Skip to content

Instantly share code, notes, and snippets.

@nakst
Last active January 20, 2023 16:40
Show Gist options
  • Star 5 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save nakst/f9c00ef6969fd6ad380bcbae2e10e64a to your computer and use it in GitHub Desktop.
Save nakst/f9c00ef6969fd6ad380bcbae2e10e64a to your computer and use it in GitHub Desktop.
// NOTE Compile without fast math flags.
/*
This is free and unencumbered software released into the public domain.
Anyone is free to copy, modify, publish, use, compile, sell, or
distribute this software, either in source code form or as a compiled
binary, for any purpose, commercial or non-commercial, and by any
means.
In jurisdictions that recognize copyright laws, the author or authors
of this software dedicate any and all copyright interest in the
software to the public domain. We make this dedication for the benefit
of the public at large and to the detriment of our heirs and
successors. We intend this dedication to be an overt act of
relinquishment in perpetuity of all present and future rights to this
software under copyright law.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
OTHER DEALINGS IN THE SOFTWARE.
For more information, please refer to http://unlicense.org/
*/
union ConvertDoubleInteger {
double d;
uint64_t i;
};
union ConvertFloatInteger {
float f;
uint64_t i;
};
double doubleToInteger = 1.0 / 2.22044604925031308085e-16;
double Floor(double x) {
if (x == 0) return x;
ConvertDoubleInteger convert = {x};
uint64_t sign = convert.i & 0x8000000000000000;
int exponent = (int) ((convert.i >> 52) & 0x7FF) - 0x3FF;
if (exponent >= 52) {
// There aren't any bits representing a fractional part.
return x;
} else if (exponent >= 0) {
// Positive exponent.
double y = sign ? (x - doubleToInteger + doubleToInteger - x) : (x + doubleToInteger - doubleToInteger - x);
return y > 0 ? x + y - 1 : x + y;
} else if (exponent < 0) {
// Negative exponent.
return sign ? -1.0 : 0.0;
}
return 0;
}
float FloorFloat(float x) {
ConvertFloatInteger convert = {x};
uint32_t sign = convert.i & 0x80000000;
int exponent = (int) ((convert.i >> 23) & 0xFF) - 0x7F;
if (exponent >= 23) {
// There aren't any bits representing a fractional part.
} else if (exponent >= 0) {
// Positive exponent.
uint32_t mask = 0x7FFFFF >> exponent;
if (!(mask & convert.i)) return x; // Already an integer.
if (sign) convert.i += mask;
convert.i &= ~mask; // Mask out the fractional bits.
} else if (exponent < 0) {
// Negative exponent.
return sign ? -1.0 : 0.0;
}
return convert.f;
}
#define D(x) (((ConvertDoubleInteger) { .i = (x) }).d)
#define F(x) (((ConvertFloatInteger) { .i = (x) }).f)
double _Sine(double x) {
// Calculates sin(x) for x in [0, pi/4].
double x2 = x * x;
return x * (D(0x3FF0000000000000) + x2 * (D(0xBFC5555555555540) + x2 * (D(0x3F8111111110ED80) + x2 * (D(0xBF2A01A019AE6000)
+ x2 * (D(0x3EC71DE349280000) + x2 * (D(0xBE5AE5DC48000000) + x2 * D(0x3DE5D68200000000)))))));
}
float _SineFloat(float x) {
// Calculates sin(x) for x in [0, pi/4].
float x2 = x * x;
return x * (F(0x3F800000) + x2 * (F(0xBE2AAAA0) + x2 * (F(0x3C0882C0) + x2 * F(0xB94C6000))));
}
double _ArcSine(double x) {
// Calculates arcsin(x) for x in [0, 0.5].
double x2 = x * x;
return x * (D(0x3FEFFFFFFFFFFFE6) + x2 * (D(0x3FC555555555FE00) + x2 * (D(0x3FB333333292DF90) + x2 * (D(0x3FA6DB6DFD3693A0)
+ x2 * (D(0x3F9F1C608DE51900) + x2 * (D(0x3F96EA0659B9A080) + x2 * (D(0x3F91B4ABF1029100)
+ x2 * (D(0x3F8DA8DAF31ECD00) + x2 * (D(0x3F81C01FD5000C00) + x2 * (D(0x3F94BDA038CF6B00)
+ x2 * (D(0xBF8E849CA75B1E00) + x2 * D(0x3FA146C2D37F2C60))))))))))));
}
float _ArcSineFloat(float x) {
// Calculates arcsin(x) for x in [0, 0.5].
