- Boost - lists.preview.boost.org

[interprocess] Assign to shared_lock from scoped_lock
by Chard 08 Apr '09

08 Apr '09

If I have (template args removed for brevity): shared_lock sh_lock(mutex); scoped_lock sc_lock(move(sh_lock), try_to_lock); // Try for exclusive if (sc_lock) { ... sh_lock = move(sc_lock); // Drop back to shared *** not available } Should that assignment operator be available? Currently I'm working around this with: sh_lock.swap(shared_lock(move(sc_lock))); [or, alternatively: sh_lock = move(shared_lock(move(sc_lock))); ]

2 1

Re: [boost] [Any] Reference counting
by Michael Howell 07 Apr '09

07 Apr '09

Steven Watanabe <watanabesj(a)gmail.com> wrote: > Have you read http://www.gotw.ca/gotw/045.htm? I haven't read it, and have looked into atomic numbers, but couldn't find the "Boost Way" to do it (except perhaps poking around with shared_ptr innards). I wanted to use intrusive counts, since we have control over the object we are sharing. Thorsten Ottosen <thorsten.ottosen(a)dezide.com> wrote: >If you want to improve the performance of any, the thing to do is to add > a small buffer optimization such that all buildin types (or equally > small UDT) are created on the stack. I'll look into that. Mathias Gaunard <mathias.gaunard(a)ens-lyon.org> wrote: > Also, any is a value type. We do not want it to have reference > semantics. So you'd need to use COW. Actually, copy on write/implicit sharing is what I was proposing, if you had looked at the code. > Hence this is fairly useless, especially since we're getting move > semantics soon. "Soon" is relative. The trunk GCC may support it instantly (may already), but Debian Stable won't, nor will MSVC (both which will be used for a good while). Also, you usually copy the value when you store it, which may or may not involve writing. For example: #include <boost/any.hpp> using namespace boost; struct example { any variable; }; int main() { example object; example object2; any value = "I Love Boost!"; object.variable = value; object2.variable = value; return 0; } Thanks for replying, everyone. -- Michael Howell mhowell123(a)gmail.com

4 5

"[gsoc] Thread scheduler support for boost" [Re: for vicente.botet]
by Indradip Ghosh 07 Apr '09

07 Apr '09

Hi Vicente, Sorry for having missed your question earlier on my intended GSOC proposal. I actually intended to design the fine-grained scheduler for threadpool class itself. My detailed application is listed here in the googlesites: http://sites.google.com/site/threadpoolschedulergsoc2009/ Please let me know if you have any questions/ comments regarding my proposal. Thanks and Regards, Indradip. > Message: 7 > Date: Mon, 30 Mar 2009 08:36:36 +0200 > From: "vicente.botet" <vicente.botet(a)wanadoo.fr> > Subject: Re: [boost] "[gsoc] Thread scheduler support for boost" > To: <boost(a)lists.boost.org> > Message-ID: <C1B76DB53B1544F9B2C5618DB7A708EF@viboes1> > Content-Type: text/plain; charset="iso-8859-1" > > Hi, > > I'm wondering this shouldn't be considered as an extension of the Boost.Thread library more that the Boost.ThreadPool library. The ThreadPool library can be improved once these extension are in place, but also other application or libraries using threads. > > Let me know if I'm missing someething? > > Best, > Vicente

1 0

Performance measurement
by Govindan, Nandagopal (GE Healthcare) 07 Apr '09

07 Apr '09

Hi, I would like to know how performance measurement( time & memory ) done in boost. Do boost has some following APIs to measure performance?? Profiler.start() Code to be measured Profiler.stop() Printstatistics() Regards Nandagopal

3 3

[gsoc] improving any and variant
by Mathias Gaunard 06 Apr '09

06 Apr '09

A few little ideas that could eventually be joined to be a single GSoC project: - Allow visitation for Any. This can actually be generalized to any object that exposes a type_info, such as polymorphic objects. Techniques, including non-portable ones based on name mangling, could be experimented to get more efficient dispatch. - Add SBO to any. - Improve variant so that the functor doesn't require to expose a result_type. The result_type can be automatically deduced by "joining" all the possible return types obtained by boost::result_of. - Allow the easy construction of a visitor from lambda expressions, with a syntax like variant<int, std::string> a = "foo"; int b = match(a).with( args<int>(_1), args<std::string>(size(_1)) ); instead of the potentially more verbose struct my_visitor : static_visitor<int> { int operator()(int a) { return a; } int operator()(const std::string& b) { return b.size(); } }; variant<int, std::string> a = "foo";; int b = apply_visitor(my_visitor(), a);

