- Boost - lists.boost.org

accurate sum accumulator (kahan)
by Gaetano Mendola 27 Apr '11

27 Apr '11

Hi all, I had to implement in terms of boost accumulators the Kahan summation algorithm. I will paste at the end of this message the accumulator, the feature and the extractor source code. This accumulator reduces the numerical error obtained with standard sequential sum, as example the sum of 1e6f elements equal to 1e-6f should give 1.0f, this is the results obtained with sum and sum_kahan: tag::sum: 1.009038925 tag::sum_kahan: 1.000000000 I'm not sure if that sum_kahan is worth to be inserted in the accumulators provided with boost, I'm not an expert of boost:accumulators (I just followed the guide to make this new accumulator) so not sure if you prefer to have a fast_sum and an accurate_sum instead of sum and sum_kahan. Regards Gaetano Mendola // File kahan_sum.hpp #ifndef KAHAN_SUM #define KAHAN_SUM #include <boost/accumulators/framework/accumulator_base.hpp> #include <boost/accumulators/framework/parameters/sample.hpp> #include <boost/numeric/conversion/cast.hpp> namespace boost { namespace accumulators { namespace impl { template<typename Sample> struct sum_kahan_accumulator : accumulator_base { typedef Sample result_type; template<typename Args> sum_kahan_accumulator(Args const & args) : sum(args[sample | Sample()]), compensation(boost::numeric_cast<Sample>(0.0)) { } template<typename Args> void operator ()(Args const & args) { const Sample myTemp1 = args[sample] - this->compensation; const Sample myTemp2 = this->sum + myTemp1; this->compensation = (myTemp2 - this->sum) - myTemp1; this->sum = myTemp2; } result_type result(dont_care) const { return this->sum; } private: Sample sum; Sample compensation; }; }}} namespace boost { namespace accumulators { namespace tag { struct sum_kahan : depends_on< > { typedef impl::sum_kahan_accumulator< mpl::_1 > impl; }; }}} namespace boost { namespace accumulators { namespace extract { extractor< tag::sum_kahan > const sum_kahan = {}; } using extract::sum_kahan; }} #endif

21 68

[Config] A proposal to Boost.Config.
by Kohei Takahashi 27 Apr '11

27 Apr '11

Hi, Boost.Config provides BOOST_PLATFORM macro (it will expand to strings). However, it cannot use to switch including header like followings. #include <boost/config.hpp> #include "somedir/" BOOST_PLATFORM "/someheader.hpp" So, I propose a new macro BOOST_PLATFORM_LABEL (change to more suitable name). That provide not stringized platform name like this. #define BOOST_PLATFORM_LABEL linux By using such macro, will be able to switch including header. #include <boost/config.hpp> #include <boost/preprocessor/stringize.hpp> #include BOOST_PP_STRINGIZE( somedir/BOOST_PLATFORM_LABEL/someheader.hpp ) However, such macro has a problem. For example, it fail including when *linux* (or another label) defined to other symbol. #include <boost/config.hpp> #include <boost/preprocessor/stringize.hpp> #define linux others #include BOOST_PP_STRINGIZE( somedir/BOOST_PLATFORM_LABEL/someheader.hpp ) // Oops! Expand to somedir/others/someheader.hpp in linux. How do you feel this macro? Thanks. -- TAKAHASHI Kohei College of Information Science, University of Tsukuba private mail: flast(a)flast.jp others: flast(a)ac-room.org s0911476(a)coins.tsukuba.ac.jp

2 2

[vault] Boost Vault is down
by Edward Diener 27 Apr '11

27 Apr '11

It appears the link to the Boost vault is down. Attempting http://www.boostpro.com/vault/index.php gives the "Server not found" message. Can this be fixed fairly ASAP ? The distribution for the next review, starting tomorrow, is in the vault ( no surprise there ), and since I am the review manager I have already gotten an e-mail pointing out that the distribution can not be accessed.

4 4

[GSoC] Results
by Andrew Sutton 26 Apr '11

26 Apr '11

All, I'd like to thank all of the students who submitted proposals this year. We had a number of excellent proposals, many more than last year. Ultimately, we were only able to accept 10 proposals and so some deserving proposals could not be accepted. If your proposal was not accepted, and you would like to know why, please send me an email. I will be more than happy to discuss the results with you. For the students whose proposals were accepted, tomorrow starts the "Getting to know you" phase of GSoC. Assuming that you will be done with exams soon, you should be getting started with the Boost infrastructure, build system, and documentation tools. Questions about any of the above can be directed to your mentors or the mailing list --- hopefully both. I will send out some instructions this week for getting started with the Boost SVN repository. I'd also like to thank all of the mentors who contributed to the reviewing process. The reviewing seems to have gone (more or less) smoothly, and the quality of the reviews was excellent. Thanks again, Andrew

1 0

Re: [boost] [MSM] Interrupt anonymous transitions or blocking process_events?
by Christophe Henry 26 Apr '11

26 Apr '11

>> Depending on how you use it, yes. I'm a bit worried with race >> conditions in case your state machine is also used in another thread, >> which would also call process_event. > > Yes, I protect every call to process_event with a mutex. Since I don't > have that many events I have created a member function that wraps and > protects the process_event call. (One member function per event.) Usually, the destructor gets tricky. Not only, but this one is hard to get right. >> Alternatively (if you want to avoid locking/deadlocking/rc), you could >> also move the worker thread out of the fsm and have both communicate >> with synchronized message queues. You might want to have a look at >> boost::circular_buffer, which provides an example >> (http://www.boost.org/doc/libs/1_46_1/libs/circular_buffer/doc/circular_buff…) > > Yes, that is an interesting idea. But how do you solve the problem with > passing events on to the message queue? The events are all of different > types. Do you mean that there should be some parsing of strings or some > other identifier? You can make a circular buffer of, say, boost::function<void()>, this allows you to anonymize functors, for example this can be assigned into this boost::function: template <class Fsm, class Event> struct process_some_event { process_some_event(Fsm* fsm): fsm_(fsm){} void operator()() { fsm_->process_event(Event()); } Fsm* fsm_; }; Or as a variant, you can make a boost::function<void(my_fsm&)> with the corresponding operator(). Whatever fits better in your case. Regards, Christophe

