Boost C++ Libraries: Ticket #7906: Very bad performance of generic implementation of shared_mutex on windows

Sat, 19 Jan 2013 07:38:58 GMT

My proposal is to fix bugs 7720 and 7755 in win32 implementation and leave it default in boost 1.53.

component, severity changed; keywords, owner set

Sat, 19 Jan 2013 07:40:17 GMT

keywords thread shared_mutex added
owner set to Anthony Williams
component None → thread
severity Problem → Regression

viboes — Sat, 19 Jan 2013 10:44:43 GMT

I don't know if you are aware but BOOST_THREAD_PROVIDES_GENERIC_SHARED_MUTEX_ON_WIN will not be defined by default in 1.53.

Anyway, I have reopened the tickets that were closed as duplicated #7720 and #7755.

BTW, thanks for providing a performance test.

owner, status, severity changed

viboes — Sat, 19 Jan 2013 11:21:13 GMT

owner changed from Anthony Williams to viboes
status new → assigned
severity Regression → Optimization

Moved to optimization as BOOST_THREAD_PROVIDES_GENERIC_SHARED_MUTEX_ON_WIN is not defined by default.

attachment set

viboes — Sun, 27 Jan 2013 21:14:57 GMT

attachment → 7906_it.patch

viboes — Sun, 27 Jan 2013 21:19:54 GMT

One of the factors making the difference is that the win32 doesn't mask the interruption.

Once the attached patch is applied the measures are closer even if the win32 is yet better.

Please could you try the attached patch on trunk? BTW, give the measure for the best sample as done on the performance test example/perf_shared_mutex.cpp

viboes — Thu, 07 Feb 2013 18:50:18 GMT

Replying to viboes:

One of the factors making the difference is that the win32 doesn't mask the interruption.

Once the attached patch is applied the measures are closer even if the win32 is yet better.

Please could you try the attached patch on trunk? BTW, give the measure for the best sample as done on the performance test example/perf_shared_mutex.cpp

Andrey, any news on the measures?

Fri, 08 Feb 2013 12:32:20 GMT

I am going to test you patch on next week. Sorry for the delay.

viboes — Sat, 09 Feb 2013 14:00:09 GMT

Replying to Andrey <nikolay@…>:

I am going to test you patch on next week. Sorry for the delay.

After thinking a little bit more on this subject, I'm not sure shared_mutex should be an interruption point or not. With the pthread implementation most of the operations are interruption points as they use condition_variables, but the win32 implementation don't use condition_variables. What do you think?

In any case the patch is incomplete as it only disable interrupts, but doesn't adds calls to this_thread::interruption_point().

The test I also did is to compile without interruptions. Please could you give the performances when you enable/disable interruptions (BOOST_THREAD_PROVIDES_INTERRUPTIONS/BOOST_THREAD_DONT_PROVIDE_INTERRUPTIONS)?

Tue, 12 Feb 2013 05:01:08 GMT

Results (average for 5 runs) on my new PC:

shared_mutex_speed_win32_no_patch: Time spent:1910861 nanoseconds

shared_mutex_speed_generic_no_patch: Time spent:6831144 nanoseconds

shared_mutex_speed_win32_patch: Time spent:1969341 nanoseconds

shared_mutex_speed_generic_patch: Time spent:6782312 nanoseconds

shared_mutex_speed_generic_patch_no_interruptions (defined BOOST_THREAD_DONT_PROVIDE_INTERRUPTIONS): Time spent:6798547 nanoseconds

shared_mutex_speed_win32_patch_no_interruptions (defined BOOST_THREAD_DONT_PROVIDE_INTERRUPTIONS): Time spent:1998582 nanoseconds

In generic implementation most of CPU time is spent on boost::mutex::scoped_lock lk(state_change);. Win32 implementation is lock free and it is a reason why this implementation is very fast.

viboes — Sun, 17 Feb 2013 16:47:45 GMT

I don't know what I can do to improve this situation. I'm working on fixing some of the shared_mutex fixes on win specific implementation. The generic implementation provides more services, e.g. timed operations which are missing on the win specific implementation. Could I close the ticket?

status changed; resolution set; milestone deleted

viboes — Mon, 18 Feb 2013 20:26:40 GMT

status assigned → closed
resolution → wontfix
milestone To Be Determined