boost.atomic C++0x atomic library for boost Fix typo in interlocked.hpp 2011-11-05T16:00:51Z Helge Bahmann hcb@chaoticmind.net 2011-11-05T16:00:51Z aaa5f08d75db8aa329b7856d890363bb904f4c09 Fix accidental call qualification of memory_order_seq_cst in interlocked.hpp. Reported-by: Eichhorn Mike-Joachim <mike.eichhorn@tu-ilmenau.de> 
Fix accidental call qualification of memory_order_seq_cst in
interlocked.hpp.

Reported-by: Eichhorn Mike-Joachim <mike.eichhorn@tu-ilmenau.de>
Add (tentative) sparcv9 support 2011-10-26T20:39:32Z Helge Bahmann hcb@chaoticmind.net 2011-10-26T20:39:32Z 59791b540d8fe9b65d80240080082c9bdf257ecc Add support for sparcv9 atomic operations. Only 32 bit mode supported so far, and I am not yet sure about the memory barriers. 
Add support for sparcv9 atomic operations. Only 32 bit mode
supported so far, and I am not yet sure about the memory barriers.
Update documentation to describe unit tests 2011-10-16T14:12:09Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T14:12:09Z defaea9f432fdcf6f196dc99b6799e33d28196c4 Update documentation to explain what the various test programs are supposed to do. 
Update documentation to explain what the various test programs
are supposed to do.
Add test for memory ordering 2011-10-16T13:48:35Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T13:48:35Z 79d2b8a47f54dce13e3c237ce5a58fe2d2dfe7f2 Add a test program to verify that memory ordering works as expected. Note that this program cannot possibly fail except on true SMP platforms, so do not expect anything from it when running it on uniprocessor systems or inside an emulator. 
Add a test program to verify that memory ordering works as expected.
Note that this program cannot possibly fail except on true SMP
platforms, so do not expect anything from it when running it on
uniprocessor systems or inside an emulator.
Add platform_fence_after_store 2011-10-16T13:46:14Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T13:46:14Z 94d0ec6ab8346fadc3834158bd5205940445a795 Add platform_fence_after_store that allows insertion of platform- specific synchronization after store. In contrast to "platform_fence_after" generally nothing is required here except for "seq_cst" so this allows somewhat more relaxed but still correct memory ordering. 
Add platform_fence_after_store that allows insertion of platform-
specific synchronization after store. In contrast to
"platform_fence_after" generally nothing is required here except
for "seq_cst" so this allows somewhat more relaxed but still
correct memory ordering.
Add test for expectation of bool lock freedom 2011-10-16T12:32:03Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T12:32:03Z 38ad4a16623700ee032960a03dced2287f9c7891 Test whether lock freedom of bool matches expectations for supported platforms. 
Test whether lock freedom of bool matches expectations for
supported platforms.
Override atomicity of bool for platforms 2011-10-16T12:30:28Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T12:30:28Z 89a20cdbe33545282ce3214784d12cbc7110853f Properly mark "bool" as lock free for all platforms supplying native atomic operations. 
Properly mark "bool" as lock free for all platforms supplying
native atomic operations.
Add test macro for bool lock freedom 2011-10-16T12:25:59Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T12:25:59Z 36b7be2f607c1fd7ecebeba64cc1fb488bb74eb3 Add macro BOOST_ATOMIC_BOOL_LOCK_FREE. 
Add macro BOOST_ATOMIC_BOOL_LOCK_FREE.
Mark bool as "generic" type 2011-10-16T12:19:43Z Helge Bahmann hcb@chaoticmind.net 2011-10-16T12:19:43Z 0b0deb62df6a8dcec9eee4a288dbd98575ffd757 Mark bool as generic non-integral type so that is automatically mapped to a platform-specific atomic operation. 
Mark bool as generic non-integral type so that is automatically
mapped to a platform-specific atomic operation.
Fix out-of-line code for Apple 2011-10-15T11:40:34Z Helge Bahmann hcb@chaoticmind.net 2011-10-15T11:40:34Z 088187d176c1bee36de361d5af864dde4824e53a Apple's assembler apparently does not understand the section flags that are used to move asm code out-of-line. Therefore make this feature conditional on !defined(__APPLE__). 
Apple's assembler apparently does not understand the section
flags that are used to move asm code out-of-line. Therefore make
this feature conditional on !defined(__APPLE__).