January 2012 - Boost-users - lists.preview.boost.org

[random] saving state of generator
by Thomas Mang 31 Jan '12

31 Jan '12

Hi, Consider an application intended to run for weeks and where a deterministic sequence of random numbers is needed (deterministic in order to make results 100% reproduceable). For hardware-related reasons that application might get halted (killed). In order to smoothly restart the application I would like to regularly save the complete system state and can hence use that as restore point. That complete save works for everything except the random number generator - I am using boost::random::mt19937. The mt199937 class does not seem to offer such an option (full state save). Using serialization also does not sound great in my eyes as I am, of course, not familiar with the internal representation of mt19937. So what is the best option? Dissecting mt19937 (intrusive changes to class template mersenne_twister_engine) and making my own serialization routine? Any other / more straightforward ideas? thanks Thomas

2 1

dijkstra-example.cpp
by fergie 31 Jan '12

31 Jan '12

I’ve been reviewing the dijkstra-example.cpp that is provided as part of boost Graph. I think I understand most of it and have been able to demonstrate the application working to my boss. I have one question about the code. When output the .dot file near the end of the program the results are shaded either grey or black. What is the significance of the black and grey lines? I’ve looked through the documentation and am also confused about the source and the target since we have an iterator going through the entire graph. What defines a source and target vertex? Thanks, D -- View this message in context: http://boost.2283326.n4.nabble.com/dijkstra-example-cpp-tp4334386p4334386.h… Sent from the Boost - Users mailing list archive at Nabble.com.

3 5

MPL newbie : what is the difference between mpl::true_ and true_type
by Linh Ha 31 Jan '12

31 Jan '12

I'm a newbie for both boost and MPL. My question is regarding about the differences between mpl::true_ and true_type. While from my point of view they are both structure of constant boolean. Even in the implementation true_type is derived from mpl:true_. So why we need both data structure instead of only single one. When we should use mpl::true_ instead and true_type and vice versa Thanks

3 2

ASIO: HTTP Client Issues with \r vs \n
by Fitzsimmons, Bo CIV SPAWARSYSCEN-PACIFIC, 81320 31 Jan '12

31 Jan '12

Hello, I'm trying to use the HTTP client example code (sync_cleint.cpp) to retrieve a jpg. I'm using VS2008 on a Windows XP machine and trying to access a Linux server. The resulting data I get back does not start the new lines at the same place as data I retrieve by simply downloading. I believe this is an issue with how windows interprets a new line, \r\n, versus how unix/linux interprets a new line, \n. The example code (http://www.boost.org/doc/libs/1_48_0/doc/html/boost_asio/example/http/clien…) uses \r\n, \n, and \r, so I'm a little confused on how to rectify the problem. Any suggestions to correct this for my case (hard code) or automatically detect would be greatly appreciated. P.S. I'm using boost 1.48.0 Thank you, Bo

2 1

Raw sockets for Layer 2 communications
by Carlos Guimarães 31 Jan '12

31 Jan '12

Hi, I've been looking for any way to send L2 messages (with no transport protocol) using Boost. As long as I could find, Boost libraries support raw sockets, however it must have a transport protocol (like IP) associated. Is this right?There is any way to send a pure L2 packet? Best regards, Carlos Guimarães

1 0

boost 1.49 beta tar problems
by Avi Bahra 31 Jan '12

31 Jan '12

I am trying boost 1.49 beta1 on the different platforms. This builds fine on linux variants However on AIX and HPUX I get. "directory checksum error" when untarring ? I can untar boost 1.48 and all previous boost version on these platforms, so guessing there been a change in the tar used ? Best regards, Ta, Avi

1 0

Boost.Regex: regex_replace with named captures/groups
by norulez＠me.com 31 Jan '12

