Adding some 8-bit atomic operations for future use

llvm-svn: 231774

diff --git a/openmp/runtime/src/kmp_os.h b/openmp/runtime/src/kmp_os.h
index 1112096d9ee2d369578fc77d6ca41e5c41501476..20868cd5b4ebcd36c8968790964cb677110f540e 100644 (file)
--- a/openmp/runtime/src/kmp_os.h
+++ b/openmp/runtime/src/kmp_os.h
@@ -458,6 +458,9 @@ enum kmp_mem_fence_type {
 # define KMP_TEST_THEN_DEC_ACQ32(p)             InterlockedExchangeAdd( (volatile long *)(p), -1 )
 # define KMP_TEST_THEN_ADD32(p, v)              InterlockedExchangeAdd( (volatile long *)(p), (v) )
 
+extern kmp_int8 __kmp_test_then_add8( volatile kmp_int8 *p, kmp_int8 v );
+extern kmp_int8 __kmp_test_then_or8( volatile kmp_int8 *p, kmp_int8 v );
+extern kmp_int8 __kmp_test_then_and8( volatile kmp_int8 *p, kmp_int8 v );
 # define KMP_COMPARE_AND_STORE_RET32(p, cv, sv) InterlockedCompareExchange( (volatile long *)(p),(long)(sv),(long)(cv) )
 
 # define KMP_XCHG_FIXED32(p, v)                 InterlockedExchange( (volatile long *)(p), (long)(v) )
@@ -494,8 +497,11 @@ extern kmp_int32 __kmp_xchg_fixed32( volatile kmp_int32 *p, kmp_int32 v );
 extern kmp_int64 __kmp_xchg_fixed64( volatile kmp_int64 *p, kmp_int64 v );
 extern kmp_real32 __kmp_xchg_real32( volatile kmp_real32 *p, kmp_real32 v );
 extern kmp_real64 __kmp_xchg_real64( volatile kmp_real64 *p, kmp_real64 v );
+# define KMP_TEST_THEN_ADD8(p, v)              __kmp_test_then_add8( (p), (v) )
 
 //# define KMP_TEST_THEN_INC32(p)                 __kmp_test_then_add32( (p), 1 )
+# define KMP_TEST_THEN_OR8(p, v)               __kmp_test_then_or8( (p), (v) )
+# define KMP_TEST_THEN_AND8(p, v)              __kmp_test_then_and8( (p), (v) )
 //# define KMP_TEST_THEN_INC_ACQ32(p)             __kmp_test_then_add32( (p), 1 )
 # define KMP_TEST_THEN_INC64(p)                 __kmp_test_then_add64( (p), 1LL )
 # define KMP_TEST_THEN_INC_ACQ64(p)             __kmp_test_then_add64( (p), 1LL )
@@ -544,9 +550,12 @@ extern kmp_real64 __kmp_xchg_real64( volatile kmp_real64 *p, kmp_real64 v );
 
 
 #elif (KMP_ASM_INTRINS && KMP_OS_UNIX) || !(KMP_ARCH_X86 || KMP_ARCH_X86_64)
+# define KMP_TEST_THEN_ADD8(p, v)               __sync_fetch_and_add( (kmp_int8 *)(p), (v) )
 
 /* cast p to correct type so that proper intrinsic will be used */
 # define KMP_TEST_THEN_INC32(p)                 __sync_fetch_and_add( (kmp_int32 *)(p), 1 )
+# define KMP_TEST_THEN_OR8(p, v)                __sync_fetch_and_or( (kmp_int8 *)(p), (v) )
+# define KMP_TEST_THEN_AND8(p, v)               __sync_fetch_and_and( (kmp_int8 *)(p), (v) )
 # define KMP_TEST_THEN_INC_ACQ32(p)             __sync_fetch_and_add( (kmp_int32 *)(p), 1 )
 # define KMP_TEST_THEN_INC64(p)                 __sync_fetch_and_add( (kmp_int64 *)(p), 1LL )
 # define KMP_TEST_THEN_INC_ACQ64(p)             __sync_fetch_and_add( (kmp_int64 *)(p), 1LL )
@@ -586,6 +595,9 @@ extern kmp_real64 __kmp_xchg_real64( volatile kmp_real64 *p, kmp_real64 v );
 #define KMP_XCHG_FIXED32(p, v)                  __sync_lock_test_and_set( (volatile kmp_uint32 *)(p), (kmp_uint32)(v) )
 #define KMP_XCHG_FIXED64(p, v)                  __sync_lock_test_and_set( (volatile kmp_uint64 *)(p), (kmp_uint64)(v) )
 
