- Boost-users - lists.preview.boost.org

[MPL] Fold a vector of pairs
by Peter Nyssen 14 Mar '12

14 Mar '12

Hi all, I'm trying to understand how to use mpl::fold to transform a vector like this: vector< pair<tag1, A> pair<tag2, B> pair<tag1, C> > into the following: pair< vector<A,C> vector<B> > Following test program fails to compile and I don't understand why. If you could explain me what I'm missing here, you will gain my eternal respect. Kind regards, Peter //-------------------------------------------- #include <boost/mpl/if.hpp> #include <boost/mpl/int.hpp> #include <boost/mpl/fold.hpp> #include <boost/mpl/pair.hpp> #include <boost/mpl/push_back.hpp> #include <boost/mpl/vector.hpp> namespace mpl = boost::mpl; struct tag1 {}; struct tag2 {}; template <typename T> struct is_tag1 : boost::false_type {}; template <> struct is_tag1<tag1> : boost::true_type {}; int main() { typedef mpl::vector< mpl::pair<tag1, mpl::int_<1> >, mpl::pair<tag2, mpl::int_<2> >, mpl::pair<tag1, mpl::int_<3> > > initial_vector; typedef typename mpl::fold< initial_vector, mpl::pair<mpl::vector<>, mpl::vector<> >, mpl::if_< is_tag1<mpl::first<mpl::_2>::type >, mpl::push_back< mpl::first<mpl::_1>::type, mpl::second<mpl::_2>::type >, mpl::push_back< mpl::second<mpl::_1>::type, mpl::second<mpl::_2>::type > > >::type sorted_pair; return 0; } //-------------------------------------------- *$* g++ -Wall -I ../boost/ -o test test.cpp In file included from test.cpp:5:0: ../boost/boost/mpl/pair.hpp: In instantiation of 'boost::mpl::first<mpl_::arg<2> >': test.cpp:32:40: instantiated from here ../boost/boost/mpl/pair.hpp:43:31: fout: no type named 'first' in 'struct mpl_::arg<2>' test.cpp: In functie 'int main()': test.cpp:32:47: fout: template argument 1 is invalid In file included from test.cpp:5:0: ../boost/boost/mpl/pair.hpp: At global scope: ../boost/boost/mpl/pair.hpp: In instantiation of 'boost::mpl::first<mpl_::arg<1> >': test.cpp:34:36: instantiated from here ../boost/boost/mpl/pair.hpp:43:31: fout: no type named 'first' in 'struct mpl_::arg<1>' ../boost/boost/mpl/pair.hpp: In instantiation of 'boost::mpl::second<mpl_::arg<2> >': test.cpp:35:37: instantiated from here ../boost/boost/mpl/pair.hpp:56:32: fout: no type named 'second' in 'struct mpl_::arg<2>' test.cpp: In functie 'int main()': test.cpp:36:13: fout: template argument 1 is invalid test.cpp:36:13: fout: template argument 2 is invalid In file included from test.cpp:5:0: ../boost/boost/mpl/pair.hpp: At global scope: ../boost/boost/mpl/pair.hpp: In instantiation of 'boost::mpl::second<mpl_::arg<1> >': test.cpp:38:37: instantiated from here ../boost/boost/mpl/pair.hpp:56:32: fout: no type named 'second' in 'struct mpl_::arg<1>' test.cpp: In functie 'int main()': test.cpp:40:13: fout: template argument 1 is invalid test.cpp:40:13: fout: template argument 2 is invalid test.cpp:41:9: fout: template argument 1 is invalid test.cpp:41:9: fout: template argument 2 is invalid test.cpp:41:9: fout: template argument 3 is invalid test.cpp:42:5: fout: template argument 3 is invalid test.cpp:42:24: fout: invalid type in declaration before ';' token **