31 Jan '12

Hi @all, does boost's regex_replace not recognize named captures/groups or did I something wrong? std::string date = "2012/01/30"; std::string replace = "\5\g{separator}\4\g{separator}\1\2"; // should be => 30/01/2012 boost::regex rx("^(19|20)(\d\d)(?<separator>[/.])(0[1-9]|1[012])\g{separator}(0[1-9]|[12][0-9]|3[01])$"); std::string replaced = boost::regex_replace(date, rx, replace); The content in "replaced" is now "30\g{separator}01\g{separator}2012" Thanks in advance Best Regards NoRulez

4 3

[Thread] code segfaults/hangs when being called from matlab, not when run standalone
by Dougal Sutherland 31 Jan '12

31 Jan '12

Hi all, I have some code using boost::thread that works fine when it's called directly from a main() function, but (almost) always segfaults and/or hangs forever when called from a Matlab mex interface. I don't know whether that's due to the Matlab interface breaking something about boost::thread or if the different environment is revealing a flaw with my code. Here's a simplified version of the code I'm using. (This same question is also posted on stackoverflow at http://stackoverflow.com/q/9011780/344821-- the code is prettier, and you can get some stackoverflow reputation out of me if you answer there.) #include <cstdlib> #include <queue> #include <vector> #include <boost/thread.hpp> #include <boost/utility.hpp> #ifndef NO_MEX #include "mex.h" #endif class Worker : boost::noncopyable { boost::mutex &jobs_mutex; std::queue<size_t> &jobs; boost::mutex &results_mutex; std::vector<double> &results; public: Worker(boost::mutex &jobs_mutex, std::queue<size_t> &jobs, boost::mutex &results_mutex, std::vector<double> &results) : jobs_mutex(jobs_mutex), jobs(jobs), results_mutex(results_mutex), results(results) {} void operator()() { size_t i; float r; while (true) { // get a job { boost::mutex::scoped_lock lk(jobs_mutex); if (jobs.size() == 0) return; i = jobs.front(); jobs.pop(); } // do some "work" r = rand() / 315.612; // write the results { boost::mutex::scoped_lock lk(results_mutex); results[i] = r; } } } }; std::vector<double> doWork(size_t n) { std::vector<double> results; results.resize(n); boost::mutex jobs_mutex, results_mutex; std::queue<size_t> jobs; for (size_t i = 0; i < n; i++) jobs.push(i); Worker w1(jobs_mutex, jobs, results_mutex, results); boost::thread t1(boost::ref(w1)); Worker w2(jobs_mutex, jobs, results_mutex, results); boost::thread t2(boost::ref(w2)); t1.join(); t2.join(); return results; } #ifdef NO_MEX int main() { #else void mexFunction(int nlhs, mxArray **plhs, int nrhs, const mxArray **prhs) { #endif std::vector<double> results = doWork(10); for (size_t i = 0; i < results.size(); i++) printf("%g ", results[i]); printf("\n"); } If I compile and run this directly (with gcc or clang and -DNO_MEX), it always completes successfully. If I run this through Matlab's mex interface and link to a release-variant boost (I've tried 1.48.0 on OSX and both 1.40 and 1.33.1 on CentOS), I get a segfault. On 1.48.0, this is trying to access a pointer that's near 0 inside boost::thread::start_thread from inside of t1's constructor; on 1.33.1 and 1.40, it's inside pthread_mutex_lock being called from t1.join(). If I instead link to a debug-variant boost (1.48 on OSX or 1.33.1 on CentOS -- the machine with 1.40 installed doesn't have a debug variant), it hangs forever inside t1.join(), though the threads have completed successfully, with results containing the expected elements that were calculated as rinside Worker. (On OSX, these are different numbers than are printed by the standalone version, but they're still consistent between threads.) Am I doing something stupid that's causing this, or is this somehow due to something on the matlab end? I'm pretty new to C++ and boost and so it might well be something dumb that I did, but I don't know what that could be. I also don't really have any ideas on how to proceed from here in terms of debugging, especially since the segfaults only occur when there are no debugging symbols in the boost libraries themselves around to help figure out what's going on. My only real idea is to try this without boost::thread and using pthreads directly, but even if that worked it's nonideal since I'd also like to support Windows. Thanks, Dougal