5 5

Intrusive list/ilist and rbtree marshalling support
by Brian Wood 06 Apr '09

06 Apr '09

I'm happy to announce that version 1.8 of the C++ Middleware Writer is available on line -- http://webEbenezer.net/cgi-bin/samb.cgi<http://webebenezer.net/cgi-bin/samb.cgi>. This version adds support for the following: * Exclusive marshalling support of boost::intrusive::list and boost::intrusive::rbtree. Neither Boost Intrusive nor Boost Serialization offer serialization support of the Intrusive containers. Rather than having two types called 'list,' we refer to boost::intrusive::list as an ilist. An example of this support is here -- http://webEbenezer.net/rbtree_marshalling.html<http://webebenezer.net/rbtree_marshalling.html>. * Several new options on the user interface including: + An "Error Handling" option that indicates whether the code should be written using return codes or exceptions. The exception support is new in this release. This option affects the signatures of the generated marshalling functions. + A "Code Structure" option that indicates whether the code should be written in header-only format or separated interface and implementation files. The separated interface and implementation support is new in this release. This option also affects the signatures of the generated marshalling functions. + A choice between unsigned short or unsigned int as the type for constants written by the C++ Middleware Writer. Constants that identify messages and user defined types are written by the software. Previously the constants were always unsigned ints. * A mechanism to turn off automatic code generation for a type. Adding a // comment within a class declaration with "hand_written_marshalling_code" in the text of the comment will disable default code generation for that type. For example, class bag { // hand_written_marshalling_code ... }; As the comment indicates, you will provide the implementation of the marshalling functions for this class. This option is not available with /* */ comments. Additionally, I would like to say that I find the Intrusive library to be sound and am using it internally in the C++ Middleware Writer. Thanks to the authors: Olaf Krzikalla and Ion Gaztañaga. Brian Wood Ebenezer Enterprises www.webEbenezer.net <http://www.webebenezer.net/> "Then Samuel took a rock and set it up between Mizpah and Shen. He named it Ebenezer [Rock of Help] and said, 'Until now the L-RD has helped us.'"

1 0

[program_options] multitoken arguments
by Carlo Medas 06 Apr '09

06 Apr '09

Dears, we've been using boost::program_options and it works great. We just found a minor thing, which could be an issue, and I'm reporting it to you: when specifying and using a multitoken argument, it actually works only if put as last parameter of command line; in any other case, it eats all remaining parameters. My specification is as follows: desc.add_options() ("output,o", bpo::value<std::vector<std::string> >()->multitoken(), "use output setup for discovery/scan"); If then I want to use it like follows, it works: > fing -p1 -p2 -p3 -o 1 2 3 But if the user inverts the order, it does not work anymore, meaning that p3 is eaten by -o option. > fing -p1 -p2 -o 1 2 3 -p3 Is this a known limitation? Or maybe my specification above is not correct? Thanks in advance, Carlo Medas

2 4

Re: [boost] Shorter link to getting started
by Anders Dalvander 06 Apr '09

06 Apr '09

ghost(a)cs.msu.su wrote: > I would like to modify the configure script to mention getting started, > so that user knows where to look for information, but unfortunately, > the current URL is a bit long: > > http://www.boost.org/doc/libs/1_38_0/more/getting_started/unix-variants.html > > It is 77 characters, so it does not leave any space for any other text. > > Thanks, > Volodya What happens after 77 characters, do characters start falling off the edge of the screen? Hasn't that theory been disproved yet? J/K. It would be nice to have shorter URLs, the non version specific URL is even a character longer: http://www.boost.org/doc/libs/release/more/getting_started/unix-variants.ht… Regards, Anders

3 2

"[gsoc] Thread scheduler support for boost"
by Indradip Ghosh 06 Apr '09

06 Apr '09

Hi , I am writing to poll for suggestions/ questions/ comments regarding my GSOC application listed here : http://sites.google.com/site/threadpoolschedulergsoc2009/ Please let me know if I could clarify any concerns. Thanks, Indradip.

