- Boost - lists.preview.boost.org

[release] Boost 1.63.0 - fixes
by Oliver Kowalke 22 Nov '16

22 Nov '16

unfortunately beta 1 does not contain fixes from boost.context and boost.fiber. could you add the fixes to the 1.63 release, please? Oliver

3 4

[sort] Parallel sorting sub-library mini-review starts tomorrow, 11 November
by Steven Ross 22 Nov '16

22 Nov '16

Dear Boost community, The mini-review of Francisco Tapia's Parallel sorting library begins tomorrow November 11, and ends November 20th. The purpose of this review is to assess whether the sub-library is useful and up to Boost software standards. If the Boost community agrees on both, Francisco and I will integrate it into the existing Boost.sort library as a separate sub-library from Spreadsort. In doing so we'll integrate the documentation with the existing documentation for Boost.Sort, making it more consistent. If you review Francisco's library, please make sure you're using a compiler that supports C++11, and answer each of these questions: 1. Were you able to run it and see accurate and fast results? Please specify your compiler, OS, and processor type, and any problems (or slowdowns) encountered. 2. Would you use this library if accepted? Why or why not? 3. What is your evaluation of the design? 4. What is your evaluation of the implementation? 5. How much effort did you put into your evaluation? A glance? A quick reading? In-depth study? The library is here: https://github.com/fjtapia/sort_parallel This library provides both stable and unstable sorting algorithms, in single threaded and parallel versions. - All the algorithms have average and worst case runtime of O(NLogN). - These algorithms do not use any other library or utility. To compile and run these you will need a *C++11 compliant compiler.* - *This library is *include only, only requiring* the files in the boost/sort/parallel folder, with no external dependencies.* - The algorithms use a comparison object, in the same fashion as std::sort, defaulting to the std::less object, which uses the < operator internally for comparisons. - The algorithms are exception safe, meaning that the exceptions generated by the algorithms guarantee the integrity of the objects to sort, but not their relative order. If the exception is generated inside the objects (in the move or in the copy constructor) the results can be unpredictable. The library has a a new parallel_sort algorithm *( internally named Block Indirect)*, for processors connected with shared memory*,* invented for the library. This new algorithm combines the speed of algorithms like the GCC parallel sort (based on OpenMP), or the Microsoft parallel buffered sort, with the small memory consumption of algorithms like the TBB sort and the Microsoft parallel Sort. Results from running on an I7 5820 with 12 threads: GCC parallel_sort 1,25 secs 1560 MB TBB parallel_sort 1,64 secs 783 MB Boost parallel_sort 1,08 secs 786 MB In a separate repository (because it uses licence-restricted software like TBB and Microsoft parallel sort, and can’t be included in the Boost library), you can find all the code needed to run the benchmarks, the library code, the benchmark programs, and shell scripts to compile and run it on your machine in Linux and Windows. For the Linux benchmarks you will need to install TBB. (https://github.com/fjtapia/sort_parallel_benchmark) We look forward to your feedback!

4 5

Re: [boost] [sort] Parallel sorting sub-library mini-review. Performance tests.
by Christophe Henry 22 Nov '16

