- Boost-users - lists.boost.org

Segmentation fault in lagged fibonacci random number generator
by Moritz Beber 21 Jul '10

21 Jul '10

Dear boost users, I have a very strange segmentation fault in lagged fibonacci and unfortunately, I was not able to reduce the problem to a simple example yet. I know that stinks of the problem coming from my code but please bear with me for a moment. I have two classes that both use the same other class for some random number generation. The two classes differ only in one unrelated function. If I run my program with one class everything works just fine, if I run it with the other after a few iterations the error occurs. I have no idea why, since they both use the exact same class from which the error occurs. I use the following code to get my generators: typedef boost::uniform_smallint<short unsigned> Small; typedef boost::variate_generator<rfn::RandGen&, Small> Small_Number; Small* _us_dist = new Small(1, 20); Small_Number* _len_gen = new Small_Number(rfn::Mutation::_parameters->rng(), *_us_dist); Where rfn::Mutation::_parameters->rng() gives me a reference from a lagged_fibonacci object held in that class. I then try to get a number in the following way: short unsigned length = (*_len_gen)(); I have attached a gdb run of the program crashing. If you have any ideas where I can look for this error please tell me. It's also no problem for me to send you the complete sources but it's a rather big project. Thank you for your help, Moritz

3 3

true typedef (true_typedef) implemented in boost?
by alfC 20 Jul '10

20 Jul '10

Hi, I stumbled upon the concept of true typedef, or strong typedef in opposition to usual typedef which are weak. http://www.drdobbs.com/184401633, there is a true_typedef implemented in stlsoft http://www.stlsoft.org/doc-1.9/classstlsoft_1_1true__typedef.html I was wondering if this is already implemented in some corner of Boost, or as a detail implementation. If it is really a useful concept (I don't know) then, I though, Boost should have it somewhere. And if it doesn't have it then it must mean that Boosters come out with something even better. Did they? Thank you, Alfredo

6 10

[pool] Threadsafety
by Ruediger Berlich 20 Jul '10

20 Jul '10

Hi there, just to clarify: The pool documentation says in the "Object Usage vs. Singleton Usage" section (see here: http://www.boost.org/doc/libs/1_43_0/libs/pool/doc/interfaces.html) "Pool objects with Singleton Usage may be shared; thus, Singleton Usage implies thread-safety as well." So, with the singleton pool interface I can safely do concurrent mallocs and frees on all supported OSs (using Boost.Thread for the concurrency part), albeit at the expense of having a YASWIOI (yet another singleton with its own interface) in my programs, with all pros and cons that singletons have ? Thanks and Kind Regards, Ruediger Berlich

1 0

First set of Boostcon 2010 videos are available
by Marshall Clow 20 Jul '10

20 Jul '10

Almost all of the Transactional Memory sessions from Boostcon are now available, with the exception of Tatiana Shpeisman's presentation on the "Draft Specification of Transactional Language Constructs for C++". That talk should (hopefully) be available tomorrow. Here are the links Maurice Herlihy : Transactional Memory Today - Keynote: http://blip.tv/file/3900047/ Justin Gottschlich:TBoost.STM Engine: http://blip.tv/file/3900167/ Tatiana Shpeisman: Draft Specification of Transactional Language Constructs for C++: <coming soon> Michael Wong: IBM's TM Research: http://blip.tv/file/3900242/ Mark Moir: Sun's / Oracle's TM Research: http://blip.tv/file/3905166/ Tatiana Shpeisman: Intel's TM Research: http://blip.tv/file/3905189/ Hartmut - I apologize for removing your (very nice) introduction for Maurice from the video. Unfortunately, Maurice was wearing the microphone, and none of what you said could be heard. -- Marshall

2 1

[asio] suitable chunked circular buffer implementation?
by Steve Lorimer 20 Jul '10

