- Boost - lists.boost.org

personal samsung
by Fernando Cacciola 02 Apr '09

02 Apr '09

11 6 100 6359

1 1

[gsoc] Proposal - Boost.Matrix - DRAFT RFC
by Kornel Kisielewicz 01 Apr '09

01 Apr '09

Hello all, Following the discussion in the threads: "[boost][gsoc] Interest check for 3d geometry proposal" "[boost] [gsoc] fixed size matrix class?" I post a draft of my proposal with a request for comments. Following the suggestions I decided to scale down the proposal to a polished and production-quality fixed-size Boost.Matrix library, with a big emphasis on correctness and performance. The application follows: Abstract ========= uBLAS is Boost's general linear algebra library, and it is a fine library indeed. However, in some cases it's design decisions, while quite rational for the general use case, are restrictive. This proposal is about a new library that would address some of these issues in the specific situation of vector and matrix operations on fixed sized structures. Suggested is a new library boost::matrix. The library would introduce a single container class that allows representation of fixed size vectors and matrices, and their respective standard operations. The library would define the following template template<typename T, std::size_t N, std::size_t M, class Traits> matrix and standard operations on it. The class would make heavy use of template specialization to allow efficient vector operations ( matrices with size 1xN ), and provide efficient operations for the common sizes (2-4). The class will provide the interface of STL containers with the exact concept depending on the Traits used. Care will be taken to allow transparent usage within STL and other boost libraries. Rationale ========== Boost.Matrix aims to be for uBLAS, what Boost.Array is for vectors -- a fixed size version to remove the overhead. The need is however greater, because while uBLAS provides a template for a fixed-sized version of it's arrays, the template is just a cover for the original uBLAS matrices. While the performance is comparable, the size includes a unnecessary field size. In cases that make heavy use of fixed sized matrices and vectors ( like real-time computer graphics ), such a tradeoff is unacceptable -- due to increased memory consumption as well as effective vectorized math operations, and alignment of structures passed ( a single 4D vector should effectively fit a 128 bit register, and a single 128 bit move operation ). Also, knowing the dimensions of a matrix/vector at compile time, allows for a more efficient operation implementation, both from the programmers side ( better algorithms ), as well as from the compilers side ( optimization ). Primary deliverables ===================== Boost.Matrix library, with the following features: a) a boost::matrix base container class for 2-dimensional matrices, with the proper behavious and semantics of a container, written following the boost coding guidelines, and with full portability in mind b) specializations for efficient vector (1xN matrix) handling c) operators for readable but efficient vector/matrix math d) operator specializations for matrices and vectors of the most common size (2-4) for most efficient implementation Secondary deliverables ======================= a) performance test suite, for testing optimizations b) unit tests for the whole library providing full operation coverage and code-coverage c) documentation for the whole library in the BoostBook format d) efficient implementation of interoperability of Boost.Matrix with other boost libraries (uBLAS in particular) Bonus deliverables =================== ( if time allows these will be delivered during program timeline, if not, they will be implemented afterwards ) a) optional (compile-time controlled) integration with Boost.SIMD to boost (no pun intended) performance b) performance tests on different compilers/architectures, further optimizations c) support for further boost libraries ( Boost.Assign, sandbox libraries, etc. ) Design issues ============== a) compile-time concept checking to only generare functions that make sense for the given type ( no floating point operations for integral types ) b) usage of templates to unroll the operations at compile-time to allow the compiler to optimize them ( needs to be tested case by case with the performance test suite ) c) usage of expression templates to eliminate temporaries in expressions d) care to keep most useful memory alignment ( to provide ease of use with direct usage in architecture dependent pipelines ) e) design with architecture-dependent optimizations in mind ( ease of expansion ) f) keeping all the standard boost practices, including workarounds for non-conformant compilers Project schedule ================= I plan to do at least weekly progress reports with emphasised design decisions (however well designed in the first phase, reality will probably change the design afterwards at least a little). Progress reports will be directed to the medium established with the mentor (blog, list or only direct communication). This will be my only job once summer starts, and will be treated as such during program coding timeline ( full-time schedule ) -- a week may fall out for a short vacation. Preplanning (April 20 - May 23) -------------------------------- Most of the major design decisions need to be done before the program starts. A end-user interface for the library should be established before the program starts. This will be summed up in a report that will be sent to boost-dev with a request for comments. Phase 1 (Correctness) - (May 24 - June 14) ------------------------------------------- A working basic (non-optimized) implementation, and a full test suite for the basic operations will be created. Goal for this phase is to provide a correct and working implementation. Basic usage document. Phase 2 (Functionality) - (June 14 - July 5) --------------------------------------------- Performance test suite. Compiler-level optimizations ( expression templates, unrolling, etc ). Basic library compatibility ( STL container, uBLAS ). Concept checking. Phase 3 (Optimization) - (July 6 - July 26) -------------------------------------------- Optimizations for vectors, optimizations for specific vector and matrix sizes, optimization checking through performance test suite. Code review for eventual optional architecture optimization ( Boost.SIMD for example ). General optimization of code. Update of documentation. Phase 4 (Finalization and bonuses) - (July 27 - August 10+) ------------------------------------------------------------ Testing on as many architectures as possible ( probably will need some help here ). Fixes. Cleanup of the code where needed. A report of the implementation that again will be posted to boost-dev with a request for comments. Here is included also buffer time if there will be delays in one of the former phases. If not, work will continue with the Bonus deliverables. Brief resume ============= I'm a Computer Science PhD student (2nd year) at the University of Wroclaw (Poland), in the field of Computer Graphics. My interests lie in Game Programming and Design, with a special hunch for Real-time Proceduraly Generated Content. I actively run self-designed courses and lectures at the University ( 3ds MAX course last year, Game Programming lecture this semester, and probably an Advanced C++ course next semester ). Currently, until the end of the semester, I'm working full-time at Tieto corporation, doing a large scale project for Nokia. Previously I worked at a smaller computer game company working on a massive multiplayer game. I've successfully taken part in two last GSoC's creating a Procedural City Generator for the BZFlag project. I consider myself a fairly competent C++ programmer, although I only recently discovered its true strengths ( learned to properly use and write in the style of STL, started reading Boost sources, and got hooked into C++0x ). I'm currently using the language both at work and for my own projects. The latter include a 3D game engine (in progress), and an unannounced space combat/trading game. I am a big fan of roguelike (ASCII) games, and managed to write several roguelikes up to date, the most famous one of them being DoomRL ( http://doom.chaosforge.org ) that did even get mentions in contemporary gaming magazines. Homepage : http://chaosforge.org/ IRC nick : Epyon E-mail : kornel.kisielewicz at gmail.com / epyon at cs.uni.wroc.pl -- regards, Kornel Kisielewicz