22 Nov '16

Hi Boost Community, I did not manage yet to write a full review. One week is a bit short so a few more days will be necessary. I provide some performance tests first. To check the library performance, I pitted it against my own versions of a parallel mergesort and quicksort provided by the asynchronous library ( https://github.com/henry-ch/asynchronous) which is now ready for review, so we have a stable state for testing. I also added a parallel version inside the asynchronous library using Francisco's single-thread implementations to compare with his parallel ones. I tested on a i7-5960X. Due to time limits, I could not test on more interesting NUMA platforms (I did not get the reviewed library to compile for my KNC) so the tests do not pretend to have value outside the i7 architecture. I linked with tbb_malloc_proxy to limit memory influence. In a nutshell. I'm not a big fan of the parallel version of the algorithms. It seems to be based on std::async, so that a lot of threads are started and joined at every call. I would suggest using a threadpool. OTOH the single-threaded ones are interesting, especially the stable_sort and intro_sort for cases where usage of spreadsort is not possible. Cheers, Christophe A short summary: 100000000 uint64_t elements already sorted: OMP parallel sort : 0.390899 secs Boost parallel sort : 0.064965 secs OMP parallel stable sort : 1.06128 secs Boost sample sort : 0.0695357 secs Boost parallel stable sort : 0.0662401 secs Asynchronous parallel sort : 0.0167134 secs (same with other algorithms) Asynchronous provides a special optimization for this case. I added this one: 100000000 uint64_t elements reverse sorted OMP parallel sort : 0.317039 secs Boost parallel sort : 0.581381 secs OMP parallel stable sort : 1.06448 secs Boost sample sort : 0.524746 secs Boost parallel stable sort : 0.73036 secs Asynchronous parallel sort : 0.0478701 secs Asynchronous provides a special optimization for this case. I this the library should do it too. This one is pretty common and a popular DOS attack. 100000000 uint64_t elements randomly filled OMP parallel sort : 1.03594 secs Boost parallel sort : 0.796447 secs OMP parallel stable sort : 1.28601 secs Boost sample sort : 0.818954 secs Boost parallel stable sort : 1.13604 secs Asynchronous parallel quickspreadsort: 0.587432 secs Asynchronous quick_intro_sort : 0.728393 secs This mixing of quicksort degrading into a spreadsort works here best. The parallel adaptation of intro_sort is not bad either and best of other library algorithms. Asynchronous parallel stable sort : 1.26141 secs Asynchronous boost::stable sort : 0.804814 secs Interesting. The stable version of the library easily beats std::stable_sort I used until now. 10000000 strings randomly filled OMP parallel sort : 1.05803 secs Boost parallel sort : 0.933055 secs OMP parallel stable sort : 1.12269 secs Boost sample sort : 0.889216 secs Boost parallel stable sort : 1.56564 secs Asynchronous parallel quickspreadsort: 0.788856 secs Asynchronous quick_intro_sort : 0.893652 secs Asynchronous parallel stable sort : 1.23495 secs Asynchronous boost::stable sort : 1.21817 secs Similar results. Let's move on to big objects: 1562500 elements of size 512 randomly filled H E A V Y C O M P A R I S O N OMP parallel sort : 0.308923 secs Boost parallel sort : 0.342992 secs Asynchronous parallel_quicksort : 0.246709 secs Asynchronous quick_intro_sort : 0.269666 secs Both quicksorts are best, with a slight advantage to intro_sort. The light comparison is similar. My test code (the modified benchmark.cpp provided by the library): https://github.com/henry-ch/asynchronous/blob/master/libs/asynchronous/test… The full test results: https://github.com/henry-ch/asynchronous/blob/master/libs/asynchronous/test…

3 2

[process] daemon ideas
by Dracon Pern 21 Nov '16

21 Nov '16

I am not sure if [process] is the right library, but I am thinking about building something to manage daemon install/uninstall across platforms, for example Windows's service, OS X's launchd, systemd, and SysV init. Every operating system offers some way to launch a daemon at boot, and also the ability to stop, recover, etc. The issue is they all do it slightly differently, unlike having a common main(). My thinking is to provide a thin layer for install/remove and service notification. I have looked at Boost.Application, but it's lacking install/uninstall, and also seems overly complicated. Thoughts? Thanks, -Eric

