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

synchronize threads from a thread pool (using asio::io_service)
by Daren Lee 05 Aug '11

05 Aug '11

Hi - I'm using asio::io_service as a dispatcher for a thread pool. I'm wondering how I can synchronize the threads after I post the tasks to the io_service. I tried using a barrier but it doesn't always work correctly. The barrier waits indefinitely; although I verified the post() function gets called, the io_service doesn't seem to dispatch anything. I've also thread using the boost::threadpool library but the overall speed is slower. I'd appreciate any insight on what I'm doing wrong with the barrier or any other high performance thread pooling alternatives. Thanks, Daren <code> void handler ( int i, int nSize, double *pArray, double *pTemp, boost::barrier *pBarrier ) { // compute on pArrray --> write result to pTemp[i] .... pBarrier->wait(); } double doCalc ( int nSize, double *pArray, double *pTemp ) { unsigned int nNumThreads = 8; int nSubSize = nSize / nNumThreads; boost::barrier barrier( nNumThreads + 1 ); for( int i=0;i<nNumThreads;i++ ) { if( i == nNumThreads-1 ) { m_io_service.post( boost::bind( &handler, i, nSize-(i*nSubSize), pArray+(i*nSubSize), pTemp, &barrier )); } else { m_io_service.post( boost::bind( &handler, i, nSubSize, pArray+(i*nSubSize), pTemp, &barrier )); } } barrier.wait(); // sum results double fSum = 0; for( int i=0;i<nNumThreads;i++ ) { fSum += pTemp[i]; } return fSum; } I initialize the io_service with work elsewhere as: boost::asio::io_service::work m_work( m_io_service ); boost::thread_group m_threads; for (std::size_t i = 0; i < m_nNumThreads; ++i) m_threads.create_thread(boost::bind(&boost::asio::io_service::run, &m_io_service)); </code>

1 0

What's wrong with this simple multithreaded function that uses regex?
by Andrew Yancy 05 Aug '11

05 Aug '11

The code snippet below is the core of a function mythreadfunc() that is run concurrently for several threads using the Win32 MSVC functions CreateThread() and WaitForMultipleObjects(). Everything seems to work fine except for actually using boost::regex_match(). For example, if that line below is changed to if(true), the entire multithreaded function works perfectly. Similarly, if mythreadfunc() as shown below is run without multiple threads, it works perfectly. But if mythreadfunc() as shown below is run with multiple threads, the entire program crashes with return value 0xc0000005, which is some kind of memory access violation. This leads me to believe that boost::regex does not play well with MSVC threads, but according to the Boost site, there aren't any bugs related to this issue for Boost 1.45. As far as I can tell I'm using Boost in thread safe mode (BOOST_HAS_THREADS is defined). I'm probably doing something really dumb related to MSVC threads or related to the way I configured Boost, but I'm not sure what it is! I could switch to using Boost threads, but I would like to minimize the changes to the code. Thanks very much for any help~ ... static const boost::regex expression("^[0-9]+"); ifstream myfile(x); //x is a different file name for every thread string line; if(myfile.is_open()) { while(myfile.good()) { getline(myfile, line); if(boost::regex_match(line, expression)) { //do stuff } } myfile.close(); } ...

5 7

[Filesystem] There must be a better way to do path decomposition
by Ramon F Herrera 04 Aug '11

04 Aug '11

This code at least works: string directory_path(argv[1]); path p(directory_path); ostringstream component; string lastComponent; for (path::iterator it = p.begin(); it != p.end(); it++) { component.str(""); // Hack, to be replaced by better way component << *it; // to grab `*it' in a string lastComponent = component.str(); } lastComponent = lastComponent.substr(1, lastComponent.size()-2); According to the documentation examples, the item pointed to by `*it' is of the following type: path::iterator::value_type is path::string_type, and iteration treats path as a container of filenames. I tried to capture `*it' in a string and in a string_type but the compiler does not like the '=' operator. TIA, -RFH

4 5

[iostreams] Unresolved external symbols gzip_footer::process etc.
by Clifford Cary 04 Aug '11

