Improve ethread atomics

The ethread atomics API now also provide double word size atomics.
Double word size atomics are implemented using native atomic
instructions on x86 (when the cmpxchg8b instruction is available)
and on x86_64 (when the cmpxchg16b instruction is available). On
other hardware where 32-bit atomics or word size atomics are
available, an optimized fallback is used; otherwise, a spinlock,
or a mutex based fallback is used.

The ethread library now performs runtime tests for presence of
hardware features, such as for example SSE2 instructions, instead
of requiring this to be determined at compile time.

There are now functions implementing each atomic operation with the
following implied memory barrier semantics: none, read, write,
acquire, release, and full. Some of the operation-barrier
combinations aren't especially useful. But instead of filtering
useful ones out, and potentially miss a useful one, we implement
them all.

A much smaller set of functionality for native atomics are required
to be implemented than before. More or less only cmpxchg and a
membar macro are required to be implemented for each atomic size.
Other functions will automatically be constructed from these. It is,
of course, often wise to implement more that this if possible from a
performance perspective.

5 files changed, 164 insertions, 180 deletions

diff --git a/erts/include/internal/sparc32/atomic.h b/erts/include/internal/sparc32/atomic.h
index c297522ab1..fe1daaa9cf 100644
--- a/erts/include/internal/sparc32/atomic.h
+++ b/erts/include/internal/sparc32/atomic.h
@@ -35,23 +35,16 @@
 
 #ifdef ETHR_INCLUDE_ATOMIC_IMPL__
 
-#ifndef ETHR_SPARC_V9_ATOMIC_COMMON__
-#define ETHR_SPARC_V9_ATOMIC_COMMON__
-
-#define ETHR_MEMORY_BARRIER \
-  __asm__ __volatile__("membar #LoadLoad|#LoadStore|#StoreLoad|#StoreStore\n" \
-                       : : : "memory")
-
-#endif /* ETHR_SPARC_V9_ATOMIC_COMMON__ */
-
 #if ETHR_INCLUDE_ATOMIC_IMPL__ == 4
 #define ETHR_HAVE_NATIVE_ATOMIC32 1
+#define ETHR_NATIVE_ATOMIC32_IMPL "ethread"
 #define ETHR_NATMC_FUNC__(X) ethr_native_atomic32_ ## X
 #define ETHR_ATMC_T__ ethr_native_atomic32_t
 #define ETHR_AINT_T__ ethr_sint32_t
 #define ETHR_CAS__ "cas"
 #elif ETHR_INCLUDE_ATOMIC_IMPL__ == 8
 #define ETHR_HAVE_NATIVE_ATOMIC64 1
+#define ETHR_NATIVE_ATOMIC64_IMPL "ethread"
 #define ETHR_NATMC_FUNC__(X) ethr_native_atomic64_ ## X
 #define ETHR_ATMC_T__ ethr_native_atomic64_t
 #define ETHR_AINT_T__ ethr_sint64_t
@@ -66,17 +59,23 @@ typedef struct {
 
 #if defined(ETHR_TRY_INLINE_FUNCS) || defined(ETHR_ATOMIC_IMPL__)
 
+#if ETHR_INCLUDE_ATOMIC_IMPL__ == 4
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC32_ADDR 1
+#else
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC64_ADDR 1
+#endif
+
 static ETHR_INLINE ETHR_AINT_T__ *
 ETHR_NATMC_FUNC__(addr)(ETHR_ATMC_T__ *var)
 {
     return (ETHR_AINT_T__ *) &var->counter;
 }
 
-static ETHR_INLINE void
-ETHR_NATMC_FUNC__(init)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ i)
-{
-    var->counter = i;
-}
+#if ETHR_INCLUDE_ATOMIC_IMPL__ == 4
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC32_SET 1
+#else
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC64_SET 1
+#endif
 
 static ETHR_INLINE void
 ETHR_NATMC_FUNC__(set)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ i)
@@ -84,182 +83,49 @@ ETHR_NATMC_FUNC__(set)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ i)
     var->counter = i;
 }
 
+#if ETHR_INCLUDE_ATOMIC_IMPL__ == 4
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC32_READ 1
+#else
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC64_READ 1
+#endif
+
 static ETHR_INLINE ETHR_AINT_T__
 ETHR_NATMC_FUNC__(read)(ETHR_ATMC_T__ *var)
 {
     return var->counter;
 }
 
