mirror of
https://github.com/CraftyBoss/SuperMarioOdysseyOnline.git
synced 2024-11-29 06:33:06 +00:00
565 lines
11 KiB
C++
565 lines
11 KiB
C++
#pragma once
|
|
|
|
#include <basis/seadRawPrint.h>
|
|
#include <cmath>
|
|
#ifndef SEAD_MATH_MATH_CALC_COMMON_H_
|
|
#include <math/seadMathCalcCommon.h>
|
|
#endif
|
|
#include <prim/seadBitUtil.h>
|
|
#include <type_traits>
|
|
|
|
namespace sead
|
|
{
|
|
template <>
|
|
const MathCalcCommon<float>::SinCosSample MathCalcCommon<float>::cSinCosTbl[];
|
|
|
|
template <>
|
|
const MathCalcCommon<float>::AtanSample MathCalcCommon<float>::cAtanTbl[];
|
|
|
|
template <>
|
|
const MathCalcCommon<float>::ExpSample MathCalcCommon<float>::cExpTbl[];
|
|
|
|
template <>
|
|
const MathCalcCommon<float>::LogSample MathCalcCommon<float>::cLogTbl[];
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::sign(T value)
|
|
{
|
|
return value < 0 ? -1 : 1;
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::sqrt(T t)
|
|
{
|
|
return std::sqrt(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::rsqrt(T t)
|
|
{
|
|
return 1 / std::sqrt(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::pow(T x, T y)
|
|
{
|
|
return std::pow(x, y);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::sin(T t)
|
|
{
|
|
if constexpr (std::is_same_v<T, float>)
|
|
{
|
|
const auto as_int = BitUtil::bitCast<u32>(t);
|
|
const SinCosSample& sample = cSinCosTbl[as_int >> 18];
|
|
return sample.sin_val + sample.sin_delta * (as_int & 0xFFFFFFu) * 0x1p-24f;
|
|
}
|
|
else
|
|
{
|
|
static_assert(!std::is_same<T, T>(), "Unsupported type");
|
|
}
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::cos(T t)
|
|
{
|
|
if constexpr (std::is_same_v<T, float>)
|
|
{
|
|
const auto as_int = BitUtil::bitCast<u32>(t);
|
|
const SinCosSample& sample = cSinCosTbl[as_int >> 18];
|
|
return sample.cos_val + sample.cos_delta * (as_int & 0xFFFFFFu) * 0x1p-24f;
|
|
}
|
|
else
|
|
{
|
|
static_assert(!std::is_same<T, T>(), "Unsupported type");
|
|
}
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::tan(T t)
|
|
{
|
|
return std::tan(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::asin(T t)
|
|
{
|
|
if constexpr (std::is_floating_point<T>())
|
|
SEAD_ASSERT_MSG(-1.0 <= t && t <= 1.0, "t(%f) is not in [-1, 1].", t);
|
|
return std::asin(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::acos(T t)
|
|
{
|
|
if constexpr (std::is_floating_point<T>())
|
|
SEAD_ASSERT_MSG(-1.0 <= t && t <= 1.0, "t(%f) is not in [-1, 1].", t);
|
|
return std::acos(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::atan(T t)
|
|
{
|
|
return std::atan(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::atan2(T y, T x)
|
|
{
|
|
return std::atan2(y, x);
|
|
}
|
|
|
|
template <>
|
|
inline f32 MathCalcCommon<f32>::sinIdx(u32 idx)
|
|
{
|
|
u32 index = (idx >> 24) & 0xff;
|
|
u32 rest = idx & 0xffffff;
|
|
|
|
return cSinCosTbl[index].sin_val + cSinCosTbl[index].sin_delta * rest / 0x1000000;
|
|
}
|
|
|
|
template <>
|
|
inline f32 MathCalcCommon<f32>::cosIdx(u32 idx)
|
|
{
|
|
