October 2006 - Boost-users - lists.boost.org

graph-tool and boost/random
by Jens Müller 05 Oct '06

05 Oct '06

I cannot compile graph-tool on amd64/gcc4.1.1/Gentoo and Boost 1.33.1, while Tiago can with the same gcc. Maybe someone can take a look? See http://news.gmane.org/gmane.comp.science.graph.graph-tool.general/cutoff=9 ...

1 1

Problem with boost serialization with an existing 'struct load'
by Jeff Epler 05 Oct '06

05 Oct '06

Please send me a copy of any replies to this message. The following code compiles with boost_1_31_0 but fails with 1_32_0 and 1_33_1 (tested with gcc 3.2.2 and gcc 4.0.3 on Linux systems): // ------------------------------------------------------------------- #include <boost/archive/xml_iarchive.hpp> #include <boost/serialization/vector.hpp> #include <boost/serialization/split_member.hpp> #include <iostream> struct K { friend class boost::serialization::access; template<class Archive> void save(Archive &ar, const unsigned int version) const {} template<class Archive> void load(Archive &ar, const unsigned int version) {} BOOST_SERIALIZATION_SPLIT_MEMBER() }; struct load { int x; }; int main(void) { std::vector<K> v; boost::archive::xml_iarchive ia( std::cin ); ia & boost::serialization::make_nvp("v", v); return 0; } // ------------------------------------------------------------------- We would like to move our application from boost 1_31 but this problem with 'struct load' has stopped us so far. Is this a bug in BOOST serialization in 1_32 and 1_33 that will be corrected in the future, or do we have no choice but to rename our 'struct load'? If this is the correct behavior of the serialization library, is it noted anywhere in the documentation? I was unable to find such a note. Jeff

2 4

Bug? lambda+function composition not working together
by Johan Torp 05 Oct '06

05 Oct '06

Operator , (nullary composition) does not seem work between boost::function and boost::lambda. Is this a bug? Is there any way to work around this? void foo() { std::cout << "foo"; } boost::function<void()> f(boost::lambda::bind(&foo)); (f,f)(); // Outputs foo (boost::lambda::bind(&foo),f)(); // Outputs foo (f,boost::lambda::bind(&foo))(); // Outputs foo (boost::lambda::bind(&foo), boost::lambda::bind(&foo))(); // Outputs foofoo I am using boost 1.32 and gcc 3.4.4. Best Regards, Johan Torp -- System Developer TouchTable AB Tel: +46 (0)31 773 68 12 johan.torp(a)touchtable.se

2 1

Boost string allocator
by Thompson, Ian 05 Oct '06

05 Oct '06

Hello, I have recently used the fast_pool_allocator as the allocator for std::map with excellent results wrt performance. My application also uses a lot of strings (basic_string) and I was seeking some advice on changing the basic_string allocator to a "boost allocator" . Can anyone offer some advice in this regard? Thanks in advance, Ian Commonwealth Bank Ian Thompson Developer Risk Systems Project Global Markets Risk Management ************** IMPORTANT MESSAGE ************** This e-mail message is intended only for the addressee(s) and contains information which may be confidential. If you are not the intended recipient please advise the sender by return email, do not use or disclose the contents, and delete the message and any attachments from your system. Unless specifically indicated, this email does not constitute formal advice or commitment by the sender or the Commonwealth Bank of Australia (ABN 48 123 123 124) or its subsidiaries. We can be contacted through our web site: commbank.com.au. If you no longer wish to receive commercial electronic messages from us, please reply to this e-mail by typing Unsubscribe in the subject line. ***************************************************************

2 1

Using ofstream with xml_oarchive to write a file larger than 2GB
by james.jones＠firstinvestors.com 04 Oct '06

04 Oct '06

From: Thomas D'Silva <twdsilva(a)gmail.com> > Hello, > > I am trying to serialize data using boost::archive::xml_oarchive. When the > resulting file is larger than 2GB, I get an error saying "File size limit > exceeded". I am using gcc version 3.2.2. I was wondering if there is a way to > use ofstream with boost::archive::xml_oarchive to write files larger than 2GB > using gcc version 3.2.2. Does the underlying filesystem support >2GB files? Many don't by default. Can you create a large file (using dd, for example) in the shell? (I'm assuming you're using some flavor of Unix here...) - James Jones Administrative Data Mgmt. Webmaster 375 Raritan Center Pkwy, Suite A Data Architect Edison, NJ 08837

1 0

[Smart Pointers] Casts for shared_array
by Christian Henning 04 Oct '06

04 Oct '06

Hi there, don't know if this ever came up. I could not find anything in the archives. I would like to cast a shared_array<POD> into shared_array<unsigned char>. Since there is no operator for that I wonder what the reason is? Thanks, Christian

1 0

Using ofstream with xml_oarchive to write a file larger than 2GB
by Thomas D'Silva 04 Oct '06

04 Oct '06

Hello, I am trying to serialize data using boost::archive::xml_oarchive. When the resulting file is larger than 2GB, I get an error saying "File size limit exceeded". I am using gcc version 3.2.2. I was wondering if there is a way to use ofstream with boost::archive::xml_oarchive to write files larger than 2GB using gcc version 3.2.2. Thank you, Thomas D'Silva

