[Bf-blender-cvs] [c1ca667] master: Fix compilation error on Armel architecture
Sergey Sharybin
noreply at git.blender.org
Mon May 2 09:56:23 CEST 2016
Commit: c1ca667d4a9f42cda9dd7486f9826db07baacd2a
Author: Sergey Sharybin
Date: Mon May 2 09:55:08 2016 +0200
Branches: master
https://developer.blender.org/rBc1ca667d4a9f42cda9dd7486f9826db07baacd2a
Fix compilation error on Armel architecture
__GCC_HAVE_SYNC_COMPARE_AND_SWAP_n are not defined on this architecture
for some reason, however those functions are available.
Adding a workaround for newly used __sync_fetch_and_and() function.
===================================================================
M intern/atomic/atomic_ops.h
===================================================================
diff --git a/intern/atomic/atomic_ops.h b/intern/atomic/atomic_ops.h
index dc06a51..06eb8f2 100644
--- a/intern/atomic/atomic_ops.h
+++ b/intern/atomic/atomic_ops.h
@@ -45,6 +45,7 @@
* arm7 architecture does have both 32 and 64bit atomics, however
* it's gcc doesn't have __GCC_HAVE_SYNC_COMPARE_AND_SWAP_n defined.
*/
+# define JE_FORCE_SYNC_COMPARE_AND_SWAP_1
# define JE_FORCE_SYNC_COMPARE_AND_SWAP_8
# define JE_FORCE_SYNC_COMPARE_AND_SWAP_4
#endif
@@ -399,6 +400,12 @@ atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b)
return _InterlockedAnd8((char *)p, (char)b);
#endif
}
+#elif defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_1)
+ATOMIC_INLINE uint8_t
+atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b)
+{
+ return __sync_fetch_and_and(p, b);
+}
#else
# error "Missing implementation for 8-bit atomic operations"
#endif
More information about the Bf-blender-cvs
mailing list