04 Aug '11

We recently upgraded from Boost 1.34.1 to 1.46.1 and moved from Visual Studio 2005 to VS 2010. We have an application that uses the Gzip decompression feature. The implementation includes these lines: Declaration: #include <boost/iostreams/device/file_descriptor.hpp> #include <boost/iostreams/filtering_stream.hpp> #include <boost/iostreams/device/file.hpp> #include <boost/iostreams/filter/gzip.hpp> boost::iostreams::filtering_streambuf<boost::iostreams::input> m_filter_stream; Executed code: m_filter_stream.push(io::gzip_decompressor()); m_filter_stream.push(m_file); m_input_stream = new istream(&m_filter_stream); m_input_stream->read((char*)m_buffer, BUFSIZE); The application linked without errors against the 1.34.1 library using VS 2005, but with 1.46.1 and VS 2010 we get four unresolved external symbols: public: void __cdecl boost::iostreams::detail::gzip_footer::process(char) public: void __cdecl boost::iostreams::detail::gzip_footer::reset(void) public: void __cdecl boost::iostreams::detail::gzip_header::process(char) public: void __cdecl boost::iostreams::detail::gzip_header::reset(void) The build is done on Windows XP x64 using the Boost 64-bit Windows binaries downloaded from http://boost.teeks99.com<http://boost.teeks99.com/>. The zlib binaries are also linked. Can someone suggest what the problem might be?

1 0

date _facet segmentation fault
by Kraus Philipp 04 Aug '11

04 Aug '11

Hello, I use this little code for formating a date value: std::stringstream l_stream; boost::gregorian::date_facet* l_format = new boost::gregorian::date_facet(); l_format->format("%Y-%m-%d"); l_stream.imbue(std::locale(std::locale::classic(), l_format)); If I call after the imbue a delete (delte l_format) to remove the date_facet from the heap the program creates a segmentation fault value. Does the std::stringstream remove the date_facet object from the heap? If not, how can I do a correct clear of the object? I hope, that question is not off topic. I would like to format a date value in a stringstream Thanks alot Phil

3 2

[test][1.47] Test cases segfault after completion
by Thomas Taylor 04 Aug '11

04 Aug '11

Hi! When compiling my test cases with 1.47 (no matter if dynamically or statically linked) they fail. Not the test cases themselves, but a segfault after the test case is done. The very same code compiled with r73480 executes as expected. HTH, Thomas Please find relevant valgrind output attached: Running 1 test case... *** No errors detected ==11042== Invalid write of size 8 ==11042== at 0x481152: execution_monitor_jumping_signal_handler (in /discs/rest/home/anubis/CPP/mdy- svn/gepetto/branches/build/g2_147/source/hoost/3d/test-unit__vec3d) ==11042== by 0x851380F: ??? (in /lib/libc-2.14.so) ==11042== by 0x85B3F44: shutdown (in /lib/libc-2.14.so) ==11042== by 0xC6FE47F: ??? (in /usr/lib/openmpi/openmpi/mca_oob_tcp.so) ==11042== by 0x78491EE: mca_oob_base_close (in /usr/lib/openmpi/libmpi.so.1.0.1) ==11042== by 0x788D61A: mca_base_components_close (in /usr/lib/openmpi/libmpi.so.1.0.1) ==11042== by 0x7861C09: orte_rml_base_close (in /usr/lib/openmpi/libmpi.so.1.0.1) ==11042== by 0x7831B69: orte_ess_base_app_finalize (in /usr/lib/openmpi/libmpi.so.1.0.1) ==11042== by 0xC4F579D: ??? (in /usr/lib/openmpi/openmpi/mca_ess_singleton.so) ==11042== by 0x7805067: orte_finalize (in /usr/lib/openmpi/libmpi.so.1.0.1) ==11042== by 0x77B1174: ??? (in /usr/lib/openmpi/libmpi.so.1.0.1) ==11042== by 0x70B4881: boost::mpi::environment::~environment() (in /usr/lib/libboost_mpi.so.1.47.0) ==11042== Address 0xa98 is not stack'd, malloc'd or (recently) free'd

2 2

Boost-log Formater with Syslog
by Shah, Gaurav N 04 Aug '11

04 Aug '11

Hi, I have any application running. I want to use boost.log for "syslog". How do I achive this? The log format I need is: <hostname> <LOGLEVEL> <appname> <timestamp> <threadid>message..... How to customize log format as above for syslog? Thanks for help in advance. _ Gaurav This email is confidential and subject to important disclaimers and conditions including on offers for the purchase or sale of securities, accuracy and completeness of information, viruses, confidentiality, legal privilege, and legal entity disclaimers, available at http://www.jpmorgan.com/pages/disclosures/email.