2 2

[lexical_cast][variant] Compile failed when included boost.variant.
by Tianjiao Yin 31 Jan '12

31 Jan '12

Hello, all: I met a strange compile error, this code will reproduce the problem:: #include <boost/lexical_cast.hpp> #include <boost/variant.hpp> #include <iostream> namespace foo { struct bar {}; } std::ostream& operator << (std::ostream& out, foo::bar) { return out << "BAR"; } int main() { boost::lexical_cast<std::string>(foo::bar()); } However, if I just remove that unused include boost.variant, it works:: #include <boost/lexical_cast.hpp> //#include <boost/variant.hpp> #include <iostream> namespace foo { struct bar {}; } std::ostream& operator << (std::ostream& out, foo::bar) { return out << "BAR"; } int main() { boost::lexical_cast<std::string>(foo::bar()); } Another solution is put that "operator<<" in namespace foo:: #include <boost/lexical_cast.hpp> #include <boost/variant.hpp> #include <iostream> namespace foo { struct bar {}; std::ostream& operator << (std::ostream& out, foo::bar) { return out << "BAR"; } } // namespace foo int main() { boost::lexical_cast<std::string>(foo::bar()); } Anyway, what is boost.variant's business here? Won't that fail to compile, or am I missing something? BTW, I'm using boost 1.48.0 and gcc 4.6.2 in Linux 3.2.2. Regards, Tianjiao

2 1

I need a sample Asio asynchronous client to read&write into sockets and stdin&stdout.
by vincegata 30 Jan '12

30 Jan '12

Hello, I am writing a client application that reads and writes data into a remote server. Currently the communication is done using synchronous (blocking) sockets. I am running Ubuntu and using C++. I need to re-write this app so it will asynchronously read & write into sockets and stdin/stdout, I also need to send a heartbeat signal if there wasn't any communication with the server for 60 seconds. I am new to Asio, I went through tutorial and looked at the examples such as asynchronous timer and client chat but they are not all that I need. Could someone point me to a sample Asio application that does what I need. Thank you!

2 1

Conversion from numeric type to const char*
by Claude 30 Jan '12

30 Jan '12

What is the better way for convert a numeric type into a const char *? I use: (boost::lexical_cast<std::string>(myNumericVar)).c_str() It is a good idea? -- View this message in context: http://boost.2283326.n4.nabble.com/Conversion-from-numeric-type-to-const-ch… Sent from the Boost - Users mailing list archive at Nabble.com.

4 7

[Property_tree] using find to find a key
by Sean Farrow 30 Jan '12

30 Jan '12

Hi: I am declaring a property_tree, pulating this with an ini file. I then need to find out whether a section (key) is present, so I'm doing: boost::property_tree::ptree::assoc_iterator it =IniTree.find(key); if (it !=boost::property_tree::ptree::not_found) using visual c++ 2010 I get an error c2678, telling me that no operand is able to use these two values. Can someone please tell me what I'm doing wrong, as I can normally do this for std::vectors etc! Cheers Sean.

2 2

multi_array: How to get the extends?
by Helmut Zeisel 30 Jan '12

30 Jan '12

How can I get the extends of a multi_array? E.g. I have boost::multi_array<double, 3> a(boost::extents[n0][n1][n2]); Are there functions like a.extend(0) returning n0, a.extend(1) returning n1, and a.extend(2) returning n2 (or similar)? Helmut -- "Feel free" - 10 GB Mailbox, 100 FreeSMS/Monat ... Jetzt GMX TopMail testen: http://www.gmx.net/de/go/topmail

2 1

copy a range to ostream
by Neal Becker 29 Jan '12

29 Jan '12

With all the nice boost libraries, is there a modern idiom to copy a range to an ostream? I was thinking range/algorithm/copy could be useful. What I'd really like to do is: std::vector<some type> x; range::copy (x, std::cout); Of course there would also need to be a way to specify formatting, at least insertion of whitespace.

