- Boost-users - lists.boost.org

Re: [Boost-users] [Thread] Beginner question regarding thread groups
by Michel Lestrade 25 Aug '08

25 Aug '08

Hi, Actually, I probably was not clear before: only the refracted ray will be a new ray on the queue: the reflected ray continues on its merry way until it decays. That is probably the same thing as what you are suggesting. The reason I was vague before is that I have two options: follow the reflected ray and spawn new refracted rays (as above) or follow the refracted ray ans spawn the reflected rays. I expect solution #1 is better in terms of locality but #2 might result in a smaller queue since internal reflection coefficients are small (i.e. you often wind up dropping the reflected ray since it no longer carries any significant power). Right now, the code is using #1 but I was considering switching to see what effects it might have... > It may be possible to do fairly well on locality by on a split, rather > than putting both new rays into the queue and doing a new fetch from > the queue, continuing on with the ray most similar to the incident > ray, only adding the other ray(s?) to the queue, so the frame of > reference only jumps when a ray dies. (Not unlike half-tailing > quicksort.) > =========================== We actually almost always develop in Fortran except for the GUI parts. That comes from our company background in physics and the large amount of existing code we have ... But yes, the code is old and mostly written in fixed-form F77. I am at least trying to modernize some parts to use free-form F90/F95 syntax :) I actually have not taken a look recently at these papers since I always worked directly on the code. Hopefully, there is something of interest in there. > Thanks, I always ask people to post links to their work since it makes things more interesting > and I'll look as soon as I can. I thought the code may be a bit old given the language Regards, Michel Lestrade Crosslight Software

2 1

Synchronization support for boost objects
by Zach Turner 25 Aug '08

25 Aug '08

I am wondering if there are any plans, either in the near or distant future, to ever add synchronization support for arbitrary objects. The interface could be simple, every class takes one additional template argument, the class name of a locking object which defaults to some class that provides stub methods so that by default all classes are unsychronized. Then, if I want my object to become synchronized, I can pass it whatever type of lock I deem appropriate. Of course any object can be synchronized currently by making a class that exposes the same interface and acquires a lock before forwarding to the internal boost implementation, but this is inefficient, as the internal details of the object's implementation are not always such that an entire method needs to be locked. Perhaps only a single line of code inside a long method needs to be locked. Has this ever been considered in the past?

2 1

[Thread] Beginner question regarding thread groups
by Michel Lestrade 25 Aug '08

25 Aug '08

Hi, I am considering rewriting part of our ray tracing code to use the Boost thread library. As many of you might know, ray tracing is a task that lends itself well to a parallel approach since each ray is independent from the other. However, the code right now is in Fortran which doesn't support task level parallelism easily ... The thread_group class seems ideally suited to my needs (one thread == one ray) but there is at least one problem I envision that I would like to solve before embarking on this fairly time-consuming rewrite. Just to give you some background, what we are doing is not ray tracing in the visualization/render sense: rather, it is a scientific application used to determine the light emission pattern of certain devices without waveguides (such as LEDs). So keep in mind that not all the conventional ray tracing approaches will work for me. I also come from a physics/engineering background so I am not up to date on all the newer object-oriented techniques and concepts. Now to the heart of the matter: in our approach, each of the initial rays is split into a reflected/refracted ray every time a material interface is reached which generates a secondary ray. Each secondary ray can generate other secondary rays when they hit an interface and so on ... This goes on until the power in each ray has decayed sufficiently (the material is usually absorbing), all rays exit the device or a maximum number of rays has been reached. That means that I do not know ahead of time how many rays (or threads) that I will have. This brings me to my question: can I add new threads to thread group while it is running ? All of the code samples for thread_group seem to call add_thread() on a fixed number of objects before calling join_all(). So it seems that the execution of the thread does not start at the moment it is added to the group but at the moment join_all() is called. In my case, I would have a join_all() first and then periodically add new threads: so this is basically a dynamic queue with new rays being pushed at the end. How can I start these threads automatically if add_thread does not start the execution ? Should I call join() on the new threads before adding them to the thread group instead of using join_all() ? Should I give up on thread_group and store my thread objects in some other container ? Any alternative solutions or a code sample with a dynamically changing thread_group would be appreciated. I am hoping someone has faced a similar problem in another context ... Regards, Michel Lestrade Crosslight Software P.S. Since I am a new user of Boost, I will go ahead and use the latest version (1.36.0). I already built the binaries for my tool set (VS 2008).