1 0

[boost test] boost testing framework xml schema‏
by Daniel Carrara 06 Apr '09

06 Apr '09

I would like to know where I can find the xml schema for the test logs produced at the end of the testing with --log_level=all and _log_format=XML. Thanks, Daniel _________________________________________________________________ Windows Live™ SkyDrive: Get 25 GB of free online storage. http://windowslive.com/online/skydrive?ocid=TXT_TAGLM_WL_skydrive_032009

2 1

[mpl][doc]bug: prior.html should have Bidirectional Iterator requirement.
by Larry Evans 05 Apr '09

05 Apr '09

The prior.html document: http://www.boost.org/doc/libs/1_38_0/libs/mpl/doc/refmanual/prior.html says prior requirement is that Iterator be a Forward Iterator; yet, http://www.boost.org/doc/libs/1_38_0/libs/mpl/doc/refmanual/forward-iterato… makes no mention of prior. Also, compiling prior on a list iterator shows: /home/evansl/prog_dev/boost-svn/ro/boost-trunk/boost/mpl/next_prior.hpp:40: error: no type named 'prior' in 'struct boost::mpl::l_iter<boost::mpl::list1<char> >' Hence, the prior requirement should be Bidirectional instead of just Forward Iterator. -regards, Larry

2 1

My ideas for GSoC project
by Balša Raičević 05 Apr '09

05 Apr '09

Hello, I'm a Computer Science student from Faculty of Mathematics, University of Belgrade. I am interested in working on The BGL. I have a lot of experience with C++ and code refactoring also, since I've done a lot of tutoring and encountered a lot of strange students' code :) I have taken a deep look at BGL ( which is why I'm writing my proposal so late :) ) and to some extend its extension - The Parallel Boost Graph Library. I am mostly interested in graph partitioning, as I find them very useful for parallel computing. I understood the importance of it when I had to do some major calculations for numerical approximations and wanted to use all the CPU power available :) In the project idea list ( https://svn.boost.org/trac/boost/wiki/soc2009#Graphpartitioning) it's said that it would be good to implement algorithms similar to those in METIS ( http://www.cs.umn.edu/~metis <http://www.cs.umn.edu/%7Emetis>). I've read some of George Karypis' "Publications Related to Graph Partitioning", so I was wondering if I could discuss with someone which algorithms would be good to implement, and of course, how. I found little, if any, discussion about that in mailing list archive. Other thing I would really like to see is hypergraph partitioning. I see that implementing hypergraphs in the BGL is one of GSoC projects, so I would like to know would it be possible, if that project is not awarded to anyone for this summer, to do basic implementation of a hypergraph, or (since I see there is already a discussion about that) if somebody else would do that project, to collaborate on hypergraph partitioning algorithms. Last, but not least, an important question: I would like to use this work as part of my Master's Thesis. I looked at GSoC FAQ ( http://socghop.appspot.com/document/show/program/google/gsoc2009/faqs#cours…) so I guess it is OK with the program, but I would still like to discuss it with potential mentor. Thanks ahead, Balsa

1 1

boost::shared_mutex performance
by Tim Blechmann 05 Apr '09

05 Apr '09

hi all, i did some performance tests of boost::shared_mutex and a wrapper class to posix pthread_rwlock_t, which provides most of the interface from the sharedlockable concept (without the timed locking, which should be easy to add). i am not sure, whether the boost::shared_mutex class is supposed to give some guaranties (like fairness), that my wrapper implementation doesn't provide ... it does outperform the boost::shared_mutex implementation by a factor of about 2.5, though (linux, x86_64) benchmark code and wrapper class are attached. seeing these speedups, would it be reasonable to introduce a boost::native_shared_mutex class, based on a native implementation? best, tim -- tim(a)klingt.org http://tim.klingt.org The first question I ask myself when something doesn't seem to be beautiful is why do I think it's not beautiful. And very shortly you discover that there is no reason. John Cage.