4 3

Re: [boost] [iterator] Orbiting iterators
by Jeremy Murphy 20 Nov '16

20 Nov '16

Hi Steve, On 21 Nov 2016 3:01 AM, "Steve M. Robbins" <steve(a)sumost.ca> wrote: > > On Sunday, November 20, 2016 8:25:20 PM CST Jeremy Murphy wrote: > > Hi, > > > > just wondering what the mood in Boost.Iterator is for new features? At work > > we recently found use for an orbiting iterator, specifically a circular > > one, which effectively wraps at the start/end of a bounded range. (The > > domain is digital signal processing.) > > That sounds like what CGAL calls a Circulator; c.f. http://doc.cgal.org/ > latest/Circulator/classCirculator.html Right, yeah. Unfortunately I can't view the documentation properly on my phone so I'll take a look later, but I imagine the principle is the same although the interface is not. > > Implementation details aside, does this feature sound desirable to the > > library? > > I've found these handy when iterating through graph edges incident on a > vertex, when you need to start on a specific edge. It's been ages since I've > used them so I can't recall the exact use case. Makes me wonder whether > something like this already exists in boost graph. As you pointed out with CGAL, domain-specific implementations do exist. Boost.Geometry has something too. I guess I'm interested in whether this need for circular iteration can be abstracted back to the plain old iterator. Cheers. Jeremy

1 0

[outcome] Requesting pre-review of Boost.Outcome tutorial
by Matus Chochlik 20 Nov '16

20 Nov '16

On Fri, Nov 11, 2016 at 8:27 PM, Niall Douglas <s_sourceforge(a)nedprod.com> wrote: > Dear Boost, > > I am hoping to bring Boost.Outcome, a C++ 14 library providing a > factory and family of policy driven lightweight monadic > value-or-error transports, into the Boost peer review queue by March > 2017 followed shortly thereafter by an ACCU talk in April if my talk > proposal passes their programme committee. To that end, I've written > an explanation and a sort of tutorial explaining the history and > purpose of Outcome at https://ned14.github.io/boost.outcome/ and I'd > greatly appreciate if people could tell me: > > 1. Does it make sense? > Yes, the explanation is useful and does make sense. > > 2. Do you think you could use Outcome in your own code after reading > it? > > 3. Do you think you would want to use Outcome in your own code after > reading it? > I'm already using something similar in my oglplus2 OpenGL wrapper and some other projects and I really like the technique. IMO the C++ community would benefit from having something like this in Boost. In my implementation the outcome is a pair of the "real" result and an "error info" which can be more elaborate than an error code + a policy which by default throws an exception from the error info when the outcome is destroyed. The function caller can explicitly cancel the throwing and just examine the error information. All functions returning outcome can be noexcept. > > 4. Are there any missing parts which are showstoppers? > I didn't try to use your implementation, but I don't see anything obvious. But a shorter initial tutorial would be helpful. > > > My thanks in advance. Be aware the install stuff described at the > start doesn't work yet. The reference documentation is also quite > patchy thanks to doxygen not understanding CRTP. All that stuff and a > fair bit more remains to fix before it's ready to enter the Boost > peer review queue. > BR, Matus

2 1

Info on Sprout libraries (related to Boost?)
by charleyb123 . 19 Nov '16

19 Nov '16

Looking at the (very impressive) Sprout libraries that distributes using the Boost license: Welcome to the Sprout C++ Libraries <http://bolero-murakami.github.io/Sprout/docs/index.html#welcome-to-the-spro…> C++11 constexpr based Containers, Algorithms, Random numbers, Parsing, Ray tracing, Synthesizer, and others. https://github.com/bolero-MURAKAMI/Sprout ...It appears to be actively maintained, and very interesting. Anybody know the background and roadmap on these libraries (they are not in Boost now, right?) I'm particularly interesting in more 'constexpr' and compile-time libraries showing up, where C++14 (and C++17) wholly "throws-open-the-door" on what you can do at compile time (for example, see the C++14 constexpr used as a full C compiler: https://github.com/kw-udon/constexpr-8cc ) Crazy stuff. --charley

3 2

[fiber] maintenance status
by Tatsuyuki Ishi 18 Nov '16

18 Nov '16

The fiber library seems to be still updated, but I can't get into the author to do a feature proposal or request a refactor. Please take a look at this: https://github.com/boostorg/fiber/issues/89 I have pinged Oliver 3 times, and also sent some emails to his git committer addresses (he uses two identities). If you (the maintainer) are aware of this issue, please respond; I'm willing to do code contributions, but I must have some opinions about that before I get on to the work.