3 3

Re: [Boost-users] [Thread] Beginner question regarding thread groups
by Michel Lestrade 25 Aug '08

25 Aug '08

Hi, I actually looked at that first as well as another Intel Fortran solution called TBB (Thread Building Blocks). It seems to me that these approaches are made for instruction-level parallelism (i.e. processing a long loop with parallel_do or different but independent sections of code). I have a ton of other Fortran code that would benefit from this but ray tracing is not one of them. >From what I read so far, what I am looking for is task-level parallelism, not instruction-level, which is why I am looking into threads. > If you want some ideas for reference, I did a quick check on the intel > site and they have some papers and apparently even an openMP Fortran > compiler if that helps you, > [ obviously their comments are specific to their products but quite generally useful > esp if you are looking for ideas ] > http://cache-www.intel.com/cd/00/00/21/92/219292_hyperthreading_extract.pdf > > ( http://www.google.com/search?hl=en&q=site%3Aintel.com+openmp+performance+op… In our approach, light emission is isotropic (physically, it corresponds to spontaneous emission in a semiconductor) and we want to know what the extraction efficiency (how much light actually gets out) and what the output distribution is going to look like. This is affected by die shaping, external packaging, refractive index of cladding layers, position and reflectivity of metal contacts, TE/TM emission, etc... So the difference between nearby rays is critical, unlike a visualization/CG render situation. Just FYI but here is a quick reference to a few papers by my predecessor who wrote this Fortran code in the first place: http://scholar.google.ca/scholar?hl=en&lr=&q=author%3AShmatov+ray+tracing&b… We already have locality to a certain extent since each ray is bouncing around in its own "box" with a fixed number of facets and only interacts with certain predefined neighboring "boxes". So I am not too worried about cache at this point. I may have to change my mind later on though if this turns out to be a bottleneck. > While you say that rays are independent, if you do classical physical optics, > nearby rays tend to have similar trajectories etc. Rather than let an ignorant > but fair thread scheduler decide what piece of memory to access next, if you > are cache aware, you could even consider something like sorting the rays > to get the best locality and making them dependent with a transform > scheme that recognizes they are similar if nearby etc. Regards, Michel Lestrade Crosslight Software

2 1

Problem with <boost/test/unit_test.hpp> in precompiled header.
by Neutrino 25 Aug '08

