- Boost-users - lists.boost.org

[multiprecission] Large divides layered on 128 bit divides
by Neill Clift 21 Aug '21

21 Aug '21

Hi, The architecture of cpp_int being build on 64 bit arithmetic using 128 bit double_limb_type is interesting. I have a question on the large divide (divide_unsigned_helper). It uses the upper portions of the large integers to get an estimation of the quotient. Subtracts out a multiple of that quotient and repeats. It does this in 128 bit values if available from the compiler: double_limb_type a = (static_cast<double_limb_type>(prem[r_order]) << CppInt1::limb_bits) | prem[r_order - 1]; double_limb_type b = py[y_order]; double_limb_type v = a / b; The compiler emulates this operation in the routine __udivmodti4 which itself uses an iterative approach. It seems to use a pretty basic shift and subtract algorithm mind you. As a general rule for multiprecision is it OK to layer the Knuth like algorithm D on top of each other this way. I have no idea myself but wonder if this is a known issue. I would have guessed that it made sense to do a 64 by 64 bit divide to guess the quotient and repeat. Thoughts? Thanks. Neill.

2 2

[multiprecission] some thoughts on add_unsigned.hpp
by Neill Clift 19 Aug '21

19 Aug '21

Hi, In my program I see a lot of time taken up by add/subtract_unsigned. I have a couple of thoughts on the code: Carry propagation outside the range of the smallest number can be improved: for (; i < x && carry; ++i) carry = ::boost::multiprecision::detail::addcarry_limb(carry, pa[i], 0, pr + i); Maybe this: for (; i < x && carry; ++i) carry = ::boost::multiprecision::detail::addcarry_limb(0, pa[i], 1, pr + i); This generates just an inc on my machine but unfortunately it doesn't eliminate the loads and stores. I think on average the carry chain must be small though. You have the same situation on subtract. add_unsigned and subtract_unsigned are inconsistent at the tail end after carry propagation: add_unsigned: if (i == x && carry) { // We overflowed, need to add one more limb: result.resize(x + 1, x + 1); if (result.size() > x) result.limbs()[x] = static_cast<limb_type>(1u); } else std::copy(pa + i, pa + x, pr + i); <--- if I == x we do the copy. We also do the copy even if the output and input overlap subtract_unsigned: // Any remaining digits are the same as those in pa: if ((x != i) && (pa != pr)) <-------- we do these checks here and avoid the copy std_constexpr::copy(pa + i, pa + x, pr + i); I am seeing a lot of time being used by that general copy above in add_unsigned when it can be avoided. This stuff is totally awesome and works well in my program. So thanks for working on this stuff. Neill.

2 1

Boost Installation in Codeblocks
by Alexander Tanseco 18 Aug '21

18 Aug '21

To Gavin Lambert, many thanks for your help. I finally got my program running without errors with the ff settings: Global variable Lib: c:\boost_1_77_0\stage\lib Program's build options: Linker settings/Link libraries: blank Search Directories/Compiler: $(#boost) Search Directories/Linker: $(#boost.lib).

1 0

Boost uses unoptimized subtraction because of typo in intel_intrinsics.hpp
by Neill Clift 18 Aug '21

18 Aug '21

Hi, In my program I am seeing my code use the slower (I assume) subtract_unsigned_constexpr rather than subtract_unsigned with he borrow chain handled intrinsics. This seems to be happening because of these lines in intel_intrinsics.hpp: #if defined(__clang__) && (__clang__ < 9) // We appear to crash the compiler if we try to use these intrinsics? #undef BOOST_MP_HAS_IMMINTRIN_H #endif __clang__ is not the version but just the value 1. I expect this should be the major version instead: #define __clang__ 1 #define __clang_major__ 12 #define __clang_minor__ 0 As an aside 128bit support does work in windows with clang. I have to force it on like this: #define __STRICT_ANSI__ #define BOOST_HAS_INT128 #define BOOST_MP_HAS_IMMINTRIN_H #include <boost/multiprecision/cpp_int.hpp> Thanks. Neill.