2 1

BOOST_NO_CXX11_HDR_CODECVT confusion
by Robert Ramey 17 Nov '16

17 Nov '16

I use BOOST_NO_CXX11_HDR_CODECVT from config.hpp in order to selectively enable code which uses the built-in codecvt facets as opposed to the boost "home brew" one we've been using for a long time. I do similar things with other macros such as BOOST_NO_CXX11_HDR_FORWARD_LIST. Problems arise when I use a recent compiler such as gcc 6.2 with the switch -std=c++03 as in "/hpc-6.2/usr/local/bin/g++" -ftemplate-depth-128 -std=c++03 -O0 -fno-inline -Wall -g -dynamic -gdwarf-2 -fexceptions -fPIC -arch x86_64 -Wno-unused-variable -Wno-long-long -ftemplate-depth-255 -DBOOST_ALL_NO_LIB=1 -DBOOST_FILESYSTEM_DYN_LINK=1 -DBOOST_LIB_DIAGNOSTIC=1 -DBOOST_SERIALIZATION_DYN_LINK=1 -DBOOST_SYSTEM_DYN_LINK=1 -DBOOST_WSERIALIZATION_DYN_LINK=1 -I"../../.." -c -o "../../../bin.v2/libs/serialization/test/test_iterators.test/darwin-6.2~03/debug/test_iterators.o" "test_iterators.cpp" In this ease I would expect to find that BOOST_NO_CXX11_HDR_CODECVT is defined as C++ standard does not include <codecvt> headers for versions prior to C++ 11. But the test matrix http://www.boost.org/development/tests/develop/developer/output/Sandia-gcc-… for this compilers shows that BOOST_NO_CXX11_HDR_CODECVT is NOT defined when no -std=c++03 is used. OK - I see that for gcc 6.2 the default is C++ 14 if no switch is used. So the test matrix is OK. I'm not sure what the column is meant to test as it does specify a particular version as the other columns do. So I think the test for gcc 6.2 - and all the g++/clang compiler would be improved by specifying which version of the C++ standard they are actually testing. But I still have a problem. When I use the swithc -std=c++03 I expect that BOOST_NO_CXX11_HDR_CODECVT be defined as C++ versions prior to C++ 11 don't have header <codecvt>. Same goes for other headers required only for more recent compilers. But my expectations aren't realized as BOOST_NO_CXX11_HDR_CODECVT is not defined in this case. The leads to selection of library code which invokes an #error saying that it's not supported. Perhaps definition of BOOST_NO_CXX11_HDR_CODECVT is conditioned solely on the existence of <codecvt> rather than the existence of <codecvt> and the compiler running in C++ 11+ "mode". I think this is what is messing things up for me. I might be able to work around this if I config.hpp included a macro which indicated which standard version it's intended to support. It doesn't have this. But I wouldn't need it if the BOOST_NO... worked as I would expect. Robert Ramey

3 3

MSVC opensource deployment.
by Robert Ramey 16 Nov '16

16 Nov '16

This just showed up in my junk mail box. Vcpkg updates: Static linking is now available https://github.com/Microsoft/vcpkg With 90+ libraries in its catalog, Vcpkg makes it easier to find the right open source library for Visual C++ 2015. The latest update now supports static linking. This kind of interesting: a) Does this constitute and open-source source code deployment system that perhaps we might learn from/exploit? b) I'm intrigued with the comment under "License": "Code licensed under MIT License" Boost is in this repository - I don't remember changing the Boost license. Robert Ramey

4 5