Rewrite ethread library

Large parts of the ethread library have been rewritten. The
ethread library is an Erlang runtime system internal, portable
thread library used by the runtime system itself.

Most notable improvement is a reader optimized rwlock
implementation which dramatically improve the performance of
read-lock/read-unlock operations on multi processor systems by
avoiding ping-ponging of the rwlock cache lines. The reader
optimized rwlock implementation is used by miscellaneous
rwlocks in the runtime system that are known to be read-locked
frequently, and can be enabled on ETS tables by passing the
`{read_concurrency, true}' option upon table creation. See the
documentation of `ets:new/2' for more information.

The ethread library can now also use the libatomic_ops library
for atomic memory accesses. This makes it possible for the
Erlang runtime system to utilize optimized atomic operations
on more platforms than before. Use the
`--with-libatomic_ops=PATH' configure command line argument
when specifying where the libatomic_ops installation is
located. The libatomic_ops library can be downloaded from:
http://www.hpl.hp.com/research/linux/atomic_ops/

The changed API of the ethread library has also caused
modifications in the Erlang runtime system. Preparations for
the to come "delayed deallocation" feature has also been done
since it depends on the ethread library.

Note: When building for x86, the ethread library will now use
instructions that first appeared on the pentium 4 processor. If
you want the runtime system to be compatible with older
processors (back to 486) you need to pass the
`--enable-ethread-pre-pentium4-compatibility' configure command
line argument when configuring the system.

1 files changed, 55 insertions, 12 deletions

diff --git a/erts/include/internal/tile/atomic.h b/erts/include/internal/tile/atomic.h
index 5e4c7ac9fe..69569d82d1 100644
--- a/erts/include/internal/tile/atomic.h
+++ b/erts/include/internal/tile/atomic.h
@@ -34,7 +34,7 @@ typedef struct {
 
 #define ETHR_MEMORY_BARRIER __insn_mf()
 
-#ifdef ETHR_TRY_INLINE_FUNCS
+#if defined(ETHR_TRY_INLINE_FUNCS) || defined(ETHR_AUX_IMPL__)
 
 static ETHR_INLINE void
 ethr_native_atomic_init(ethr_native_atomic_t *var, long i)
@@ -45,7 +45,6 @@ ethr_native_atomic_init(ethr_native_atomic_t *var, long i)
 static ETHR_INLINE void
 ethr_native_atomic_set(ethr_native_atomic_t *var, long i)
 {
-    __insn_mf();
     atomic_exchange_acq(&var->counter, i);
 }
 
@@ -58,28 +57,24 @@ ethr_native_atomic_read(ethr_native_atomic_t *var)
 static ETHR_INLINE void
 ethr_native_atomic_add(ethr_native_atomic_t *var, long incr)
 {
-    __insn_mf();
     atomic_add(&var->counter, incr);
 }      
        
 static ETHR_INLINE void
 ethr_native_atomic_inc(ethr_native_atomic_t *var)
 {
-    __insn_mf();
     atomic_increment(&var->counter);
 }
 
 static ETHR_INLINE void
 ethr_native_atomic_dec(ethr_native_atomic_t *var)
 {
-    __insn_mf();
     atomic_decrement(&var->counter);
 }
 
 static ETHR_INLINE long
 ethr_native_atomic_add_return(ethr_native_atomic_t *var, long incr)
 {
-    __insn_mf();
     return atomic_exchange_and_add(&var->counter, incr) + incr;
 }
 
@@ -98,33 +93,81 @@ ethr_native_atomic_dec_return(ethr_native_atomic_t *var)
 static ETHR_INLINE long
 ethr_native_atomic_and_retold(ethr_native_atomic_t *var, long mask)
 {
-    /* Implement a barrier suitable for a mutex unlock. */
-    __insn_mf();
     return atomic_and_val(&var->counter, mask);
 }
 
 static ETHR_INLINE long
 ethr_native_atomic_or_retold(ethr_native_atomic_t *var, long mask)
 {
-    __insn_mf();
     return atomic_or_val(&var->counter, mask);
 }
 
 static ETHR_INLINE long
 ethr_native_atomic_xchg(ethr_native_atomic_t *var, long val)
 {   
-    __insn_mf();
     return atomic_exchange_acq(&var->counter, val);
 } 
 
 static ETHR_INLINE long
 ethr_native_atomic_cmpxchg(ethr_native_atomic_t *var, long new, long expected)
 {
-    /* Implement a barrier suitable for a mutex unlock. */
-    __insn_mf();
     return atomic_compare_and_exchange_val_acq(&var->counter, new, expected);
 }
 
+/*
+ * Atomic ops with at least specified barriers.
+ */
+
+static ETHR_INLINE long
+ethr_native_atomic_read_acqb(ethr_native_atomic_t *var)
+{
+    long res = ethr_native_atomic_read(var);
+    ETHR_MEMORY_BARRIER;
+    return res;
+}
+
+static ETHR_INLINE long
+ethr_native_atomic_inc_return_acqb(ethr_native_atomic_t *var)
+{
+    long res = ethr_native_atomic_inc_return(var);
+    ETHR_MEMORY_BARRIER;
+    return res;
+}
+
+static ETHR_INLINE void
+ethr_native_atomic_set_relb(ethr_native_atomic_t *var, long val)
+{
+    ETHR_MEMORY_BARRIER;
+    ethr_native_atomic_set(var, val);
+}
+
+static ETHR_INLINE void
+ethr_native_atomic_dec_relb(ethr_native_atomic_t *var)
+{
+    ETHR_MEMORY_BARRIER;
+    ethr_native_atomic_dec(var);
+}
+
+static ETHR_INLINE long
+ethr_native_atomic_dec_return_relb(ethr_native_atomic_t *var)
+{
+    ETHR_MEMORY_BARRIER;
+    return ethr_native_atomic_dec_return(var);
+}
+
+static ETHR_INLINE long
+ethr_native_atomic_cmpxchg_acqb(ethr_native_atomic_t *var, long new, long exp)
+{
+    return ethr_native_atomic_cmpxchg(var, new, exp);
+}
+
+static ETHR_INLINE long
+ethr_native_atomic_cmpxchg_relb(ethr_native_atomic_t *var, long new, long exp)
+{
+    ETHR_MEMORY_BARRIER;
+    return ethr_native_atomic_cmpxchg(var, new, exp);
+}
+
 #endif /* ETHR_TRY_INLINE_FUNCS */
 
 #endif /* ETHREAD_TILE_ATOMIC_H */