float x2 = x * x;
return x * (F(0x3F800004) + x2 * (F(0x3E2AA130) + x2 * (F(0x3D9B2C28) + x2 * (F(0x3D1C1800) + x2 * F(0x3D5A97C0)))));
}
double _ArcTangent(double x) {
// Calculates arctan(x) for x in [0, 0.5].
double x2 = x * x;
return x * (D(0x3FEFFFFFFFFFFFF8) + x2 * (D(0xBFD5555555553B44) + x2 * (D(0x3FC9999999803988) + x2 * (D(0xBFC249248C882E80)
+ x2 * (D(0x3FBC71C5A4E4C220) + x2 * (D(0xBFB745B3B75243F0) + x2 * (D(0x3FB3AFAE9A2939E0)
+ x2 * (D(0xBFB1030C4A4A1B90) + x2 * (D(0x3FAD6F65C35579A0) + x2 * (D(0xBFA805BCFDAFEDC0)
+ x2 * (D(0x3F9FC6B5E115F2C0) + x2 * D(0xBF87DCA5AB25BF80))))))))))));
}
float _ArcTangentFloat(float x) {
// Calculates arctan(x) for x in [0, 0.5].
float x2 = x * x;
return x * (F(0x3F7FFFF8) + x2 * (F(0xBEAAA53C) + x2 * (F(0x3E4BC990) + x2 * (F(0xBE084A60) + x2 * F(0x3D8864B0)))));
}
double _Cosine(double x) {
// Calculates cos(x) for x in [0, pi/4].
double x2 = x * x;
return D(0x3FF0000000000000) + x2 * (D(0xBFDFFFFFFFFFFFA0) + x2 * (D(0x3FA555555554F7C0) + x2 * (D(0xBF56C16C16475C00)
+ x2 * (D(0x3EFA019F87490000) + x2 * (D(0xBE927DF66B000000) + x2 * D(0x3E21B949E0000000))))));
}
float _CosineFloat(float x) {
// Calculates cos(x) for x in [0, pi/4].
float x2 = x * x;
return F(0x3F800000) + x2 * (F(0xBEFFFFDA) + x2 * (F(0x3D2A9F60) + x2 * F(0xBAB22C00)));
}
double _Tangent(double x) {
// Calculates tan(x) for x in [0, pi/4].
double x2 = x * x;
return x * (D(0x3FEFFFFFFFFFFFE8) + x2 * (D(0x3FD5555555558000) + x2 * (D(0x3FC1111110FACF90) + x2 * (D(0x3FABA1BA266BFD20)
+ x2 * (D(0x3F9664F30E56E580) + x2 * (D(0x3F822703B08BDC00) + x2 * (D(0x3F6D698D2E4A4C00)
+ x2 * (D(0x3F57FF4F23EA4400) + x2 * (D(0x3F424F3BEC845800) + x2 * (D(0x3F34C78CA9F61000)
+ x2 * (D(0xBF042089F8510000) + x2 * (D(0x3F29D7372D3A8000) + x2 * (D(0xBF19D1C5EF6F0000)
+ x2 * (D(0x3F0980BDF11E8000)))))))))))))));
}
float _TangentFloat(float x) {
// Calculates tan(x) for x in [0, pi/4].
float x2 = x * x;
return x * (F(0x3F800001) + x2 * (F(0x3EAAA9AA) + x2 * (F(0x3E08ABA8) + x2 * (F(0x3D58EC90)
+ x2 * (F(0x3CD24840) + x2 * (F(0x3AC3CA00) + x2 * F(0x3C272F00)))))));
}
double Sine(double x) {
bool negate = false;
// x in -infty, infty
if (x < 0) {
x = -x;
negate = true;
}
// x in 0, infty
x -= 2 * M_PI * Floor(x / (2 * M_PI));
// x in 0, 2*pi
if (x < M_PI / 2) {
} else if (x < M_PI) {
x = M_PI - x;
} else if (x < 3 * M_PI / 2) {
x = x - M_PI;
negate = !negate;
} else {
x = M_PI * 2 - x;
negate = !negate;
}
// x in 0, pi/2
double y = x < M_PI / 4 ? _Sine(x) : _Cosine(M_PI / 2 - x);
return negate ? -y : y;
}
float SineFloat(float x) {
bool negate = false;
// x in -infty, infty
if (x < 0) {
x = -x;
negate = true;
}
// x in 0, infty
x -= 2 * M_PI * FloorFloat(x / (2 * M_PI));
// x in 0, 2*pi
if (x < M_PI / 2) {
} else if (x < M_PI) {
x = M_PI - x;
} else if (x < 3 * M_PI / 2) {
x = x - M_PI;
negate = !negate;
} else {
x = M_PI * 2 - x;
negate = !negate;
}
// x in 0, pi/2
float y = x < M_PI / 4 ? _SineFloat(x) : _CosineFloat(M_PI / 2 - x);
return negate ? -y : y;
}
double Cosine(double x) {
