August 2011 - Boost-users - lists.boost.org

boost::thread or openMP for low-latency multi-threading
by Jack Bryan 19 Aug '11

19 Aug '11

Hi, I am trying to do parallelization for a computing intensive problem. I am working on a Linux cluster where each node is a multicore processor. e.g. 2 or 4 quad-core processor per node. I want to reduce latency and improve performance as much as possible. I plan to use multiprocessing and multithreading at the same. Each process run on a distinct node and each process spawn many threads on each node. This is a 2 level parallelism. For multiprocessing, I would like to choose MPI. For multithreading, I have two choices: openMP or boost::thread (pthread). Which one has lower latency and higher performance ? It seems that openMP coding is easier (without touching low level thread managing manually). But, it seems that openMP has higher overhead than boost::thread (pthread). Any help is really appreciated. thanks

1 0

Scoped Pairs of functions?
by Brian Allison 19 Aug '11

19 Aug '11

Hi all, I'm on a project which uses a C++ compiler, but with a lot of C-isms. One of those is that there are objects with pairs of functions which allocate/deallocate resources. I don't have time to fix all of the design issues with the classes that do this (yet), but within one compilation unit I'd like to remove the chance for leaks. So - if I wanted to use boost types to wrap around 2 functions f() and g(), where each function - has a void or primitive return type - takes zero or 1 argument - where f() is called at some point in a scope, and then g() is called *on exiting the scope *from which f() is called and only if f() doesn't throw.. - has its return value ignored. [Gahh....] Is there a boost::bind (or lambda) expression that I can wrap in a type? I'm not adverse to using several template parameters. If it were just nullary functions I could just use: struct Reference {}; struct TypeWrapper{}; // Signify something like shared_ptr, a T which wraps around a pointer to an object of the type containing the functions, where T has a typedef value_type to refer to the wrapped type, and a get() to obtain a pointer to the wrapped pointer. template<class T, type R1, type R2, class P = Reference> struct ScopedDuals { typedef R1 (T::*f_t)(); typedef R2 (T::*g_t)(); ScopedDuals(T&t, f_t f, g_t g) : t_(t), g_(g) { (t_.f)(); } ~ScopedDuals() { (t_.*g_)(); } private: T& t_; g_t g_; }; template<class T, type R1. type R2> struct ScopedDuals<T, R1, R2, TypeWrapper> { typedef typname T::value_type Inner_t; typedef R1 (Inner_t::*f_t)(); typedef R2 (Inner_t::*g_t)(); ScopedDuals(T&t, f_t f, g_t g) : t_(t), g_(g) { (t_.get()->*f)(); } ~ScopedDuals() { (t_.get()->*g_)(); } private: T& t_; g_t g_; }; How do I generalize to include the possibility of a single argument? and.... does anyone have a better suggestion for a name than "ScopedDuals"? While *I *think it's precise, it may not be useful to my teammates. thanks, Brian

2 3

dynamically linked libraries looking for static libraries which don't exist
by Pally Sandher 19 Aug '11

19 Aug '11

I've built debug binaries of boost 1.47.0 on Windows with MSVC 10 using the following options b2.exe variant=debug link=shared threading=multi runtime-link=shared which as far as I understand should build dynamically linked libraries (which it does, I get a bunch of .lib files & .dll files as expected). However when I try to link some projects in Visual Studio 2010 which import boost_filesystem-vc100-mt-gd-1_47.lib & boost_system-vc100-mt-gd-1_47.lib I get the following errors 1>LINK : fatal error LNK1104: cannot open file 'libboost_filesystem-vc100-mt-gd-1_47.lib' & 1>LINK : fatal error LNK1104: cannot open file 'libboost_system-vc100-mt-gd-1_47.lib' I can stop these errors by using the /NODEFAULTLIB option on both those libraries but I would like to stop them in the first instance. Does anyone have any ideas why the libraries which should be dynamically linked are looking for statically linked libraries which don't exist? Palbinder Sandher Software Deployment Engineer T: +44 (0) 141 945 8500 F: +44 (0) 141 945 8501 http://www.iesve.com **Design, Simulate + Innovate with the <Virtual Environment>** Integrated Environmental Solutions Limited. Registered in Scotland No. SC151456 Registered Office - Helix Building, West Of Scotland Science Park, Glasgow G20 0SP Email Disclaimer

