Boost C++ Libraries: Ticket #6308: Add sp_counted_base_aix.hpp using AIX atomic operations

attachment set

aaron.riekenberg@… — Tue, 20 Dec 2011 23:53:09 GMT

attachment → sp_counted_base_aix.patch

Patch

attachment set

aaron.riekenberg@… — Tue, 20 Dec 2011 23:54:03 GMT

attachment → test.cpp

Benchmark program

Peter Dimov — Wed, 21 Dec 2011 00:36:57 GMT

(In [76086]) Apply AIX patch from #6308. Refs #6308.

status changed; resolution set

Peter Dimov — Wed, 21 Dec 2011 00:43:58 GMT

status new → closed
resolution → fixed

(In [76087]) Merge [76086] to release. Fixes #6308.

attachment set

aaron.riekenberg@… — Thu, 22 Dec 2011 21:05:45 GMT

attachment → sp_counted_base_aix_memory_barrier.patch

Patch to add memory barrier to sp_counted_base_aix.hpp

status changed; resolution deleted

aaron.riekenberg@… — Thu, 22 Dec 2011 21:07:25 GMT

status closed → reopened
resolution fixed

As documented in smart_ptr/detail/atomic_count.hpp, in cases when the reference count is decremented to 0 we need a memory barrier before destroying the pointed-to object.

sp_counted_base_aix_memory_barrier.patch adds this using the isync instruction.

Peter Dimov — Thu, 22 Dec 2011 22:12:24 GMT

If fetch_and_add doesn't contain any memory barriers, a trailing isync is not enough, we need a leading (lw)sync as well. See sp_counted_base_gcc_ppc.hpp. Does AIX only work on PPC?

attachment set

aaron.riekenberg@… — Fri, 23 Dec 2011 13:12:15 GMT

attachment → sp_counted_base_aix_122311.hpp

aaron.riekenberg@… — Fri, 23 Dec 2011 13:18:36 GMT

AIX only runs on POWER and PowerPC.

fetch_and_add is not documented to do any memory barrier. I looked at the instructions for fetch_and_add with listi in dbx. It does lwarx and stwcx but does not do lwsync, sync, or isync. So I believe it does not contain a barrier.

I am attaching a new version of sp_counted_base_aix.hpp (sp_counted_base_aix_122311.hpp). This version is written more in the style of sp_counted_base_gcc_ppc.hpp. I added a leading sync instruction to atomic_decrement and a trailing isync for a full barrier.

The performance of this version is a bit worse than my original no-barrier version. Each iteration of the test program now takes about 8.8 microseconds. This is still 2.5X faster than using pthread_mutex.

Thanks for your patience in getting this correct.

Peter Dimov — Fri, 23 Dec 2011 15:03:41 GMT

(In [76119]) Add memory barriers to sp_counted_base_aix.hpp. Refs #6308.

Peter Dimov — Fri, 23 Dec 2011 15:07:21 GMT

lwsync should be faster. We didn't use it in the PPC version because, as I recall, GCC didn't support it. I'm not sure what compiler are you targeting on AIX.

Peter Dimov — Fri, 23 Dec 2011 15:10:53 GMT

(In [76120]) Merge [76119] to release. Refs #6308.

attachment set

aaron.riekenberg@… — Fri, 23 Dec 2011 17:23:14 GMT

attachment → sp_counted_base_aix_builtin_ns.patch

aaron.riekenberg@… — Fri, 23 Dec 2011 17:28:15 GMT

I experimented with lwsync. It was slightly faster than sync in my tests (around 300 nanoseconds). lwsync does not enforce ordering of stores followed by loads (http://www.ibm.com/developerworks/systems/articles/powerpc.html). So I think I'll stick with sync for this reason, and to be consistent with sp_counted_base_gcc_ppc.hpp.

I am attaching one last (hopefully!) patch which does the following:

Uses builtin functions sync() and isync() instead of inline asm.

Puts the atomic_* functions inside the boost::detail namespace - I accidentally put them in the boost namespace previously.

Thanks again for all your help on this.

Peter Dimov — Fri, 23 Dec 2011 17:35:43 GMT

Replying to aaron.riekenberg@…:

I experimented with lwsync. It was slightly faster than sync in my tests (around 300 nanoseconds). lwsync does not enforce ordering of stores followed by loads (http://www.ibm.com/developerworks/systems/articles/powerpc.html).

Yes, I know what lwsync does. You don't need sync here.

attachment set

aaron.riekenberg@… — Fri, 23 Dec 2011 20:45:04 GMT

attachment → sp_counted_base_aix_lwsync.patch

Switch from sync to lwsync

aaron.riekenberg@… — Fri, 23 Dec 2011 20:49:01 GMT

I convinced myself you are correct about lwsync.

Average performance of all approaches in test.cpp per iteration:

Original pthread_mutex implementation: 23 microseconds Atomic operations, barrier using sync: 8.8 microseconds Atomic operations, barrier using lwsync: 8.2 microseconds Atomic operations, no barrier (broken): 4.7 microseconds

Peter Dimov — Fri, 23 Dec 2011 23:10:39 GMT

(In [76123]) sp_counted_base_aix.hpp: switch to lwsync and builtins. Refs #6308.

Peter Dimov — Fri, 23 Dec 2011 23:12:53 GMT

(In [76124]) Merge [76123] to release. Refs #6308.

vfried@… — Thu, 01 Nov 2012 17:21:27 GMT

What is the status of sp_counted_base_aix.hpp? It is not #included in sp_counted_base.hpp, but it is a part of the official sources and there is no warning that it is e.g. experimental or for testing only.

Peter Dimov — Thu, 01 Nov 2012 17:28:40 GMT

Hmm. It seems that the patch in #6667 removed the #ifdef _AIX part of sp_counted_base.hpp for some reason and I didn't notice.

Peter Dimov — Thu, 01 Nov 2012 17:50:44 GMT

(In [81134]) Add back _AIX-specific #ifdef that was mistakenly removed. Refs #6308. Refs #6667.

status changed; resolution set

Peter Dimov — Wed, 14 Nov 2012 12:09:32 GMT

status reopened → closed
resolution → fixed

(In [81336]) Merged [81134] from trunk. Fixes #6308. Refs #6667.