July 2004 - Boost-users - lists.boost.org

Sun Solaris Forte 8 and boost_1_31_0
by Tom Aratari 02 Jul '04

02 Jul '04

Hello, Has anyone gotten boost_1_31 to compile with Forte 8.0 (__SUNPRO_CC = x550) ? I get errors for the boost::iterator. Also, does anyone know if the next release of boost will be compatible with forte 8.0? Thanks Tom

5 4

Boost Test Tutorial
by Sebastian Stein 02 Jul '04

02 Jul '04

I have written a small tutorial how to use boost test within the KDE project. If you are interested, you will find it under: http://www.hpfsc.de/boosttest/ Sebastian -- http://www.hpfsc.de/ - die Seite rund um: Assembler, Bundeswehr, TFT LCDs, Halle/Saale, Fahrradtouren, Neuseeland, Wanderstaat Mauma, Raumschiff USS Nathan, Enemy Room, MLCAD Tutorial

1 0

standard way to specify undefined values
by Erik Thiele 02 Jul '04

02 Jul '04

hi, i have a userdefined type "foobar". it has a special value that means "undefined". is there a standard way to specify which value is undefined? i do not want to use boost::optional here, as foobar itself can be "undefined". you know: class foobar { static const foobar undefined(blablabla); int a,b,c; } but is this "standard"? cya erik -- Erik Thiele

1 0

RE: [Boost-users] Re: [mpl] Alexandrescu's hierarchy generators
by Klaus Nowikow 02 Jul '04

02 Jul '04

David Abrahams wrote: > [..] > template<class T, class Base> > struct node; > > typedef mpl::vector<int, char, long> types; > > // node<int, node<char, node<long, mpl::empty_base> > > > typedef mpl::inherit_linearly< > types > , node<_2,_1> > >::type lin; > [...] > HTH. That helps a lot. Thank you! -- Klaus Nowikow

1 0

iterator_adaptor problem (mpl?)
by Angus Leeming 01 Jul '04

01 Jul '04

Consider a container such as std::vector<std::pair<int, std::string> > my_container; I want to be able to iterate over the ints of this container so, following the tutorial to the iterator_adaptor docs, came up with template <typename BaseIterator> class first_iter : public boost::iterator_adaptor< first_iter<BaseIterator>, BaseIterator, boost::use_default, boost::use_default, typename BaseIterator::value_type::first_type & > { // snip body }; This works fine so long as BaseIterator::value_type is not const, but falls over when it is because I've defined the 'super_t::reference' type to be mutable. I guess that I need a piece of metafunction magic, but my naive implementation fails typename boost::mpl::apply_if< boost::is_const<typename BaseIterator::value_type> , typename BaseIterator::value_type::first_type const & , typename BaseIterator::value_type::first_type & >::type trial.C:38: error: no type named `type' in `struct boost::mpl::apply_if< boost::is_const<std::pair<int, std::string> >, const int&, int&>' I must admit I'm a bit baffled. There plainly is a 'type' in the apply_if class template. Any pointers? I attach the code as I have it. While I'm at it, is there not something in the MPL that will return the appropriate const-qualified reference for me? None of the header file names in boost/mpl jump out and grab me, but I may well be looking in the wrong place. Regards, Angus

