RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-29 17:35:06 +00:00
Code Issues Activity Actions

Everywhere: Add -Wdouble-promotion warning

Browse source 
This warning informs of float-to-double conversions. The best solution
seems to be to do math *either* in 32-bit *or* in 64-bit, and only to
cross over when absolutely necessary.
This commit is contained in:
Nicholas-Baron 2021-04-15 00:36:14 -07:00 • committed by Andreas Kling
parent 6606d70826
commit 73dd293ec4
26 changed files with 105 additions and 98 deletions
Download patch file Download diff file Expand all files Collapse all files

16
Userland/Libraries/LibM/math.cpp
View file

@ -172,9 +172,9 @@ static FloatType internal_to_integer(FloatType x, RoundingMode rounding_mode)
// We could do this ourselves, but this saves us from manually
// handling overflow.
if (extractor.sign)
extractor.d -= 1.0;
extractor.d -= static_cast<FloatType>(1.0);
else
extractor.d += 1.0;
extractor.d += static_cast<FloatType>(1.0);
}
return extractor.d;
@ -339,7 +339,7 @@ static FloatT internal_gamma(FloatT x) NOEXCEPT
}
// Stirling approximation
return sqrtl(2.0 * M_PI / x) * powl(x / M_E, x);
return sqrtl(2.0 * M_PI / static_cast<long double>(x)) * powl(static_cast<long double>(x) / M_E, static_cast<long double>(x));
}
extern "C" {
@ -386,7 +386,7 @@ double cos(double angle) NOEXCEPT
float cosf(float angle) NOEXCEPT
{
return sinf(angle + M_PI_2);
return sinf(angle + static_cast<float>(M_PI_2));
}
long double sinl(long double angle) NOEXCEPT
@ -636,7 +636,7 @@ double fmod(double index, double period) NOEXCEPT
float fmodf(float index, float period) NOEXCEPT
{
return index - trunc(index / period) * period;
return index - truncf(index / period) * period;
}
// FIXME: These aren't exactly like fmod, but these definitions are probably good enough for now
@ -818,7 +818,7 @@ double acos(double x) NOEXCEPT
float acosf(float x) NOEXCEPT
{
return M_PI_2 - asinf(x);
return static_cast<float>(M_PI_2) - asinf(x);
}
long double fabsl(long double value) NOEXCEPT
@ -1104,9 +1104,9 @@ double lgamma_r(double value, int* sign) NOEXCEPT
float lgammaf_r(float value, int* sign) NOEXCEPT
{
if (value == 1.0 || value == 2.0)
if (value == 1.0f || value == 2.0f)
return 0.0;
if (isinf(value) || value == 0.0)
if (isinf(value) || value == 0.0f)
return INFINITY;
float result = logf(internal_gamma(value));
*sign = signbit(result) ? -1 : 1;