1 0

Naga Rama Lokesh Gopu, student - link id zaq1xsw2 , Google Summer of Code
by Lokesh Gopu 05 Apr '09

05 Apr '09

Hi sir , My name is Naga Rama Lokesh Gopu, My link id is zaq1xsw2, I have submitted a proposal on creating new math library with advanced functions, But i have submitted in the last hour before deadline. That s the reason i cant mention clearly in that proposal. i don't know how to prove my self that i am eligible for this project, i submitted the details of my project proposal. In rush you may made my proposal ineligible. I am ready to do code almost all the day more than 3 months. my contact number is (+91 9246476974) I cant find the individual email address who like to do job of mentor. I am ready to do any mathematical project of our wish , Because i am damn good in those topics of good depth. I am not writing any story here, I just want to make u understand my self. If any one like to do mentoring for my project . Please make my proposal eligible , and make me to work harder in this summer. the polynomial function in the boost libraries are not in good depth , i can do more work on the all such topics. I cant clearly mention the work i can do.. Please reply me your opinion about my project on creating new math library. Please understand me i am very enthusiastic to do that in this summer . Finally make my proposal eligible, and mentor my project . Reply is must, even a string of reply! Thank you for understanding me

1 0

[Any] Reference counting
by Michael Howell 05 Apr '09

05 Apr '09

The attached patch add reference counting "implicit sharing" to Boost.Any. It is my first patch to Boost, so it will probably need some changes. -- Michael Howell mhowell123(a)gmail.com

5 5

[gsoc] Students who have an application pending
by troy d. straszheim 04 Apr '09

04 Apr '09

Please try to keep your application up to date with mailing list archive links (via gmane or whatever) to the heads of threads about your proposals. Just add something like a "links to discussion" section at the bottom. This will really help us. If no discussion has taken place here about your proposals, if you update your proposal with a link to this message, you'll still get points for having seen this message at all... :) -t

5 7

[gsoc] Range/Iterator in Boost.Crypto
by Kasra 04 Apr '09

04 Apr '09

Hi all, I have got more excited about GSoc so I have been working ahead of schedule to get boost.crypto going. At this time I have completed my schedule until 05/05/2009. Here I need some advice on the design principles for my interface. I have decided to implemented a generic buffer called crypto_buffer, this buffer is intended to implement bufferings for a stream (stream ciphers/hash algorithms). The class head is as follows: // this class manages 1 block of state_size that the stream is deposited on // also arbitary numbers of buffers that the underlying transformer // wished to include. // when the buffer # 1 is filled in with data the transformer is called // to process the buffer and this is continue untill the end of buffer template< class Transformer, // the algorithm that will transform the buffer size_t Buffers, // the # of buffers this transformer will manage bool FixedOutput, // if the output buffer is constant bool BitBuffered // byte or bit buffered? > class crypto_buffer { public: typedef Transformer transformer_type; typedef BOOST_DEDUCED_TYPENAME transformer_type::size_type size_type; ... // the size of input block which ensured optimal buffering performance static constexpr size_type transform_size = transformer_type::transform_size; // the size of blocks which is stored between transition states static constexpr size_type state_size = transformer_type::state_size; private: transformer_type m_transformer; public: // create a new stream void create(); // transform all but the final stream block template<CryptoBuffer bufferT> size_type transform(bufferT& , const bufferT&); // transform the final block of our stream template<CryptoBuffer bufferT> size_type transform_final(bufferT& , const bufferT&); }; Notice the 'CryptoBuffer' concept. The question is how to implement this concept. Iterators and Ranges is suggested by other previously, however, how could we impelemnt them for the above interface? The reason for my question is I am not sure if I could implement a library that doesn't require for example a block-cipher to have 1-block of input in contiguous memory. For hash functions and stream ciphers is much easier since the contiguity requirement could be a byte or a bit which could be done easily. Any suggestions?? With Best Regards -- Kasra Nassiri

1 0

[GSOC] Regarding the project Boost.DeVector
by Satyam Shekhar 04 Apr '09

04 Apr '09

Hi, I am interested in doing the Boost.DeVector project for GSOC 2009. I had certain queries regarding this project: 1. Does the standard specify any sort of implementation for deque, or does it only specify the requirements and interface? I ask this because if we assume a dynamic array implementation which grows on both the sides, implementing 'reserve', and specifying the size of the chunk should be a trivial task. 2. What are the examples of the low-level operations for which this container is over-encapsulated? Regards Satyam

6 14

[gsoc] Application period ends today / review period begins.
by troy d. straszheim 03 Apr '09

03 Apr '09

Student application period ends today at 19:00 UTC, 12 Noon PDT. Then we have until the 15th to review and rank all (over 50 as of now) student proposals and match them with mentors. -t

4 7

[serialization] serializing std::bitset
by Kenny Riddile 03 Apr '09

03 Apr '09

I need to serialize a std::bitset. I found an implementation by Brian Ravnsgaard Riis from 2006 in the vault that doesn't work with xml archives out of the box since it fails to use nvp wrappers internally. Should I just add the nvp wrappers or is there anything more current for serializing std::bitsets?

