- Boost - lists.boost.org

[msm] scalability
by Christophe Henry 12 Jan '10

12 Jan '10

Hi Simon, > I think it would be worthwhile talking a little more about compile time performance in the documentation. You're right, I could explain more in the documentation. Unfortunately, the answer is not straightforward. The compilation time (and how far the compiler plays with us) depends on a number of factors: 1) the compiler 2) the compiler version 3) even your hardware (I get very different compile times on 2 different machines with the same processor) 4) the state machine structure and the chosen front-end 1) usually, g++ crashes later than VC and needed fewer workarounds 2) Surprisingly, g++4.2 > g++ 4.4. VC10 > VC9 > VC8 3) you will need RAM, especially with VC8 4) the compile-time is proportional to the number of transitions in the transition table and to the depth of the state machine (meaning submachine of submachine of...). EUML will cost you much more with VC than g++ as it implies a metaprogramming layer above the standard metaprogramming of the standard front-end and because the underlying decltype/typeof seems much better implemented with g++ (if you don't count the generated code size). A small test with VC9 in Release mode (and no other code) gave me: 20 transitions => 11s 30 transitions => 16s 40 transitions => 22s 50 transitions => 31s This is just to give you an idea as it can greatly depend on the factors named above. > Also I'd like to see some suggestions of how to improve compile times (and how factors including the number of entries in a transition table affect the compile time). There is just one solution: reduce the number of transitions. The best way to do it is using orthogonal regions. For example, if you see in your diagram many transitions to a single state based on the same event, you should factor them all into a second region. Adding submachines could help in some cases. On one hand, the compiler now has to generate code for a second machine, OTOH it reduces the number of transitions in the main one. This will only help if your submachine handles a small number of events, as Msm will add a row in the main table for every event of the submachines. If you can in exchange reduce the number of transitions, it can help. > Is there provision to split a machine across multiple compilation units (I assume not)? You unfortunately assume correctly, as what costs compile-time really is the heavy metaprogramming. > You include instructions for how to increase the maximum size of an MPL vector, but this only works up to 50 entries. What do we need to do to increase this further? This is pretty simple, we need to: - add a vector60/70/80....hpp in mpl/vector. - add a map60/70....hpp in mpl/map. - if more than 50 states, add vector60/70....hpp in fusion/vector This is pretty simple. If that helps, I can add them into the Msm sandbox as I already wrote them. We could also request a feature in the MPL. You will however not manage to increase the size of the transition table forever. At some point will almost every compiler crash in a big cloud of dark smoke. To give you an idea, I usually manage: - 60-80 transitions max with VC9. The exact point is variable on a number of different factors which are hard to understand. - probably less with VC8. VC8 will also quite fast need > 2GB RAM. - 80 transitions with g++4.2. It could probably manage more but I ran out of RAM. - 50-60 with g++4.4. Why 4.4 is in this case less good than 4.2 is a mystery to me. - In Debug mode, you'll see the compiler give up faster - Euml will also have an influence. > Euml looks really cool, but I'm getting an ICE when using it with 36 transitions in EumlSimple.cpp (12 and 24 work fine though) VC8 only has limited support for eUML for 3 reasons: - less than perfect metaprogramming capabilities - bad support for SFINAE - VC problems with decltype/typeof So yes, I would advise eUML for VC8 for only small state machines and without using complicated functors as actions/guards. VC9 would work much better, VC10 better than VC9 too (so Microsoft seems to be working on it). >One of my favourite features of boost::statechart is state local storage. It looks like this would have to be done manually in msm? Yes, this is not supported yet. However, Msm has always been developed very user-oriented and what users ask, they usually get it. Msm has from the beginning greatly gained from clever people suggesting new features and I will not stop the development after the review, whatever the outcome is. If you'd like a feature, I'd be happy to discuss a solution with you and implement it in the next version, thus making Msm more useful. > The library is looking really powerful and expressive (not to mention fast!) for small state machines. Thanks. This is the reason for the compile-times ;-) Christophe

