- Boost - lists.boost.org

[1.47.0] Release branch now closed
by Rene Rivera 19 Jun '11

19 Jun '11

The release branch is now close, i.e. the release is frozen, for further changes. This means that commits to the release branch will fail without explicit authorization. -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org (msn) - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim,yahoo,skype,efnet,gmail

12 34

[mpl] string to type conversion
by Bartosz Marcinkiewicz 19 Jun '11

19 Jun '11

Hello, I would like to have a macro which could be used like this: STRING_TO_TYPE( MyType ) which would yield: typedef mpl::string<'M','y','T','y','p','e'> MyType_id_t; Any ideas how to implement something like that? Thanks for any help, bma. -- Bartosz Marcinkiewicz tel. (+48) 516 034 243 e-mail: bma(a)epcon.pl, bartosz.marcinkiewicz(a)epcon.pl EPCON Sp. z o.o. www.epcon.pl adres: 53-407 Wrocław, ul. Gajowicka 114/116 tel. (+48) 71 361 75 55 ext.36 fax (+48) 71 361 74 44 NIP PL 899-24-92-498 KRS 0000190187 REGON 932963731 Kapitał zakładowy: 500 000 PLN

4 4

Request for write access to Boost Sandbox
by Brian Smith 18 Jun '11

18 Jun '11

Hi I've written a library for the development of multi-dimensional array applications that includes scalar and fixed-size vector and matrix classes with an expression template implementation of operations. The library currently resides in the containers directory of Boost's Vault. However, I would like to submit the library for inclusion in Boost and its been suggested there's a better chance of this happening if it is moved to Boost's Sandbox. I therefore request write access to Boost's Sandbox in order to upload the files to there. Regards Brian -- www.maidsafe.net

2 2

[1.47.0][MPI] Showstopper
by KTC 18 Jun '11

18 Jun '11

Both the release & trunk version of MPI no longer compile on Microsoft Windows with Microsoft Compute Cluster Pack. With 1.46.1: ===============MPI Auto-configuration=============== Found Microsoft Compute Cluster Pack: C:\Program Files\Microsoft Compute Cluster Pack MPI build features: <include>/C:/Program Files/Microsoft Compute Cluster Pack/Include <address-model>64:<library-path>/C:/Program Files/Microsoft Compute Cluster Pack/Lib/amd64 <library-path>/C:/Program Files/Microsoft Compute Cluster Pack/Lib/i386 <find-static-library>msmpi <toolset>msvc:<define>_SECURE_SCL=0 MPI launcher: "C:\Program Files\Microsoft Compute Cluster Pack\Bin\mpiexec.exe" -n ==================================================== .... Currently: D:/src/boost-release-test/boost/tools/build/v2/tools\mpi.jam:249: in mpi.init rule path.make unknown in module mpi. D:/src/boost-release-test/boost/tools/build/v2/build\toolset.jam:38: in toolset.using D:/src/boost-release-test/boost/tools/build/v2/build\project.jam:888: in using C:\Users\KTC\user-config.jam:5: in modules.load D:/src/boost-release-test/boost/tools/build/v2\build-system.jam:257: in load-config D:/src/boost-release-test/boost/tools/build/v2\build-system.jam:399: in load-configuration-files D:/src/boost-release-test/boost/tools/build/v2\build-system.jam:555: in load D:\src\boost-release-test\boost\tools\build\v2/kernel\modules.jam:283: in import D:\src\boost-release-test\boost\tools\build\v2\kernel\bootstrap.jam:142: in boost-build D:\src\boost-release-test\boost\boost-build.jam:17: in module scope KTC Created ticket on Trac: <https://svn.boost.org/trac/boost/ticket/5618>

3 5

[unordered] equal_range is O(bucket_count()), not O(size())
by Brad Higgins 18 Jun '11

