September 2010 - Boost-users

Boost & Pthreads
by ef 13 Sep '10

13 Sep '10

Hello, I want to build a bunch of different graphs at once,and it looks like the best way to do this is with pthreads. I need to be able to communicate information between the graphs (sadly in real time). Is Boost pthread friendly? I am just thinking of calling read_graphviz per thread, and when a vertex is waiting for data from another graph it just sits and waits. Communication should be easy and done by shared memory. I dont see any problems with this. I am just curious before I take this path that Boost is pthread friendly, or atleast in the way that I described? Thanks, EF

2 1

Re: [Boost-users] MSM & serialization
by Christophe Henry 13 Sep '10

13 Sep '10

>>> I'm planing to use the new MetaStateMachine library for a game. But i >>> want to create savegames with boost::serialization so i would need to >>> serialize the state machine. >>> There shouln't be too much to safe (the current state basically), but I >>> can't see how I can do this. >>> So, how can I do this? >> >> Hi, >> >> For the moment, unfortunately not at all, though it is something I >> want to do since a while and pretty high on my internal priority >> queue. >> How urgently do you need it? >> In the worst case, it would be easy to temporarily add a setter for >> the current state (we have a getter). >> >> Regards, >> Christophe > > Hi, > for the current project i just need 2 small state machines, so i can > hardcode them. But in the future it would really like to use MSM, at > thats impossible for me as long as i can't safe them. > So it's not very urgent, but if i could use it in 2 month, it would be > great. Hi, No, it won't be said so easily that MSM forces its users to hardcode state machines ;-) I just put something together and committed (without guarantee ;-) ) into the trunk. I also added into libs/msm/doc/HTML an example (Serialize.cpp) where you can serialize a state machine and even optionally states themselves (using a do_serialize typedef). It's not tested thoroughly in the little time I spent on it tonight and there are a few things I need to think about more but I hope it'll help you. Please keep me informed about your experience with this brand new feature ;-) Regards, Christophe

1 0

MSM & serialization
by Christophe Henry 13 Sep '10

13 Sep '10

> I'm planing to use the new MetaStateMachine library for a game. But i > want to create savegames with boost::serialization so i would need to > serialize the state machine. > There shouln't be too much to safe (the current state basically), but I > can't see how I can do this. > So, how can I do this? Hi, For the moment, unfortunately not at all, though it is something I want to do since a while and pretty high on my internal priority queue. How urgently do you need it? In the worst case, it would be easy to temporarily add a setter for the current state (we have a getter). Regards, Christophe

2 1

[Interprocess] Compile error when simply including managed_mapped_file.hpp in 1.44 using VC10
by Phil Ratzloff 13 Sep '10

13 Sep '10

Simply including managed_mapped_file.hpp in an empty project will generate an error in boost 1.44 using VC 2010. This doesn't occur in 1.43 and is a showstopper to me. Any ideas to work around this? I created a Console App using the New Project wizard. Here's the whole program: #include "stdafx.h" #include <boost/interprocess/managed_mapped_file.hpp> int _tmain(int argc, _TCHAR* argv[]) { return 0; } The following output log is generated: 1>------ Build started: Project: boost144Err, Configuration: Debug Win32 ------ 1> boost144Err.cpp 1>c:\lib\boost_1_44_0\boost\interprocess\detail\move.hpp(342): error C2440: 'return' : cannot convert from 'boost::interprocess::mapped_region' to 'boost::interprocess::mapped_region &&' 1> You cannot bind an lvalue to an rvalue reference 1> c:\lib\boost_1_44_0\boost\interprocess\mapped_region.hpp(159) : see reference to function template instantiation 'boost::interprocess::mapped_region &&boost::interprocess::move<boost::interprocess::mapped_region&>(T)' being compiled 1> with 1> [ 1> T=boost::interprocess::mapped_region & 1> ] 1>c:\lib\boost_1_44_0\boost\interprocess\detail\move.hpp(342): error C2440: 'return' : cannot convert from 'boost::interprocess::detail::file_wrapper' to 'boost::interprocess::detail::file_wrapper &&' 1> You cannot bind an lvalue to an rvalue reference 1> c:\lib\boost_1_44_0\boost\interprocess\detail\file_wrapper.hpp(63) : see reference to function template instantiation 'boost::interprocess::detail::file_wrapper &&boost::interprocess::move<boost::interprocess::detail::file_wrapper&>(T)' being compiled 1> with 1> [ 1> T=boost::interprocess::detail::file_wrapper & 1> ] 1>c:\lib\boost_1_44_0\boost\interprocess\detail\move.hpp(342): error C2440: 'return' : cannot convert from 'boost::interprocess::file_mapping' to 'boost::interprocess::file_mapping &&' 1> You cannot bind an lvalue to an rvalue reference 1> c:\lib\boost_1_44_0\boost\interprocess\file_mapping.hpp(62) : see reference to function template instantiation 'boost::interprocess::file_mapping &&boost::interprocess::move<boost::interprocess::file_mapping&>(T)' being compiled 1> with 1> [ 1> T=boost::interprocess::file_mapping & 1> ] ========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========

2 2

[fusion] Non-default initialisation of sequence components.
by Hugh Dickinson 13 Sep '10