2 1

Re: [Boost-users] [Graph] On Dijkstra SSSP performance
by Leo Hidd 13 Mar '12

13 Mar '12

sry, I could not make a reply in the post through http://thread.gmane.org/gmane.comp.lib.boost.user/73084/focus=73086 If i use the "Follow up" action and post it I have this message: "You seem to be top-posting. Don't do that." That's why I'm doing this now. Hope it works. Anyways... thx Jeremiah for the inputs. I tested what you said for a mesh of 3 million vertices (then, 9 million edges), and here is a summary of the different configs: 1- graph type: out-edge as listS; predecessors: vector<vertex_descriptor> p(num_vertices(g)); distances: vector<double> d(num_vertices(g)); NDEBUG defined (release configuration); no optimizations: - dijkstra_shortest_paths elapsed time: 11970ms 2- graph type: out-edge as listS; predecessors: vertex_descriptor* p = new vertex_descriptor[num_vertices(g)]; distances: double* d = new double[num_vertices(g)]; NDEBUG defined (release configuration); no optimizations: - dijkstra_shortest_paths elapsed time: 9446ms 3- graph type: out-edge as vecS; predecessors: vertex_descriptor* p = new vertex_descriptor[num_vertices(g)]; distances: double* d = new double[num_vertices(g)]; NDEBUG defined (release configuration); no optimizations: - dijkstra_shortest_paths elapsed time: 8436ms 4- graph type: out-edge as listS; predecessors: vertex_descriptor* p = new vertex_descriptor[num_vertices(g)]; distances: double* d = new double[num_vertices(g)]; NDEBUG defined (release configuration); with optimizations (option /Ox) - dijkstra_shortest_paths elapsed time: 5741ms 5- my implementation without optimizations takes 3289ms and with optimizations takes 2860ms. Using predecessors and distances storage as simple arrays instead of std::vector have a real gain (although optimization does even more). Is it possible to use arrays (maybe array of arrays) instead of vecS for OutEdgeList and VertexList? Or put arrays inside some kind of class with functions allowing compatibility/interfacing with the Dijkstra algorithm (like insert, sort, remove, or whatever it needs)? Since the Dijkstra algorithm does more read access to the edge list than the distance or path arrays, it could really accelerate things. thx again, leo

2 10

error compiling under Solaris10 in release mode
by Magallo 13 Mar '12

13 Mar '12

Hi all, I am new to boost libraries and I find them very powerful. I'm using the Boost.Interprocess library to make some interprocess communication and synchronization. I'm using mutexes and I was able to let my program build and work as expected. This is good when I build it in Debug mode, when I switch to release mode, i.e. using the -O2 or -O3 flag to optimize performances, the compiler returns this error: "boost/interprocess/detail/atomic.hpp:316: internal compiler error: in rewrite_stmt, at tree-into-ssa.c:1380" Please submit a full bug report, with preprocessed source if appropriate. When I use a debug compilation, without -O2, it works fine. Why that? I read somewhere that I could try to define BOOST_SP_USE_PTHREADS but I tryed and nothing changes. Is it something related to the operating system? I see that the error brings me to inline boost::uint32_t atomic_cas32 (volatile boost::uint32_t *mem, boost::uint32_t with, boost::uint32_t cmp) { return __sync_val_compare_and_swap(const_cast<boost::uint32_t *>(mem), cmp, with); } in the atomic.hpp file. The compiler is: g++ --version sparc-sun-solaris2.10-g++ (GCC) 4.2.1 (20090222) (gccfss) Copyright (C) 2007 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. What could I do? Please help. Cheers.

1 0

[Graph] How to do multi source shortest path
by Leo Hidd 13 Mar '12

13 Mar '12

Hi all, I have a graph and multiple sources (please read this thread: http://thread.gmane.org/gmane.comp.lib.boost.user/73084/focus=73086 first paragraph, for the characteristics of the graph, if it matters). So, instead of running the single source shortest path (SSSP) algorithm (e.g., dijkstra_shortest_paths()) multiple times, what I do is that I specify several sources instead of only one (so all the sources are initialized with distance 0 and put into the queue) and than I run Dijkstra as one usually does for SSSP. Doing so, the multi source runs almost as fast as a single source instead of running NbOfSouces*ElapsedTimeForSingleSource. Then, I'd like to know if it's possible to use boost algorithms for multi-source shortest path, like specifying several sources or anything alike. And, in the case it is possible, it would be possible to have an array with the index of the source the closest to a given node? All pairs is not a solution because the number of sources is way less than the number of nodes in the graph and because all sources use lots of memory for storing the distances (for a 4 million nodes graph I'd need a 4*10^12 elements table, don't?) thx for your time, leo

4 14

Re: [Boost-users] Boost interprocess containers - performance issue
by Naveen Santhapuri 12 Mar '12