25 Aug '08

Hi. I'd like a little guidance regarding the use of <boost/test/unit_test.hpp> please. I'm using mingw--gcc-3.4.5 with CodeBlocks 8.02. I'm dynamically linking to boost_unit_test_framework-mgw34-d-1_36.dll via boost_unit_test_framework-mgw34-d-1_36.lib. I have BOOST_TEST_DYN_LINK defined and I've worked out that I need to have #define BOOST_TEST_MODULE <whatever> defined in one source file prior to #include <boost/test/unit_test.hpp>. That's all good so far and the following example compiles and runs. ------------------------ // Main.cpp #define BOOST_TEST_MODULE Main #include <boost/test/unit_test.hpp> ------------------------ // Test1.cpp #include <boost/test/unit_test.hpp> BOOST_AUTO_TEST_CASE(test1) { BOOST_CHECK(true); } ------------------------ // Test2.cpp #include <boost/test/unit_test.hpp> BOOST_AUTO_TEST_CASE(test2) { BOOST_CHECK(true); } ------------------------ My problem arises when I try to include <boost/test/unit_test.hpp> in a precompiled header. ------------------------ // Pch.h #ifndef PCH_H_INCLUDED #define PCH_H_INCLUDED // Boost headers. #include <boost/test/unit_test.hpp> #endif // PCH_H_INCLUDED ------------------------ // Main.cpp #define BOOST_TEST_MODULE Main #include <boost/test/unit_test.hpp> ------------------------ // Test1.cpp #include "Pch.h" BOOST_AUTO_TEST_CASE(test1) { BOOST_CHECK(true); } ------------------------ // Test2.cpp #include "Pch.h" BOOST_AUTO_TEST_CASE(test2) { BOOST_CHECK(true); } ------------------------ The compiler is exiting with no errors but no output binary is produced. The build log reports: Info: resolving vtable for boost::unit_test::unit_test_log_tby linking to __imp___ZTVN5boost9unit_test15unit_test_log_tE (auto-import) obj\Debug\Mnf\Mnf.EventTest.o: In function `ZN5boost6detail17sp_counted_impl_pINS_9unit_test9ut_detail16callback0_impl_tINS3_6unusedEPFvvEEEED0Ev': F:/My Documents/Development/Libraries/boost_1_36_0/boost/detail/sp_counted_base_gcc_x86.hpp:(.bss+0x0): multiple definition of `boost::unit_test::(anonymous namespace)::unit_test_log' obj\Debug\Mnf\Mnf.BoostFunctionBindTest.o:F:/My Documents/Development/Libraries/boost_1_36_0/boost/detail/sp_counted_base_gcc_x86.hpp:(.bss+0x0): first defined here obj\Debug\Mnf\Mnf.EventTest.o: In function `ZN5boost6detail17sp_counted_impl_pINS_9unit_test9ut_detail16callback0_impl_tINS3_6unusedEPFvvEEEED0Ev': F:/My Documents/Development/Libraries/boost_1_36_0/boost/detail/sp_counted_base_gcc_x86.hpp:(.bss+0x4): multiple definition of `boost::test_tools::(anonymous namespace)::dummy_cond' obj\Debug\Mnf\Mnf.BoostFunctionBindTest.o:F:/My Documents/Development/Libraries/boost_1_36_0/boost/detail/sp_counted_base_gcc_x86.hpp:(.bss+0x4): first defined here obj\Debug\Pch.o: In function `ZN5boost9unit_test13test_observer12test_abortedEv': F:/My Documents/Development/Libraries/boost_1_36_0/boost/test/test_observer.hpp:(.bss+0x0): multiple definition of `boost::unit_test::(anonymous namespace)::unit_test_log' obj\Debug\Mnf\Mnf.BoostFunctionBindTest.o:F:/My Documents/Development/Libraries/boost_1_36_0/boost/detail/sp_counted_base_gcc_x86.hpp:(.bss+0x0): first defined here obj\Debug\Pch.o: In function `ZN5boost9unit_test13test_observer12test_abortedEv': F:/My Documents/Development/Libraries/boost_1_36_0/boost/test/test_observer.hpp:(.bss+0x4): multiple definition of `boost::test_tools::(anonymous namespace)::dummy_cond' obj\Debug\Mnf\Mnf.BoostFunctionBindTest.o:F:/My Documents/Development/Libraries/boost_1_36_0/boost/detail/sp_counted_base_gcc_x86.hpp:(.bss+0x4): first defined here I do not understand why the header file <boost/test/unit_test.hpp> objects to being included via a precompiled header file in this way. The following simple case also works fine. ------------------------ // Pch.h #ifndef PCH_H_INCLUDED #define PCH_H_INCLUDED // Boost headers. #define BOOST_TEST_MODULE Main #include <boost/test/unit_test.hpp> #endif // PCH_H_INCLUDED ------------------------ // Main.cpp #include "Pch.h" BOOST_AUTO_TEST_CASE(test1) { BOOST_CHECK(true); } ------------------------ I'll start having a look at the Boost source and see if I can work out what I'm doing wrong, but if anyone can see a problem with what I'm trying to do here I'd be grateful if they could point it out for me. Thanks Neutrino. -- View this message in context: http://www.nabble.com/Problem-with-%3Cboost-test-unit_test.hpp%3E-in-precom… Sent from the Boost - Users mailing list archive at Nabble.com.

2 1

Detecting memfun staticness
by joaquin＠tid.es 25 Aug '08

25 Aug '08

Hello, although this is not strictly a Boost-related question I wonder whether some of the utilities in Boost (enable_if, type_traits) can help me solve the problem: Say I've got the following template function template<typename T> void call_f(int x) { T t; t.f(x); } whose purpose is to invoke t.f(x) on an object of type T: struct foo { void f(int){} }; struct bar { static void f(int){}; }; ... call_f<foo>(5); call_f<bar>(5); And now I'd like to optimize call_f's implementation by avoiding the creation of the T object when it is not necessary (as in bar, whose memfun f is static). The following will do for simple cases like foo and bar above: template<typename T,typename R,typename A> void call_f_impl(int x,R(*)(A)) { T::f(x); } template<typename T,typename Q> void call_f_impl(int x,Q) { T t; t.f(x); } template<typename T> void call_f(int x) { call_f_impl<T>(x,&T::f); } but fails when f is a memfun template: struct baz { template<typename T> void f(T){} }; ... call_f<baz>(5); For instance, Comeau says: "ComeauTest.c", line 17: error: no instance of overloaded function "call_f_impl" matches the argument list The argument types that you used are: (int, <unknown-type>) call_f_impl<T>(x,&T::f); ^ detected during instantiation of "void call_f<T>(int) [with T=baz]" at line 42 1 error detected in the compilation of "ComeauTest.c". Any idea about how to solve this in a more comprehensive way? Thank you in advance, Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