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(add_return)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ incr)
-{
-    ETHR_AINT_T__ old, tmp;
-
-    __asm__ __volatile__("membar #LoadLoad|#StoreLoad\n" : : : "memory");
-    do {
-	old = var->counter;
-	tmp = old+incr;
-	__asm__ __volatile__(
-	    ETHR_CAS__ " [%2], %1, %0"
-	    : "=&r"(tmp)
-	    : "r"(old), "r"(&var->counter), "0"(tmp)
-	    : "memory");
-    } while (__builtin_expect(old != tmp, 0));
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore" : : : "memory");
-    return old+incr;
-}
-
-static ETHR_INLINE void
-ETHR_NATMC_FUNC__(add)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ incr)
-{
-    (void)ETHR_NATMC_FUNC__(add_return)(var, incr);
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(inc_return)(ETHR_ATMC_T__ *var)
-{
-    return ETHR_NATMC_FUNC__(add_return)(var, 1);
-}
-
-static ETHR_INLINE void
-ETHR_NATMC_FUNC__(inc)(ETHR_ATMC_T__ *var)
-{
-    (void)ETHR_NATMC_FUNC__(add_return)(var, 1);
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(dec_return)(ETHR_ATMC_T__ *var)
-{
-    return ETHR_NATMC_FUNC__(add_return)(var, -1);
-}
-
-static ETHR_INLINE void
-ETHR_NATMC_FUNC__(dec)(ETHR_ATMC_T__ *var)
-{
-    (void)ETHR_NATMC_FUNC__(add_return)(var, -1);
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(and_retold)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ mask)
-{
-    ETHR_AINT_T__ old, tmp;
-
-    __asm__ __volatile__("membar #LoadLoad|#StoreLoad\n" : : : "memory");
-    do {
-	old = var->counter;
-	tmp = old & mask;
-	__asm__ __volatile__(
-	    ETHR_CAS__ " [%2], %1, %0"
-	    : "=&r"(tmp)
-	    : "r"(old), "r"(&var->counter), "0"(tmp)
-	    : "memory");
-    } while (__builtin_expect(old != tmp, 0));
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore" : : : "memory");
-    return old;
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(or_retold)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ mask)
-{
-    ETHR_AINT_T__ old, tmp;
-
-    __asm__ __volatile__("membar #LoadLoad|#StoreLoad\n" : : : "memory");
-    do {
-	old = var->counter;
-	tmp = old | mask;
-	__asm__ __volatile__(
-	    ETHR_CAS__ " [%2], %1, %0"
-	    : "=&r"(tmp)
-	    : "r"(old), "r"(&var->counter), "0"(tmp)
-	    : "memory");
-    } while (__builtin_expect(old != tmp, 0));
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore" : : : "memory");
-    return old;
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(xchg)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ val)
-{
-    ETHR_AINT_T__ old, new;
-
-    __asm__ __volatile__("membar #LoadLoad|#StoreLoad" : : : "memory");
-    do {
-	old = var->counter;
-	new = val;
-	__asm__ __volatile__(
-	    ETHR_CAS__ " [%2], %1, %0"
-	    : "=&r"(new)
-	    : "r"(old), "r"(&var->counter), "0"(new)
-	    : "memory");
-    } while (__builtin_expect(old != new, 0));
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore" : : : "memory");
-    return old;
-}
+#if ETHR_INCLUDE_ATOMIC_IMPL__ == 4
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC32_CMPXCHG 1
+#else
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC64_CMPXCHG 1
+#endif
 
 static ETHR_INLINE ETHR_AINT_T__
 ETHR_NATMC_FUNC__(cmpxchg)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ new, ETHR_AINT_T__ old)
 {
-    __asm__ __volatile__("membar #LoadLoad|#StoreLoad\n" : : : "memory");
     __asm__ __volatile__(
       ETHR_CAS__ " [%2], %1, %0"
       : "=&r"(new)
       : "r"(old), "r"(&var->counter), "0"(new)
       : "memory");
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore" : : : "memory");
     return new;
 }
 