bool negate = false;
// x in -infty, infty
if (x < 0) {
x = -x;
}
// x in 0, infty
x -= 2 * M_PI * Floor(x / (2 * M_PI));
// x in 0, 2*pi
if (x < M_PI / 2) {
} else if (x < M_PI) {
x = M_PI - x;
negate = !negate;
} else if (x < 3 * M_PI / 2) {
x = x - M_PI;
negate = !negate;
} else {
x = M_PI * 2 - x;
}
// x in 0, pi/2
double y = x < M_PI / 4 ? _Cosine(x) : _Sine(M_PI / 2 - x);
return negate ? -y : y;
}
float CosineFloat(float x) {
bool negate = false;
// x in -infty, infty
if (x < 0) {
x = -x;
}
// x in 0, infty
x -= 2 * M_PI * FloorFloat(x / (2 * M_PI));
// x in 0, 2*pi
if (x < M_PI / 2) {
} else if (x < M_PI) {
x = M_PI - x;
negate = !negate;
} else if (x < 3 * M_PI / 2) {
x = x - M_PI;
negate = !negate;
} else {
x = M_PI * 2 - x;
}
// x in 0, pi/2
float y = x < M_PI / 4 ? _CosineFloat(x) : _SineFloat(M_PI / 2 - x);
return negate ? -y : y;
}
double Tangent(double x) {
bool negate = false;
// x in -infty, infty
if (x < 0) {
x = -x;
negate = !negate;
}
// x in 0, infty
x -= M_PI * Floor(x / M_PI);
// x in 0, pi
if (x > M_PI / 2) {
x = M_PI - x;
negate = !negate;
}
// x in 0, pi/2
double y = x < M_PI / 4 ? _Tangent(x) : (1.0 / _Tangent(M_PI / 2 - x));
return negate ? -y : y;
}
float TangentFloat(float x) {
bool negate = false;
// x in -infty, infty
if (x < 0) {
x = -x;
negate = !negate;
}
// x in 0, infty
x -= M_PI * FloorFloat(x / M_PI);
// x in 0, pi
if (x > M_PI / 2) {
x = M_PI - x;
negate = !negate;
}
// x in 0, pi/2
float y = x < M_PI / 4 ? _TangentFloat(x) : (1.0 / _TangentFloat(M_PI / 2 - x));
return negate ? -y : y;
}
double ArcSine(double x) {
bool negate = false;
if (x < 0) {
x = -x;
negate = true;
}
double y;
if (x < 0.5) {
y = _ArcSine(x);
} else {
y = M_PI / 2 - 2 * _ArcSine(sqrt(0.5 - 0.5 * x));
}
return negate ? -y : y;
}
float ArcSineFloat(float x) {
bool negate = false;
if (x < 0) {
x = -x;
negate = true;
}
float y;
if (x < 0.5) {
y = _ArcSineFloat(x);
} else {
y = M_PI / 2 - 2 * _ArcSineFloat(sqrtf(0.5 - 0.5 * x));
}
return negate ? -y : y;
}
double ArcCosine(double x) {
return ArcSine(-x) + M_PI / 2;
}
float ArcCosineFloat(float x) {
return ArcSineFloat(-x) + M_PI / 2;
}
double ArcTangent(double x) {
bool negate = false;
if (x < 0) {
x = -x;
negate = true;
}
bool reciprocalTaken = false;
if (x > 1) {
x = 1 / x;
reciprocalTaken = true;
}
double y;
if (x < 0.5) {
y = _ArcTangent(x);
} else {
y = 0.463647609000806116 + _ArcTangent((2 * x - 1) / (2 + x));
}
if (reciprocalTaken) {
y = M_PI / 2 - y;
}
return negate ? -y : y;
}
float ArcTangentFloat(float x) {
bool negate = false;
if (x < 0) {
x = -x;
negate = true;
}
bool reciprocalTaken = false;
if (x > 1) {
x = 1 / x;
reciprocalTaken = true;
}
float y;
if (x < 0.5f) {
y = _ArcTangentFloat(x);
} else {
y = 0.463647609000806116f + _ArcTangentFloat((2 * x - 1) / (2 + x));
}
if (reciprocalTaken) {
y = M_PI / 2 - y;
}
return negate ? -y : y;
}
double ArcTangent2(double y, double x) {
if (x == 0) return y > 0 ? M_PI / 2 : -M_PI / 2;
else if (x > 0) return ArcTangent(y / x);
else if (y >= 0) return M_PI + ArcTangent(y / x);
