Re: [Boost-bugs] [Boost C++ Libraries] #7422: Provide a condition variable with zero-overhead performance penality

Subject: Re: [Boost-bugs] [Boost C++ Libraries] #7422: Provide a condition variable with zero-overhead performance penality
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2012-12-14 07:40:52


#7422: Provide a condition variable with zero-overhead performance penality
------------------------------------------+---------------------------------
  Reporter: maxim.yegorushkin@… | Owner: viboes
      Type: Feature Requests | Status: assigned
 Milestone: To Be Determined | Component: thread
   Version: Boost 1.51.0 | Severity: Regression
Resolution: | Keywords: condition_variable
------------------------------------------+---------------------------------

Comment (by viboes):

 I have run the no-interruptions benchmark (modified to compile on MacOS)
 using clang-3.2 without setting any special conditions changing 3
 parameters:

 CONSUMER_MAX: from 20 to 10
 test iterations: from 3 to 100
 shared_data iteration : from 100000 to 10000.

 here are the results:


 {{{
 consumers,(std-boost)/std,std,boost
 1,-5.063747,44009306,46237826
 2,-1.764249,72169561,73442812
 3,5.327530,94552244,89514945
 4,0.767002,121827567,120893147
 5,1.826591,141517038,138932100
 6,0.471998,171223003,170414834
 7,-0.937626,187902198,189664017
 8,-0.840567,220015472,221864850
 9,-3.291703,236884513,244682047
 10,0.397024,272506100,271424186
 }}}

 I'm sure that the setting of your special condition would change these
 results.
 From these results, I don't see a loss of 5% of Boost respect to std.
 Please could you run the benchmark with these parameters ans your
 settings:
 * running the benchmark with real-time priority FIFO 99,
 * disable CPU frequency scaling to prevent the CPU from shifting gears
 while benchmarking

 How do you do this?

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/7422#comment:19>
Boost C++ Libraries <http://www.boost.org/>
Boost provides free peer-reviewed portable C++ source libraries.

This archive was generated by hypermail 2.1.7 : 2017-02-16 18:50:11 UTC