2 1

boostrap fails with intel-linux
by Alain O' Miniussi 18 Aug '21

18 Aug '21

Hi I have a C++11 compiler: [alainm@zztop boost_1_77_0]$ icpc --version icpc (ICC) 2021.3.0 20210609 Copyright (C) 1985-2021 Intel Corporation. All rights reserved. [alainm@zztop boost_1_77_0]$ I can use it to build b2: [alainm@zztop boost_1_77_0]$ icpc --version icpc (ICC) 2021.3.0 20210609 Copyright (C) 1985-2021 Intel Corporation. All rights reserved. [alainm@zztop boost_1_77_0]$ ./tools/build/src/engine/build.sh --cxx=icpc ... ### Using 'cxx' toolset. >>> Why cxx ? ... icpc (ICC) 2021.3.0 20210609 Copyright (C) 1985-2021 Intel Corporation. All rights reserved. ..... > icpc -DNDEBUG builtins.cpp class.cpp command.cpp compile.cpp constants.cpp cwd.cpp debug.cpp debugger.cpp execcmd.cpp execnt.cpp execunix.cpp filesys.cpp filent.cpp fileunix.cpp frames.cpp function.cpp glob.cpp hash.cpp hcache.cpp hdrmacro.cpp headers.cpp jam_strings.cpp jam.cpp jamgram.cpp lists.cpp make.cpp make1.cpp md5.cpp mem.cpp modules.cpp native.cpp object.cpp option.cpp output.cpp parse.cpp pathnt.cpp pathsys.cpp pathunix.cpp regexp.cpp rules.cpp scan.cpp search.cpp startup.cpp subst.cpp sysinfo.cpp timestamp.cpp variable.cpp w32_getreg.cpp modules/order.cpp modules/path.cpp modules/property-set.cpp modules/regex.cpp modules/sequence.cpp modules/set.cpp -o b2 > cp b2 bjam [alainm@zztop boost_1_77_0]$ But bootstrap.sh fails to see it: alainm@zztop boost_1_77_0]$ ./bootstrap.sh --with-toolset=intel-linux Building B2 engine.. A C++11 capable compiler is required for building the B2 engine. Toolset 'intel-linux' does not appear to support C++11. ... ** Note, the C++11 capable compiler is _only_ required for building the B2 ** engine. The B2 build system allows for using any C++ level and any other ** supported language and resource in your projects. ... You can specify the toolset as the argument, i.e.: ./build.sh [options] gcc Toolsets supported by this script are: acc, clang, como, gcc, intel-darwin, intel-linux, kcc, kylix, mipspro, pathscale, pgi, qcc, sun, sunpro, tru64cxx, vacpp ## And then a lot of comment regarding build.sh which we did not called ourselve so not very usefull Failed to build B2 build engine [alainm@zztop boost_1_77_0]$ Any idea of what went wrong ? Thanks ---- Alain Miniussi DSI, Pôles Calcul et Genie Log. Observatoire de la Côte d'Azur Tél. : +33492003009 (Mont-Gros) +33483618544 (Sophia Antipolis) +33609650665

1 0

Boost Installation in Codeblocks
by Alexander Tanseco 17 Aug '21

17 Aug '21

I successfully built Boost in my directory c:\boost_1_77_0. After which, I set up my global variable boost with base c:\boost_1_77_0. For the built-in field lib, I entered c:\boost_1_77_0\stage\lib (see attached screenshot). For the program's build options, under Linker settings/Link libraries: $(#boost.stage.lib) Under Search Directories/Compiler I entered $(#boost) and under Search Directories/Linker I entered $(#boost.stage.lib). When I built a typical Hello World program (see attached screenshot), I get an error message (see attached screenshot). Where did I go wrong? Somebody, please help.

2 1

[release]Boost 1.77.0
by Marshall Clow 14 Aug '21

14 Aug '21

Release 1.77.0 of the Boost C++ Libraries is now available. These open-source libraries work well with the C++ Standard Library, and are usable across a broad spectrum of applications. The Boost license encourages both commercial and non-commercial use. This release contains two new libraries and numerous enhancements and bug fixes for existing libraries. * Describe: A C++14 reflection library, from Peter Dimov. Provides macros for describing enumerators and struct/class members, and primitives for querying this information. * Lambda2: A C++14, dependency-free, single header lambda library, from Peter Dimov. For details, including download links, see <http://www.boost.org/users/news/version_1.77.0> You can also download directly from: <https://boostorg.jfrog.io/artifactory/main/release/1.77.0/> To install this release on your system, see <http://www.boost.org/doc/libs/release/more/getting_started/index.html> Thanks to everyone who participated in this release. Thanks, -- The Boost Release Team Marshall Clow, Michael Caisse, Glen Fernandes

2 1

Replace icpx with icpc
by Borchert, Christopher B ERDC-RDE-ITL-MS CIV 10 Aug '21

10 Aug '21

I'm trying to build Boost 1.76 from source on SLES 15 using Intel 19.1.3.304 (and GCC 10.2.0 underneath). Simply, how to I get b2 not to use icpx? With icpc it builds but icpx gives ./b2 toolset=intel-linux icpx: error while loading shared libraries: libimf.so: cannot open shared object file: No such file or directory I've tried everything and resorted to altering $PATH to remove icpx, which is how I know that's the problem. Thanks, Chris