u32 index = (idx >> 24) & 0xff;
|
|
u32 rest = idx & 0xffffff;
|
|
|
|
return cSinCosTbl[index].cos_val + cSinCosTbl[index].cos_delta * rest / 0x1000000;
|
|
}
|
|
|
|
template <>
|
|
inline f32 MathCalcCommon<f32>::tanIdx(u32 idx)
|
|
{
|
|
u32 index = (idx >> 24) & 0xff;
|
|
f32 rest = static_cast<f32>(idx & 0xffffff) / 0x1000000;
|
|
const SinCosSample& sample = cSinCosTbl[index];
|
|
|
|
return (sample.sin_val + sample.sin_delta * rest) / (sample.cos_val + sample.cos_delta * rest);
|
|
}
|
|
|
|
template <>
|
|
inline u32 MathCalcCommon<f32>::asinIdx(f32 s)
|
|
{
|
|
SEAD_ASSERT_MSG(s <= 1 && s >= -1, "s(%f) is not in [-1, 1].", s);
|
|
|
|
const f32 rsqrt_2 = 0.7071067690849304f; // rsqrt(2)
|
|
|
|
if (s >= 0)
|
|
{
|
|
if (s > rsqrt_2)
|
|
return 0x40000000 - atanIdx_(sqrt(1 - s * s) / s);
|
|
|
|
else
|
|
return atanIdx_(s / sqrt(1 - s * s));
|
|
}
|
|
else
|
|
{
|
|
if (s < -rsqrt_2)
|
|
return 0xC0000000 + atanIdx_(-sqrt(1 - s * s) / s);
|
|
|
|
else
|
|
return -atanIdx_(-s / sqrt(1 - s * s));
|
|
}
|
|
}
|
|
|
|
template <>
|
|
inline u32 MathCalcCommon<f32>::acosIdx(f32 c)
|
|
{
|
|
SEAD_ASSERT_MSG(c <= 1 && c >= -1, "c(%f) is not in [-1, 1].", c);
|
|
|
|
const f32 rsqrt_2 = 0.7071067690849304f; // rsqrt(2)
|
|
|
|
if (c >= 0)
|
|
{
|
|
if (c > rsqrt_2)
|
|
return atanIdx_(sqrt(1 - c * c) / c);
|
|
|
|
else
|
|
return 0x40000000 - atanIdx_(c / sqrt(1 - c * c));
|
|
}
|
|
else
|
|
{
|
|
if (c < -rsqrt_2)
|
|
return 0x80000000 - atanIdx_(-sqrt(1 - c * c) / c);
|
|
|
|
else
|
|
return 0x40000000 + atanIdx_(-c / sqrt(1 - c * c));
|
|
}
|
|
}
|
|
|
|
template <>
|
|
inline u32 MathCalcCommon<f32>::atanIdx(f32 t)
|
|
{
|
|
if (t >= 0)
|
|
{
|
|
if (t > 1)
|
|
return 0x40000000 - atanIdx_(1 / t);
|
|
|
|
else
|
|
return atanIdx_(t);
|
|
}
|
|
else
|
|
{
|
|
if (t < -1)
|
|
return 0xC0000000 + atanIdx_(-1 / t);
|
|
|
|
else
|
|
return -atanIdx_(-t);
|
|
}
|
|
}
|
|
|
|
template <>
|
|
inline u32 MathCalcCommon<f32>::atan2Idx(f32 y, f32 x)
|
|
{
|
|
if (x == 0 && y == 0)
|
|
return 0;
|
|
|
|
if (x >= 0)
|
|
{
|
|
if (y >= 0)
|
|
{
|
|
if (x >= y)
|
|
return atanIdx_(y / x);
|
|
|
|
else
|
|
return 0x40000000 - atanIdx_(x / y);
|
|
}
|
|
else
|
|
{
|
|
if (x >= -y)
|
|
return -atanIdx_(-y / x);
|
|
|
|
else
|
|
return 0xC0000000 + atanIdx_(x / -y);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
if (y >= 0)
|
|
{
|
|
if (-x >= y)
|
|
return 0x80000000 - atanIdx_(y / -x);
|
|
|
|
else
|
|
return 0x40000000 + atanIdx_(-x / y);
|
|
}
|
|
else
|
|
{
|
|
if (x <= y)
|
|
return 0x80000000 + atanIdx_(y / x);
|
|
|
|
else
|
|
return 0xC0000000 - atanIdx_(x / y);
|
|
}
|
|
}
|
|
}
|
|
|
|
template <>
|
|
inline void MathCalcCommon<f32>::sinCosIdx(f32* pSin, f32* pCos, u32 idx)
|
|
{
|
|
u32 index = (idx >> 24) & 0xff;
|
|
f32 rest = static_cast<f32>(idx & 0xffffff) / 0x1000000;
|
|
const SinCosSample& sample = cSinCosTbl[index];
|
|
|
|
/*if (pSin != NULL)*/ *pSin = sample.sin_val + sample.sin_delta * rest;
|
|
/*if (pCos != NULL)*/ *pCos = sample.cos_val + sample.cos_delta * rest;
