Skip to content

Commit

Permalink
Merge pull request #303 from ckormanyos/update_from_embedded
Browse files Browse the repository at this point in the history 
Disable warnings embedded systems update
  • Loading branch information
@ckormanyos
ckormanyos authored Jul 1, 2024
2 parents 60af5f4 + 598ee88 commit dfd4dba
Show file tree
Hide file tree
Showing 3 changed files with 184 additions and 150 deletions.
150 changes: 11 additions & 139 deletions math/wide_decimal/decwide_t.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -69,15 +69,6 @@
namespace math { namespace wide_decimal { // NOLINT(modernize-concat-nested-namespaces)
#endif

// Forward declaration of the decwide_t template class.
template<const std::int32_t ParamDigitsBaseTen,
typename LimbType = std::uint32_t,
typename AllocatorType = std::allocator<void>,
typename InternalFloatType = double,
typename ExponentType = std::int64_t,
typename FftFloatType = double>
class decwide_t;

// Forward declarations of various decwide_t namespace functions.
template<const std::int32_t ParamDigitsBaseTen, typename LimbType, typename AllocatorType, typename InternalFloatType, typename ExponentType, typename FftFloatType>
constexpr auto zero() -> decwide_t<ParamDigitsBaseTen, LimbType, AllocatorType, InternalFloatType, ExponentType, FftFloatType>;
Expand Down Expand Up @@ -605,134 +596,6 @@
};

private:
template<typename FloatingPointType>
class native_float_parts final
{
public:
// Emphasize: This template class can be used with native
// floating-point types like float, double and long double.

explicit native_float_parts(FloatingPointType f)
{
using native_float_type = FloatingPointType;

// For long double, verify that the mantissa fits in std::uintmax_t.

static_assert(std::numeric_limits<native_float_type>::digits <= std::numeric_limits<std::uintmax_t>::digits,
"Error: The width of the mantissa does not fit in std::uintmax_t");

const native_float_type ff = ((f < static_cast<native_float_type>(0)) ? static_cast<native_float_type>(-f) : f);

if(ff < (std::numeric_limits<native_float_type>::min)())
{
return;
}

using std::frexp;

// Get the fraction and base-2 exponent.

// TBD: Need to properly handle frexp when GCC's __float128
// is active (in case of -std=gnu++XX).
// This happens when native_float_type is of type __float128.

auto man = static_cast<native_float_type>(frexp(static_cast<long double>(f), &my_exponent_part));

auto n2 = static_cast<unsigned>(UINT8_C(0));

for(auto i = static_cast<std::uint_fast16_t>(UINT8_C(0));
i < static_cast<std::uint_fast16_t>(std::numeric_limits<native_float_type>::digits);
++i)
{
// Extract the mantissa of the floating-point type in base-2
// (one bit at a time) and store it in a std::uintmax_t.

man *= 2;

n2 = static_cast<unsigned>(man);
man -= static_cast<native_float_type>(n2);

if(n2 != static_cast<unsigned>(UINT8_C(0)))
{
my_mantissa_part |= static_cast<unsigned>(UINT8_C(1));
}

if(i < static_cast<unsigned>(std::numeric_limits<native_float_type>::digits - 1))
{
my_mantissa_part <<= static_cast<unsigned>(UINT8_C(1));
}
}

// Ensure that the value is normalized and adjust the exponent.

// TBD: Need to properly handle this left shift amount
// when GCC's __float128 is active (in case of -std=gnu++XX).
// This happens when native_float_type is of type __float128.

static_assert(std::numeric_limits<native_float_type>::digits <= std::numeric_limits<long double>::digits,
"Error: The instantiated native float type does not fit in long double");

constexpr auto max_left_shift_amount =
static_cast<int>
(
(std::numeric_limits<native_float_type>::digits < std::numeric_limits<long double>::digits)
? std::numeric_limits<native_float_type>::digits
: std::numeric_limits<long double>::digits
);

#if (defined(__GNUC__) && !defined(__clang__))
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wshift-count-overflow"
#endif

my_mantissa_part |= static_cast<std::uintmax_t>(UINTMAX_C(1) << static_cast<unsigned>(max_left_shift_amount - 1));

#if (defined(__GNUC__) && !defined(__clang__))
#pragma GCC diagnostic pop
#endif

my_exponent_part -= 1;
}

constexpr native_float_parts(const native_float_parts& other) noexcept
: my_mantissa_part(other.my_mantissa_part),
my_exponent_part(other.my_exponent_part) { }

constexpr native_float_parts(native_float_parts&& other) noexcept
: my_mantissa_part(other.my_mantissa_part),
my_exponent_part(other.my_exponent_part) { }

native_float_parts() = delete;

