lihao/jemalloc
1
0
Fork 0
mirror of https://github.com/jemalloc/jemalloc.git synced 2026-07-24 05:33:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Introduce a backport of C11 atomics

Browse Source 
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.
This commit is contained in:
David Goldblatt
2017-01-25 09:54:27 -08:00
committed by David Goldblatt
parent 957b8c5f21
commit d4ac7582f3
15 changed files with 947 additions and 672 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
configure.ac
View File

@@ -550,7 +550,7 @@ case "${host}" in
AC_DEFINE([JEMALLOC_HAS_ALLOCA_H])
AC_DEFINE([JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY], [ ])
AC_DEFINE([JEMALLOC_THREADED_INIT], [ ])
AC_DEFINE([JEMALLOC_C11ATOMICS])
AC_DEFINE([JEMALLOC_C11_ATOMICS])
force_tls="0"
default_munmap="0"
;;
@@ -1730,36 +1730,44 @@ JE_COMPILABLE([C11 atomics], [
volatile atomic_uint_least64_t *a = (volatile atomic_uint_least64_t *)p;
uint64_t r = atomic_fetch_add(a, x) + x;
return r == 0;
], [je_cv_c11atomics])
if test "x${je_cv_c11atomics}" = "xyes" ; then
AC_DEFINE([JEMALLOC_C11ATOMICS])
], [je_cv_c11_atomics])
if test "x${je_cv_c11_atomics}" = "xyes" ; then
AC_DEFINE([JEMALLOC_C11_ATOMICS])
fi
dnl ============================================================================
dnl Check for atomic(9) operations as provided on FreeBSD.
dnl Check for GCC-style __atomic atomics.
JE_COMPILABLE([atomic(9)], [
#include <sys/types.h>
#include <machine/atomic.h>
#include <inttypes.h>
JE_COMPILABLE([GCC __atomic atomics], [
], [
{
uint32_t x32 = 0;
volatile uint32_t *x32p = &x32;
atomic_fetchadd_32(x32p, 1);
}
{
unsigned long xlong = 0;
volatile unsigned long *xlongp = &xlong;
atomic_fetchadd_long(xlongp, 1);
}
], [je_cv_atomic9])
if test "x${je_cv_atomic9}" = "xyes" ; then
AC_DEFINE([JEMALLOC_ATOMIC9])
int x = 0;
int val = 1;
int y = __atomic_fetch_add(&x, val, __ATOMIC_RELAXED);
int after_add = x;
return after_add == 1;
], [je_cv_gcc_atomic_atomics])
if test "x${je_cv_gcc_atomic_atomics}" = "xyes" ; then
AC_DEFINE([JEMALLOC_GCC_ATOMIC_ATOMICS])
fi
dnl ============================================================================
dnl Check for GCC-style __sync atomics.
JE_COMPILABLE([GCC __sync atomics], [
], [
int x = 0;
int before_add = __sync_fetch_and_add(&x, 1);
int after_add = x;
return (before_add == 0) && (after_add == 1);
], [je_cv_gcc_sync_atomics])
if test "x${je_cv_gcc_sync_atomics}" = "xyes" ; then
AC_DEFINE([JEMALLOC_GCC_SYNC_ATOMICS])
fi
dnl ============================================================================
dnl Check for atomic(3) operations as provided on Darwin.
dnl We need this not for the atomic operations (which are provided above), but
dnl rather for the OSSpinLock type it exposes.
JE_COMPILABLE([Darwin OSAtomic*()], [
#include <libkern/OSAtomic.h>