2 1

Darwin Clang help needed
by John Maddock 26 Apr '11

26 Apr '11

There are some strange failures in Boost.Math for Clang on Darwin, for example: boost::math::binomial_coefficient<long double>(42, 15); Appears to be returning zero. The code is pretty simple, so is anyone able to step through it and see what's going wrong (it seems to be working OK for type double BTW). Thanks, John.

2 2

[asio] [date_time] boost::asio::deadline_timer performance
by Isidor Zeuner 25 Apr '11

25 Apr '11

Hi all, while profiling a boost::asio-based application with heavy use of timers, I noticed many CPU cycles being burnt inside a gregorian::date::date constructor, which I traced down as being called from asio::deadline_timer::expires_from_now. As asio::deadline_timer just specializes asio::basic_deadline_timer using posix_time::ptime, I assume this can be tuned by using another time implementation. Is there a particular implementation targeted at low overhead, ideally with high accuracy? Are there other ways of tuning timer performance? Best regards, Isidor -- Isidor Zeuner http://quidecco.de/cms/blog/

1 0

[accumulators]
by Edward Diener 25 Apr '11

25 Apr '11

Attempting to run the accumulators regression test on the trunk, using 'bjam toolset=gcc-4.5.2' ( MingW/gcc-4.5.2 ) is giving me a number of pop-up Windows in Vista saying that "cc1plus.exe has stopped working". Cc1plus.exe is the gcc compiler under MingW. In the output there are a number of failures. Does anybody know why this is occuring ?

1 0

[Pack Pointer] Review Request
by Phil Bouchard 25 Apr '11

25 Apr '11

Greetings, I've followed what users were dissatisfied with and I turned out renaming Shifted Pointer to Pack Pointer. Furthermore I've added a small tutorial that can be found at: https://svn.boost.org/svn/boost/sandbox/pack_ptr/libs/smart_ptr/doc/tutoria… Once again the main documentation is at: https://svn.boost.org/svn/boost/sandbox/pack_ptr/libs/smart_ptr/doc/index.h… I still need cleaning up some code but it works well with complex examples and I would like queuing the library up for a review if possible. Regards, -Phil

2 2

[build] bjam library path on gcc
by Lorenzo Caminiti 25 Apr '11

25 Apr '11

Hello all, How do I setup gcc library path (LD_LIBRARY_PATH) for bjam? For example I get this error when using bjam: lcaminiti@lcaminiti-PORTEGE-R700:~/sandbox/boost-sandbox-local/libs/local/example$ bjam -q profile_boost_local toolset=gcc ...patience... ...found 790 targets... ...updating 2 targets... gcc.compile.c++ bin/gcc-4.5.1/debug/profile_boost_local.o gcc.link bin/gcc-4.5.1/debug/profile_boost_local bin/gcc-4.5.1/debug/profile_boost_local.o: In function `main': /home/lcaminiti/sandbox/boost-sandbox-local/libs/local/example/profile_boost_local.cpp:17: undefined reference to `boost::chrono::system_clock::now()' /home/lcaminiti/sandbox/boost-sandbox-local/libs/local/example/profile_boost_local.cpp:23: undefined reference to `boost::chrono::system_clock::now()' /home/lcaminiti/sandbox/boost-sandbox-local/libs/local/example/profile_boost_local.cpp:31: undefined reference to `boost::chrono::system_clock::now()' /home/lcaminiti/sandbox/boost-sandbox-local/libs/local/example/profile_boost_local.cpp:33: undefined reference to `boost::chrono::system_clock::now()' bin/gcc-4.5.1/debug/profile_boost_local.o: In function `__static_initialization_and_destruction_0': /home/lcaminiti/sandbox/boost-trunk/boost/system/error_code.hpp:214: undefined reference to `boost::system::generic_category()' /home/lcaminiti/sandbox/boost-trunk/boost/system/error_code.hpp:215: undefined reference to `boost::system::generic_category()' /home/lcaminiti/sandbox/boost-trunk/boost/system/error_code.hpp:216: undefined reference to `boost::system::system_category()' collect2: ld returned 1 exit status "g++" -o "bin/gcc-4.5.1/debug/profile_boost_local" -Wl,--start-group "bin/gcc-4.5.1/debug/profile_boost_local.o" -Wl,-Bstatic -Wl,-Bdynamic -Wl,--end-group -g ...failed gcc.link bin/gcc-4.5.1/debug/profile_boost_local... ...failed updating 1 target... ...updated 1 target... But I can tell bjam to add -L and -l options as in the following it would work: lcaminiti@lcaminiti-PORTEGE-R700:~/sandbox/boost-sandbox-local/libs/local/example$ g++ -o "bin/gcc-4.5.1/debug/profile_boost_local" -Wl,--start-group "bin/gcc-4.5.1/debug/profile_boost_local.o" -Wl,-Bstatic -Wl,-Bdynamic -Wl,--end-group -g -L/home/lcaminiti/sandbox/boost-trunk/stage/lib -lboost_chrono -lboost_system Thank you. -- Lorenzo

2 1