20 Jul '10

Hi folks I'm trying (in vain so far) to find a suitable circular_buffer that allows chunked reads/writes which I can use in boost asio. This is my use case: I have 2 threads, Thread 1 reads data from a socket (boost::asio::tp::tcp::socket) and writes it into a buffer (boost::asio::mutable_buffer) Thread 2 reads the buffer and parses out individual messages. Thread 2 shouldn't care about what Thread 1 is doing, specifically whether or not it has completed (or is in the process of) more writes than what Thread 2 has read. All Thread 2 should receive is a small packet of information (the interface being perhaps only iterator begin() and size_t size() ) relating to the chunk of data that was written into the buffer. Reading from the iterator should be safe (ie: it should wrap when it reaches the end of the buffer) Writing into the buffer shouldn't invalidate the iterators that were created previously (which were passed to the 2nd thread) I have an implementation from before I started using boost::asio, which is below. It's quite clearly not standard compliant, but it worked for me. Is there anything out there that works like I've described? Is there an alternative method which I should pursue rather than "chunked circular buffers"? Will I be able to create something like I've pasted below successfully to work with boost asio, or are there issues I'm unaware of? (The reason I ask this question is because I find it surprising that something like this doesn't exist already) Any and all help greatly appreciated. TIA Steve #ifndef _ring_buf_h_ #define _ring_buf_h_ #include <errno.h> #include <boost/thread/mutex.hpp> namespace ring { typedef unsigned char byte; static const uint32_t RING_SIZE = 2 * 1024 * 1024; // 2 meg class iterator { private: const byte* buf; int cur; size_t buf_size; public: iterator() : buf(), cur(), buf_size() {} iterator(const byte* _buf, int _cur, size_t _buf_size) : buf(_buf), cur(_cur), buf_size(_buf_size) {} const byte& operator*() const { return buf[cur % RING_SIZE]; } const byte& operator[](int n) const { return buf[(cur + n) % RING_SIZE]; } iterator& operator+=(int n) { cur += n; buf_size -= n; return *this; } iterator& operator++() { ++cur; --buf_size; return *this; } // post-inc iterator operator++(int) { iterator tmp(*this); ++cur; --buf_size; return tmp; } // pre-inc bool operator==(iterator b) const { return cur == b.cur; } bool operator!=(iterator b) const { return cur != b.cur; } size_t size() const { return buf_size; } int offset() const { return cur; } }; //---------------------------------------------------------------- inline iterator operator+(iterator a, int b) { return a += b; } inline iterator operator+(int b, iterator a) { return a += b; } class buffer { mutable boost::mutex mtx; byte buf[RING_SIZE]; unsigned beg; unsigned end; // may wrap, wrapping is blessed for unsigned types only public: buffer() : beg(), end() { memset(buf, 0, RING_SIZE); } // create an iterator object whose size = unread data in buf iterator data() const { boost::mutex::scoped_lock l(mtx); return iterator(buf, beg % RING_SIZE, end - beg); } // free space in ring unsigned size() const { boost::mutex::scoped_lock l(mtx); return end - beg; } // frees up space in the ring (called once data has been read) void adv(unsigned n) { boost::mutex::scoped_lock l(mtx); beg += n; } // read data from a file descriptor into the buffer // returns: // -errno for error // 0 for EOF // > 0 for successful reads int read(int fd) { mtx.lock(); unsigned free = RING_SIZE - (end - beg); if (!free) { mtx.unlock(); return -ENOMEM; // must never get here } unsigned free_end = std::min(free, RING_SIZE - end % RING_SIZE); unsigned free_beg = free - free_end; iovec io[] = { { (char*)buf + end % RING_SIZE, free_end }, { (char*)buf, free_beg } }; mtx.unlock(); ssize_t n = readv(fd, io, sizeof io / sizeof *io); if (n == -1) return -errno; mtx.lock(); end += n; mtx.unlock(); return n; } }; //---------------------------------------------------------------- } // namespace ring #endif