-/*
- * Atomic ops with at least specified barriers.
- */
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(read_acqb)(ETHR_ATMC_T__ *var)
-{
-    ETHR_AINT_T__ res = ETHR_NATMC_FUNC__(read)(var);
-    __asm__ __volatile__("membar #LoadLoad|#LoadStore" : : : "memory");
-    return res;
-}
-
-static ETHR_INLINE void
-ETHR_NATMC_FUNC__(set_relb)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ i)
-{
-    __asm__ __volatile__("membar #LoadStore|#StoreStore" : : : "memory");
-    ETHR_NATMC_FUNC__(set)(var, i);
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(inc_return_acqb)(ETHR_ATMC_T__ *var)
-{
-    ETHR_AINT_T__ res = ETHR_NATMC_FUNC__(inc_return)(var);
-    return res;
-}
-
-static ETHR_INLINE void
-ETHR_NATMC_FUNC__(dec_relb)(ETHR_ATMC_T__ *var)
-{
-    ETHR_NATMC_FUNC__(dec)(var);
-}
-
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(dec_return_relb)(ETHR_ATMC_T__ *var)
-{
-    return ETHR_NATMC_FUNC__(dec_return)(var);
-}
+#if ETHR_INCLUDE_ATOMIC_IMPL__ == 4
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC32_CMPXCHG_ACQB 1
+#else
+#  define ETHR_HAVE_ETHR_NATIVE_ATOMIC64_CMPXCHG_ACQB 1
+#endif
 
 static ETHR_INLINE ETHR_AINT_T__
 ETHR_NATMC_FUNC__(cmpxchg_acqb)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ new, ETHR_AINT_T__ old)
 {
     ETHR_AINT_T__ res = ETHR_NATMC_FUNC__(cmpxchg)(var, new, old);
+    ETHR_MEMBAR(ETHR_StoreLoad|ETHR_StoreStore);
     return res;
 }
 
-static ETHR_INLINE ETHR_AINT_T__
-ETHR_NATMC_FUNC__(cmpxchg_relb)(ETHR_ATMC_T__ *var, ETHR_AINT_T__ new, ETHR_AINT_T__ old)
-{
-    return ETHR_NATMC_FUNC__(cmpxchg)(var, new, old);
-}
-
 #endif /* ETHR_TRY_INLINE_FUNCS */
 
 #undef ETHR_NATMC_FUNC__
diff --git a/erts/include/internal/sparc32/ethr_membar.h b/erts/include/internal/sparc32/ethr_membar.h
new file mode 100644
index 0000000000..6eb0c5a1d6
--- /dev/null
+++ b/erts/include/internal/sparc32/ethr_membar.h
@@ -0,0 +1,115 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2011. All Rights Reserved.
+ *
+ * The contents of this file are subject to the Erlang Public License,
+ * Version 1.1, (the "License"); you may not use this file except in
+ * compliance with the License. You should have received a copy of the
+ * Erlang Public License along with this software. If not, it can be
+ * retrieved online at http://www.erlang.org/.
+ *
+ * Software distributed under the License is distributed on an "AS IS"
+ * basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
+ * the License for the specific language governing rights and limitations
+ * under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+/*
+ * Description: Memory barriers for sparc-v9
+ * Author: Rickard Green
+ */
+
+#ifndef ETHR_SPARC_V9_MEMBAR_H__
+#define ETHR_SPARC_V9_MEMBAR_H__
+
+#if defined(ETHR_SPARC_TSO)
+/* --- Total Store Order ------------------------------------------------ */
+
+#define ETHR_LoadLoad	0
+#define ETHR_LoadStore	0
+#define ETHR_StoreLoad	1
+#define ETHR_StoreStore	0
+
+static __inline__ void
+ethr_cb__(void)
+{
+    __asm__ __volatile__ ("" : : : "memory");
+}
+
+static __inline__ void
+ethr_StoreLoad__(void)
+{
+    __asm__ __volatile__ ("membar #StoreLoad\n\t" : : : "memory");
+}
+
+
+#define ETHR_MEMBAR(B) \
+  ETHR_CHOOSE_EXPR((B), ethr_StoreLoad__(), ethr_cb__())
+
+#elif defined(ETHR_SPARC_PSO)
+/* --- Partial Store Order ---------------------------------------------- */
+
+#define ETHR_LoadLoad	0
+#define ETHR_LoadStore	0
+#define ETHR_StoreLoad	(1 << 0)
+#define ETHR_StoreStore	(1 << 1)
+
+static __inline__ void
+ethr_cb__(void)
+{
+    __asm__ __volatile__ ("" : : : "memory");
+}
+
+static __inline__ void
+ethr_StoreLoad__(void)
+{
+    __asm__ __volatile__ ("membar #StoreLoad\n\t" : : : "memory");
+}
+
+static __inline__ void
+ethr_StoreStore__(void)
+{
+    __asm__ __volatile__ ("membar #StoreStore\n\t" : : : "memory");
+}
+
+static __inline__ void
+ethr_StoreLoad_StoreStore__(void)
+{
+    __asm__ __volatile__ ("membar #StoreLoad|StoreStore\n\t" : : : "memory");
+}
+
+#define ETHR_MEMBAR(B)					\
+  ETHR_CHOOSE_EXPR(					\
+      (B) == ETHR_StoreLoad,				\
+      ethr_StoreLoad__(),				\
+      ETHR_CHOOSE_EXPR(					\
+	  (B) == ETHR_StoreStore,			\
+	  ethr_StoreStore__(),				\
+	  ETHR_CHOOSE_EXPR(				\
+	      (B) == (ETHR_StoreLoad|ETHR_StoreStore),	\
+	      ethr_StoreLoad_StoreStore__(),		\
+	      ethr_cb__())))
+
+#elif defined(ETHR_SPARC_RMO)
+/* --- Relaxed Memory Order --------------------------------------------- */
+
+#  define ETHR_LoadLoad #LoadLoad
+#  define ETHR_LoadStore #LoadStore
+#  define ETHR_StoreLoad #StoreLoad
+#  define ETHR_StoreStore #StoreStore
+
+#  define ETHR_MEMBAR_AUX__(B) \
+     __asm__ __volatile__("membar " #B "\n\t" : : : "memory")
+
+#  define ETHR_MEMBAR(B) ETHR_MEMBAR_AUX__(B)
+
+#else
+
+#  error "No memory order defined"
+
+#endif
+
+#endif
diff --git a/erts/include/internal/sparc32/ethread.h b/erts/include/internal/sparc32/ethread.h
index aea9794390..5ad92d3da7 100644
--- a/erts/include/internal/sparc32/ethread.h
+++ b/erts/include/internal/sparc32/ethread.h
@@ -1,7 +1,7 @@
 /* 
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2005-2010. All Rights Reserved.
+ * Copyright Ericsson AB 2005-2011. All Rights Reserved.
  * 
  * The contents of this file are subject to the Erlang Public License,
  * Version 1.1, (the "License"); you may not use this file except in
@@ -24,6 +24,7 @@
 #ifndef ETHREAD_SPARC32_ETHREAD_H
 #define ETHREAD_SPARC32_ETHREAD_H
 
