X-Git-Url: https://git.donarmstrong.com/?p=rsem.git;a=blobdiff_plain;f=boost%2Fmath%2Fspecial_functions%2Fnext.hpp;fp=boost%2Fmath%2Fspecial_functions%2Fnext.hpp;h=38f521649f5ede686a616e612163cbd5170fdbdd;hp=0000000000000000000000000000000000000000;hb=2d71eb92104693ca9baa5a2e1c23eeca776d8fd3;hpb=da57529b92adbb7ae74a89861cb39fb35ac7c62d diff --git a/boost/math/special_functions/next.hpp b/boost/math/special_functions/next.hpp new file mode 100644 index 0000000..38f5216 --- /dev/null +++ b/boost/math/special_functions/next.hpp @@ -0,0 +1,490 @@ +// (C) Copyright John Maddock 2008. +// Use, modification and distribution are subject to the +// Boost Software License, Version 1.0. (See accompanying file +// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) + +#ifndef BOOST_MATH_SPECIAL_NEXT_HPP +#define BOOST_MATH_SPECIAL_NEXT_HPP + +#ifdef _MSC_VER +#pragma once +#endif + +#include +#include +#include +#include + +#ifdef BOOST_MSVC +#include +#endif + +namespace boost{ namespace math{ + +namespace detail{ + +template +inline T get_smallest_value(mpl::true_ const&) +{ + // + // numeric_limits lies about denorms being present - particularly + // when this can be turned on or off at runtime, as is the case + // when using the SSE2 registers in DAZ or FTZ mode. + // + static const T m = std::numeric_limits::denorm_min(); + return ((tools::min_value() - m) == tools::min_value()) ? tools::min_value() : m; +} + +template +inline T get_smallest_value(mpl::false_ const&) +{ + return tools::min_value(); +} + +template +inline T get_smallest_value() +{ +#if defined(BOOST_MSVC) && (BOOST_MSVC <= 1310) + return get_smallest_value(mpl::bool_::is_specialized && (std::numeric_limits::has_denorm == 1)>()); +#else + return get_smallest_value(mpl::bool_::is_specialized && (std::numeric_limits::has_denorm == std::denorm_present)>()); +#endif +} + +// +// Returns the smallest value that won't generate denorms when +// we calculate the value of the least-significant-bit: +// +template +T get_min_shift_value(); + +template +struct min_shift_initializer +{ + struct init + { + init() + { + do_init(); + } + static void do_init() + { + get_min_shift_value(); + } + void force_instantiate()const{} + }; + static const init initializer; + static void force_instantiate() + { + initializer.force_instantiate(); + } +}; + +template +const typename min_shift_initializer::init min_shift_initializer::initializer; + + +template +inline T get_min_shift_value() +{ + BOOST_MATH_STD_USING + static const T val = ldexp(tools::min_value(), tools::digits() + 1); + min_shift_initializer::force_instantiate(); + + return val; +} + +template +T float_next_imp(const T& val, const Policy& pol) +{ + BOOST_MATH_STD_USING + int expon; + static const char* function = "float_next<%1%>(%1%)"; + + int fpclass = (boost::math::fpclassify)(val); + + if((fpclass == FP_NAN) || (fpclass == FP_INFINITE)) + { + if(val < 0) + return -tools::max_value(); + return policies::raise_domain_error( + function, + "Argument must be finite, but got %1%", val, pol); + } + + if(val >= tools::max_value()) + return policies::raise_overflow_error(function, 0, pol); + + if(val == 0) + return detail::get_smallest_value(); + + if((fpclass != FP_SUBNORMAL) && (fpclass != FP_ZERO) && (fabs(val) < detail::get_min_shift_value()) && (val != -tools::min_value())) + { + // + // Special case: if the value of the least significant bit is a denorm, and the result + // would not be a denorm, then shift the input, increment, and shift back. + // This avoids issues with the Intel SSE2 registers when the FTZ or DAZ flags are set. + // + return ldexp(float_next(T(ldexp(val, 2 * tools::digits())), pol), -2 * tools::digits()); + } + + if(-0.5f == frexp(val, &expon)) + --expon; // reduce exponent when val is a power of two, and negative. + T diff = ldexp(T(1), expon - tools::digits()); + if(diff == 0) + diff = detail::get_smallest_value(); + return val + diff; +} + +} + +template +inline typename tools::promote_args::type float_next(const T& val, const Policy& pol) +{ + typedef typename tools::promote_args::type result_type; + return detail::float_next_imp(static_cast(val), pol); +} + +#if 0 //def BOOST_MSVC +// +// We used to use ::_nextafter here, but doing so fails when using +// the SSE2 registers if the FTZ or DAZ flags are set, so use our own +// - albeit slower - code instead as at least that gives the correct answer. +// +template +inline double float_next(const double& val, const Policy& pol) +{ + static const char* function = "float_next<%1%>(%1%)"; + + if(!