else return -M_PI + ArcTangent(y / x);
}
float ArcTangent2Float(float y, float x) {
if (x == 0) return y > 0 ? M_PI / 2 : -M_PI / 2;
else if (x > 0) return ArcTangentFloat(y / x);
else if (y >= 0) return M_PI + ArcTangentFloat(y / x);
else return -M_PI + ArcTangentFloat(y / x);
}
double Exponential2(double x) {
double a = Floor(x * 8);
int64_t ai = a;
if (ai < -1024) {
return 0;
}
double b = x - a / 8;
double y = D(0x3FF0000000000000) + b * (D(0x3FE62E42FEFA3A00) + b * (D(0x3FCEBFBDFF829140)
+ b * (D(0x3FAC6B08D73C4A40) + b * (D(0x3F83B2AB53873280) + b * (D(0x3F55D88F363C6C00)
+ b * (D(0x3F242C003E4A2000) + b * D(0x3EF0B291F6C00000)))))));
const double m[8] = {
D(0x3FF0000000000000),
D(0x3FF172B83C7D517B),
D(0x3FF306FE0A31B715),
D(0x3FF4BFDAD5362A27),
D(0x3FF6A09E667F3BCD),
D(0x3FF8ACE5422AA0DB),
D(0x3FFAE89F995AD3AD),
D(0x3FFD5818DCFBA487),
};
y *= m[ai & 7];
ConvertDoubleInteger c;
c.d = y;
c.i += (ai >> 3) << 52;
return c.d;
}
float Exponential2Float(float x) {
float a = FloorFloat(x);
int32_t ai = a;
if (ai < -128) {
return 0;
}
float b = x - a;
float y = F(0x3F7FFFFE) + b * (F(0x3F31729A) + b * (F(0x3E75E700)
+ b * (F(0x3D64D520) + b * (F(0x3C128280) + b * F(0x3AF89400)))));
ConvertFloatInteger c;
c.f = y;
c.i += ai << 23;
return c.f;
}
double Logarithm2(double x) {
ConvertDoubleInteger c;
c.d = x;
int64_t e = ((c.i >> 52) & 2047) - 0x3FF;
c.i = (c.i & ~(0x7FFL << 52)) + (0x3FFL << 52);
x = c.d;
double a;
if (x < 1.125) {
a = 0;
} else if (x < 1.250) {
x *= 1.125 / 1.250;
a = D(0xBFC374D65D9E608E);
} else if (x < 1.375) {
x *= 1.125 / 1.375;
a = D(0xBFD28746C334FECB);
} else if (x < 1.500) {
x *= 1.125 / 1.500;
a = D(0xBFDA8FF971810A5E);
} else if (x < 1.625) {
x *= 1.125 / 1.625;
a = D(0xBFE0F9F9FFC8932A);
} else if (x < 1.750) {
x *= 1.125 / 1.750;
a = D(0xBFE465D36ED11B11);
} else if (x < 1.875) {
x *= 1.125 / 1.875;
a = D(0xBFE79538DEA712F5);
} else {
x *= 1.125 / 2.000;
a = D(0xBFEA8FF971810A5E);
}
double y = D(0xC00FF8445026AD97) + x * (D(0x40287A7A02D9353F) + x * (D(0xC03711C58D55CEE2)
+ x * (D(0x4040E8263C321A26) + x * (D(0xC041EB22EA691BB3) + x * (D(0x403B00FB376D1F10)
+ x * (D(0xC02C416ABE857241) + x * (D(0x40138BA7FAA3523A) + x * (D(0xBFF019731AF80316)
+ x * D(0x3FB7F1CD3852C200)))))))));
return y - a + e;
}
float Logarithm2Float(float x) {
ConvertFloatInteger c;
c.f = x;
int32_t e = ((c.i >> 23) & 255) - 0x7F;
c.i = (c.i & ~(0xFF << 23)) + (0x7F << 23);
x = c.f;
double y = F(0xC05B5154) + x * (F(0x410297C6) + x * (F(0xC1205CEB)
+ x * (F(0x4114DF63) + x * (F(0xC0C0DBBB) + x * (F(0x402942C6)
+ x * (F(0xBF3FF98A) + x * (F(0x3DFE1050) + x * F(0xBC151480))))))));
return y + e;
}
double Power(double x, double y) {
return Exponential2(y * Logarithm2(x));
}
float PowerFloat(float x, float y) {
return Exponential2Float(y * Logarithm2Float(x));
}
double Modulo(double x, double y) {
return x - y * Floor(x / y);
}
float ModuloFloat(float x, float y) {
return x - y * FloorFloat(x / y);
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment