
David Major cd455b826e Bug 860254 - Part 4: Enable poisoning. r=glandium 2014-02-13 20:11:15 -05:00
David Major 445940d0b6 Bug 860254 - Part 2: Split junk settings into opt_junk (allocate) versus opt_poison (deallocate). r=glandium 2014-01-14 11:56:18 -08:00
David Major 9989920680 Bug 860254 - Part 1: #define MALLOC_FILL but keep the options const-false. r=glandium 2014-02-13 20:10:35 -05:00
Marcin Juszkiewicz 420122fd2c Bug 963028 - AArch64 support for jemalloc. r=froydnj 2014-01-23 16:34:31 -05:00
Jon Coppeard 8e8c44704e Bug 956501 - Update mozjemalloc chunk_alloc_mmap functions to latest upstream version r=glandium 2014-01-14 10:06:25 +00:00
Kai-Zhen Li 5979fa3eb3 Bug 957450 - fix compile error in gonk-kk. r=iacobcatalin 2014-01-10 14:14:53 +08:00
Mike Hommey 62d390485a Bug 945042 - Move NO_PROFILE_GUIDED_OPTIMIZE to moz.build. r=gps,r=njn 2013-12-09 13:39:26 +09:00
Mike Hommey 8fd06cf41b Bug 874266 - Move all DEFINES that can be moved to moz.build. r=mshal 2013-11-27 22:55:07 +09:00
Mike Hommey b32a4ed166 Backout changeset 3fd4b546eed4 (bug 874266) and changeset a35d2e3a872f (bug 942043) for ASAN build bustage and Windows test bustage
--HG--
extra : amend_source : f20d09aeff1c8b5cbd0f1d24c7ce04e86f3aed1d
2013-11-28 14:24:05 +09:00
Mike Hommey 682364d535 Bug 874266 - Move all DEFINES that can be moved to moz.build. r=mshal 2013-11-28 13:08:16 +09:00
Mike Hommey e80e877ab7 Bug 939044 - Remove most definitions of MODULE. r=mshal 2013-11-19 11:47:39 +09:00
David Major 06ff88fdbc Bug 931196 - Remove MOZ_TEMP_INVESTIGATION poisoning code. r=glandium 2013-10-31 11:52:08 -04:00
Cykesiopka d733533b8f Bug 914270 - Part 2: Manual moves. r=joey 2013-10-24 18:52:00 +01:00
Mike Hommey b000a846c2 Bug 929905 - Consolidate sources in moz.build. r=gps 2013-10-25 08:23:05 +09:00
Nathan Froyd 23a4aa4d48 Bug 925169 - part 1 - don't use MOZ_MEMORY_SIZEOF_PTR_2POW in jemalloc.c; r=glandium 2013-10-09 19:04:43 -04:00
Mike Hommey f8bc7fa754 Bug 912293 - Remove now redundant boilerplate from Makefile.in. r=gps 2013-09-05 09:01:46 +09:00
Douglas Crosher ea8a7bdc6c Bug 908971 - jemalloc: fix uninitialized 'mapped' variable. r=jlebar 2013-08-24 13:35:08 +10:00
Justin Lebar 1cfc0ff873 Bug 903420 - Two fixes to jemalloc's memory reporters. r=glandium
1) We were counting "dirty" pages in "waste", when we shouldn't have
   been.  This was causing the assertion at the end of jemalloc_stats()
   which checks that mapped memory is greater than committed memory to
   fail.

2) jemalloc_stats used stats_chunks.curchunks to measure the number of
   mapped pages.  This was problematic for two reasons.

   a) stats_chunks.curchunks was not locked when it was modified in
      chunk_{de}alloc(), so its value could be garbage.

   b) Even if it had been locked properly, it was possible for an
      allocation to occur during a call to jemalloc_stats which would
      cause the measured amount of allocated memory to exceed the
      measured amount of mapped memory, tripping the assertion we
      tripped in (1).

   We fixed these issues by deleting stats_chunks entirely, and by
   introducing huge_mapped, which measures the amount of memory mapped
   by huge allocations (and is properly protected by huge_mtx).

   We now measure the amount of mapped memory by adding huge_mapped and
   each arena's mapped memory, and we do this in such a way that even if
   an allocation occurs during our call to jemalloc_stats, we'll still
   get a consistent result (where mapped >= committed).
2013-08-15 11:15:04 -07:00
Mike Hommey f33621ca8f Bug 901211 - Don't use static page size on ia64, sparc and mips. r=jlebar 2013-08-05 09:10:35 +09:00
Ms2ger a4e67abf5e Bug 897921 - Remove some dead assignments in makefiles; r=mshal 2013-08-02 09:03:55 +02:00
Justin Lebar e67b738ee5 Bug 898558 - Rework jemalloc_stats so it exposes how much memory is used for bookkeeping. r=glandium
This patch also gets rid of the redundant "committed" entry, so now
there's no confusion as to which entries overlap.

--HG--
extra : rebase_source : 429f3d44011f02dda43aa10b077c917c4d02fe50
2013-07-29 09:10:53 -07:00
Brian O'Keefe f11b19a860 Bug 875934 - Move LIBRARY_NAME to moz.build (batch #2); r=mshal
--HG--
extra : rebase_source : 555b28e5c3412ffc210c60b6fe2fee6f053fd587
2013-06-18 08:13:42 -04:00
Joey Armstrong 05956ba0cd bug 870406: move CSRCS to mozbuild (config batch #2) r=mshal 2013-06-04 11:08:44 -04:00