12 Mar '12

> Date: Sun, 11 Mar 2012 12:21:36 +0100 > From: Ion Gazta?aga <igaztanaga(a)gmail.com> > To: Boost User List <boost-users(a)lists.boost.org> > Subject: Re: [Boost-users] Boost interprocess containers - performance issue > Message-ID: <4F5C8AC0.9020508(a)gmail.com> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed >El 11/03/2012 1:00, Naveen Santhapuri escribi?: >> Hello all, >> >> I am experimenting with Boost interprocess containers and like the >> ease of using them. I want to use these in place of the "linux shared >> memory+shared memory allocator + std containers". However, I am seeing >> some performance hit when using boost. I measured the times for vector >> access from shared memory. I define access as indexing into the vector >> (v[i]) (all times in milli seconds): >> >> Num elements Local heap | Boost ipc | Shmat() >> 100,000 0.23 0.39 0.27 >> 1,000,000 2.3 3.85 2.7 > What does Shmat mean? Are you putting a std::vector directly in a shared > memory segment? > Obviously if you put boost containers in shared memory you don't store > raw pointers (as every process can map the memory in a different > address, making raw pointers unusable). When using offset_ptr we have a > performance hit as we need to calculate the real address computing a > pointer arithmetic (this + stored offset). Apart from that, as > offset_ptr relies on undefined behaviour, the offset calculation must be > marked as non-inlineable in most compilers, increasing the performance > hit. I think it can be improved using some kind of compiler-specific > annotations or assembler, but I have no skills to do that nowadays. > > Best, > > Ion Thank you for the reply Ion. Re: shmat() - Yes, I am creating a std::vector with shared allocator in shared memory. My use case: One executable (lets call it shvec) writes std containers to shared memory. Another instance of shvec reads those containers from shared memory. I want to do this to preload a large amount of data into memory, before my service starts, so that the start time of the service can be cut down. I want to use boost but I am afraid of the runtime hit. Even with the offset pointer, do the numbers above make sense (as much as 70% extra time)? My alternative strategy without boost ipc: I will be running on 64bit linux and hoping to map the shared memory segment to the same location in all instances using shmat(). To achieve that, the first instance gets the shared memory segment address, which will be used by the other instances that read from the shared memory. If I can map them to the same address, can I use raw pointers without any problem? Naveen

2 1

Boost timer initialization hangs in win_therad.ipp
by chaitanya sonavale 12 Mar '12

