| /* boost random/detail/large_arithmetic.hpp header file |
| * |
| * Copyright Steven Watanabe 2011 |
| * Distributed under 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) |
| * |
| * See http://www.boost.org for most recent version including documentation. |
| * |
| * $Id: large_arithmetic.hpp 71018 2011-04-05 21:27:52Z steven_watanabe $ |
| */ |
| |
| #ifndef NDNBOOST_RANDOM_DETAIL_LARGE_ARITHMETIC_HPP |
| #define NDNBOOST_RANDOM_DETAIL_LARGE_ARITHMETIC_HPP |
| |
| #include <ndnboost/cstdint.hpp> |
| #include <ndnboost/integer.hpp> |
| #include <ndnboost/limits.hpp> |
| #include <ndnboost/random/detail/integer_log2.hpp> |
| |
| #include <ndnboost/random/detail/disable_warnings.hpp> |
| |
| namespace ndnboost { |
| namespace random { |
| namespace detail { |
| |
| struct div_t { |
| ndnboost::uintmax_t quotient; |
| ndnboost::uintmax_t remainder; |
| }; |
| |
| inline div_t muldivmod(ndnboost::uintmax_t a, ndnboost::uintmax_t b, ndnboost::uintmax_t m) |
| { |
| static const int bits = |
| ::std::numeric_limits< ::ndnboost::uintmax_t>::digits / 2; |
| static const ::ndnboost::uintmax_t mask = (::ndnboost::uintmax_t(1) << bits) - 1; |
| typedef ::ndnboost::uint_t<bits>::fast digit_t; |
| |
| int shift = std::numeric_limits< ::ndnboost::uintmax_t>::digits - 1 |
| - detail::integer_log2(m); |
| |
| a <<= shift; |
| m <<= shift; |
| |
| digit_t product[4] = { 0, 0, 0, 0 }; |
| digit_t a_[2] = { digit_t(a & mask), digit_t((a >> bits) & mask) }; |
| digit_t b_[2] = { digit_t(b & mask), digit_t((b >> bits) & mask) }; |
| digit_t m_[2] = { digit_t(m & mask), digit_t((m >> bits) & mask) }; |
| |
| // multiply a * b |
| for(int i = 0; i < 2; ++i) { |
| digit_t carry = 0; |
| for(int j = 0; j < 2; ++j) { |
| ::ndnboost::uint64_t temp = ::ndnboost::uintmax_t(a_[i]) * b_[j] + |
| carry + product[i + j]; |
| product[i + j] = digit_t(temp & mask); |
| carry = digit_t(temp >> bits); |
| } |
| if(carry != 0) { |
| product[i + 2] += carry; |
| } |
| } |
| |
| digit_t quotient[2]; |
| |
| if(m == 0) { |
| div_t result = { |
| ((::ndnboost::uintmax_t(product[3]) << bits) | product[2]), |
| ((::ndnboost::uintmax_t(product[1]) << bits) | product[0]) >> shift, |
| }; |
| return result; |
| } |
| |
| // divide product / m |
| for(int i = 3; i >= 2; --i) { |
| ::ndnboost::uintmax_t temp = |
| ::ndnboost::uintmax_t(product[i]) << bits | product[i - 1]; |
| |
| digit_t q = digit_t((product[i] == m_[1]) ? mask : temp / m_[1]); |
| |
| ::ndnboost::uintmax_t rem = |
| ((temp - ::ndnboost::uintmax_t(q) * m_[1]) << bits) + product[i - 2]; |
| |
| ::ndnboost::uintmax_t diff = m_[0] * ::ndnboost::uintmax_t(q); |
| |
| int error = 0; |
| if(diff > rem) { |
| if(diff - rem > m) { |
| error = 2; |
| } else { |
| error = 1; |
| } |
| } |
| q -= error; |
| rem = rem + error * m - diff; |
| |
| quotient[i - 2] = q; |
| product[i] = 0; |
| product[i-1] = (rem >> bits) & mask; |
| product[i-2] = rem & mask; |
| } |
| |
| div_t result = { |
| ((::ndnboost::uintmax_t(quotient[1]) << bits) | quotient[0]), |
| ((::ndnboost::uintmax_t(product[1]) << bits) | product[0]) >> shift, |
| }; |
| return result; |
| } |
| |
| inline ndnboost::uintmax_t muldiv(ndnboost::uintmax_t a, ndnboost::uintmax_t b, ndnboost::uintmax_t m) |
| { return detail::muldivmod(a, b, m).quotient; } |
| |
| inline ndnboost::uintmax_t mulmod(ndnboost::uintmax_t a, ndnboost::uintmax_t b, ndnboost::uintmax_t m) |
| { return detail::muldivmod(a, b, m).remainder; } |
| |
| } // namespace detail |
| } // namespace random |
| } // namespace ndnboost |
| |
| #include <ndnboost/random/detail/enable_warnings.hpp> |
| |
| #endif // NDNBOOST_RANDOM_DETAIL_LARGE_ARITHMETIC_HPP |