+extern kmp_int8 __kmp_test_then_add8( volatile kmp_int8 *p, kmp_int8 v );
+extern kmp_int8 __kmp_test_then_or8( volatile kmp_int8 *p, kmp_int8 v );
+extern kmp_int8 __kmp_test_then_and8( volatile kmp_int8 *p, kmp_int8 v );
 inline kmp_real32 KMP_XCHG_REAL32( volatile kmp_real32 *p, kmp_real32 v)
 {
     kmp_int32 tmp = __sync_lock_test_and_set( (kmp_int32*)p, *(kmp_int32*)&v);
@@ -621,9 +633,12 @@ extern kmp_int16 __kmp_xchg_fixed16( volatile kmp_int16 *p, kmp_int16 v );
 extern kmp_int32 __kmp_xchg_fixed32( volatile kmp_int32 *p, kmp_int32 v );
 extern kmp_int64 __kmp_xchg_fixed64( volatile kmp_int64 *p, kmp_int64 v );
 extern kmp_real32 __kmp_xchg_real32( volatile kmp_real32 *p, kmp_real32 v );
+# define KMP_TEST_THEN_ADD8(p, v)               __kmp_test_then_add8( (p), (v) )
 extern kmp_real64 __kmp_xchg_real64( volatile kmp_real64 *p, kmp_real64 v );
 
 # define KMP_TEST_THEN_INC32(p)                 __kmp_test_then_add32( (p), 1 )
+# define KMP_TEST_THEN_OR8(p, v)                __kmp_test_then_or8( (p), (v) )
+# define KMP_TEST_THEN_AND8(p, v)               __kmp_test_then_and8( (p), (v) )
 # define KMP_TEST_THEN_INC_ACQ32(p)             __kmp_test_then_add32( (p), 1 )
 # define KMP_TEST_THEN_INC64(p)                 __kmp_test_then_add64( (p), 1LL )
 # define KMP_TEST_THEN_INC_ACQ64(p)             __kmp_test_then_add64( (p), 1LL )
@@ -713,6 +728,8 @@ extern kmp_real64 __kmp_xchg_real64( volatile kmp_real64 *p, kmp_real64 v );
 # define KMP_LD_ACQ64(A)        ( *(A) )
 #endif
 
+#define TCR_1(a)            (a)
+#define TCW_1(a,b)          (a) = (b)
 /* ------------------------------------------------------------------------ */
 //
 // FIXME - maybe this should this be

diff --git a/openmp/runtime/src/z_Linux_util.c b/openmp/runtime/src/z_Linux_util.c
index a7b4eb12757ef4fa3bf14a58c2c57677b89f0973..4e3236db564aaa9c641e6fac2ad0f0bdb44f5133 100644 (file)
--- a/openmp/runtime/src/z_Linux_util.c
+++ b/openmp/runtime/src/z_Linux_util.c
@@ -437,6 +437,40 @@ __kmp_change_thread_affinity_mask( int gtid, kmp_affin_mask_t *new_mask,
         if ( old_mask != NULL ) {
             status = __kmp_get_system_affinity( old_mask, TRUE );
             int error = errno;
+kmp_int8
+__kmp_test_then_or8( volatile kmp_int8 *p, kmp_int8 d )
+{
+    kmp_int8 old_value, new_value;
+
+    old_value = TCR_1( *p );
+    new_value = old_value | d;
+
+    while ( ! KMP_COMPARE_AND_STORE_REL8 ( p, old_value, new_value ) )
+    {
+        KMP_CPU_PAUSE();
+        old_value = TCR_1( *p );
+        new_value = old_value | d;
+    }
+    return old_value;
+}
+
+kmp_int8
+__kmp_test_then_and8( volatile kmp_int8 *p, kmp_int8 d )
+{
+    kmp_int8 old_value, new_value;
+
+    old_value = TCR_1( *p );
+    new_value = old_value & d;
+
+    while ( ! KMP_COMPARE_AND_STORE_REL8 ( p, old_value, new_value ) )
+    {
+        KMP_CPU_PAUSE();
+        old_value = TCR_1( *p );
+        new_value = old_value & d;
+    }
+    return old_value;
+}
+
             if ( status != 0 ) {
                 __kmp_msg(
                     kmp_ms_fatal,
@@ -472,6 +506,23 @@ __kmp_change_thread_affinity_mask( int gtid, kmp_affin_mask_t *new_mask,
 