3 3

Starting...
by Giordano Bruno 19 Aug '11

19 Aug '11

Hi people, I want to use boost_pp in my embedded systems development using avr-gcc (C language). I was trying to understand how boostpp works, but it looks fuzzy to me. Can you explain me what is this "0" in the line 40 from preprocessor/comparison/less.hpp: define BOOST_PP_LESS_D(d, x, y) BOOST_PP_IIF(BOOST_PP_NOT_EQUAL(x, y), BOOST_PP_LESS_EQUAL_D, 0 BOOST_PP_TUPLE_EAT_3)(d, x, y) http://www.boost.org/doc/libs/1_43_0/boost/preprocessor/comparison/less.hpp thanks Giordano >

2 1

Re: [Boost-users] : dynamically linked libraries looking for static
by Alex Perry 18 Aug '11

18 Aug '11

Pally Sandher <pally.sandher(a)iesve.com> wrote on Thu, 18 Aug 2011 15:49:07 +0000 > > I've built debug binaries of boost 1.47.0 on Windows with MSVC 10 using the > following options > snip > I can stop these errors by using the /NODEFAULTLIB option on both those > libraries but I would like to stop them in the first instance. Does anyone have > any ideas why the libraries which should be dynamically linked are looking for > statically linked libraries which don't exist? Are you using the appropriate #define in your consuming code Eg #ifndef BOOST_REGEX_DYN_LINK #define BOOST_REGEX_DYN_LINK #endif #include <boost/regex.hpp> Which would ask to link to the dynamic link version of regex library not the static one - then all you need to do is make sure that the path to the boost libraries directory is added rather than having to add specific libraries Similar defines exist for the other libraries which are not header only. HTH Alex

1 0

Why is asio clobbering my file before it can be sent?
by Andrew Yancy 18 Aug '11

18 Aug '11

I've used the same boilerplate boost::asio code for years to send a plain text file "foo.txt" from a server back to a client and it has always worked fine. Suddenly for no apparent reason, the same code I've always used with no problems now deletes the contents of foo.txt before it can actually be sent to the client. I'm creating foo.txt with a different program now, but I don't see how that could make any difference on Win32. Below is the relevant part of the code just before I use boost::asio::write to send the contents of foo.txt back to the client. As shown by the two printf statements, the contents of the file are somehow lost when the file name and size are sent to the client. Anyone know what's going on? Is there any other information I can provide to shed light on the problem? Thanks very much for any help~ std::ifstream source_file("foo.txt", std::ios_base::binary | std::ios_base::ate); if (!source_file) return __LINE__; size_t file_size = source_file.tellg(); source_file.seekg(0); printf("foo.txt size: %d\n", (int) file_size); //size is OK here boost::asio::streambuf request; std::ostream request_stream(&request); request_stream << return_fn << "\n" << file_size << "\n\n"; boost::asio::write(socket, request); file_size = source_file.tellg(); printf("foo.txt size: %d\n", (int) file_size); //size is now 0!

3 4

[IOStreams] Is such a behaviour considered as bug
by Ovanes Markarian 18 Aug '11

18 Aug '11

Hello *, I really enjoyed working with Iostreams, while testing some of my components. Using IOStreams with array_device seemed to be a very powerfull approach to test some of my stream based algorithms. But currently I face some issues which seem to be really weired to me. I produced some code snippet to better explain it. I used Boost.Test library to verify my assumptions. using namespace boost; namespace io=boost::iostreams; typedef unsinged char byte; typedef array<byte, 2> arr_buffer; //boost::array typedef io::basic_array<byte> arr_dev; arr_buffer buffer_ = {{}}; arr_dev dev_(buffer_.begin(), buffer_.end()); io::stream<arr_dev> ios_(dev_); arr_dev::pair_type range = dev_.input_sequence(); BOOST_MESSAGE("size: " << distance(range.first, range.second)); // output: size is 2 BOOST_CHECK_EQUAL(0u, ios_.tellg()); //stream is at position 0 as expected ios_.seekg(1); BOOST_CHECK_EQUAL(1u, ios_.tellg()); // stream is at position 1 as expected ios_.seekg(2); BOOST_MESSAGE("current pos: " << ios_.tellg()); // !!! stream is at position 2 ios_.seekg(3); BOOST_MESSAGE("current pos: " << ios_.tellg()); // stream is at position -1 BOOST_MESSAGE("eof: " << ios_.eof()); // !!! but eof is false If the defined stream is only 2 bytes long why is 2 a valid stream position? Why is eof always false? Do you think this behavior is buggy? I changed the same behavior to read bytes from stream and created an array of length 3, but passed the range of 2 bytes to the device ctor. This works as expected: arr_buffer buffer_ = {{'a', 'b', 'c'}}; arr_dev dev_(buffer_.begin(), buffer_.begin()+2); //range of 2 bytes is passed io::stream<arr_dev> ios_(dev_); arr_dev::pair_type range = dev_.input_sequence(); BOOST_MESSAGE("size: " << distance(range.first, range.second)); byte bt; BOOST_MESSAGE("current pos: " << ios_.tellg()); // prints 0 BOOST_CHECK_EQUAL(0u, ios_.tellg()); ios_.read(&bt, 1); BOOST_MESSAGE("retrieved: " << bt); // prints 'a' BOOST_MESSAGE("current pos: " << ios_.tellg()); //prints 1 BOOST_CHECK_EQUAL(1u, ios_.tellg()); ios_.read(&bt, 1); BOOST_MESSAGE("retrieved: " << bt); // prints 'b' BOOST_MESSAGE("current pos: " << ios_.tellg()); // !!! prints 2, why not -1??? ios_.read(&bt, 1); BOOST_MESSAGE("retrieved: " << bt); // still prints 'b' => nothing retrieved BOOST_MESSAGE("current pos: " << ios_.tellg()); // !!! now it prints -1 BOOST_MESSAGE("eof: " << ios_.eof()); // !!! now eof is true Any ideas why it behaves as described? With Kind Regards, Ovanes