+#include "ethr_membar.h"
 #define ETHR_ATOMIC_WANT_32BIT_IMPL__
 #include "atomic.h"
 #if ETHR_SIZEOF_PTR == 8
@@ -33,8 +34,4 @@
 #include "spinlock.h"
 #include "rwlock.h"
 
-#define ETHR_HAVE_NATIVE_ATOMICS 1
-#define ETHR_HAVE_NATIVE_SPINLOCKS 1
-#define ETHR_HAVE_NATIVE_RWSPINLOCKS 1
-
 #endif /* ETHREAD_SPARC32_ETHREAD_H */
diff --git a/erts/include/internal/sparc32/rwlock.h b/erts/include/internal/sparc32/rwlock.h
index 465ec96866..8b6f2e9c57 100644
--- a/erts/include/internal/sparc32/rwlock.h
+++ b/erts/include/internal/sparc32/rwlock.h
@@ -1,7 +1,7 @@
 /* 
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2005-2010. All Rights Reserved.
+ * Copyright Ericsson AB 2005-2011. All Rights Reserved.
  *
  * The contents of this file are subject to the Erlang Public License,
  * Version 1.1, (the "License"); you may not use this file except in
@@ -24,6 +24,9 @@
 #ifndef ETHREAD_SPARC32_RWLOCK_H
 #define ETHREAD_SPARC32_RWLOCK_H
 
+#define ETHR_HAVE_NATIVE_RWSPINLOCKS 1
+#define ETHR_NATIVE_RWSPINLOCK_IMPL "ethread"
+
 /* Unlocked if zero, read-locked if positive, write-locked if -1. */
 typedef struct {
     volatile int lock;
@@ -42,7 +45,7 @@ ethr_native_read_unlock(ethr_native_rwlock_t *lock)
 {
     unsigned int old, new;
 