12 Mar '12

Hi, I have a C++ DLL which has exported functions which is used in a C# application via DllImport. The C++ DLL is linked to various .lib files which use boost functionality (thread, asio etc) internally and are compiled with BOOST_ALL_DYN_LINK option. When I run the application it just hangs with following stack trace. I get the same results if I convert this to a C++\CLI DLL and use it directly from the C# application. Stack Trace: ------------------------------------------------ OneComShim.dll!boost::asio::detail::win_thread::start_thread(boost::asio::detail ::win_thread::func_base * arg=0x035574f0, unsigned int stack_size=0x00010000) Line 98 + 0xe bytes C++ OneComShim.dll!boost::asio::detail::win_thread::win_thread<boost::asio::detail:: win_iocp_io_service::timer_thread_function> (boost::asio::detail::win_iocp_io_service::timer_thread_function f={...}, unsigned int stack_size=0x00010000) Line 72 C++ OneComShim.dll!boost::asio::detail::win_iocp_io_service::do_add_timer_queue(boos t::asio::detail::timer_queue_base & queue={...}) Line 471 + 0x31 bytes C++ OneComShim.dll!boost::asio::detail::win_iocp_io_service::add_timer_queue<boost:: asio::time_traits<boost::posix_time::ptime> > (boost::asio::detail::timer_queue<boost::asio::time_traits<boost::posix_time::pt ime> > & queue={...}) Line 78 C++ OneComShim.dll!boost::asio::detail::deadline_timer_service<boost::asio::time_tra its<boost::posix_time::ptime> >::deadline_timer_service<boost::asio::time_traits<boost::posix_time::ptime> > (boost::asio::io_service & io_service={...}) Line 68 C++ OneComShim.dll!boost::asio::deadline_timer_service<boost::posix_time::ptime,boos t::asio::time_traits<boost::posix_time::ptime> >::deadline_timer_service<boost::posix_time::ptime,boost::asio::time_traits<boos t::posix_time::ptime> >(boost::asio::io_service & io_service={...}) Line 73 C++ OneComShim.dll!boost::asio::detail::service_registry::create<boost::asio::deadli ne_timer_service<boost::posix_time::ptime,boost::asio::time_traits<boost::posix_ time::ptime> > >(boost::asio::io_service & owner={...}) Line 81 + 0x2f bytes C++ OneComShim.dll!boost::asio::detail::service_registry::do_use_service(const boost::asio::io_service::service::key & key={...}, boost::asio::io_service::service * (boost::asio::io_service &)* factory=0x100c770d) Line 123 + 0xc bytes C++ OneComShim.dll!boost::asio::detail::service_registry::use_service<boost::asio::d eadline_timer_service<boost::posix_time::ptime,boost::asio::time_traits<boost::p osix_time::ptime> > >() Line 49 C++ OneComShim.dll!boost::asio::use_service<boost::asio::deadline_timer_service<boos t::posix_time::ptime,boost::asio::time_traits<boost::posix_time::ptime> > > (boost::asio::io_service & ios={...}) Line 34 C++ OneComShim.dll!boost::asio::basic_io_object<boost::asio::deadline_timer_service< boost::posix_time::ptime,boost::asio::time_traits<boost::posix_time::ptime> > >::basic_io_object<boost::asio::deadline_timer_service<boost::posix_time::ptime, boost::asio::time_traits<boost::posix_time::ptime> > >(boost::asio::io_service & io_service={...}) Line 91 + 0x4b bytes C++ OneComShim.dll!boost::asio::basic_deadline_timer<boost::posix_time::ptime,boost: :asio::time_traits<boost::posix_time::ptime>,boost::asio::deadline_timer_service <boost::posix_time::ptime,boost::asio::time_traits<boost::posix_time::ptime> > >::basic_deadline_timer<boost::posix_time::ptime,boost::asio::time_traits<boost: :posix_time::ptime>,boost::asio::deadline_timer_service<boost::posix_time::ptime ,boost::asio::time_traits<boost::posix_time::ptime> > >(boost::asio::io_service & io_service={...}) Line 147 C++ OneComShim.dll!CORE::SERVER_COMMUNICATOR::CurlHandler::CurlHandler(CORE::SERVER_ COMMUNICATOR::ServerAgent * serverAgent=0x1029c950) Line 37 + 0x118 bytes C++ OneComShim.dll!CORE::SERVER_COMMUNICATOR::ServerAgent::ServerAgent() Line 22 C++ OneComShim.dll!`dynamic initializer for 'CORE::SERVER_COMMUNICATOR::ServerAgent::agent''() Line 17 + 0x28 bytes C++ TestWin.exe!TestWin.Program.Main(string[] args = {string[0x00000000]}) Line 16 + ------------------------------------------------ As it can be seen, it stops in start_thread inside win_thread.ipp at following lines on WaitForSingleObject if (entry_event) { ::WaitForSingleObject(entry_event, INFINITE); ::CloseHandle(entry_event); } I am just creating a timer object in the code and do not understand why it leads to this situation. Also everything works fine when I use a LIB instead of DLL and all code is in C++. Environment details are as follows OS: Win XP IDE: Visual Studio 2010 Boost version: boost_1_48_0 Regards, Chaitanya