~native_float_parts() noexcept = default; // LCOV_EXCL_LINE

auto operator=(const native_float_parts& other) noexcept -> native_float_parts& // NOLINT(cert-oop54-cpp)
{
if(this != &other)
{
my_mantissa_part = other.my_mantissa_part;
my_exponent_part = other.my_exponent_part;
}

return *this;
}

auto operator=(native_float_parts&& other) noexcept -> native_float_parts&
{
my_mantissa_part = other.my_mantissa_part;
my_exponent_part = other.my_exponent_part;

return *this;
}

WIDE_DECIMAL_NODISCARD constexpr auto get_mantissa() const noexcept -> std::uintmax_t { return my_mantissa_part; }
WIDE_DECIMAL_NODISCARD constexpr auto get_exponent() const noexcept -> int { return my_exponent_part; }

private:
std::uintmax_t my_mantissa_part { }; // NOLINT(readability-identifier-naming)
int my_exponent_part { }; // NOLINT(readability-identifier-naming)
};

#if !defined(WIDE_DECIMAL_DISABLE_CACHED_CONSTANTS)
// LCOV_EXCL_START
// Static data initializer
Expand Down Expand Up @@ -2882,7 +2745,7 @@
{
const auto b_neg = (flt < static_cast<FloatingPointType>(0.0F));

const native_float_parts<FloatingPointType> ld_parts((!b_neg) ? flt : -flt);
const detail::template native_float_parts<FloatingPointType> ld_parts((!b_neg) ? flt : -flt);

// Create a decwide_t from the fractional part of the
// mantissa expressed as an unsigned long long.
Expand Down Expand Up @@ -4415,7 +4278,12 @@
friend constexpr auto half() -> decwide_t<OtherMyDigits10, OtherLimbType, OtherAllocatorType, OtherInternalFloatType, OtherExponentType, OtherFftFloatType>; // NOLINT(readability-redundant-declaration)
};


#if ((defined(__GNUC__) && (__GNUC__ >= 12)) && !defined(__clang__))
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Warray-bounds"
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wstringop-overread"
#endif
template<const std::int32_t ParamDigitsBaseTen, typename LimbType, typename AllocatorType, typename InternalFloatType, typename ExponentType, typename FftFloatType>
constexpr auto zero() -> decwide_t<ParamDigitsBaseTen, LimbType, AllocatorType, InternalFloatType, ExponentType, FftFloatType>
{
Expand All @@ -4426,6 +4294,10 @@

return other_wide_decimal_type::from_lst( { static_cast<other_limb_type>(UINT8_C(0)) } );
}
#if ((defined(__GNUC__) && (__GNUC__ >= 12)) && !defined(__clang__))
#pragma GCC diagnostic pop
#pragma GCC diagnostic pop
#endif

template<const std::int32_t ParamDigitsBaseTen, typename LimbType, typename AllocatorType, typename InternalFloatType, typename ExponentType, typename FftFloatType>
constexpr auto one() -> decwide_t<ParamDigitsBaseTen, LimbType, AllocatorType, InternalFloatType, ExponentType, FftFloatType>
Expand Down
164 changes: 163 additions & 1 deletion math/wide_decimal/decwide_t_detail.h
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
///////////////////////////////////////////////////////////////////
// Copyright Christopher Kormanyos 1999 - 2023. //
// Copyright Christopher Kormanyos 1999 - 2024. //
// 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) //
Expand Down Expand Up @@ -60,12 +60,174 @@

WIDE_DECIMAL_NAMESPACE_BEGIN

#if(__cplusplus >= 201703L)
namespace math::wide_decimal {
#else
namespace math { namespace wide_decimal { // NOLINT(modernize-concat-nested-namespaces)
#endif

// Forward declaration of the decwide_t template class.
template<const std::int32_t ParamDigitsBaseTen,
typename LimbType = std::uint32_t,
typename AllocatorType = std::allocator<void>,
typename InternalFloatType = double,
typename ExponentType = std::int64_t,
typename FftFloatType = double>
class decwide_t;

#if(__cplusplus >= 201703L)
} // namespace math::wide_decimal
#else
} // namespace wide_decimal
} // namespace math
#endif

#if(__cplusplus >= 201703L)
namespace math::wide_decimal::detail {
#else
namespace math { namespace wide_decimal { namespace detail { // NOLINT(modernize-concat-nested-namespaces)
#endif

template<typename FloatingPointType>
class native_float_parts final
{
public:
native_float_parts() = delete;

~native_float_parts() noexcept = default; // LCOV_EXCL_LINE

private:
// Emphasize: This template class can be used with native
// floating-point types like float, double and long double.