2 1

mpi::allocator::construct function in gcc 4.6.2
by Hicham Mouline 29 Jan '12

29 Jan '12

Hello, I get the following error: In file included from /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/vector: 64:0, from /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/ra ndom.h:34, from /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/random: 50, from /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/st l_algo.h:67, from /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/algorit hm:63, from /usr/include/boost/utility/swap.hpp:24, from /usr/include/boost/swap.hpp:10, from /usr/include/boost/array.hpp:44, from /usr/include/boost/serialization/array.hpp:26, from /usr/include/boost/archive/detail/oserializer.hpp:58, from /usr/include/boost/mpi/detail/mpi_datatype_oarchive.hpp:14, from /usr/include/boost/mpi/detail/mpi_datatype_cache.hpp:13, from /usr/include/boost/mpi/datatype.hpp:27, from /usr/include/boost/mpi/packed_iarchive.hpp:21, from /home/hich/trunk/serialization/contract.cpp:3: /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/st l_uninitialized.h: In function 'void std::__uninitialized_default_n_a(_ForwardIterator, _Size, _Allocator&) [with _ForwardIterator = char*, _Size = long unsigned int, _Allocator = boost::mpi::allocator<char>]': /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/ve ctor.tcc:477:8: instantiated from 'void std::vector<_Tp, _Alloc>::_M_default_append(std::vector<_Tp, _Alloc>::size_type) [with _Tp = char, _Alloc = boost::mpi::allocator<char>, std::vector<_Tp, _Alloc>::size_type = long unsigned int]' /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/st l_vector.h:592:4: instantiated from 'void std::vector<_Tp, _Alloc>::resize(std::vector<_Tp, _Alloc>::size_type) [with _Tp = char, _Alloc = boost::mpi::allocator<char>, std::vector<_Tp, _Alloc>::size_type = long unsigned int]' /usr/include/boost/mpi/detail/packed_iprimitive.hpp:57:23: instantiated from here /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/st l_uninitialized.h:576:6: error: no matching function for call to 'boost::mpi::allocator<char>::construct(char*)' /usr/lib/gcc/x86_64-redhat-linux/4.6.2/../../../../include/c++/4.6.2/bits/st l_uninitialized.h:576:6: note: candidate is: /usr/include/boost/mpi/allocator.hpp:168:8: note: void boost::mpi::allocator<T>::construct(boost::mpi::allocator<T>::pointer, const T&) [with T = char, boost::mpi::allocator<T>::pointer = char*] /usr/include/boost/mpi/allocator.hpp:168:8: note: candidate expects 2 arguments, 1 provided I am using c++0x. Should mpi::allocator::construct() be different? Is it required to provide a 1-arg construct()? Regards, MM

2 2

Re: [Boost-users] [boost] numeric_cast_traits/cstdint: request for fix in boost1.49
by John Maddock 29 Jan '12

29 Jan '12

> I've put a fix for this up on the trunk. I've run my unit tests on vc100 > and gcc-4.5.3 cygwin. Could some of you with the other platforms give it a > try? Confirmed that existing tests fail with old code (including with VC8). Tested with new code and VC8,9,10 and Intel-11.1, 12.1 all OK. On Linux gcc-4.4, 4.6 intel-11.1, 12.1 and clang all OK. Sun on Linux still fails, but appears to be a lexical_cast issue and not your problem - will report that separately. HTH, John.

1 0

boost for S60 - what's the current status?
by Igor R 29 Jan '12

29 Jan '12

Hello, I'd appreciate if anyone could sched some light on the subj. In particular, what boost libs are known to *not* compile on Symbian? Thanks!

