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

[Serialization] Boost serialization how to speed it up?
by tang qingxiong 13 Jun '11

13 Jun '11

I am currently trying to use boost serialization in my code and is encountering some performance issues mainly with speed of serialization and de-serialization. I am coding such that it is able to choose to serialize in both binary and XML format. My binary size is currently about 180+MB which takes a while to de-serialize Below is a sample of my code for the serialization of a struct. I have quite a few of them but they are all similarly implemented namespace boost { namespace serialization { template<class Archive> void serialize( Archive& theArchive, sim::SpeedStruct& theValue, const unsigned int theVersion ) { theArchive & BOOST_SERIALIZATION_NVP(theValue.XValue); theArchive & BOOST_SERIALIZATION_NVP(theValue.YValue); theArchive & BOOST_SERIALIZATION_NVP(theValue.ZValue); } } } BOOST_CLASS_VERSION( SpeedStruct, 1 ); BOOST_CLASS_EXPORT_GUID( SpeedStruct, "SpeedStruct" ); and the speed struct is just a simple struct with 3 double values of X,Y and Z. MY question is I read the FAQs and there is indication of various ways to speed up the process, but I am unclear of how to implement it in my code , specifically how to use BOOST_IS_BITWISE_SERIALIZABLE and also removing tracking and versioning. Thanks! Lester

1 0

how to change boost thread priority ?
by chyujian 12 Jun '11

12 Jun '11

I am new using boost thread, in my project I need to change the priority of the threads, but I cannot find help in the documentation.

2 1

Last of the first batch of Boostcon videos
by Marshall Clow 12 Jun '11

12 Jun '11

Gordon (Thursday): http://blip.tv/boostcon/introducing-mpl-graph-5262812 Bryce (Friday): http://blip.tv/boostcon/ast-construction-with-the-universal-tree-5266608 I am looking into why Bartoz' Haskell preso is not showing up. You can download the mp4 directly here: http://blip.tv/file/get/Boostcon-HaskellThePseudocodeLanguageForCTemplateMe… but the conversion to H.264 seems to time out (I've tried several times) I'll be contacting blip.tv about this in the morning. -- Marshall Marshall Clow Idio Software <mailto:mclow.lists@gmail.com> A.D. 1517: Martin Luther nails his 95 Theses to the church door and is promptly moderated down to (-1, Flamebait). -- Yu Suzuki

1 0

Re: [Boost-users] [boost] Even more Boostcon videos
by Marshall Clow 12 Jun '11

12 Jun '11

On Jun 11, 2011, at 9:29 AM, John Bytheway wrote: > On 10/06/11 16:56, Marshall Clow wrote: >> Joel (Thursday): >> http://blip.tv/boostcon/practical-simd-acceleration-with-boost-simd-5261117 > > This one doesn't entirely work for me; the audio cuts off at about > 1:10:43. Fails at the same point in mplayer and VLC. I downloaded it > twice with the same result. > > $ md5sum < Boostcon-PracticalSIMDAccelerationWithBoostSIMD525.mp4 > 56c08b18dc5141a383736e1456d9c48b - I went back and checked the tapes - the audio is missing there, too It appears that the battery on the wireless mic that we were using ran out. :-( Thanks for the heads-up, but I don't think there's anything that I can do to fix this. -- Marshall Marshall Clow Idio Software <mailto:mclow.lists@gmail.com> A.D. 1517: Martin Luther nails his 95 Theses to the church door and is promptly moderated down to (-1, Flamebait). -- Yu Suzuki

1 0

Re: [Boost-users] boost MSDN
by Peter 12 Jun '11

12 Jun '11

On Sat, Jun 11, 2011 at 06:54:57AM -0400, Edward Diener wrote: > > > On 6/10/2011 5:51 AM, Klaim - Jo?l Lamotte wrote: >> > >> On Fri, Jun 10, 2011 at 05:39, Peter<lq.yan(a)siat.ac.cn >> > >> I'm a newer to boost library. when i try to use the boost library to >> > >> get the files of current directory, I found it is so difficult to >> > >> get some help from boost HTML document. >> > >> Is there any document could teach me to use boost library API? Most >> > >> of time, i use visual studio, MSDN teach me how to use API. >> > >> So i think Boost maybe has similar documents to help me improving >> > >> coding speed. >> > >> -- >> > >> >> > >> >> > >> Hi Peter and welcome, >> > >> >> > >> Boost is a set of libraries, not only one library. You can think >> > >> Boost is a "quality label" to make it more clear. >> > >> >> > >> That means that each library have it own purpose, might be totally >> > >> independant of other Boost libraries and have an interface >> > >> specifically thought to solve a specific kind of problem. So not all >> > >> libraries can share the same interface. >> > >> >> > >> That also mean that each library have to have it's documentation >> > >> separate from the others. >> > >> >> > >> As Boost is totally independant from companies like Microsoft, you >> > >> can't >> > >> (AFAIK) automatically provide a way to get the Boost documentation in >> > >> MSDN browser. > > I believe you lads are misunderstanding his original question. My interpretation is the following: > > * When he refers to "MSDN teach me", he's referring to the ample amounts > of examples, snippets and proper articles present in the APIs documented > on the MSDN website/documentation. > > * Most Boost library documentation is a bit sparse and unstructured when > it comes to the prose, and samples tend to just be code dumps, while on > MSDN they're typically interleaved with prose, describing things in more > detail. > > You make some points about that each Boost library having its own documentation style, which makes it > harder to find the relevant sections of the documentation you're interested in. > > I know I personally avoid the iostreams and serialization docs solely due to the very hard to use expandos > in the tree structure, and the impossibility to link to particular pages while retaining said navigation tree. > > While it would be Good if all Boost docs was under the same documentation system (which some are), > porting old docs to whatever system is in use now would probably be rather costly for little gain. > > The IBD (Improving Boost Docs) got some progress in the past, but I'm unsure what the end results of their > effort were. It may not *yet* be apparent, but most new and upcoming libraries come with *much* better documentation, especially those using the Quickbook tool chain. (The Improving Boost docs program is only now yielding fruit, and the recently accepted AutoIndex will also increasingly provide the benefit of indexes quite soon). IMO, some Boost libraries are examples of the best documentation - often much better than MSDN. Having spent some time reworking an old library (Boost.Pool), I say that it was time-consuming and at the same time both mind-numbingly boring and yet testing ones knowledge of the details of the library (I found myself out of my depth!). So if anyone is an reasonably expert user of an existing library and has some time to spare, now is the time to volunteer to upgrade its docs! Assistance getting started can be provided. Paul Hi Paul and Lars, Exactly, i agree with lars. Yes, if there is a document for all the libraries; there are some simple examples are included, and the corresponding sections could be listed. The document will be wonderful. *Regards, ??? /Peter,YAN/ * ?????????? ????????????????? Email: lq.yan(a)siat.ac.cn TEL: +86 755 8639 2301

1 0

Even more Boostcon videos
by Marshall Clow 12 Jun '11

12 Jun '11

Bojan (Thursday): http://blip.tv/boostcon/fast-numerical-computation-in-c-expression-template… Joel (Thursday): http://blip.tv/boostcon/practical-simd-acceleration-with-boost-simd-5261117 Matt (Thursday): http://blip.tv/boostcon/boost-generic-concepts-without-concepts-5262126 I will post a couple more over the weekend, then be on holiday for a couple weeks, then post the rest in early July. -- Marshall Marshall Clow Idio Software <mailto:mclow.lists@gmail.com> A.D. 1517: Martin Luther nails his 95 Theses to the church door and is promptly moderated down to (-1, Flamebait). -- Yu Suzuki