2 4

RE: [Boost-users] [pool] Speed of pool_allocator<>
by Klaus Nowikow 01 Jul '04

01 Jul '04

Peter Dimov wrote: > Klaus Nowikow wrote: > > > > I changed the test program (see below). Now I use object_pool<> > > to create and destroy a number of elements (I use a vector<C*> to > > store them in the meantime). The results are even worse than before: > > > > new/delete: 0.08s > > object_pool: 25.20s > > [...] > > > template <template <class, class> class Pool> > > double Test() > > { > > Pool<C> P(CHUNK_SIZE); > > boost::timer Timer; > > > > std::vector<C*> V; > > for(int i = 0; i < NUM_ELEMENTS; ++i) > > { > > V.push_back(P.construct()); > > } > > for(std::vector<C*>::iterator pos = V.begin(), end = V.end(); pos != > > end; ++pos) { > > P.destroy(*pos); > > } > > > > return Timer.elapsed(); > > } > > Your test is still very unrealistic. You create a large > number of objects in > one batch, then destroy them afterwards. A more realistic > scenario would > have creation and destruction interleaved. > > The most accurate test would be to try Boost.Pool with your actual > application and see how it affects its real-world > performance. It's still > possible for it to be slower, of course, but at least you'd > measure the real > thing. Of course you're right. Obviously I broke the "don't optimize what does not have to be optimized" rule ;-). When creating and deleting 100000 Objects with new/delete takes only 0.1 seconds I won't have to look for something more efficient. I will do what you suggest: write my code and see if it needs optimization. Regards, Klaus

1 0

[mpl] Alexandrescu's hierarchy generators
by Klaus Nowikow 01 Jul '04

01 Jul '04

Greetings! Does boost.mpl provide type hierarchy generators like GenScatterHierarchy and GenLinearHierarchy as described in "Modern C++ Design"? I must confess I do not understand the mpl completely yet ;-) Best regards, Klaus

2 1

RE: [Boost-users] [pool] Speed of pool_allocator<>
by Klaus Nowikow 01 Jul '04

01 Jul '04

Peter Dimov wrote: > Klaus Nowikow wrote: > > I am in the process of writing a program which has to > > handle large numbers (~200000 - 400000) objects of the same class. > > These objects will be created and deleted frequently, so I thought > > some kind of specialized memory management would be handy. > > > > I wrote a test program (see below) to check what boost.pool > would gain > > me. The results make me think that I have done something wrong: > > > > [...] > > > > i.e., std::allocator was more than 3 times faster than > > boost::pool_allocator (Win2k, 256MB). Any ideas what i have done > > wrong? > > [...] > > > template <class Allocator> > > double Test() > > { > > std::vector<Allocator::value_type, Allocator> Vector; > > boost::timer Timer; > > > > for(int i = 0; i < NUM_ELEMENTS; ++i) > > { > > Vector.push_back(i); > > } > > > > return Timer.elapsed(); > > } > > If I read the code correctly, you don't create or delete > objects. You create > and delete (large) arrays of objects. Hmm. I'll have to think about this. Maybe the test was a bit over-simplified. The real program would have a vector<C*> where C is my class. I'll change the example (using object_pool) and post the results.

1 0

[pool] Speed of pool_allocator<>
by Klaus Nowikow 01 Jul '04

01 Jul '04

Hi! I am in the process of writing a program which has to handle large numbers (~200000 - 400000) objects of the same class. These objects will be created and deleted frequently, so I thought some kind of specialized memory management would be handy. I wrote a test program (see below) to check what boost.pool would gain me. The results make me think that I have done something wrong: debug mode: std::allocator 2.17s boost::pool_allocator 7.18s release mode: std::allocator 0.11s boost::pool_allocator 0.36s i.e., std::allocator was more than 3 times faster than boost::pool_allocator (Win2k, 256MB). Any ideas what i have done wrong? See the test program below. // MemoryPoolTest.cpp // Boost includes ------------------------------------------------------------- #include <boost/timer.hpp> #include <boost/pool/pool_alloc.hpp> // Std includes --------------------------------------------------------------- #include <vector> #include <iostream> enum { NUM_ELEMENTS = 3000000 }; template <class Allocator> double Test() { std::vector<Allocator::value_type, Allocator> Vector; boost::timer Timer; for(int i = 0; i < NUM_ELEMENTS; ++i) { Vector.push_back(i); } return Timer.elapsed(); } int main(int argc, char* argv[]) { std::cout << "Starting memory pool test" << std::endl; double Time; std::cout << "Allocating with boost::pool_allocator<>: "; Time = Test<boost::pool_allocator<int> >(); std::cout << Time << "s" << std::endl; std::cout << "Allocating with std::allocator<> : "; Time = Test<std::allocator<int> >(); std::cout << Time << "s" << std::endl; return 0; }

2 1