1 0

boost with arm-none-eabi
by Artur Stępniak 28 Jan '12

28 Jan '12

Hi, I'm trying to compile a program with boost for my Cortex M3 based board. I don't run any OS in there. Is it possible to build program with boost on such board? My problem is that boost seems like to consume all the memory. The program builds without boost but after including shared_ptr.hpp I get error: section `.text' will not fit in region `rom'. I try to run configure script for boost: sh ./configure --host=arm but then I get arm-none-eabi/bin/ld: cannot find -lrt. Since this library supports POSIX, then I'm afraid it's impossible to compile programs with boost with no-OS systems. But maybe there's a switch that enables it? Thanks in advance, best regards, Artur Stepniak

1 0

numeric_cast_traits/cstdint: request for fix in boost 1.49
by Avi Bahra 28 Jan '12

28 Jan '12

Currently HPUX/acc compiler fails in boost 1.48, because boost::int8_t was being interpreted as a char, This causes duplicate class definition in the file ./boost/numeric/conversion/detail/preprocessed/numeric_cast_traits.hpp and hence many of compiled boost libs fail. See:: https://svn.boost.org/trac/boost/attachment/ticket/6158/cstdint_patch.diff The posted fix was to modify boost/cstdint.hpp: 103c103,107 2 < using ::int8_t; 3 --- 4 > #if defined(sun) || defined(__sun) 5 > typedef signed char int8_t; 6 > #else 7 > using ::int8_t; 8 > #endif Can I request a similar fix be added for the HPUX/acc compiler in the next boost version. Best regards, Ta, Avi

3 2

[1.49.0] Beta 1 available
by Beman Dawes 28 Jan '12

28 Jan '12

Boost release 1.49.0 beta 1 is now available from SourceForge See http://sourceforge.net/projects/boost/files/ This release contains a new library, Boost.Heap priority queues, and numerous enhancements and bug fixes for existing libraries. For details of what's in the release, see http://beta.boost.org/users/news/version_1_49_0. Note that the links to files on this web page are for the final release - use the SourceForge link above to get the beta files. Please download the beta, give it a try, and report any problems you encounter. Thanks, -- The Boost Release Team

1 0

status of Boost.LockFree and Boost.Atomic. Boost.Endian libraries?
by Igor Hersht 28 Jan '12

28 Jan '12

Could you tell me what is the status of Boost.LockFree and Boost.Atomic <https://svn.boost.org/trac/boost/wiki/ReviewScheduleLibraries#Boost.Atomic> . Boost.Endian libraries? Boost.LockFree and Boost.Atomic <https://svn.boost.org/trac/boost/wiki/ReviewScheduleLibraries#Boost.Atomic> accepted a while ago, but I didn't find the libraries in the latest 1.49 beta. Similar question about Boost.Endian ($(boost)\spirit\home\support\detail\endian\endian.hpp). When/if the Boost.Endian is going to be accepted as a separate library.

2 1

[interprocess] compilation errors with 1.48 on 64-bit Centos 5.5
by Daniel Herring 27 Jan '12

27 Jan '12

Hi, I started updating an application from Boost 1.35. It had a few compiler errors with 1.44, and things got worse with 1.48. While tracing down the new issues, problems were found with the distributed examples. For example, the following commands pass with 1.44 and 1.47 but fail with 1.48 (see attached logs). $ cd libs/interprocess/example $ ../../../bjam doc_managed_allocation_command > doc_managed_allocation_command.log $ ../../../bjam doc_managed_raw_allocation > doc_managed_raw_allocation.log All 1.44, 1.47, and 1.48 were built locally from the release source tarballs using a simple `./bootstrap.sh; ./bjam`. The example sources haven't changed during this period; so this may be a portability issue with the interprocess library itself. Here's some more info on the system. $ cat /etc/redhat-release CentOS release 5.5 (Final) $ g++ --version g++ (GCC) 4.1.2 20080704 (Red Hat 4.1.2-48) Copyright (C) 2006 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. $ rpm -qi gcc-c++ Name : gcc-c++ Relocations: (not relocatable) Version : 4.1.2 Vendor: CentOS Release : 48.el5 Build Date: Wed 31 Mar 2010 12:09:02 PM EDT Install Date: Mon 20 Sep 2010 06:02:55 AM EDT Build Host: builder17.centos.org Group : Development/Languages Source RPM: gcc-4.1.2-48.el5.src.rpm Size : 7824005 License: GPL Signature : DSA/SHA1, Mon 26 Apr 2010 07:41:02 PM EDT, Key ID a8a447dce8562897 URL : http://gcc.gnu.org Summary : C++ support for GCC Description : This package adds C++ support to the GNU Compiler Collection. It includes support for most of the current C++ specification, including templates and exception handling. The above commands do pass on release 1.48 when compiled using g++44 from the gcc44-c++-4.4.0-6.el5 package. However, that might not be an option for some of my users. Please let me know if more information is needed to debug the problem, if this is a known problem with gcc, etc. Thanks, Daniel