4 3

Retrieving edge weights
by David Doria 11 Jun '11

11 Jun '11

I found one way to get edge weights: typedef boost::property_map<Graph, boost::edge_weight_t>::type EdgeWeightMap; EdgeWeightMap edgeWeightMap = get(boost::edge_weight, g); std::pair<Graph::edge_descriptor, bool> edgePair = boost::edge(v0, v1, g); Graph::edge_descriptor edge = edgePair.first; That seems like a lot of work though. It looks like there is a operator[](edge_descriptor), but I don't understand how to use it: I tried: g[edge]->edge_weight but there are compiler errors. The whole demonstration is here: http://programmingexamples.net/index.php?title=CPP/Boost/BGL/Directed/Weigh… Any suggestions? Or is the first way the only/best way? Thanks, David

2 1

boost MSDN
by Peter 11 Jun '11

11 Jun '11

Dear all, I'm a newer to boost library. when i try to use the boost library to get the files of current directory, I found it is so difficult to get some help from boost HTML document. Is there any document could teach me to use boost library API? Most of time, i use visual studio, MSDN teach me how to use API. So i think Boost maybe has similar documents to help me improving coding speed. -- *Regards, ??? /Peter,YAN/ * ?????????? ????????????????? Email: lq.yan(a)siat.ac.cn TEL: +86 755 8639 2301

5 4

boost::thread and boost::asio::io_service questions
by Christopher Pisz 11 Jun '11

11 Jun '11

I am trying to understand both of these items and how they are executing in our code, so I made a small demo to debug through. 1) Why can't I pass boost::io_service by reference in the call, boost::thread workerThread(DoWork, 1, ioService); with a function signature of void DoWork (const unsigned int threadNum, boost::asio::io_service * ioService) ? I changed it to a pointer and it works, but I don't know why I can't do it by reference. 2) If boost::thread is not copyable, how can I store a collection of "thread handles" or boost::thread objects like the variable workerThread? I can't use the std collections I would usually use... 3) When stepping through the code, the debugger jumps between the main thread, the worker thread, and a call stack in the depths of boost code where one item points to the line where the call to ioService->run() is. The first two make sense. I don't understand the latter. The call stack looks very much like one we get in production code and is all I have to go on in trying to debug a problem. The challenge is trying to determing who created the thread and find that line in source, because the production code is keeps creating threads until the program crashes. #include <iostream> #include <boost/asio.hpp> #include <boost/bind.hpp> #include <boost/date_time/posix_time/posix_time.hpp> #include <boost/thread.hpp> //----------------------------------------------------------------------------- /// This function is the entry point for each worker thread void DoWork(const unsigned int threadNum, boost::asio::io_service * ioService) { std::cout << "Doing work in worker thread: " << threadNum << "\n"; // Perform the work // Multiple threads may call io_service::run() to set up a pool // of threads from which completion handlers may be invoked. ioService->run(); } //----------------------------------------------------------------------------- /// This function gets called when the asynchronous operation completes /// Guarantee - A completion handler will only get called from the same thread that called boost::asio::io_service::run() void OnCompletion(const boost::system::error_code & e, const unsigned * itemNum) { std::cout << "Completed item # " << *itemNum << " has completed. \n"; } //----------------------------------------------------------------------------- int main() { boost::asio::io_service ioService; boost::asio::deadline_timer timer(ioService, boost::posix_time::seconds(5)); const unsigned maxItems = 5; // Queue up some work for(unsigned int itemNum = 0; itemNum < maxItems; ++itemNum) { // The work to do is an asynchronous wait timer.async_wait(boost::bind(OnCompletion, // Method to callback boost::asio::placeholders::error, // &itemNum)); // Arg 1 to pass } // Create a thread to do the work boost::thread workerThread(DoWork, 1, &ioService); // Do other stuff in the main thread for(unsigned i = 0; i < 100; ++i) { std::cout << "Doing other work in main thread\n"; } // Done return 0; }

2 1

enable_if and separated defintion
by Olaf Peter 11 Jun '11

11 Jun '11

Hello, in this example I separate the definition from declaration. Is there are short-hand form for the enable template parameter? Or even a better way to do this? (Also, it's expect a type ...)

1 0