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::exp(T t)
|
|
{
|
|
return std::exp(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::log(T t)
|
|
{
|
|
return std::log(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::log2(T n)
|
|
{
|
|
return std::log2(n);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::log10(T t)
|
|
{
|
|
return std::log10(t);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::minNumber()
|
|
{
|
|
return std::numeric_limits<T>::min();
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::maxNumber()
|
|
{
|
|
return std::numeric_limits<T>::max();
|
|
}
|
|
|
|
template <>
|
|
inline float MathCalcCommon<float>::minNumber()
|
|
{
|
|
return -std::numeric_limits<float>::max();
|
|
}
|
|
|
|
template <>
|
|
inline float MathCalcCommon<float>::maxNumber()
|
|
{
|
|
return std::numeric_limits<float>::max();
|
|
}
|
|
|
|
template <>
|
|
inline double MathCalcCommon<double>::minNumber()
|
|
{
|
|
return -std::numeric_limits<double>::max();
|
|
}
|
|
|
|
template <>
|
|
inline double MathCalcCommon<double>::maxNumber()
|
|
{
|
|
return std::numeric_limits<double>::max();
|
|
}
|
|
|
|
template <>
|
|
inline long double MathCalcCommon<long double>::minNumber()
|
|
{
|
|
return -std::numeric_limits<long double>::max();
|
|
}
|
|
|
|
template <>
|
|
inline long double MathCalcCommon<long double>::maxNumber()
|
|
{
|
|
return std::numeric_limits<long double>::max();
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::infinity()
|
|
{
|
|
return std::numeric_limits<T>::infinity();
|
|
}
|
|
|
|
template <>
|
|
inline f32 MathCalcCommon<f32>::nan()
|
|
{
|
|
return BitUtil::bitCast<f32>(0x7FFFFFFF);
|
|
}
|
|
|
|
template <>
|
|
inline f64 MathCalcCommon<f64>::nan()
|
|
{
|
|
return BitUtil::bitCast<f64>(0x7FFFFFFFFFFFFFFF);
|
|
}
|
|
|
|
template <>
|
|
inline s32 MathCalcCommon<s32>::abs(s32 x)
|
|
{
|
|
return (x ^ x >> 31) - (x >> 31);
|
|
}
|
|
|
|
template <>
|
|
inline u32 MathCalcCommon<u32>::abs(u32 x)
|
|
{
|
|
return x;
|
|
}
|
|
|
|
#ifdef cafe
|
|
|
|
template <>
|
|
inline f32 MathCalcCommon<f32>::abs(f32 x)
|
|
{
|
|
return std::abs(x);
|
|
}
|
|
|
|
template <>
|
|
inline f64 MathCalcCommon<f64>::abs(f64 x)
|
|
{
|
|
return std::abs(x);
|
|
}
|
|
|
|
#endif // cafe
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::max(T a, T b)
|
|
{
|
|
return a < b ? b : a;
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::min(T a, T b)
|
|
{
|
|
return a < b ? a : b;
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::max3(T a, T b, T c)
|
|
{
|
|
return max(max(a, b), c);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::min3(T a, T b, T c)
|
|
{
|
|
return min(min(a, b), c);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::deg2rad(T deg)
|
|
{
|
|
return deg * (numbers::pi_v<T> / static_cast<T>(180));
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::rad2deg(T rad)
|
|
{
|
|
return rad * (static_cast<T>(180) / numbers::pi_v<T>);
|
|
}
|
|
|
|
template <typename T>
|
|
inline u32 MathCalcCommon<T>::deg2idx(T a)