2 3

Re: [Boost-users] mpi::allocator::construct function in gcc 4.6.2
by MM 27 Jan '12

27 Jan '12

> -----Original Message----- > From: MM [mailto:finjulhich@gmail.com] > Sent: 27 January 2012 18:54 > To: 'boost-users(a)lists.boost.org' > Subject: mpi::allocator::construct function in gcc 4.6.2 > > Hello, > > I get the following error: > > In file included from /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/vector:64:0, > from /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/random.h:34, > from /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/random:50, > from /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/stl_algo.h:67, > from /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/algorithm:63, > from /usr/include/boost/utility/swap.hpp:24, > from /usr/include/boost/swap.hpp:10, > from /usr/include/boost/array.hpp:44, > from /usr/include/boost/serialization/array.hpp:26, > from > /usr/include/boost/archive/detail/oserializer.hpp:58, > from > /usr/include/boost/mpi/detail/mpi_datatype_oarchive.hpp:14, > from > /usr/include/boost/mpi/detail/mpi_datatype_cache.hpp:13, > from /usr/include/boost/mpi/datatype.hpp:27, > from /usr/include/boost/mpi/packed_iarchive.hpp:21, > from /home/hich/trunk/serialization/contract.cpp:3: > /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/stl_uninitialized.h: In > function 'void std::__uninitialized_default_n_a(_ForwardIterator, > _Size, _Allocator&) [with _ForwardIterator = char*, _Size = long > unsigned int, _Allocator = boost::mpi::allocator<char>]': > /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/vector.tcc:477:8: > instantiated from 'void std::vector<_Tp, > _Alloc>::_M_default_append(std::vector<_Tp, _Alloc>::size_type) [with > _Tp = char, _Alloc = boost::mpi::allocator<char>, std::vector<_Tp, > _Alloc>::size_type = long unsigned int]' > /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/stl_vector.h:592:4: > instantiated from 'void std::vector<_Tp, > _Alloc>::resize(std::vector<_Tp, _Alloc>::size_type) [with _Tp = char, > _Alloc = boost::mpi::allocator<char>, std::vector<_Tp, > _Alloc>::size_type = long unsigned int]' > /usr/include/boost/mpi/detail/packed_iprimitive.hpp:57:23: > instantiated from here > /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/stl_uninitialized.h:576: > 6: error: no matching function for call to > 'boost::mpi::allocator<char>::construct(char*)' > /usr/lib/gcc/x86_64-redhat- > linux/4.6.2/../../../../include/c++/4.6.2/bits/stl_uninitialized.h:576: > 6: note: candidate is: > /usr/include/boost/mpi/allocator.hpp:168:8: note: void > boost::mpi::allocator<T>::construct(boost::mpi::allocator<T>::pointer, > const T&) [with T = char, boost::mpi::allocator<T>::pointer = char*] > /usr/include/boost/mpi/allocator.hpp:168:8: note: candidate expects 2 > arguments, 1 provided > > I am using c++0x. > > Should mpi::allocator::construct() be different? Is it required to > provide a 1-arg construct()? > > > > Regards, > > MM I have added a 1-arg construct() that constructs default T()to mpi::allocator and that seems to compile correctly ,