 #if KMP_OS_LINUX && (KMP_ARCH_X86 || KMP_ARCH_X86_64 || KMP_ARCH_ARM || KMP_ARCH_AARCH64) && !KMP_OS_CNK
 
+kmp_int8
+__kmp_test_then_add8( volatile kmp_int8 *p, kmp_int8 d )
+{
+    kmp_int8 old_value, new_value;
+
+    old_value = TCR_1( *p );
+    new_value = old_value + d;
+
+    while ( ! KMP_COMPARE_AND_STORE_REL8 ( p, old_value, new_value ) )
+    {
+        KMP_CPU_PAUSE();
+        old_value = TCR_1( *p );
+        new_value = old_value + d;
+    }
+    return old_value;
+}
+
 int
 __kmp_futex_determine_capable()
 {

diff --git a/openmp/runtime/src/z_Windows_NT-586_util.c b/openmp/runtime/src/z_Windows_NT-586_util.c
index 7b99f75e0f75d1b79d183ca779f96f77835d7ffa..3aeafae9102b10b0b4f497541aff92264da93e68 100644 (file)
--- a/openmp/runtime/src/z_Windows_NT-586_util.c
+++ b/openmp/runtime/src/z_Windows_NT-586_util.c
@@ -20,6 +20,40 @@
  * use compare_and_store for these routines
  */
 
+kmp_int8
+__kmp_test_then_or8( volatile kmp_int8 *p, kmp_int8 d )
+{
+    kmp_int8 old_value, new_value;
+
+    old_value = TCR_1( *p );
+    new_value = old_value | d;
+
+    while ( ! __kmp_compare_and_store8 ( p, old_value, new_value ) )
+    {
+        KMP_CPU_PAUSE();
+        old_value = TCR_1( *p );
+        new_value = old_value | d;
+    }
+    return old_value;
+}
+
+kmp_int8
+__kmp_test_then_and8( volatile kmp_int8 *p, kmp_int8 d )
+{
+    kmp_int8 old_value, new_value;
+
+    old_value = TCR_1( *p );
+    new_value = old_value & d;
+
+    while ( ! __kmp_compare_and_store8 ( p, old_value, new_value ) )
+    {
+        KMP_CPU_PAUSE();
+        old_value = TCR_1( *p );
+        new_value = old_value & d;
+    }
+    return old_value;
+}
+
 kmp_int32
 __kmp_test_then_or32( volatile kmp_int32 *p, kmp_int32 d )
 {
@@ -34,7 +68,6 @@ __kmp_test_then_or32( volatile kmp_int32 *p, kmp_int32 d )
         old_value = TCR_4( *p );
         new_value = old_value | d;
     }
-
     return old_value;
 }
 
@@ -55,6 +88,22 @@ __kmp_test_then_and32( volatile kmp_int32 *p, kmp_int32 d )
     return old_value;
 }
 
+kmp_int8
+__kmp_test_then_add8( volatile kmp_int8 *p, kmp_int8 d )
+{
+    kmp_int64 old_value, new_value;
+
+    old_value = TCR_1( *p );
+    new_value = old_value + d;
+    while ( ! __kmp_compare_and_store8 ( p, old_value, new_value ) )
+    {
+        KMP_CPU_PAUSE();
+        old_value = TCR_1( *p );
+        new_value = old_value + d;
+    }
+    return old_value;
+}
+
 #if KMP_ARCH_X86
 kmp_int64
 __kmp_test_then_add64( volatile kmp_int64 *p, kmp_int64 d )
@@ -69,7 +118,6 @@ __kmp_test_then_add64( volatile kmp_int64 *p, kmp_int64 d )
         old_value = TCR_8( *p );
         new_value = old_value + d;
     }
-
     return old_value;
 }
 #endif /* KMP_ARCH_X86 */