template<const std::int32_t ParamDigitsBaseTen,
typename LimbType,
typename AllocatorType,
typename OtherFloatingPointType,
typename ExponentType,
typename FftFloatType>
#if defined(WIDE_DECIMAL_NAMESPACE)
friend class WIDE_DECIMAL_NAMESPACE::math::wide_decimal::decwide_t;
#else
friend class ::math::wide_decimal::decwide_t;
#endif

explicit native_float_parts(FloatingPointType f)
{
using native_float_type = FloatingPointType;

// For long double, verify that the mantissa fits in std::uintmax_t.

static_assert(std::numeric_limits<native_float_type>::digits <= std::numeric_limits<std::uintmax_t>::digits,
"Error: The width of the mantissa does not fit in std::uintmax_t");

const native_float_type ff = ((f < static_cast<native_float_type>(0)) ? static_cast<native_float_type>(-f) : f);

if(ff < (std::numeric_limits<native_float_type>::min)())
{
return;
}

using std::frexp;

// Get the fraction and base-2 exponent.

// TBD: Need to properly handle frexp when GCC's __float128
// is active (in case of -std=gnu++XX).
// This happens when native_float_type is of type __float128.

auto man = static_cast<native_float_type>(frexp(static_cast<long double>(f), &my_exponent_part));

auto n2 = static_cast<unsigned>(UINT8_C(0));

for(auto i = static_cast<std::uint_fast16_t>(UINT8_C(0));
i < static_cast<std::uint_fast16_t>(std::numeric_limits<native_float_type>::digits);
++i)
{
// Extract the mantissa of the floating-point type in base-2
// (one bit at a time) and store it in a std::uintmax_t.

man *= 2;

n2 = static_cast<unsigned>(man);
man -= static_cast<native_float_type>(n2);

if(n2 != static_cast<unsigned>(UINT8_C(0)))
{
my_mantissa_part |= static_cast<unsigned>(UINT8_C(1));
}

if(i < static_cast<unsigned>(std::numeric_limits<native_float_type>::digits - 1))
{
my_mantissa_part <<= static_cast<unsigned>(UINT8_C(1));
}
}

// Ensure that the value is normalized and adjust the exponent.

// TBD: Need to properly handle this left shift amount
// when GCC's __float128 is active (in case of -std=gnu++XX).
// This happens when native_float_type is of type __float128.

static_assert(std::numeric_limits<native_float_type>::digits <= std::numeric_limits<long double>::digits,
"Error: The instantiated native float type does not fit in long double");

constexpr auto max_left_shift_amount =
static_cast<int>
(
(std::numeric_limits<native_float_type>::digits < std::numeric_limits<long double>::digits)
? std::numeric_limits<native_float_type>::digits
: std::numeric_limits<long double>::digits
);

#if (defined(__GNUC__) && !defined(__clang__))
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wshift-count-overflow"
#endif

my_mantissa_part |= static_cast<std::uintmax_t>(UINTMAX_C(1) << static_cast<unsigned>(max_left_shift_amount - 1));

#if (defined(__GNUC__) && !defined(__clang__))
#pragma GCC diagnostic pop
#endif

my_exponent_part -= 1;
}

constexpr native_float_parts(const native_float_parts& other) noexcept
: my_mantissa_part(other.my_mantissa_part),
my_exponent_part(other.my_exponent_part) { }

constexpr native_float_parts(native_float_parts&& other) noexcept
: my_mantissa_part(other.my_mantissa_part),
my_exponent_part(other.my_exponent_part) { }

auto operator=(const native_float_parts& other) noexcept -> native_float_parts& // NOLINT(cert-oop54-cpp)
{
if(this != &other)
{
my_mantissa_part = other.my_mantissa_part;
my_exponent_part = other.my_exponent_part;
}

return *this;
}

auto operator=(native_float_parts&& other) noexcept -> native_float_parts&
{
my_mantissa_part = other.my_mantissa_part;
my_exponent_part = other.my_exponent_part;

return *this;
}

WIDE_DECIMAL_NODISCARD constexpr auto get_mantissa() const noexcept -> std::uintmax_t { return my_mantissa_part; }
WIDE_DECIMAL_NODISCARD constexpr auto get_exponent() const noexcept -> int { return my_exponent_part; }

std::uintmax_t my_mantissa_part { }; // NOLINT(readability-identifier-naming)
int my_exponent_part { }; // NOLINT(readability-identifier-naming)
};

template<const std::size_t BitCount,
typename EnableType = void>
struct uint_type_helper
Expand Down
Loading

0 comments on commit dfd4dba

Please sign in to comment.