1 0

[accumulators] density histogram, what is typical way of using it?
by alfC 20 Jul '10

20 Jul '10

Hi, the documentation on accumulator::density is not very clear to me. It seems to fulfil the need for a histogram. Can anyone point at one concrete example using accumulator::density. For example, it seems to determine de bins from the first N samples accumulated. what if you pass less elements than this. Are the bins completely fixed after that? What was the logic behind that design rather than, for example, to somehow declared uniform bins in a range? What if I have to define uniform equispaced bins, do I pass a uniform sequence at the beginning? but then those elements will be counted? How the counters or probabilities are accessed at the end? What else can be done with the density accumulator object, can the cummulative or the inverse pdf be accessed easily within the library. Can it be plugged in into a random number generator as a distribution function. Thank you, Alfredo

2 1

BGL - accessing component subgraphs.
by Adam Spargo 20 Jul '10

20 Jul '10

Any tips on the best way to access a subgraph - ideally without the memory overhead of copying the subgraph to a new object. I want to pass each component of my graph to an algorithm separately, but in some cases the original graph will only have a single or a few component(s), in other cases there will be many. Maybe I will do it differently depending on how many components I have. Any tips welcome. Thanks, Adam. -- Dr Adam Spargo High Performance Assembly Group email: aws(a)sanger.ac.uk Wellcome Trust Sanger Institute Tel: +44 (0)1223 834244 x7728 Hinxton, Cambridge CB10 1SA Fax: +44 (0)1223 494919 -- The Wellcome Trust Sanger Institute is operated by Genome Research Limited, a charity registered in England with number 1021457 and a company registered in England with number 2742969, whose registered office is 215 Euston Road, London, NW1 2BE.

2 7

students_t distribution with mt19937
by Kraus Philipp 20 Jul '10

20 Jul '10

Hello, I use some distributions with this code: boost::normal_distribution<> l_range(p_mean, p_sd); static boost::mt19937 l_random(static_cast<std::size_t>(time(0))); boost::variate_generator<boost::mt19937&, boost::normal_distribution<> > l_noise( l_random, l_range ); I get my value with l_noise() and it works without any problem. I would like to create now a student_t distribution with boost::math::students_t_distribution<T> l_range(p_v); static boost::mt19937 l_random(static_cast<std::size_t>(time(0))); boost::variate_generator<boost::mt19937&, boost::math::students_t_distribution<> > l_noise( l_random, l_range ); If I run my l_noise() I got a compiler error: error: no match for call to '(boost::variate_generator<boost::random::mersenne_twister<uint32_t, 32, 624, 397, 31, 2567483615u, 11, 7, 2636928640u, 15, 4022730752u, 18, 3346425566u>&, boost::math::students_t_distribution<double, boost::math::policies::policy<boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy, boost::math::policies::default_policy> > >) ()' Can anyone help me please to understand my mistake while using the students_t distribution? I think I need the mt19937 for creating random values and the variate_generator to create a "distribution value" Thanks Phil

2 2

[Random] uniform_int::operator()(Engine &, result_type)
by Hossein Haeri 20 Jul '10

20 Jul '10

Dear all, I was looking at the uniform_int<>'s reference page: http://www.boost.org/doc/libs/1_43_0/doc/html/boost/uniform_int.html and realised that there is no information whatsoever about what the following overload of its operator () does: template<typename Engine> result_type operator()(Engine & eng, result_type n) ; Do I understand it correctly that this generates a number in [min, n] (probably subject to some good conditions)? TIA, --Hossein

2 2

serialization of boost::local_time::tz_database
by Michael Reichlin 20 Jul '10

20 Jul '10

Hi everybody I just started to use boost in some of my applications. I started to serialize my objects. Now I want to serialize my boost::local_time::tz_database object, but the compiler says, there is now serialize function for this object. So do I have to add this function in the header file, or is there any other way to solve this problem. Thanks Michael

1 0