1 0

Mersenne Twister 19937 suffers from "lattice effect"?
by Matthias Nagel 27 Jan '12

27 Jan '12

Hello, I tried to implement a two-dimensional multi-modal gaussian mixture and used the "mt19937" as the underlying generator. After I drew some samples (from the gaussian mixture) the outcome was not as I expected. After some bug tracking I suppose that the reason is something that is called "lattice effect" in German math literature. A pseudo-random generator suffers from a lattice effect, if one draws a sequence of samples, groups them together into N-dimensional vectors and these N-dimensional vectors are not uniformly distributed in the N-dimensional euclidean space. Does anybody know if the boost implementation has this problem? In the following, I will present my problem (and my algorithm) more elaborately, especially how I implemented my multi-model gaussian that is the origin of my problem: Lets assume we want a 2-dimensial gaussian mixture with M gaussian. 1) Create three lists (stl vector objects in my case), each with M entries. The first list stores the weight of each gaussian. The weights sum up to 1. The second list stores the mean value of each gaussian, i.e. a two-dimensional vector (in the mathematical sense, not the stl vector). The third list stores a 2-by-2 covariance matrix for each gaussian 2) Create a "mt19937" object, a "discrete_distribution" object and a "normal_distribution" object The "mt19937" is the common source of randomness for "discrete_distribution" and "normal_distribution". The "discrete_distribution" produces integers between 0 and M (number of gaussians) according the the given weights The "normal_distribution" is parameteized with mean = 0 and variance = 1, i.e. it is the standard normal distribtion. 3) Drawing one sample from the gaussian mixture - Draw one sample from the mt19937 and feed it into the discrete_distribution object. Denote the result by "index". - Draw a second sample from the mt19937 and feed it into the normal_distribution object. Denote the result by "u". - Draw a third sample from the mt19937 and feed it into the normal_distribution object. Denote the result by "v". - Pick up the mean vector and covariance matrix from the lists with respect to "index" - Transform the vector (u,v) according to the selected mean vector and covariance matrix. Denote the result "(x,y)". - Return (x,y) Lets remark that three samples are drawn from mt19937 to obtain one sample from the gaussian mixture. The result of this algorithm depends on the numbers of gaussians that build the mixture. 1) If M equals 1, the result looks fine. The (x,y) samples form an ellipsoid in the 2-dimensional euclidean space 2) If M equals 2, one gaussian looks like an ellipsoid, but it is too small. The second "gaussian" looks like a torus. Obviously all samples (u,v) with large coordinates are assigned to one gaussian, all (u,v) near zero are assigned to the other gaussian 3) If M equals 3, the "gaussians" look like segments of an ellipsoid. In each ellipsoid one third is missing. 4) For higher M the result is not correct either, but not as vivid and easily to explain as the "small" cases. As soon as I switch to random_device instead of mt19937 the phenomena disappears. ---------------------------------------------------------------------- Matthias Nagel Willy-Andreas-Allee 1, Zimmer 506 76131 Karlsruhe Telefon: +49 721 8695-1506 Telefax: +49 721 8695-4506 Mobil: +49 151 15998774 e-Mail: matthias.nagel(a)student.kit.edu ICQ: 499797758 Skype: nagmat84

3 3

BOOST_RESULT_OF_USE_DECLTYPE not default on C++11 compilers
by Jeremiah Willcock 27 Jan '12

27 Jan '12

At least on GCC 4.6 in C++0x (C++11) mode, BOOST_RESULT_OF_USE_DECLTYPE is not set by default, preventing classes such as boost::transform_iterator from working on built-in lambda expressions. Is there a reason that the use of decltype is not on by default for compilers that support it, perhaps with a flag to force emulation mode (like Boost.Move supports)? -- Jeremiah Willcock

4 7