5 5

Re: [boost] Meta State Machine (MSM) review
by Christophe Henry 12 Jan '10

12 Jan '10

Hi Barend, Sorry that I didn't have time to address earlier the questions in your review. >I would like to see a connection with Boost.Graph (I mean, see it in an >example or so, not within the kernel), to see what I'm doing with all >those transitions; I asked this on the list and this is possible, good. It is a good point and I'll try to provide this, at least for simple state machines. >1) the struct "Row" is written with a capital, this is unusual within >Boost. Words with a capital are normally template parameters. I >understand why it is done (there was already a row without a capital in >Msm 1.1), but I think it could have been solved differently (for >example: "transition", or if that is too long "tr" which could stand for >transition row or table row). I don't like to see Row where I expect to >see a normal classname. Conventions are always hard to define. I usually try to follow the following convention: - rows are usually not capitalized (a_row, g_row...) - except the row working with functors, like Row - eUML functors are also capitalized and finish with an underscore. - eUML functions are not capitalized and finish with an underscore. This tries to match the phoenix syntax. This allows differentiation between functors/non functors. I found this system practical but would not mind changing it if required. >The capitals are also used in the documentation and samples (struct >Empty) but there I've less problems with it. Maybe use of Capitals for >states is a state machine convention Yes, it is a common habit in UML to capitalize state names (though I am not aware of any written rules). >I noticed different conventions in >the source files. Many template parameters are completely capitalized. I >had done the same thing, earlier, but was hinted that this is unusual as >well. Just a matter of style (as long as library users normally don't >see it, it is less important) Yes, while the user doesn't see it, it is not very beautiful so I'll change this. >2) the class state_machine is quite long (in lines of code), containing >several sub-structs which maybe (don't know) could be moved to a >separate source file (detail). Yes, it became quite big, you're right. It's quite practical for me to have functions and helpers at the same place but some parts could be separated. >You have to call p.start() manually Yes. Calling start() will move the machine in its initial state and call on_entry on the machine and the initial state. You have to call it explicitly because sometimes you simply don't want on_entry to be called at the creation of the state machine. >Thanks for submitting this excellent library. I'm happy that you like it. Thanks for the great work you put on the review. Regards, Christophe

2 1

[Polygon] "Ambiguous use of partial specialization" in CodeWarrior
by Kef Schecter 12 Jan '10

12 Jan '10

I'm using a special version of CodeWarrior for an embedded system, but I suspect that other versions of the software will show the problem. The problem is, basically, the compiler hates polygon_traits<T>. Here's an example from polygon_90_set_traits.hpp: template <typename T> struct traits_by_concept<T, polygon_90_concept> { typedef polygon_traits<T> type; }; It chokes on the typedef, saying "Ambiguous use of partial specialization". This is the complete text of the error message; it provides no other information. Most of the surrounding definitions are very similar, but only the ones that use polygon_traits<T> -- instead of, say, rectangle_traits<T> -- raise errors. Here's another example from polygon_traits.hpp: template <typename T> struct view_of<rectangle_concept, T> { typedef typename polygon_traits<T>::coordinate_type coordinate_type; [snip] }; It chokes on the typedef and any code that tries to use the typedef. The same problem happens with all the other view_of functions and a couple of others, but all the problems seem to deal with the use of polygon_traits<T> in a template. Unfortunately, I know very little about partial specialization, traits classes, or any advanced features of generic programming in C++. My understanding of templates is limited to stuff like making basic containers, generalized sorting functions, etc. -- the easy stuff. So I find the definition of polygon_traits too complicated and am unable to fully comprehend the essence of the problem, let alone how to fix or work around it. Since I need to get this to work, if a proper solution is not obvious, I'm willing to accept any kind of clumsy hack, including disabling parts of the library that I might not need. (I use very little of the Polygon library, but I have a hard time finding an alternative to the part I do use -- namely, finding the union of polygons while allowing for holes. Alternative suggestions in that area are welcome, but I can't use GPLed code, or code that I would need to pay a license for to use commercially.) Any ideas? - Kef

4 16

[Archive] oserializer.hpp
by Richard Crossley 12 Jan '10

12 Jan '10

The following code from archive/detail/oserializer.hpp on head looks rather dubious... >line 394 // note:if this exception is thrown, be sure that derived pointer // is either registered or exported. if(NULL == true_type){ boost::serialization::throw_exception( archive_exception( archive_exception::unregistered_class, true_type->get_debug_info() // << Here ) ); } Regards, Richard.

2 1

Re: [boost] [msm] eUML guard/action location
by Christophe Henry 12 Jan '10

12 Jan '10

>> You could declare those instances at namespace scope, in the same way >> most Proto terminals are declared, non? >The user could. But not the library writer AFAIU, since these state >types are created by the user. Exactly. Of course the library could predefine some state names (the way I understand arg1, arg2, etc. are defined) but I'm afraid it'd barely be an improvement. Christophe

2 1

[persistent] Persistent library preview
by Stefan Strasser 12 Jan '10

12 Jan '10

Here's the current documentation of (Boost?).Persistent: https://svn.boost.org/svn/boost/sandbox/persistent/libs/persistent/doc/html… In short, it is a library that provides access to objects persistent on disk with an interface as close as possible to accessing regular objects in memory, and it introduces transactions to e.g. ensure consistency after a crash. See "Introduction" for more information. This is the version I plan to submit for a boost review, so if you are interested in that kind of library, please have a look at it now and don't wait for a formal review. The boost review queue has grown substantially, so it might be a long time until a formal review, which could be used for improvements based on your comments. Code will follow as soon as some issues are sorted out, especially porting the (few) platform dependent parts to windows. Regards,