13 Sep '10

Hi Everyone, I'm wondering if there is a (built-in) way in boost::fusion to instantiate a sequence of heterogeneous elements while passing an identical set of arguments to the constructor of each element? It seems like the only options are to default construct the elements or to supply references to previously constructed elements. As a semi-concrete example, is there a way to achieve the equivalent of the following? template <int Label> struct TestType { TestType(int i, double d, string s):I(i), D(d), S(s){} int I; double D; string S; }; int main(int argc, char* argv[]){ typedef vector<int, double, string> ArgVector; typedef vector<Type<1>, Type<2>, Type<3> > TypeVector; ArgVector args(1, 2.5, "hello"); // Want to pass the elements of Args as constructor arguments to each element of vec. TypeVector vec(args); } Any help much appreciated, Hugh

2 1

[Unordered] No operator[] const?
by Hossein Haeri 13 Sep '10

13 Sep '10

Dear all, Is there any specific reason why there is no operator[] const for boost::unordered_map? Or, is that only an oversight? TIA, --Hossein

3 3

MSM & serialization
by Jonathan 13 Sep '10

13 Sep '10

I'm planing to use the new MetaStateMachine library for a game. But i want to create savegames with boost::serialization so i would need to serialize the state machine. There shouln't be too much to safe (the current state basically), but I can't see how I can do this. So, how can I do this? Greetings Jona

1 0

g++ compilation is slow with Boost's constructs
by Ilya Murav'jov 13 Sep '10

13 Sep '10

Hello, I've been using several Boost libraries since 1.33 and now I have a situation that it takes over 4 seconds to compile for some files <= 40Kb (not so big!), with gcc. I've upgraded my machine recently (AMD X3 425, 2700MHz, 2GB memory) and hoped that things would change but it is still somewhat a blocker to developing. So I began to investigate and found out each invocation of known Boost constructs has its non-zero cost. I chose a source file(*), commented out all and then incrementally uncommented functions from beginning, compiled and measured times. Here is my results (compile time, not run time!): - BOOST_FOREACH costs 0.1 sec - boost::function<> invocation costs 0.05 sec - boost::lambda::bind() with placeholders costs 0.1 sec - Boost.Range adaptor transformed() costs ~0.1 too As you can see, 10 boost::lambda::bind() slow down your compilation for whole 1 second! All that is measured against gcc 4.4, Ubuntu 9.10, with the most light options I picked up: "g++ -g -O0 -pipe" (I need -g for debugging anyway). Also I found out that gcc 4.2 is the quickest from 4.1, 4.2, 4.3, 4.4 version of gcc, and gcc 4.4 is the most slow one of them (~28% slower than gcc 4.2). :( Therefore, I have a few questions. 1) Are there more means/tricks for faster compilation with g++ I miss? (what a quickest gcc version?) 2) What compiler is known to be much more speedy for now, if any (icc, clang, msvc)? 3) Did someone try to investigate (and maybe fix) that gcc bottleneck(s)? Regards, Ilya Murav'jov (*) I did measures on a source file select.cpp from my project, Bombono DVD (http://gitorious.org/bombono-dvd/bombono-dvd/blobs/master/src/mgui/editor/s…)

6 17

how could I output the data of matrix(ublas) to txt file faster?
by Samurai Deeper 13 Sep '10

13 Sep '10

Example /***********begin**************/ void testMatrixRow() { matrix<double> m(1920, 1080); clock_t begin,end; begin = clock(); for(unsigned int i = 0; i < m.size1(); ++i) for(unsigned int j = 0; j < m.size2(); j++) { m.insert_element(i, j, i*3 + j); } end = clock(); cout<<"time of inserting data : "<<(double)(end-begin)/CLOCKS_PER_SEC<<" "<<endl; ofstream ob1("testing.txt",ios::out,ios::binary); begin = clock(); for (unsigned i = 0; i < m.size1 (); ++ i) { for(unsigned int j = 0; j < m.size2(); j++) ob1<<m(i, j); } end = clock(); cout<<"time of regular output : "<<(double)(end-begin)/CLOCKS_PER_SEC<<" "<<endl; begin = clock(); for (unsigned i = 0; i < m.size1 (); ++ i) { matrix_row<matrix<double> > mr(m, i); ob1.write((const char*)&mr(0), mr.size()); ob1<<endl; } end = clock(); cout<<"time of nonregular output : "<<(double)(end-begin)/CLOCKS_PER_SEC<<" "<<endl; ob1.close(); } /***********end**************/ As expected,ob1.write is much more faster than ob1<< But the results are not I want(ob1<< is) Do I have any way to solve this problem? Thanks a lot -- View this message in context: http://old.nabble.com/how-could-I-output-the-data-of-matrix%28ublas%29-to-t… Sent from the Boost - Users mailing list archive at Nabble.com.

1 0

follow to "[regex] Perl regex special characters"
by DBW 13 Sep '10

13 Sep '10

I found a thread from 2006-09-08 titled '[regex] Perl regex special characters' about whether ']' and '}' should be included in the documentation as characters that must be escaped. After spending quite some time today to discover that my '}' character did in fact need to be escaped, my opinion is that the documentation should be modified. Thanks, Brad

2 1