1 0

Keynotes for C++ Now!
by Jon Kalb 12 Mar '12

12 Mar '12

C++ Now! (formerly BoostCon) is pleased to announce its lineup of keynote speakers for the 2012 conference: Howard Hinnant, Sean Parent, and Daveed Vandevoorde. Howard Hinnant is the principal author of libc++, the standard C++ library implementation used by the Clang compiler. During the development of C++11, Howard served as chairman of the C++ Standards Committee¹s Library Working Group, and is largely responsible for the design of what Scott Meyers has called the language¹s ³flagship new feature:² rvalue references. His keynote will demonstrate how rvalue references are used in move semantics and perfect forwarding. Sean Parent is a principal scientist and software architect for Adobe¹s mobile digital imaging group and formerly managed Adobe¹s Software Technology Lab. He was also the keynote speaker for the very first BoostCon, where he gave one of the most inspiring and memorable talks we¹ve ever had. We¹re very pleased to welcome him back for an encore. He will be speaking on what¹s next for C++ programmers: how do we write beautiful code that truly exploits today¹s hardware and infrastructure? Daveed Vandevoorde is a compiler developer with Edison Design Group, the author (with Nicolai Josuttis) of ³C++ Templates: The Complete Guide,² a co-founder of comp.lang.c++.moderated, and one of the few C++ Committee Members who is equally comfortable working on library design, language evolution, or the arcana of core language issues. Daveed is currently a leader in the effort to add true modules to the C++ language. He¹ll be talking about the design of the C++ modules feature and what it would mean for the future of the language. ³We are thrilled to have such a powerhouse lineup in the first year of the new language standard,² said Dave Abrahams the conference founder and co-chair. ³C++Now! 2012 is truly shaping up to be a feast for the C++ enthusiast.² C++ Now! 2012 will be held May 13-18 in Aspen Colorado. Registration is now open at http://cppnow.org/registration/. Keynote details at http://cppnow.org/overview/. Jon Co-Chair C++ Now!

1 0

Strange error while using BOOST_FOREACH
by John M. Dlugosz 12 Mar '12

12 Mar '12

I used BOOST_FOREACH with a collection whose iterator is derived from boost::iterator_adaptor. I got this: 7>E:\boost_1_47\boost/range/iterator.hpp(63): error C2039: 'type' : is not a member of 'boost::mpl::eval_if_c<C,F1,F2>' 7> with 7> [ 7> C=true, 7> F1=boost::range_const_iterator<VSeeClient_list>, 7> F2=boost::range_mutable_iterator<const type> 7> ] 7> E:\boost_1_47\boost/foreach.hpp(670) : see reference to class template instantiation 'boost::range_iterator<C>' being compiled 7> with 7> [ 7> C=const type 7> ] 7> AutoPerformance\APSystem.cpp(224) : see reference to function template instantiation 'boost::foreach_detail_::auto_any<T> boost::foreach_detail_::begin<VSeeClient_list,boost::mpl::false_>(boost::foreach_detail_::auto_any_t,boost::foreach_detail_::type2type<VSeeClient_list,C> *,boost::mpl::false_ *)' being compiled 7> with 7> [ 7> T=VSeeClient_list::iterator, 7> C=boost::mpl::false_ 7> ] The error points here: template< typename C > struct range_iterator { typedef BOOST_RANGE_DEDUCED_TYPENAME mpl::eval_if_c< is_const<C>::value, range_const_iterator< typename remove_const<C>::type >, range_mutable_iterator<C> >::type type; }; Now a little looking through the manual tells me that mpl::eval_if_c::type will be one of the two metafunctions, so it should not disappear completely. I don't believe I missed anything from the iterator code: the Value type is given as the 3rd template argument to the iterator_adaptor, so that should set up any metafunctions needed to find the iterator's type, right? I'm compiling using MS Visual Studio 10. —John

2 3

[Graph] Help with delta_stepping_shortest_paths()
by Leo Hidd 12 Mar '12

12 Mar '12

Hi all, I'm having a hard time trying to use the distributed SSSP algorithm delta_stepping_shortest_paths. I found this example http://code.google.com/p/boost-doc-zh/source/browse/trunk/libs/graph_parall… but it compiles with error in line 192: 'boost::property_map<Graph,Property>' : more than one partial specialization matches the template argument list (MS Visual Studio 2008, MPI from Microsoft HPC Pack 2008 R2, Boost 1.48) Do anyone have any suggestion, or any other example? thx for your time, leo

1 0

Trouble with template on boost::function and boost::bind
by pwx 11 Mar '12

11 Mar '12

I tried to create a template function that receives boost::function as parameter. Say like this: template <typename T1, typename T2> void proc(const boost::function<void(const T1&, T2&)> &p) { // ... } void f(const string& s1, string&s2) { // ... } Everything goes ok when I called proc( boost::function<void(const string&, string&)>(f) ). But when I try this: proc(boost::bind(f, _1, _2)); The compiler refused with the error: no matching function for call to ‘proc(boost::_bi::bind_t<boost::_bi::unspecified, F, boost::_bi::list2<boost::arg<1>, boost::arg<2> > >)’ Anyway to fulfill this design purpose? Thanks.

3 2