3 2

[spirit] string_to_list_of_tuples (tr1 version)
by Markus Werle 25 Aug '08

25 Aug '08

> [gmane anti-top-poster-complaint-header] Hi! [OK, Joel, it's not as trivial as I thought. Show me the way, master] Below is my first toy implementation of string_to_list_of_tuples for tr1::tuple. Unfortunately this code only parses the text, but does not store the data. Any idea how to add the correct semantic actions in a pure and elegant way? Markus ---snip--- #define BOOST_SPIRIT_DEBUG_OUT std::cerr #define BOOST_SPIRIT_DEBUG // #include "Enforce.h" #include <iostream> #include <iterator> #include <boost/tr1/tuple.hpp> #include <boost/mpl/assert.hpp> #include <boost/mpl/at.hpp> #include <boost/spirit/include/classic_spirit.hpp> #include <list> #include <string> namespace sp = boost::spirit::classic; namespace fusion = boost::fusion; namespace mpl = boost::mpl; namespace { struct parser_traits { template <typename T, int Dummy = 0> struct parser; #define PARSER_TRAITS(TYPE, PARSER_T) \ template <int Dummy> struct parser<TYPE, Dummy> \ { \ typedef PARSER_T type; \ \ static inline \ PARSER_T p() \ { \ PARSER_T result; \ return result; \ } \ } PARSER_TRAITS(int, sp::int_parser<int>); PARSER_TRAITS(unsigned int, sp::uint_parser<unsigned>); typedef sp::real_parser <double, sp::ureal_parser_policies<double> > real_parser_t; PARSER_TRAITS(float, real_parser_t); PARSER_TRAITS(double, real_parser_t); template <int Dummy> struct parser<std::string, Dummy> { typedef sp::contiguous <sp::confix_parser <sp::strlit<const char*>, sp::kleene_star <sp::anychar_parser>, sp::strlit<const char*>, sp::unary_parser_category, sp::non_nested, sp::non_lexeme> > parser_t; static inline parser_t p() { using namespace phoenix; parser_t result = sp::lexeme_d[sp::confix_p("\"" , (*sp::anychar_p) , "\"")]; return result; } static inline sp::strlit<char const *> comma() { sp::strlit<char const *> str_p(","); return str_p; } }; }; template <typename Tuple, long I> struct rule_generator_helper { typedef typename std::tr1::tuple_element<I, Tuple>::type element_type; typedef typename parser_traits::parser<element_type> pt_type; typedef typename rule_generator_helper<Tuple, (I-1)>::type element_1_t; typedef typename sp::sequence<sp::sequence<element_1_t, sp::chlit<char> >, typename pt_type::type> type; static type apply() { return rule_generator_helper<Tuple, (I-1)>::apply() >> sp::ch_p(',') >> pt_type::p() ; } }; template <typename Tuple> struct rule_generator_helper<Tuple, 0L> { typedef typename std::tr1::tuple_element<0, Tuple>::type element_type; typedef typename parser_traits::parser<element_type> pt_type; typedef typename pt_type::type type; static type apply() { return pt_type::p(); } }; template <typename Tuple> struct rule_generator { typedef std::tr1::tuple_size<Tuple> tuple_size; BOOST_MPL_ASSERT_RELATION(tuple_size::value, >, 0); typedef rule_generator_helper<Tuple, (tuple_size::value - 1)> helper_t; static inline typename helper_t::type apply() { return helper_t::apply(); } }; template <typename Tuple> struct string_to_list_of_tuples { static std::list<Tuple> apply(std::string const & s) { std::list<Tuple> result; typedef sp::rule<sp::phrase_scanner_t> rule_t; rule_t rule_tuple_rep = '(' >> rule_generator<Tuple>::apply() >> ')'; rule_t rule_tuple_list = *(rule_tuple_rep) >> sp::end_p // for trailing whitespace ; BOOST_SPIRIT_DEBUG_RULE(rule_tuple_rep); BOOST_SPIRIT_DEBUG_RULE(rule_tuple_list); sp::parse_info<> info; try { info = sp::parse(s.c_str(), rule_tuple_list, sp::space_p); } catch (std::exception const & e) { std::cerr << e.what() << std::endl; } // ENFORCE(info.full) // ("Failed to generate a list of tuples from ") // ("string representation\n'") // (s)("'\n"); return result; } }; } // namespace int main() { try { { typedef std::tr1::tuple<int, int> tuple_t; std::list<tuple_t> test = string_to_list_of_tuples<tuple_t>::apply(" ( 1, 2 ) ( 3, 4) (5,6)"); std::copy(test.begin(), test.end(), std::ostream_iterator<tuple_t>(std::cout, "\n")); } { typedef std::tr1::tuple<unsigned int, int, double> tuple_t; std::list<tuple_t> test = string_to_list_of_tuples<tuple_t>::apply (" ( 1, -2 , 5.123) ( 3, 4,7.9)(5,6,8.6789)"); std::copy(test.begin(), test.end(), std::ostream_iterator<tuple_t>(std::cout, "\n")); } } catch (std::exception const & e) { std::cerr << e.what() << std::endl; } }