|
|
{
|
|
return static_cast<u32>(a * (cHalfRoundIdx / static_cast<T>(180)));
|
|
}
|
|
|
|
template <typename T>
|
|
inline u32 MathCalcCommon<T>::rad2idx(T a)
|
|
{
|
|
return static_cast<u32>(a * (cHalfRoundIdx / pi()));
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::idx2deg(u32 a)
|
|
{
|
|
return static_cast<T>(a * (static_cast<T>(180) / cHalfRoundIdx));
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::idx2rad(u32 a)
|
|
{
|
|
return static_cast<T>(a * (pi() / cHalfRoundIdx));
|
|
}
|
|
|
|
template <typename T>
|
|
inline s32 MathCalcCommon<T>::roundOff(T val)
|
|
{
|
|
return std::floor(val + 0.5f);
|
|
}
|
|
|
|
template <>
|
|
inline s32 MathCalcCommon<s32>::roundOff(s32 val)
|
|
{
|
|
return val;
|
|
}
|
|
|
|
template <typename T>
|
|
inline s32 MathCalcCommon<T>::floor(T val)
|
|
{
|
|
return std::floor(val);
|
|
}
|
|
|
|
template <>
|
|
inline s32 MathCalcCommon<s32>::floor(s32 val)
|
|
{
|
|
return val;
|
|
}
|
|
|
|
template <typename T>
|
|
inline s32 MathCalcCommon<T>::ceil(T val)
|
|
{
|
|
return std::ceil(val);
|
|
}
|
|
|
|
template <>
|
|
inline s32 MathCalcCommon<s32>::ceil(s32 val)
|
|
{
|
|
return val;
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::roundUp(T x, s32 multNumber)
|
|
{
|
|
SEAD_ASSERT(multNumber > 0);
|
|
return (x + multNumber - 1) / multNumber * multNumber;
|
|
}
|
|
|
|
template <>
|
|
inline s32 MathCalcCommon<u32>::roundUpPow2(u32 val, s32 base)
|
|
{
|
|
SEAD_ASSERT_MSG((u32(base - 1) & u32(base)) == 0, "illegal param[val:%d, base:%d]", val, base);
|
|
return (val + base - 1) & (u32)-base;
|
|
}
|
|
|
|
template <>
|
|
inline s32 MathCalcCommon<s32>::roundUpPow2(s32 val, s32 base)
|
|
{
|
|
SEAD_ASSERT_MSG(val >= 0 && (u32(base - 1) & u32(base)) == 0, "illegal param[val:%d, base:%d]",
|
|
val, base);
|
|
return (val + base - 1) & (u32)-base;
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::clampMax(T val, T max_)
|
|
{
|
|
return min(val, max_);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::clampMin(T val, T min_)
|
|
{
|
|
return max(val, min_);
|
|
}
|
|
|
|
template <typename T>
|
|
inline T MathCalcCommon<T>::clamp(T value, T low, T high)
|
|
{
|
|
if (value < low)
|
|
value = low;
|
|
else if (value > high)
|
|
value = high;
|
|
return value;
|
|
}
|
|
|
|
template <typename T>
|
|
inline bool MathCalcCommon<T>::chase(T* value, T target, T step)
|
|
{
|
|
const T current = *value;
|
|
|
|
if (current < target)
|
|
{
|
|
const T new_value = current + step;
|
|
if (target <= new_value || new_value < current)
|
|
{
|
|
*value = target;
|
|
return true;
|
|
}
|
|
*value = new_value;
|
|
return false;
|
|
}
|
|
|
|
if (current > target)
|
|
{
|
|
const T new_value = current - step;
|
|
if (target >= new_value || current < new_value)
|
|
{
|
|
*value = target;
|
|
return true;
|
|
}
|
|
*value = new_value;
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
template <typename T, typename T2>
|
|
inline T lerp(T a, T b, T2 t)
|
|
{
|
|
return a + (b - a) * t;
|
|
}
|
|
|
|
} // namespace sead
|