(boost::math::isfinite)(val) && (val > 0)) + return policies::raise_domain_error( + function, + "Argument must be finite, but got %1%", val, pol); + + if(val >= tools::max_value()) + return policies::raise_overflow_error(function, 0, pol); + + return ::_nextafter(val, tools::max_value()); +} +#endif + +template +inline typename tools::promote_args::type float_next(const T& val) +{ + return float_next(val, policies::policy<>()); +} + +namespace detail{ + +template +T float_prior_imp(const T& val, const Policy& pol) +{ + BOOST_MATH_STD_USING + int expon; + static const char* function = "float_prior<%1%>(%1%)"; + + int fpclass = (boost::math::fpclassify)(val); + + if((fpclass == FP_NAN) || (fpclass == FP_INFINITE)) + { + if(val > 0) + return tools::max_value(); + return policies::raise_domain_error( + function, + "Argument must be finite, but got %1%", val, pol); + } + + if(val <= -tools::max_value()) + return -policies::raise_overflow_error(function, 0, pol); + + if(val == 0) + return -detail::get_smallest_value(); + + if((fpclass != FP_SUBNORMAL) && (fpclass != FP_ZERO) && (fabs(val) < detail::get_min_shift_value()) && (val != tools::min_value())) + { + // + // Special case: if the value of the least significant bit is a denorm, and the result + // would not be a denorm, then shift the input, increment, and shift back. + // This avoids issues with the Intel SSE2 registers when the FTZ or DAZ flags are set. + // + return ldexp(float_prior(T(ldexp(val, 2 * tools::digits())), pol), -2 * tools::digits()); + } + + T remain = frexp(val, &expon); + if(remain == 0.5) + --expon; // when val is a power of two we must reduce the exponent + T diff = ldexp(T(1), expon - tools::digits()); + if(diff == 0) + diff = detail::get_smallest_value(); + return val - diff; +} + +} + +template +inline typename tools::promote_args::type float_prior(const T& val, const Policy& pol) +{ + typedef typename tools::promote_args::type result_type; + return detail::float_prior_imp(static_cast(val), pol); +} + +#if 0 //def BOOST_MSVC +// +// We used to use ::_nextafter here, but doing so fails when using +// the SSE2 registers if the FTZ or DAZ flags are set, so use our own +// - albeit slower - code instead as at least that gives the correct answer. +// +template +inline double float_prior(const double& val, const Policy& pol) +{ + static const char* function = "float_prior<%1%>(%1%)"; + + if(!(boost::math::isfinite)(val) && (val < 0)) + return policies::raise_domain_error( + function, + "Argument must be finite, but got %1%", val, pol); + + if(val <= -tools::max_value()) + return -policies::raise_overflow_error(function, 0, pol); + + return ::_nextafter(val, -tools::max_value()); +} +#endif + +template +inline typename tools::promote_args::type float_prior(const T& val) +{ + return float_prior(val, policies::policy<>()); +} + +template +inline typename tools::promote_args::type nextafter(const T& val, const U& direction, const Policy& pol) +{ + typedef typename tools::promote_args::type result_type; + return val < direction ? boost::math::float_next(val, pol) : val == direction ? val : boost::math::float_prior(val, pol); +} + +template +inline typename tools::promote_args::type nextafter(const T& val, const U& direction) +{ + return nextafter(val, direction, policies::policy<>()); +} + +namespace detail{ + +template +T float_distance_imp(const T& a, const T& b, const Policy& pol) +{ + BOOST_MATH_STD_USING + // + // Error handling: + // + static const char* function = "float_distance<%1%>(%1%, %1%)"; + if(!(boost::math::isfinite)(a)) + return policies::raise_domain_error( + function, + "Argument a must be finite, but got %1%", a, pol); + if(!(boost::math::isfinite)(b)) + return policies::raise_domain_error( + function, + "Argument b must be finite, but got %1%", b, pol); + // + // Special cases: + // + if(a > b) + return -float_distance(b, a, pol); + if(a == b) + return 0; + if(a == 0) + return 1 + fabs(float_distance(static_cast((b < 0) ? T(-detail::get_smallest_value()) : detail::get_smallest_value()), b, pol)); + if(b == 0) + return 1 + fabs(float_distance(static_cast((a < 0) ? T(-detail::get_smallest_value()) : detail::get_smallest_value()), a, pol)); + if(boost::math::sign(a) != boost::math::sign(b)) + return 2 + fabs(float_distance(static_cast((b < 0) ? T(-detail::get_smallest_value()) : detail::get_smallest_value()), b, pol)) + + fabs(float_distance(static_cast((a < 0) ? T(-detail::get_smallest_value()) : detail::get_smallest_value()), a, pol)); + // + // By the time we get here, both a and b must have the same sign, we want + // b > a and both postive for the following logic: + // + if(a < 0) + return float_distance(static_cast(-b), static_cast(-a), pol); + + BOOST_ASSERT(a >= 0); + BOOST_ASSERT(b >= a); + + int expon; + // + // Note that if a is a denorm then the usual formula fails + // because we actually have fewer than tools::digits() + // significant bits in the representation: + // + frexp(((boost::math::fpclassify)(a) == FP_SUBNORMAL) ? tools::min_value() : a, &expon); + T upper = ldexp(T(1), expon); + T result = 0; + expon = tools::digits() - expon; + // + // If b is greater than upper, then we *must* split the calculation + // as the size of the ULP changes with each order of magnitude change: + // + if(b > upper) + { + result = float_distance(upper, b); + } + // + // Use compensated double-double addition to avoid rounding + // errors in the subtraction: + // + T mb, x, y, z; + if(((boost::math::fpclassify)(a) == FP_SUBNORMAL) || (b - a < tools::min_value())) + { + // + // Special case - either one end of the range is a denormal, or else the difference is. + // The regular code will fail if we're using the SSE2 registers on Intel and either + // the FTZ or DAZ flags are set. + // + T a2 = ldexp(a, tools::digits()); + T b2 = ldexp(b, tools::digits()); + mb = -(std::min)(T(ldexp(upper, tools::digits())), b2); + x = a2 + mb; + z = x - a2; + y = (a2 - (x - z)) + (mb - z); + + expon -= tools::digits(); + } + else + { + mb = -(std::min)(upper, b); + x = a + mb; + z = x - a; + y = (a - (x - z)) + (mb - z); + } + if(x < 0) + { + x = -x; + y = -y; + } + result += ldexp(x, expon) + ldexp(y, expon); + // + // Result must be an integer: + // + BOOST_ASSERT(result == floor(result)); + return result; +} + +} + +template +inline typename tools::promote_args::type float_distance(const T& a, const U& b, const Policy& pol) +{ + typedef typename tools::promote_args::type result_type; + return detail::float_distance_imp(static_cast(a), static_cast(b), pol); +} + +template +typename tools::promote_args::type float_distance(const T& a, const U& b) +{ + return boost::math::float_distance(a, b, policies::policy<>()); +} + +namespace detail{ + +template +T float_advance_imp(T val, int distance, const Policy& pol) +{ + BOOST_MATH_STD_USING + // + // Error handling: + // + static const char* function = "float_advance<%1%>(%1%, int)"; + + int fpclass = (boost::math::fpclassify)(val); + + if((fpclass == FP_NAN) || (fpclass == FP_INFINITE)) + return policies::raise_domain_error( + function, + "Argument val must be finite, but got %1%", val, pol); + + if(val < 0) + return -float_advance(-val, -distance, pol); + if(distance == 0) + return val; + if(distance == 1) + return float_next(val, pol); + if(distance == -1) + return float_prior(val, pol); + + if(fabs(val) < detail::get_min_shift_value()) + { + // + // Special case: if the value of the least significant bit is a denorm, + // implement in terms of float_next/float_prior. + // This avoids issues with the Intel SSE2 registers when the FTZ or DAZ flags are set. + // + if(distance > 0) + { + do{ val = float_next(val, pol); } while(--distance); + } + else + { + do{ val = float_prior(val, pol); } while(++distance); + } + return val; + } + + int expon; + frexp(val, &expon); + T limit = ldexp((distance < 0 ? T(0.5f) : T(1)), expon); + if(val <= tools::min_value()) + { + limit = sign(T(distance)) * tools::min_value(); + } + T limit_distance = float_distance(val, limit); + while(fabs(limit_distance) < abs(distance)) + { + distance -= itrunc(limit_distance); + val = limit; + if(distance < 0) + { + limit /= 2; + expon--; + } + else + { + limit *= 2; + expon++; + } + limit_distance = float_distance(val, limit); + if(distance && (limit_distance == 0)) + { + policies::raise_evaluation_error(function, "Internal logic failed while trying to increment floating point value %1%: most likely your FPU is in non-IEEE conforming mode.", val, pol); + } + } + if((0.5f == frexp(val, &expon)) && (distance < 0)) + --expon; + T diff = 0; + if(val != 0) + diff = distance * ldexp(T(1), expon - tools::digits()); + if(diff == 0) + diff = distance * detail::get_smallest_value(); + return val += diff; +} + +} + +template +inline typename tools::promote_args::type float_advance(T val, int distance, const Policy& pol) +{ + typedef typename tools::promote_args::type result_type; + return detail::float_advance_imp(static_cast(val), distance, pol); +} + +template +inline typename tools::promote_args::type float_advance(const T& val, int distance) +{ + return boost::math::float_advance(val, distance, policies::policy<>()); +} + +}} // namespaces + +#endif // BOOST_MATH_SPECIAL_NEXT_HPP +