18 Jun '11

TR1 specifies that unordered_multimap::equal_range(k) worst case complexity is O(size()). However, in practice, I find that the worst case performance of boost unordred_multimap's equal_range is O(bucket_count()). Is this known? I believe the issue is with the second iterator returned by equal_range(). Specifically, if there are sparse elements in the hash table, the implementation may have to traverse many buckets to find the second iterator. I have a test that demonstrates the issue. The test makes 2 runs over an unordered_multimap, each using the same number of elements. In the second run, it calls rehash() in order to grow the bucket count by a large amount. Notice that the first run is quite fast, while the second run takes more than 10 seconds. Here are the results from my test: ---- $ ./main rehash(0) insert 1000 elements built hash, with 1543 buckets and 1000 elements. Now, get some ranges time to obtain 1000 elements: 00:00:00.001366 rehash(5000000) insert 1000 elements built hash, with 6291469 buckets and 1000 elements. Now, get some ranges time to obtain 1000 elements: 00:00:10.818002 ---- Here is my code for the test: ---- #include <iostream> #include <boost/unordered_map.hpp> #include <boost/date_time/posix_time/posix_time.hpp> typedef boost::unordered_multimap<size_t, size_t> hash_int_type; typedef hash_int_type::iterator hash_int_itr; typedef std::pair<size_t, size_t> hash_insert_type; void run_test(const size_t data_size, const size_t rehash_size) { hash_int_type hash; std::cout << "rehash(" << rehash_size << ")\n"; hash.rehash(rehash_size); std::cout << "insert " << data_size << " elements\n"; for (size_t i=0; i<data_size; i++) { hash_insert_type ins(i,i); hash.insert(ins); } std::cout << "built hash, with " << hash.bucket_count() << " buckets and " << hash.size() << " elements. Now, get some ranges\n"; boost::posix_time::ptime start(boost::posix_time::microsec_clock::local_time()); for (size_t i=data_size-1; i!=0; i--) { typedef std::pair<hash_int_itr, hash_int_itr> ret_pair; ret_pair ret = hash.equal_range(i); if (ret.first != ret.second) { hash.quick_erase(ret.first); } } boost::posix_time::ptime stop(boost::posix_time::microsec_clock::local_time()); boost::posix_time::time_duration diff_time(stop-start); std::cout << "time to obtain " << data_size << " elements: " << diff_time << "\n"; } int main(int argc, char* argv[]) { run_test(1000,0); run_test(1000,5000000); return 0; } ----- Thanks, Brad

2 3

[type_traits] Reverting support for clang 3.0 intrinsics
by Daniel James 18 Jun '11

18 Jun '11

Hi, As discussed here: http://lists.boost.org/Archives/boost/2011/06/182574.php Changeset #72423 is incorrect, since it's based on the clang version number which is not a reliable indicator of feature support: https://svn.boost.org/trac/boost/changeset/72423 So I think it should be reverted. Daniel

1 0

[array] assignment and constructor coherency
by Vicente Botet 18 Jun '11

18 Jun '11

Hi, Boost.Array defines the following assignment template (* typename U *) array(* T,N *)& operator=(const array(* U, N *)&); as std::copy(rhs.begin(),rhs.end(), begin()) Even if the standard doesn't provides this conversion assignment, for coherency purposes, shouldn't we have also the copy constructor template (* typename U *) explicit array(* T,N *)(const array(* U, N *)&); Best, Vicente P.S. Please replace (* and *) by angle brackets. -- View this message in context: http://boost.2283326.n4.nabble.com/array-assignment-and-constructor-coheren… Sent from the Boost - Dev mailing list archive at Nabble.com.

2 2

C++ AMP
by Neal Becker 17 Jun '11

17 Jun '11

Interesting: http://techreport.com/discussions.x/21134

5 7

Re: [boost] [1.47.0] Release branch now closed
by Rene Rivera 17 Jun '11

17 Jun '11

Sorry this should have gone to the list... On 6/17/2011 10:05 AM, Jim Bell wrote: > > > On 6/17/2011 9:37 AM, Rene Rivera wrote: >> On 6/17/2011 8:58 AM, Jim Bell wrote: >>> >>> On 1:59 PM, Rene Rivera wrote: >>>> The release branch is now close, i.e. the release is frozen, for >>>> further changes. This means that commits to the release branch will >>>> fail without explicit authorization. >>> >>> >>> The Boost.Thread patch for Mingw-64 is crucial for that important >>> platform.<https://svn.boost.org/trac/boost/ticket/4849> >>> >>> It has been patched into the trunk for three months without complaint, >>> and just needs to be shepherded to the release branch. >>> >>> Could you please do this? >> >> I'm not sure what you are asking. > > > I'd like a maintainer or release manager to merge the patched trunk > files into the release branch. > boost/detail/interlocked.hpp > libs/thread/src/win32/thread.cpp > libs/thread/src/win32/tss_pe.cpp > > > See<https://svn.boost.org/trac/boost/changeset/70383> Since I don't know enough to judge the code itself at this instant.. Could others please comment on the above? Is the thread maintainer around? -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org (msn) - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim,yahoo,skype,efnet,gmail

3 11

[interval] implicit versus explicit conversions
by Vicente Botet 17 Jun '11

17 Jun '11

Hi, The current interface of boost::interval defines the following implicit conversions interval(T const &v); template(* class T1 *) interval(T1 const &v); template(* class T1, class Policies1 *) interval(interval(* T1, Policies1 *) const &r); I was wondering if some/all of these shouldn't be declared explicit. Any thoughts? Best, Vicente P.S. Please replace (* and *) by angle brackets. -- View this message in context: http://boost.2283326.n4.nabble.com/interval-implicit-versus-explicit-conver… Sent from the Boost - Dev mailing list archive at Nabble.com.

2 1