7 14

Extrinsec conversions (Was [safebool] Can we generalize it and put it intoutilities?)
by vicente.botet 01 Apr '09

01 Apr '09

----- Original Message ----- From: "Eugene Wee" <crystalrecursion(a)gmail.com> To: <boost(a)lists.boost.org> Sent: Wednesday, April 01, 2009 4:10 AM Subject: Re: [boost] [safebool] Can we generalize it and put it intoutilities? > > Hi, > > On Wed, Apr 1, 2009 at 6:20 AM, Vladimir Batov > <vladimir.batov(a)wrsa.com.au>wrote: > >> I glanced over "explicit conversion operators" in C++0x. It did not struck >> me as >> relevant to safe_bool. Care to elaborate? > > > It is relevant since you can say, define an explicit conversion function to > convert to bool, and yet avoid the very problems with say, also allowing > implicit conversion to int, that the safe bool idiom is used to avoid. In > other words, it is a language feature directly designed to replace the safe > bool idiom, and more. Hi, Thanks for the clarification. Even if the following is not related to the safe bool idiom it is related to conversions. The current standard allows to define intrinsec coversions, i.e. conversions that are defined either as a conversion operator on the from class or as a constructor on the to class. Vladimir StringConvert proposal seems to go towards a conversion from/to strings (If I've understood?). lexical_cast can be seen as an extrinsec converions of unrelated types via a common type stream type. I have proposed a generic convert_to template function that generalize extrinsicaly the explicit convertion of unrelated types. template <typename To, typename From> To convert_to(From&); One difference is that for C++0x explicit conversion you can write From from; To to=To(from); While for unrelated types you are forced to write From from; To to=convert_to<To>(from); The other difference is that constructors and conversion operators can be implicit, while the convert_to function can not. Do you think that it could be an interest in extending the C++0X standard to take care of these extrinsically defined conversions between unrelated types? Have a sens to have extrinsec conversions which implicitly convert? Should this convert_to<T> operator be considered as a friend class of T? I know that this is out of the scope of Boost, but I just wanted to know your opinion before posting to comp.lang.c++.moderated. Apologize if this has already been discused. Best, Vicente

3 2

[safebool] Can we generalize it and put it into utilities?
by Vladimir.Batov＠wrsa.com.au 01 Apr '09

01 Apr '09

Repeating safe-bool scaffolding over and over again is a real nuisance. It is worse for the end users as not exactly everyone is intimately familiar with the issue and the safe-bool workaround. Can we generalize it and put something like the following into utilities? struct safebool { typedef void (safebool::*unspecified_bool_type)() const; typedef unspecified_bool_type result; explicit safebool(bool v) : value_(v) {} operator result() const { return value_ ? &safebool::internal_bool : 0; } private: void internal_bool() const {}; bool value_; }; That way all we (and the users) will have to do to deploy the technique will be class Foo { ... operator safebool::result() const { return safebool(my_condition); } }; This seems simple and hassle-free. Thanks, Vladimir.