4 3

[Boost][1_33_1] Assertion during build stage: Solaris 10
by Prashant Thakre 04 Oct '06

04 Oct '06

Hi, I have encountered following error during the build process for [Boost].[1_33_1] on Solaris 10. bjam install "-sTOOLS=sunpro" "-sBUILD=release <runtime-link>dynamic/static <threading>multi" --without-python Jamfile sunpro-C++-action /export/home/user/tmp/bin/boost/libs/regex/build/ libboost_regex.so/sunpro/release/cregex.o Assertion: (../lnk/init.cc, line 1032) while processing /db2/sunstudio11/SUNWspro/prod/include/CC/Cstd/rw/collate.cc at l ine 0. ...failed sunpro-C++-action /export/home/user/tmp/bin/boost/libs/regex/build/ libboost_regex.so/sunpro/release/cregex.o... CC: Sun C++ 5.8 2005/10/13 SunOS <hostname> 5.10 Generic_118833-03 sun4u sparc SUNW,Sun-Fire-280R Thanks. regards, Prashant Thakre

3 3

Boost in PS3
by Delfin Rojas 03 Oct '06

03 Oct '06

Hello, Does anybody know of a program for the PS3 platform that uses boost, or if boost would compile in that platform "as is"? Thanks, -delfin

2 1

[BGL] Boost poor performance (against LEDA)
by Leandro Melo 03 Oct '06

03 Oct '06

Hi! I'm making some performance tests (nothing too fancy) between LEDA and Boost graphs. I'm not testing algorithms, but some basic adjacency list operations (add_vertex, add_edge, remove_vertex and remove_edge). I will not post the whole code I used here, but if anyone is interested, I can send the files. It's very surprising (taking into account the Boost claims to be a high performance library) that LEDA did better in all tests I made. Actually, in most tests LEDA was more than one order of magnitude better than Boost. For example: I generated a random (self-loops and parallel edges) directed graph with 10000 vertices and 20000 edges. The generation time for LEDA was 0.025 seconds in average and the generation time for Boost was 0.500 seconds in the average. I know the template arguments I parameterized the adjacency list are important. In this case I used the following definition (I also switched the listS for vecS, but the results were pretty much the same): typedef adjacency_list<listS, listS, directedS> Graph; I also tested remove_edge and remove_vertex operations with various template parameters. But in all situations LEDA did better than Boost. In some cases, LEDA did much better (more that one oder of magnitude again). The reason of this post is not to start some kind of arguments here (LEDA X Bosst), but to try to clarify some questions. Naturally, I'd be convenient if someone has already done such things I just did (actually, this is what I'm looking for). Anyway, I'm trying to find out: - Has anyone come up with results similar to mine's. In other words, if anyone has tested LEDA against Boost. Has LEDA done better (like in my case)? If so, how better? - Does anyone happen to know some reasons that make LEDA so efficient (assuming my results are correct in a general way)? And it's not even made using generic programming techniques (which provide some optimizations at compile time). - I made tests with versions 4.1 and 5.0 of LEDA. I believe that when the GGCL ('old' Boost) was released, LEDA was in a version close to 4. And even in LEDA's 4.1 version, the results were pretty much the same (better than Boost). However, the Boost Graph Library book presents Boost as a flexible and high performace library as 'better options' than LEDA (among others like GTL and Stanford Graph). I agree with the 'flexible' part of the statement, but according to my results that 'high performance' part is not realistic. One important thing to note is the following. As I mentioned, I did switch some template parameters of the adjacency list (basically, the edges and vertices container for both directed and undirected graphs). However, the tests I'm working on are supposed to be tests for general situations in which you might have remove_edges, remove_vertices and even other operations. Thus, it'd be more realistic to use Boost's adjacency list with template arguments that together represent the 'best' or a 'good' relationship between 'cost and benefit'. Because I'm trying to simulate a situation in which I'd need all operations with the same 'importance' (add_vertex, add_edge, remove_vertex, remove_edge, etc). Well, I'd appreciate if anyone could comment on this post. Thank you for reading. Bellow are the some of the functions I used (assume Graph is defined and all_vertices and all_edges are vectors of defined vertex and edge descriptors). void generate_random_graph(Graph& g) { for (int i = 0; i < num_v; i++) all_vertices[i] = add_vertex(g); for (int i = 0; i < num_e; i++) { vertex_desc s = all_vertices[rand() % num_v]; vertex_desc t = all_vertices[rand() % num_v]; all_edges[i] = add_edge(s, t, g).first; } } void remove_edges(Graph& g) { //I didn't test the other version of remove_edge(s, t, g) because LEDA doesn't have a similar one. for (int i = 0; i < 1000; i++) { edge_desc edge = all_edges[i]; remove_edge(edge, g); } } void remove_vertices(Graph& g) { for (int i = 0; i < 1000; i++) { vertex_desc vertex = vertices[i]; clear_vertex(vertex, g); remove_vertex(vertex, g); } } -- Leandro Terra C. Melo

5 11