Add atomic(9) implementations of atomic operations.

Add atomic(9) implementations of atomic operations.  These are used on
FreeBSD for non-x86 architectures.

configure.ac

@@ -980,6 +980,29 @@ if test "x${je_cv_function_ffsl}" != "xyes" ; then
    AC_MSG_ERROR([Cannot build without ffsl(3)])
 fi
 
+dnl ============================================================================
+dnl Check for atomic(9) operations as provided on FreeBSD.
+
+JE_COMPILABLE([atomic(9)], [
+#include <sys/types.h>
+#include <machine/atomic.h>
+#include <inttypes.h>
+], [
+	{
+		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])
+fi
+
 dnl ============================================================================
 dnl Check for atomic(3) operations as provided on Darwin.
 
@@ -1031,7 +1054,7 @@ AC_DEFUN([JE_SYNC_COMPARE_AND_SWAP_CHECK],[
   fi
 ])
 
-if test "x${je_cv_osatomic}" != "xyes" ; then
+if test "x${je_cv_atomic9}" != "xyes" -a "x${je_cv_osatomic}" != "xyes" ; then
   JE_SYNC_COMPARE_AND_SWAP_CHECK(32, 4)
   JE_SYNC_COMPARE_AND_SWAP_CHECK(64, 8)
 fi

include/jemalloc/internal/atomic.h

@@ -32,6 +32,7 @@ unsigned	atomic_sub_u(unsigned *p, unsigned x);
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ATOMIC_C_))
 /******************************************************************************/
 /* 64-bit operations. */
+#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)
 #  ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8
 JEMALLOC_INLINE uint64_t
 atomic_add_uint64(uint64_t *p, uint64_t x)
@@ -87,6 +88,28 @@ atomic_sub_uint64(uint64_t *p, uint64_t x)
 
 	return (x);
 }
+#  elif (defined(JEMALLOC_ATOMIC9))
+JEMALLOC_INLINE uint64_t
+atomic_add_uint64(uint64_t *p, uint64_t x)
+{
+
+	/*
+	 * atomic_fetchadd_64() doesn't exist, but we only ever use this
+	 * function on LP64 systems, so atomic_fetchadd_long() will do.
+	 */
+	assert(sizeof(uint64_t) == sizeof(unsigned long));
+
+	return (atomic_fetchadd_long(p, (unsigned long)x) + x);
+}
+
+JEMALLOC_INLINE uint64_t
+atomic_sub_uint64(uint64_t *p, uint64_t x)
+{
+
+	assert(sizeof(uint64_t) == sizeof(unsigned long));
+
+	return (atomic_fetchadd_long(p, (unsigned long)(-(long)x)) - x);
+}
 #  elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_8))
 JEMALLOC_INLINE uint64_t
 atomic_add_uint64(uint64_t *p, uint64_t x)
@@ -102,7 +125,6 @@ atomic_sub_uint64(uint64_t *p, uint64_t x)
 	return (__sync_sub_and_fetch(p, x));
 }
 #  else
-#  if (LG_SIZEOF_PTR == 3)
 #    error "Missing implementation for 64-bit atomic operations"
 #  endif
 #endif
@@ -164,6 +186,20 @@ atomic_sub_uint32(uint32_t *p, uint32_t x)
 
 	return (x);
 }
+#elif (defined(JEMALLOC_ATOMIC9))
+JEMALLOC_INLINE uint32_t
+atomic_add_uint32(uint32_t *p, uint32_t x)
+{
+
+	return (atomic_fetchadd_32(p, x) + x);
+}
+
+JEMALLOC_INLINE uint32_t
+atomic_sub_uint32(uint32_t *p, uint32_t x)
+{
+
+	return (atomic_fetchadd_32(p, (uint32_t)(-(int32_t)x)) - x);
+}
 #elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_4))
 JEMALLOC_INLINE uint32_t
 atomic_add_uint32(uint32_t *p, uint32_t x)

include/jemalloc/internal/jemalloc_internal.h.in

@@ -161,6 +161,10 @@ static const bool config_ivsalloc =
 #endif
     ;
 
+#ifdef JEMALLOC_ATOMIC9
+#include <machine/atomic.h>
+#endif
+
 #if (defined(JEMALLOC_OSATOMIC) || defined(JEMALLOC_OSSPIN))
 #include <libkern/OSAtomic.h>
 #endif

include/jemalloc/jemalloc_defs.h.in

@@ -47,6 +47,9 @@
  */
 #undef CPU_SPINWAIT
 
+/* Defined if the equivalent of FreeBSD's atomic(9) functions are available. */
+#undef JEMALLOC_ATOMIC9
+
 /*
  * Defined if OSAtomic*() functions are available, as provided by Darwin, and
  * documented in the atomic(3) manual page.