RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-27 23:27:35 +00:00
Code Issues Activity Actions

AK: Introduce fixed-width floating point types (f32, f64, f80 and f128)

Browse source
This commit is contained in:
Dan Klishch 2022-11-04 11:19:27 -04:00 • committed by Andrew Kaster
parent 266882937d
commit 73f4cfa930
2 changed files with 23 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

18
AK/FloatingPoint.h
View file

@ -15,13 +15,9 @@ namespace AK {
template<typename T>
union FloatExtractor;
#if ARCH(I386) || ARCH(X86_64) || ARCH(AARCH64)
// FIXME: There is no straightforward way, I can think of, to check
// in compile time that long double is really an 80-bit IEEE574 floating point number.
static_assert(__LDBL_MAX__ == 1.189731495357231765e4932L);
#ifdef AK_HAS_FLOAT_80
template<>
union FloatExtractor<long double> {
union FloatExtractor<f80> {
static constexpr int mantissa_bits = 64;
static constexpr unsigned long long mantissa_max = ~0u;
static constexpr int exponent_bias = 16383;
@ -32,12 +28,12 @@ union FloatExtractor<long double> {
unsigned exponent : 15;
unsigned sign : 1;
};
long double d;
f80 d;
};
#endif
template<>
union FloatExtractor<double> {
union FloatExtractor<f64> {
static constexpr int mantissa_bits = 52;
static constexpr unsigned long long mantissa_max = (1ull << 52) - 1;
static constexpr int exponent_bias = 1023;
@ -48,11 +44,11 @@ union FloatExtractor<double> {
unsigned exponent : 11;
unsigned sign : 1;
};
double d;
f64 d;
};
template<>
union FloatExtractor<float> {
union FloatExtractor<f32> {
static constexpr int mantissa_bits = 23;
static constexpr unsigned mantissa_max = (1 << 23) - 1;
static constexpr int exponent_bias = 127;
@ -63,7 +59,7 @@ union FloatExtractor<float> {
unsigned exponent : 8;
unsigned sign : 1;
};
float d;
f32 d;
};
template<size_t S, size_t E, size_t M>