1 0

[mpl|fusion|tr1]mpl::at_c<std::tr1::tuple<...>, I>::type fails
by Markus Werle 25 Aug '08

25 Aug '08

Hi! I am using boost_1_36_0 manually patched with the files mentioned in http://thread.gmane.org/gmane.comp.lib.boost.user/39093/focus=39106 I thought that std::tr1::tuple was a valid mpl sequence, but this code here fails. Is this by design? ---snip--- #include <boost/tr1/tuple.hpp> #include <boost/mpl/assert.hpp> #include <boost/mpl/at.hpp> template <typename Tuple, long I> struct show_problem { typedef typename boost::mpl::at_c<Tuple, I>::type type; }; int main() { typedef std::tr1::tuple<int, int> tuple_t; typedef boost::mpl::at_c<tuple_t, 1L>::type Ouch; }

2 2

[tr1] maximum number of allowed tuple elements?
by Markus Werle 25 Aug '08

25 Aug '08

Hi! Maybe I overlooked this portion of the docs, so could you please give me a hint, whether and how one can set the maximum number of tuple elements. regards, Markus

2 3

[Test] newbie linking problems
by Thijs Koerselman 25 Aug '08

25 Aug '08

Hi, I'm trying to integrate the unit test framework in a library I'm developing but I'm really stuck at something. I suspect its something simple, but I've tried everything I could think of and can't seem to figure it out. My library is loaded as a 3rd party plugin in a host application. On windows it's a dll on OSX its a loadable carbon bundle. I'm currently only developing on OSX. When the bundle is loaded, its main function registers the methods that make up the public interface. So I trigger a method from the host which starts the unit testing. It's something like this #define BOOST_TEST_NO_MAIN #undef nil // because of Carbon nil definition #include <boost/test/included/unit_test.hpp> using namespace boost::unit_test; /* UnitTest class declaration and other member functions */ void UnitTest::run() { BOOST_GLOBAL_FIXTURE( MyConfig ); // setting log output stream etc. unit_test_main( &init_mylib_test_suite, 0, 0 ); } test_suite* init_mylib_test_suite( int argc, char* argv[] ) { std::string header = "Testing MyLib "; header += MYLIB_VERSION_STRING; std::string rule = std::string(header.length(),'='); BOOST_MESSAGE(rule); BOOST_MESSAGE(header); BOOST_MESSAGE(rule); return 0; } Then I'm trying to organize it so that I have different unit test suits for each logical module within the library. I'm separating the test suits into each its own compilation unit (object file output), because I don't want to recompile all the tests if I change something in one of them. So one testing module would have a source file with something like: #define BOOST_TEST_MODULE Example #define BOOST_TEST_NO_MAIN #undef nil // because of Carbon nil definition #include <boost/test/included/unit_test.hpp> using namespace boost::unit_test; BOOST_AUTO_TEST_SUITE( "MyModuleSuite" ); BOOST_AUTO_TEST_CASE( test_case1 ) { BOOST_CHECK( true ); } BOOST_AUTO_TEST_CASE( test_case2 ) { BOOST_CHECK( false ); } BOOST_AUTO_TEST_SUITE_END() With this approach I keep getting linker errors about a duplicate symbol between these 2 object files that result from it : "duplicate symbol boost::unit_test::traverse_test_tree(boost::unit_test::test_case const&, boost::unit_test::test_tree_visitor&)in ......" I've tried linking the framework as a static and a dynamic library. I've also tried including the framework by using the "BOOST_TEST_NO_LIB" flag. All without luck. Am I approaching this the wrong way? Am I missing something obvious? I would really appreciate some help with this. Tips on how to best organize the unit testing in my library are also very welcome. Note that I'm testing the internals of my library, not just the public API, so linking against the library from a simple command line app won't work. The library is also dependant on the host API. In order for those parts to work/initialize properly I'm forced to launch my plugin inside the host at all time. Cheers, Thijs

2 2