2 2

[Date Time] format_date_parser.parse_date fails to compile with "treat wchar_t as builtin type" set to No
by Daniel Mitchell 17 Aug '11

17 Aug '11

Boost 1.47, using Visual Studio 2010 and Visual Studio 2008, with boost 1.47 both built myself from source and with boost from the boostpro installers. (and with boost 1.47 rebuilt myself with /Zwchar_t- in msvc.jam). If I set "treat wchar_t as builtin type" to "No", then the following code fails to compile: #include <boost/date_time/gregorian/gregorian.hpp> #include <boost/date_time/posix_time/posix_time.hpp> void parseDate() { std::locale locale; boost::date_time::format_date_parser<boost::gregorian::date, wchar_t> parser(L"", locale); boost::date_time::special_values_parser<boost::gregorian::date, wchar_t> svp; boost::gregorian::date date = parser.parse_date(L"", L"", svp); } It gives the following error: c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\lexical_cast.hpp(1116): error C2535: 'bool boost::detail::lexical_stream_limited_src<CharT,Base,Traits>::operator >>(unsigned short &)' : member function already defined or declared with [ CharT=char_type, Base=base, Traits=traits ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\lexical_cast.hpp(993) : see declaration of 'boost::detail::lexical_stream_limited_src<CharT,Base,Traits>::operator >>' with [ CharT=char_type, Base=base, Traits=traits ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\lexical_cast.hpp(1524) : see reference to class template instantiation 'boost::detail::lexical_stream_limited_src<CharT,Base,Traits>' being compiled with [ CharT=char_type, Base=base, Traits=traits ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\lexical_cast.hpp(1653) : see reference to function template instantiation 'Target boost::detail::lexical_cast<Target,src,false,char_type>(const std::basic_string<_Elem,_Traits,_Ax> &,CharT *,size_t)' being compiled with [ Target=short, _Elem=wchar_t, _Traits=std::char_traits<wchar_t>, _Ax=std::allocator<wchar_t>, CharT=char_type ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\lexical_cast.hpp(1641) : while compiling class template member function 'short boost::detail::lexical_cast_do_cast<Target,Source>::lexical_cast_impl(const Source &)' with [ Target=short, Source=src ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\lexical_cast.hpp(1813) : see reference to class template instantiation 'boost::detail::lexical_cast_do_cast<Target,Source>' being compiled with [ Target=short, Source=src ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\date_time\format_date_parser.hpp(72) : see reference to function template instantiation 'Target boost::lexical_cast<int_type,std::basic_string<_Elem,_Traits,_Ax>>(const Source &)' being compiled with [ Target=short, int_type=short, _Elem=wchar_t, _Traits=std::char_traits<wchar_t>, _Ax=std::allocator<wchar_t>, Source=std::basic_string<wchar_t,std::char_traits<wchar_t>,std::allocator<wchar_t>> ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\date_time\format_date_parser.hpp(365) : see reference to function template instantiation 'int_type boost::date_time::fixed_string_to_int<short,charT>(std::istreambuf_iterator<_Elem,_Traits> &,std::istreambuf_iterator<_Elem,_Traits> &,boost::date_time::parse_match_result<charT> &,unsigned int,const charT &)' being compiled with [ int_type=short, charT=wchar_t, _Elem=wchar_t, _Traits=std::char_traits<wchar_t> ] c:\users\danielm\downloads\boost_1_47_0\boost_1_47_0\boost\date_time\format_date_parser.hpp(254) : while compiling class template member function 'boost::gregorian::date boost::date_time::format_date_parser<date_type,charT>::parse_date(std::istreambuf_iterator<_Elem,_Traits> &,std::istreambuf_iterator<_Elem,_Traits> &,std::basic_string<_Elem,_Traits,_Ax>,const boost::date_time::special_values_parser<date_type,charT> &) const' with [ date_type=boost::gregorian::date, charT=wchar_t, _Elem=wchar_t, _Traits=std::char_traits<wchar_t>, _Ax=std::allocator<wchar_t> ] c:\source\dan\boost_wchar_t_test\boost_wchar_t_test\boost_wchar_t_test.cpp(18) : see reference to class template instantiation 'boost::date_time::format_date_parser<date_type,charT>' being compiled with [ date_type=boost::gregorian::date, charT=wchar_t ] If I set "wchar_t is builtin type" to Yes, then it compiles fine. Testing with #pragma message to verify things, BOOST_NO_INTRINSIC_WCHAR_T is defined (or not defined) appropriately depending on how I set that compiler option, so it looks as if Boost knows what's going on at that level. I've tried all the possible reconfigurations I can think of, but this error happens with the boostpro installs, with Boost built unchanged from the 1.47 distribution, and with Boost rebuilt with /Zwchar_t- set in msvc.jam. Is there a way to solve this? The issue originally appears when trying to build code from another library -- but this is a minimal repro it with just boost code, so I think it's a Boost problem, unless I'm missing something. thanks, -- dan