-    __asm__ __volatile__("membar #LoadLoad|#StoreLoad");
+    ETHR_MEMBAR(ETHR_LoadLoad|ETHR_StoreLoad);
     do {
 	old = lock->lock;
 	new = old-1;
@@ -70,7 +73,7 @@ ethr_native_read_trylock(ethr_native_rwlock_t *lock)
 	    : "r"(old), "r"(&lock->lock), "0"(new)
 	    : "memory");
     } while (__builtin_expect(old != new, 0));
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore");
+    ETHR_MEMBAR(ETHR_StoreLoad|ETHR_StoreStore);
     return 1;
 }
 
@@ -87,7 +90,7 @@ ethr_native_read_lock(ethr_native_rwlock_t *lock)
 	if (__builtin_expect(ethr_native_read_trylock(lock) != 0, 1))
 	    break;
 	do {
-	    __asm__ __volatile__("membar #LoadLoad");
+	    ETHR_MEMBAR(ETHR_LoadLoad);
 	} while (ethr_native_read_is_locked(lock));
    }
 }
@@ -95,7 +98,7 @@ ethr_native_read_lock(ethr_native_rwlock_t *lock)
 static ETHR_INLINE void
 ethr_native_write_unlock(ethr_native_rwlock_t *lock)
 {
-    __asm__ __volatile__("membar #LoadStore|#StoreStore");
+    ETHR_MEMBAR(ETHR_LoadStore|ETHR_StoreStore);
     lock->lock = 0;
 }
 
@@ -115,7 +118,7 @@ ethr_native_write_trylock(ethr_native_rwlock_t *lock)
 	    : "r"(old), "r"(&lock->lock), "0"(new)
 	    : "memory");
     } while (__builtin_expect(old != new, 0));
-    __asm__ __volatile__("membar #StoreLoad|#StoreStore");
+    ETHR_MEMBAR(ETHR_StoreLoad|ETHR_StoreStore);
     return 1;
 }
 
@@ -132,7 +135,7 @@ ethr_native_write_lock(ethr_native_rwlock_t *lock)
 	if (__builtin_expect(ethr_native_write_trylock(lock) != 0, 1))
 	    break;
 	do {
-	    __asm__ __volatile__("membar #LoadLoad");
+	    ETHR_MEMBAR(ETHR_LoadLoad);
 	} while (ethr_native_write_is_locked(lock));
    }
 }
diff --git a/erts/include/internal/sparc32/spinlock.h b/erts/include/internal/sparc32/spinlock.h
index 493d514210..d4e36e09cf 100644
--- a/erts/include/internal/sparc32/spinlock.h
+++ b/erts/include/internal/sparc32/spinlock.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2005-2010. All Rights Reserved.
+ * Copyright Ericsson AB 2005-2011. All Rights Reserved.
  *
  * The contents of this file are subject to the Erlang Public License,
  * Version 1.1, (the "License"); you may not use this file except in
@@ -24,6 +24,9 @@
 #ifndef ETHR_SPARC32_SPINLOCK_H
 #define ETHR_SPARC32_SPINLOCK_H
 
+#define ETHR_HAVE_NATIVE_SPINLOCKS 1
+#define ETHR_NATIVE_SPINLOCK_IMPL "ethread"
+
 /* Locked with ldstub, so unlocked when 0 and locked when non-zero. */
 typedef struct {
     volatile unsigned char lock;
@@ -40,7 +43,7 @@ ethr_native_spinlock_init(ethr_native_spinlock_t *lock)
 static ETHR_INLINE void
 ethr_native_spin_unlock(ethr_native_spinlock_t *lock)
 {
-    __asm__ __volatile__("membar #LoadStore|#StoreStore");
+    ETHR_MEMBAR(ETHR_LoadStore|ETHR_StoreStore);
     lock->lock = 0;
 }
 
@@ -51,10 +54,10 @@ ethr_native_spin_trylock(ethr_native_spinlock_t *lock)
 
     __asm__ __volatile__(
 	"ldstub [%1], %0\n\t"
-	"membar #StoreLoad|#StoreStore"
 	: "=r"(prev)
 	: "r"(&lock->lock)
 	: "memory");
+    ETHR_MEMBAR(ETHR_StoreLoad|ETHR_StoreStore);
     return prev == 0;
 }
 
@@ -71,7 +74,7 @@ ethr_native_spin_lock(ethr_native_spinlock_t *lock)
 	if (__builtin_expect(ethr_native_spin_trylock(lock) != 0, 1))
 	    break;
 	do {
-	    __asm__ __volatile__("membar #LoadLoad");
+	    ETHR_MEMBAR(ETHR_LoadLoad);
 	} while (ethr_native_spin_is_locked(lock));
     }
 }