3 5

Release schedule
by Vladimir Prus 03 Apr '09

03 Apr '09

Hi, the calendar at http://www.boost.org/development/index.html has: 1.39.0 Closed for major code changes for today and: 1.39.0 Closed except by permission for next Friday, Apr 10th. I have a few questions: 1. Is this schedule still up-to-date? I did not notice any reminder emails. 2. I assume this refers to all commits to release branch? 3. What is 'major code change'. Will wholesale merge of Boost.Build count as one (though I'm not away of any disturbing changes on trunk). 4. Would it be possible to move both deadlines over the weekend to Apr 6 and Apr 13 respectively? I'm sure many developers have more spare time on weekend, so it will make it easier to complete changes. I, in particular, have some changes in pipeline to make building Boost less confusing for newcomers, and certainly will appreciate having that weekend of Apr 11-12. Thanks, Volodya

4 4

[statechart] Bug when switching inner-states?
by George van Venrooij 03 Apr '09

03 Apr '09

Hi, I have attached a test program that gives some curious results, indicating the statemachine does not reach its proper state when performing state changes in orthogonal regions. The machine attached contains a main state that handles an "update" event. That main state contains 2 inner states, one of which also handles the "update" event. If that inner state transits to another state and back again, the "update" event no longer arrives at it, which in my opinion is a bug. If I modify the state machine so it does not have 2 but only 1 inner state, then the behavior matches my expectations. Could someone please verify if this is bug or tell me if my usage scenerio is incorrect. Regards, George van Venrooij Organic Vectory http://www.organicvectory.com #include <boost/statechart/state_machine.hpp> #include <boost/statechart/event.hpp> #include <boost/statechart/simple_state.hpp> #include <boost/statechart/custom_reaction.hpp> #include <iostream> namespace bs = boost::statechart; using namespace std; // The main event, should be handles by a sub-state and the main state struct update_event; // An event triggering the sub-state to switch struct toggle_event; // A dummy event struct other_event; // Sub-states in 2 orthogonal regions struct state_1A; struct state_1B; struct state_2A; struct state_2B; // Main state struct parent_state; // State machine consisting of the main state struct machine : bs::state_machine<machine, parent_state> { }; typedef boost::mpl::list < state_1A , state_2A > inner_states_t; // The main state has 2 orthogonal sub-states struct parent_state : bs::simple_state<parent_state, machine, inner_states_t> { typedef boost::mpl::list < bs::custom_reaction <update_event> > reactions; bs::result react(const update_event&) { cout << __FUNCTION__ << endl; return discard_event(); } }; struct update_event : bs::event<update_event> { }; struct other_event : bs::event<other_event> { }; struct toggle_event : bs::event<toggle_event> { }; struct state_1A : bs::simple_state<state_1A, parent_state::orthogonal<0>> { typedef boost::mpl::list < bs::custom_reaction <toggle_event> , bs::custom_reaction <update_event> > reactions; state_1A() { cout << __FUNCTION__ << endl; } bs::result react(const toggle_event&) { cout << __FUNCTION__ << endl; return transit<state_1B>(); } bs::result react(const update_event&) { cout << __FUNCTION__ << endl; return forward_event(); } }; struct state_1B : bs::simple_state<state_1B, parent_state::orthogonal<0>> { typedef boost::mpl::list < bs::custom_reaction <toggle_event> > reactions; state_1B() { cout << __FUNCTION__ << endl; } bs::result react(const toggle_event&) { cout << __FUNCTION__ << endl; return transit<state_1A>(); } }; struct state_2A : bs::simple_state<state_2A, parent_state::orthogonal<1>> { typedef boost::mpl::list < bs::custom_reaction <other_event> > reactions; state_2A() { cout << __FUNCTION__ << endl; } bs::result react(const other_event&) { cout << __FUNCTION__ << endl; return discard_event(); } }; struct state_2B : bs::simple_state<state_2B, parent_state::orthogonal<1>> { typedef boost::mpl::list < bs::custom_reaction <other_event> > reactions; state_2B() { cout << __FUNCTION__ << endl; } bs::result react(const other_event&) { cout << __FUNCTION__ << endl; return discard_event(); } }; int main(int argc, char* argv[]) { machine m; cout << "0. Initiating state machine:" << endl; m.initiate(); cout << endl << "1. Processing 'other_event':" << endl; m.process_event(other_event ()); cout << endl << "2. Processing 'update_event':" << endl; m.process_event(update_event()); cout << endl << "3. Processing 'toggle_event':" << endl; m.process_event(toggle_event()); cout << endl << "4. Processing 'update_event':" << endl; m.process_event(update_event()); cout << endl << "5. Processing 'toggle_event':" << endl; m.process_event(toggle_event()); cout << endl << "6. Processing 'update_event':" << endl; m.process_event(update_event()); return 0; } ////////////////////////////////////////////////////////////////////////// // Program output after building with VC 9 and Boost 1.37.0 // // Note the different event handling in step 6. as compared to step 2. // although the machine should be back in the same state. ////////////////////////////////////////////////////////////////////////// /* 0. Initiating state machine: state_1A::state_1A state_2A::state_2A 1. Processing 'other_event': state_2A::react 2. Processing 'update_event': state_1A::react parent_state::react 3. Processing 'toggle_event': state_1A::react state_1B::state_1B 4. Processing 'update_event': parent_state::react 5. Processing 'toggle_event': state_1B::react state_1A::state_1A 6. Processing 'update_event': parent_state::react */