8 19

[gsoc] Concept check
by Kasra 01 Apr '09

01 Apr '09

Hi, I have decided to use the following concept for my gsco proposal regarding boost.crypto. The dilema at the moment what is the minimum contiguous ranges concept? Any suggestiong over the design? concept CryptoBuffer // which is a variable sized buffer that // uses an allocator that securely deletes. concept BlockCipher { typename value_type; typename size_type; typename traits_type; auto key(void) const throw(); template<CryptoBuffer CB> void key(const CB& ) const throw(invalid_key_size); template<CryptoBuffer CB> void encrypt(CB&, const CB& ) const throw(bad_state); template<CryptoBuffer CB> void decrypt(CB&, const CB& ) const throw(bad_state); }; struct crypto_stream_parallel_encryption { }; struct crypto_stream_parallel_decryption { }; struct crypto_stream_parallel : crypto_stream_parallel_encryption, crypto_stream_parallel_decryption{ }; concept ModeOfOperation { typename value_type; typename size_type; typename block_cipher_type; typename category; static constexpr size_type transform_size; static constexpr size_type state_size; const auto cipher() const throw(); void create() throw(); template<CryptoBuffer CB0, CryptoBuffer CB1> size_type update(CB0& , const CB1&); template<CryptoBuffer CB0, CryptoBuffer CB1> size_type update_final_block(CB0& , const CB1&); }; concept StreamCipher { typename value_type; typename size_type; typename block_cipher_type; typename category; static constexpr size_type transform_size; static constexpr size_type state_size; const auto cipher() const throw(); void create() throw(); template<CryptoBuffer CB0, CryptoBuffer CB1> size_type update(CB0& , const CB1&); template<CryptoBuffer CB0, CryptoBuffer CB1> size_type update_final_block(CB0& , const CB1&); }; concept HashAlgorithm { typename value_type; typename size_type; typename block_cipher_type; typename category; static constexpr size_type state_size; static constexpr size_type transform_size; void create() throw(); template<CryptoBuffer CB0, CryptoBuffer CB1> size_type update(const CB1&); template<CryptoBuffer CB0, CryptoBuffer CB1> size_type finalise(const CB1&); }; Any suggestiong over the design? -- Kasra

1 0

has_operator_new
by Robert Ramey 01 Apr '09

01 Apr '09

I've made a TRAK item to request a new type traits feature - has_operator_new. I've included an implementation and a test - both type_traitsified. Hope you can see fit to check it out, and if it's OK, check it in. Let me know when you do this. Once this is in, I can clear a dependent feature request TRAK item from the serialization library. Robert Ramey

2 3

[gsoc] Interest check for 3d geometry proposal
by Kornel Kisielewicz 01 Apr '09

01 Apr '09

Hello all, I was instructed on the #boost-soc channel to post the initial interest check on the GSoC proposal on the dev list, so here it is. It was inspired by one of the project proposals from 2006 (Geometry Library). As a known fact Boost begins to be used in the gaming industry. Following the idea of code reusage, there are several generic algorithms and components reused by virtually every 3d gaming engine. Most particulary this is 3D math. Sure, Boost has a linear algebra library graph algorithms, and a proposed geometry library, yet what is missing is a high-performance low-dimensional linear algebra library ( vectors and matrices up to the 4th dimension ). Side by side we have also a need of several 3-dimensional algorithms -- including, but not limited to intersection tests, inclusion tests, volume bounding, and several computational geometry algorithms. My suggestion would be creating a 3d specific geometry library for boost, with reasonable performance and the most used algorithms. I'd be happy to work on such a library as a GSoC project. My goal is to assess the initial interest in such an GSoC proposal idea before preparing and posting a more detailed description. -- regards, Kornel Kisielewicz http://chaosforge.org/

22 56

GSoC 2009
by Julian Mautner 01 Apr '09

01 Apr '09

Hi everyone! I'm Julian, 21 years old, student at the university of Innsbruck, Austria. Currently I'm making my bachelor in physics and computer science. I really would like to participate as a GSoC-student of yours. I intend to write a proposal for the Boost.Devector project. But first I wanted to ask a few questions: 1) I didn't understand if devector is an intrusive container or it only has to use them? 2) What does this devector class should (exaclty) be capable to do? 3) Which methods are the most important to perform in O(1)? 4) I didn't used the boost-library very much in the past, is this a great problem for this project? Thanks for your time. Regards, Julian

2 1

[type-traits] compilation error in has_new_operator
by Andrey Torba 01 Apr '09

01 Apr '09

Hi All, After 52057-52061 commits the compilation error oocur: D:\c++\libs\boost_trunk\boost/type_traits/has_new_operator.hpp(40) : error C2039: 'nothrow_t' : is not a member of 'std' to reproduce try to compile: #include <boost/type_traits/has_new_operator.hpp> int main() { return 0; } Does it need #include <new>?

2 1

GSoc Proposals - some friendly advice
by Robert Ramey 01 Apr '09

01 Apr '09

I've been looking through the GSoc proposals with some bemusement. Most of the proposals are projects that will thake MUCH, MUCH more than a summer. I know "But I've almost got it working". I don't think that one can fullly appreciate the effort it takes to make a boost library. I know its very easy to get started. I'm here to tell you its very hard to finish. It's so much more than writing the code. Theres the tests, getting them to pass on all platforms, getting the design past the criticism an reveiws. So here's my advice. Pick something really small that you know you can finish in three months. If you finish too soon, we can give you something else to do. Robert Ramey

3 2