4 11

Anyone care if I merge Boost.Array fixes to the release branch?
by Marshall Clow 12 Jan '10

12 Jan '10

Since no one is maintaining Boost.Array (as far as I know), and there is at least one bug fix on the trunk that hasn't been merged to the release branch, I propose merging: boost/array.hpp libs/array/doc/array.xml to the release branch. Any objections?? If not, I'll do it about midnight EST tonight. -- Marshall

2 1

[asio] Pass some parameter to io_service::run
by Dmytro Ovdiienko 11 Jan '10

11 Jan '10

Hello All, I'm new in boost asio. I'm trying to adapt this library to my existent network application. There is a problem with current application's architecture. I see no way to pass some parameter to the completion routine from the io_service::run calling thread. Below brief application description. Hope this will help: 1. It is *low-latency* network application 2. Application processes incoming multicast UDP messages. 3. Number of multicast groups/ports is *unlimited*. Technical details: 1. Application encapsulates udp::sockets subscribed to multicast group 2. udp::socket::async_receive method used to receive data 3. io_service::run is used in the thread pool to call socket's complete routine. 4. Every thread of the pool encapsulates some object for messages processing. I need to pass reference to this object to the socket::async_receive complete routine to complete message processing. This object allocates a lot of memory and I cannot create it for every socket. It is possible to create but there is no sense. It is very important to avoid thread_local_storages and null_buffers to make application as simple as possible without performance gaps. I believe it is possible because POSIX and Windows I/O allow it. Please help with advise. Thanks. Dima

3 7

[asio] streambuf example wrong
by Michael Caisse 11 Jan '10

11 Jan '10

The example section in the documentation for boost::asio::streambuf indicates you can: boost::asio::streambuf::const_buffers_type bufs = b.prepare(512); The compiler (gcc) thinks not. You can however: boost::asio::streambuf::const_buffer_type bufs( b.prepare(512) ); Regards - michael -- ---------------------------------- Michael Caisse Object Modeling Designs www.objectmodelingdesigns.com

1 0

Regression testing on win32 platforms
by Arash Partow 11 Jan '10

11 Jan '10

Hi all, With regards to regression testing of boost libraries on the win32 platform using MSVC versions post introduction of checked iterators, are any of the regression tests compiled with the following defines? _SCL_SECURE_NO_WARNINGS _SECURE_SCL=0 As I've noticed MSVC installs with SP1 that have had patches included from the last month or so are causing this failure. #include <cstddef> #include <iostream> #include <string> #include <boost/regex.hpp> int main() { using namespace boost; std::string s = "(1)(23)(456)(7890)"; regex expr("\\(.*?\\)"); sregex_iterator itr(s.begin(),s.end(),expr); sregex_iterator it_end; std::string token; std::size_t match_count = 0; while (it_end != itr) { ++itr; ++match_count; } std::cout << "Match count: " << match_count << std::endl; return 0; } ************************************************** example.exe!_crt_debugger_hook(int _Reserved=3619112) Line 65 C example.exe!_invalid_parameter(const wchar_t * pszExpression=0x00000000, const wchar_t * pszFunction=0x00000000, const wchar_t * pszFile=0x00000000, unsigned int nLine=0, unsigned int pReserved=0) Line 112 + 0x7 bytes C++ example.exe!_invalid_parameter_noinfo() Line 125 + 0xc bytes C++ example.exe!boost::match_results<std::_String_const_iterator<char,std::char_traits<char>,std::allocator<char> >,std::allocator<boost::sub_match<std::_String_const_iterator<char,std::char_traits<char>,std::allocator<char> > > > >::set_size() + 0x5b bytes C++ example.exe!boost::re_detail::perl_matcher<std::_String_const_iterator<char,std::char_traits<char>,std::allocator<char> >,std::allocator<boost::sub_match<std::_String_const_iterator<char,std::char_traits<char>,std::allocator<char> > > >,boost::regex_traits<char,boost::w32_regex_traits<char> > >::find_imp() + 0xc9 bytes C++ ************************************************** Removing SP1 or applying SP1 with patches up until mid nov '09 slip-streamed seems to work fine, this is just a heads up for people using this particular environment. Initially I was using the BOOST Pro distribution, with static linking (1.39 and 1.40) MT and single modes, I've also done clean builds just in case something was wrong with the distrib, same problem observed in all cases. I've tested the above code on linux with gcc 4.4 and intel v11 runs fine and valgrind checks-out ok (as expected). BTW I've gone here http://www.boost.org/development/testing.html looking for any kind of documentation relating to the types of settings used, is there a more up-to-date location? Kind regards, Arash

2 2