1 0

Disabling Tokenizer in Derived Classes - after their constructors have been executed.
by asif saeed 04 Aug '11

04 Aug '11

Hi, I have boost tokenizer and its iterator stored in my base class. The tokenizer got initialized with a const std::string&. I do the tokenization in my base class and provide the iterator, pointing to the first token, to the derived classes - but only to be used in their constructors. After the derived class constructor has been executed, the tokenizer and its iterator should not remain available - what can I do to accomplish that? Best regards, Asif

2 2

[network] Bagged, Tagged, New Release 0.9.1
by Dean Michael Berris 04 Aug '11

04 Aug '11

Good day everyone! If you're not interested in cpp-netlib's progress, please stop reading now. :) I'm happy to announce that cpp-netlib (the library that's hoping to be Boost.Network someday) is now available for general availability. You can get the latest from https://github.com/cpp-netlib/cpp-netlib/downloads and find out what's new at http://cpp-netlib.github.com/0.9.1/whats_new.html . This is an updated release that's inching closer to the "1.0" release that's going to be up for review. I've also written a more detail-filled announcement that you can read over at http://cplusplus-soup.com/2011/08/04/cpp-netlib-0-9-1-released/ -- please feel free to comment and share as you see fit. I'm looking to work with Michael Caisse who has agreed to be the review manager for Boost.Network to hopefully have a review at a convenient time for both of us. Right now I'm gunning for Q4 of 2011, but that's still fairly a ways away so things may still change. Cheers everyone, I do hope this helps! -- Dean Michael Berris http://about.me/deanberris

1 0

boost::thread - usage
by Christopher Pisz 03 Aug '11

03 Aug '11

I am trying to make a thread object, so I can later have better control of threads and the resources they are using. So, I started wrapping up a boost::thread. //----------------------------------------------------------------------------- ---------------------------------------------------------------- // BaseThread.h #include <boost/thread/thread.hpp> //----------------------------------------------------------------------------- - /// \brief Thread object that performs a task that is to run once (non- looped) /// \detail Derive from this class and implement the Work method to define the /// work this thread is to perform class BaseThread { public: BaseThread(); /// \brief Deconstructor /// \detail It is required that a derived class make a call to thread_.join() /// in its deconstructor virtual ~BaseThread(); virtual void Start(); virtual int Work() = 0; protected: void Run(); boost::thread thread_; }; //----------------------------------------------------------------------------- ---------------------------------------------------------------- // BaseThread.cpp #include "BaseThread.h" //----------------------------------------------------------------------------- - BaseThread::BaseThread() { } //----------------------------------------------------------------------------- - BaseThread::~BaseThread() { // Wait for the thread to complete thread_.join(); // DEBUG - Does join wait? int x = 1; } //----------------------------------------------------------------------------- - void BaseThread::Start() { boost::thread newThread(boost::bind(&BaseThread::Run, this)); thread_.swap(newThread); } //----------------------------------------------------------------------------- - void BaseThread::Run() { this->Work(); } I am kind of winging it here just going off of the boost documentation and guessing at what I should be doing. This code has a problem in that in order for us to wait for the thread to complete, the derived class is going to have to call thread_.join(). That leaves a bad taste in my mouth as I'd rather the base class take care of all the details, and all the derived class has to worry about is providing a Work method. This class doesn't provide much that boost::thread doesn't already have, but I want to make different classes for additonal requirements. One of which I have in mind is a thread with looped work and a start, stop, method. Making an object out of it will also probably help with derving classes that take care of locking their data...or at least organize it a bit. So the question 1 is, is there a way I can avoid the requirement here that derived classes call thread_.join() and do something different in the base class? Question 2 - What happens to the execution of the thread, when an exception is thrown from the thread that owns boost::thread object or the boost::thread object is destroyed? I'll probably have more questions as I progress.

1 0