1 0

Boost.Bimap vector_of element access VS 9 compilation errors
by John Eddy 17 Aug '11

17 Aug '11

Hello, I’m having difficulty figuring out how to access elements from a vector_of view in a bimap. The following code does not compile in VS 2008. #include <string> #include <iostream> #include <boost/bimap/bimap.hpp> #include <boost/bimap/vector_of.hpp> using namespace std; using namespace boost::bimaps; int main() { typedef bimap< string, vector_of<int> > bimap_type; bimap_type bm; bm.insert(bimap_type::value_type("zero", 0)); bm.insert(bimap_type::value_type("one", 1)); const bimap_type::right_map& right = bm.right; // Both of the lines below cause compilation failure with vc++ 9.0. bm.right[0]; //bm.right.at(0); return 0; } Can anyone see what I’m doing wrong or is there some other means of accessing the elements of the map via the vector? Thanks, John

1 0

[thread] 1.47 on cygwin 1.7 -> undefined reference to boost::detail::set_tss_data::shared_ptr::tss_cleanup
by Frédéric Bron 17 Aug '11

17 Aug '11

I am trying to use boost 1.47 on cygwin 1.7 cross-compiling to win32 with i686-w64-mingw32-g++ (4.5.3). This program gives me an error: #include <boost/thread/tss.hpp> int main() { boost::thread_specific_ptr<int> ts; return 0; } $ i686-w64-mingw32-g++ -o mgw-release/obj/Tests/toto.o -c -O3 -DNDEBUG -Wall -Wextra -pipe -mthreads -I/cygdrive/d/Softs/1.7/mgw-release/boost/include -I. Tests/toto.cc In file included from /cygdrive/d/Softs/1.7/mgw-release/boost/include/boost/thread/detail/thread_heap_alloc.hpp:15:0, from /cygdrive/d/Softs/1.7/mgw-release/boost/include/boost/thread/tss.hpp:10, from Tests/toto.cc:1: /cygdrive/d/Softs/1.7/mgw-release/boost/include/boost/thread/win32/thread_heap_alloc.hpp:59:40: warning: inline function ‘void* boost::detail::allocate_raw_heap_memory(unsigned int)’ declared as dllimport: attribute ignored /cygdrive/d/Softs/1.7/mgw-release/boost/include/boost/thread/win32/thread_heap_alloc.hpp:69:39: warning: inline function ‘void boost::detail::free_raw_heap_memory(void*)’ declared as dllimport: attribute ignored $ i686-w64-mingw32-g++ -o mgw-release/bin/toto -O3 -DNDEBUG -Wall -Wextra -pipe -mthreads -I/cygdrive/d/Softs/1.7/mgw-release/boost/include mgw-release/obj/Tests/toto.o -Wl,--enable-auto-import -static -L/cygdrive/d/Softs/1.7/mgw-release/boost/lib -lboost_thread_win32 mgw-release/obj/Tests/toto.o:toto.cc:(.text+0x104): undefined reference to `__imp___ZN5boost6detail12set_tss_dataEPKvNS_10shared_ptrINS0_20tss_cleanup_functionEEEPvb' collect2: ld returned 1 exit status any help would be welcome. Regards, Frédéric

3 5