2 1

Mutex vs Semaphore
by Jean-Sebastien Stoezel 03 Apr '09

03 Apr '09

Hi, I need to synchronize 2 threads, that is once one is done doing something, it should activate another thread that had been waiting in a pending state. I see the thread library offers mutexes and nothing else. I guess mutexes would do, but I do not believe I should use mutexes for my application. Mutexes are supposed to provide mutual exclusion when critical data is accessed. Semaphores should be used for what I want to do. However, I only see semaphores available in the interprocess library. I only need inter-thread synchronization, and going the long way with the interprocess library seems a bit overkill. It may be an ideology issue here. I would appreciate your comments on what to use in this case. Thanks, Jean

4 5

[gsoc] Min Cost Flow in Boost Graph library
by shilp gupta 03 Apr '09

03 Apr '09

Hey, I am interested in contributing to the Boost Graph library with algorithms. Some of my suggestions are, - minimum cost flow - weighted bupartite matching - maximum flow (based on blocking flows) I had a few questions to ask, Is there any significant progress already in development or implementation of minimum cost flow algorithms in Boost? To the best of my knowledge, Boost does not have a structure for bipartite graphs. Would it be appreciated if I work on one? I know that these can be represented using standard graph structures without much loss, but the simplicity (and scope) of bipartite graph's can be used to reduce constants in the run times of algorithms on them - maximum flow using blocking flows infact runs in O(sqrt(V)E) time for calculating the maximum cardinality matching in the bipartite case. Thanks! Regards Shilp Gupta

2 1

BOOST::SIMD - handling double : precision vs speed
by Joel Falcou 02 Apr '09

02 Apr '09

SIMD algorithms for double precision seem to be rather hard to do right. It's difficult to get the right precision with respect to the scalar reference as scalar algorithm take advantages of the internal 80 bits floating points register, thus leading comparison between our implementation and the reference to yields things like 3000 ulp (ie 10^-13 RMS instead of 10^16). Fixing this is difficult and even if it's possible for some algorithms, the average speed-up then drop to less than 10% - ie as fast as an unrolled scalar call over the SIMD vector elements. What should we enforce : precision or speed ? Or is the 10^-13 RMS enough ? Of course, this is mostly a problem on current SIMD extension in which vector of double only have 2 elements. It may be different with upcoming AVX and Larabee featuring larger vectors. Discussions welcome. -- ___________________________________________ Joel Falcou - Assistant Professor PARALL Team - LRI - Universite Paris Sud XI Tel : (+33)1 69 15 66 35

7 6