RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-27 09:37:34 +00:00
Code Issues Activity Actions

AK: Change standard casting method of FixedPoint to truncation

Browse source 
This matches what floats do.

Also fix typo `trunk`->`trunc`
This commit is contained in:
Hendiadyoin1 2023-07-23 19:53:04 +02:00 • committed by Andrew Kaster
parent 8526791617
commit 05c959e40b
3 changed files with 29 additions and 27 deletions
Download patch file Download diff file Expand all files Collapse all files

38
AK/FixedPoint.h
View file

@ -66,19 +66,7 @@ public:
template<Integral I>
explicit constexpr operator I() const
{
I value = m_value >> precision;
// fract(m_value) >= .5?
if (m_value & (1u << (precision - 1))) {
// fract(m_value) > .5?
if (m_value & (radix_mask >> 2u)) {
// yes: round "up";
value += (m_value > 0 ? 1 : -1);
} else {
// no: round to even;
value += value & 1;
}
}
return value;
return trunc().raw() >> precision;
}
static constexpr This create_raw(Underlying value)
@ -111,9 +99,23 @@ public:
return *this;
}
// Note/FIXME: This uses round to nearest break-tie to even
// Not break-tie away from 0 as the C99's round does
constexpr This round() const
{
return This { static_cast<Underlying>(*this) };
Underlying value = m_value >> precision;
// fract(m_value) >= .5?
if (m_value & (1u << (precision - 1))) {
// fract(m_value) > .5?
if (m_value & (radix_mask >> 2u)) {
// yes: round "up";
value += (m_value > 0 ? 1 : -1);
} else {
// no: round to even;
value += value & 1;
}
}
return value;
}
constexpr This floor() const
{
@ -124,7 +126,7 @@ public:
return create_raw((m_value & ~radix_mask)
+ (m_value & radix_mask ? 1 << precision : 0));
}
constexpr This trunk() const
constexpr This trunc() const
{
return create_raw((m_value & ~radix_mask)
+ ((m_value & radix_mask)
@ -132,14 +134,14 @@ public:
: 0));
}
constexpr Underlying lround() const { return static_cast<Underlying>(*this); }
constexpr Underlying lround() const { return round().raw() >> precision; }
constexpr Underlying lfloor() const { return m_value >> precision; }
constexpr Underlying lceil() const
{
return (m_value >> precision)
+ (m_value & radix_mask ? 1 : 0);
}
constexpr Underlying ltrunk() const
constexpr Underlying ltrunc() const
{
return (m_value >> precision)
+ ((m_value & radix_mask)
@ -437,7 +439,7 @@ struct Formatter<FixedPoint<precision, Underlying>> : StandardFormatter {
if constexpr (IsSigned<Underlying>)
is_negative = value < 0;
i64 integer = value.ltrunk();
i64 integer = value.ltrunc();
constexpr u64 one = static_cast<Underlying>(1) << precision;
u64 fraction_raw = value.raw() & (one - 1);
return builder.put_fixed_point(is_negative, integer, fraction_raw, one, base, upper_case, m_zero_pad, m_use_separator, m_align, m_width.value(), m_precision.value(), m_fill, m_sign_mode, real_number_display_mode);