3 5

[boost][multiprecision/mpc][math/quadrature] Why boost::multiprecision::exp gets stuck when evaluating complex-valued integral?
by Kirill Kudashkin 06 Aug '21

06 Aug '21

Dear all, I have encountered a problem while using Boost C++ libraries ( multiprecision/mpc.hpp, math/quadrature) which I cannot resolve. This problems appears when running the following piece of code || |#include <boost/math/constants/constants.hpp> #include <boost/math/quadrature/exp_sinh.hpp> #include <boost/multiprecision/mpc.hpp> #include <boost/multiprecision/mpfr.hpp> #include <boost/math/special_functions/fpclassify.hpp> namespace mpns = boost::multiprecision; typedef mpns::number<boost::multiprecision::mpc_complex_backend <50> > mpc_type ; typedef mpc_type::value_type mp_type ; int main() { using boost::math::constants::root_pi ; mpc_type z{mp_type(1),mp_type(1)} ; auto f = [&](mp_type x) { //actually I did not test if all these conditions are needed... if (boost::math::fpclassify<mp_type> (x) == FP_ZERO) { return mpc_type(mp_type(1)) ; }; if (boost::math::fpclassify<mp_type> (x) == FP_INFINITE) { return mpc_type(mp_type(0)) ; }; mp_type x2 = mpns::pow(x,2U) ; if (boost::math::fpclassify<mp_type> (x2) == FP_ZERO) { return mpc_type(mp_type(1)) ; }; if (boost::math::fpclassify<mp_type> (x2) == FP_INFINITE) { return mpc_type(mp_type(0)) ; }; mpc_type ex2 = mpns::exp(-z*x2) ; if (boost::math::fpclassify<mpc_type> (ex2) == FP_ZERO) { return mpc_type(mp_type(0)) ; }; return ex2 ; } ; mp_type termination = boost::math::tools::root_epsilon <mp_type> () ; mp_type error; mp_type L1; size_t max_halvings = 12; boost::math::quadrature::exp_sinh<mp_type> integrator(max_halvings) ; mpc_type res = integrator.integrate(f,termination,&error,&L1) ; mpc_type div = 2U*mpns::sqrt(z) ; mpc_type result = (mpc_type(root_pi<mp_type> ())/div) - res ; return 0; } | ||Namely, the problem appears when evaluating the exponent. The code is compiled without any problems, but hangs on computing | mpc_type ex2 = mpns::exp(-z*x2) ;| It was possible to go into the function while debugging. It crashed eventually, when the compiler reached "unable to resolve non-existing file.../src/init2.c". I let it run for a while checking the call stack. It appears that the programs is stuck atcalling ___gmpn_sub_n_coreisbr(). Or better, in the call stuck it appeared frequently. I did several tests. The real-valued version (with mpfr_backend <2000>) works without any problems. cpp_complex_backend <100> works fine, but I could not test higher precision yet, since it is much slower. I found a work-around which is not ideal. Namely, mpc_type ex2 = mpc_type(mpns::exp(x2)) ; mpc_type ezx2 = mpns::pow(ex2,-z) ; works fine with mpc_complex_backend <2000>. The lambda-function as it is declared in the code above works and yields correct numbers. The problem appears when it is called from the integrator. My system. I use openSUSE Tumbeleweed. I installed boost libraries as provided by |boost_1_76_0-gnu-mpich-hpc-devel.| Complier -> g++, cppStandard gnu+17. I check that I have all libs, e.g. libgmp, libmpc, libmpfr and etc. All headers are present (the programs is complied without any warnings). Thank you in advance and Best wishes, Kirill ||

2 4

[release] Boost 1.77.0 Release Candidate 1
by Marshall Clow 06 Aug '21

06 Aug '21

The first release candidates for the 1.77.0 release are now available at: <https://boostorg.jfrog.io/artifactory/main/release/1.77.0/source/> The SHA256 checksums are as follows: fc9f85fc030e233142908241af7a846e60630aa7388de9a5fafb1f3a26840854 boost_1_77_0_rc1.tar.bz2 f2dd9722b21450daf5ce8ef82c2d99c7ec304f913ee6c6b7f047d5639503dda1 boost_1_77_0_rc1.7z 5347464af5b14ac54bb945dc68f1dd7c56f0dad7262816b956138fc53bcc0131 boost_1_77_0_rc1.tar.gz d2886ceff60c35fc6dc9120e8faa960c1e9535f2d7ce447469eae9836110ea77 boost_1_77_0_rc1.zip As always, the release managers would appreciate it if you download the candidate of your choice and give building it a try. Please report both success and failure, and anything else that is noteworthy. -- The Release managers

4 4