summaryrefslogtreecommitdiffstats
path: root/external/include/glm/detail/func_common.inl
diff options
context:
space:
mode:
Diffstat (limited to 'external/include/glm/detail/func_common.inl')
-rw-r--r--external/include/glm/detail/func_common.inl826
1 files changed, 0 insertions, 826 deletions
diff --git a/external/include/glm/detail/func_common.inl b/external/include/glm/detail/func_common.inl
deleted file mode 100644
index 8b2d0b0..0000000
--- a/external/include/glm/detail/func_common.inl
+++ /dev/null
@@ -1,826 +0,0 @@
-/// @ref core
-/// @file glm/detail/func_common.inl
-
-#include "../vector_relational.hpp"
-#include "type_vec2.hpp"
-#include "type_vec3.hpp"
-#include "type_vec4.hpp"
-#include "_vectorize.hpp"
-#include <limits>
-
-namespace glm
-{
- // min
- template<typename genType>
- GLM_FUNC_QUALIFIER genType min(genType x, genType y)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || std::numeric_limits<genType>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'min' only accept floating-point or integer inputs");
- return (y < x) ? y : x;
- }
-
- // max
- template<typename genType>
- GLM_FUNC_QUALIFIER genType max(genType x, genType y)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || std::numeric_limits<genType>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'max' only accept floating-point or integer inputs");
-
- return (x < y) ? y : x;
- }
-
- // abs
- template<>
- GLM_FUNC_QUALIFIER int32 abs(int32 x)
- {
- int32 const y = x >> 31;
- return (x ^ y) - y;
- }
-
- // round
-# if GLM_HAS_CXX11_STL
- using ::std::round;
-# else
- template<typename genType>
- GLM_FUNC_QUALIFIER genType round(genType x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'round' only accept floating-point inputs");
-
- return x < static_cast<genType>(0) ? static_cast<genType>(int(x - static_cast<genType>(0.5))) : static_cast<genType>(int(x + static_cast<genType>(0.5)));
- }
-# endif
-
- // trunc
-# if GLM_HAS_CXX11_STL
- using ::std::trunc;
-# else
- template<typename genType>
- GLM_FUNC_QUALIFIER genType trunc(genType x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'trunc' only accept floating-point inputs");
-
- return x < static_cast<genType>(0) ? -std::floor(-x) : std::floor(x);
- }
-# endif
-
-}//namespace glm
-
-namespace glm{
-namespace detail
-{
- template<typename genFIType, bool /*signed*/>
- struct compute_abs
- {};
-
- template<typename genFIType>
- struct compute_abs<genFIType, true>
- {
- GLM_FUNC_QUALIFIER static genFIType call(genFIType x)
- {
- GLM_STATIC_ASSERT(
- std::numeric_limits<genFIType>::is_iec559 || std::numeric_limits<genFIType>::is_signed || GLM_UNRESTRICTED_GENTYPE,
- "'abs' only accept floating-point and integer scalar or vector inputs");
-
- return x >= genFIType(0) ? x : -x;
- // TODO, perf comp with: *(((int *) &x) + 1) &= 0x7fffffff;
- }
- };
-
- #if GLM_COMPILER & GLM_COMPILER_CUDA
- template<>
- struct compute_abs<float, true>
- {
- GLM_FUNC_QUALIFIER static float call(float x)
- {
- return fabsf(x);
- }
- };
- #endif
-
- template<typename genFIType>
- struct compute_abs<genFIType, false>
- {
- GLM_FUNC_QUALIFIER static genFIType call(genFIType x)
- {
- GLM_STATIC_ASSERT(
- (!std::numeric_limits<genFIType>::is_signed && std::numeric_limits<genFIType>::is_integer) || GLM_UNRESTRICTED_GENTYPE,
- "'abs' only accept floating-point and integer scalar or vector inputs");
- return x;
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_abs_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return detail::functor1<L, T, T, Q>::call(abs, x);
- }
- };
-
- template<length_t L, typename T, typename U, qualifier Q, bool Aligned>
- struct compute_mix_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, U, Q> const& a)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<U>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'mix' only accept floating-point inputs for the interpolator a");
-
- return vec<L, T, Q>(vec<L, U, Q>(x) + a * vec<L, U, Q>(y - x));
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_mix_vector<L, T, bool, Q, Aligned>
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, bool, Q> const& a)
- {
- vec<L, T, Q> Result;
- for(length_t i = 0; i < x.length(); ++i)
- Result[i] = a[i] ? y[i] : x[i];
- return Result;
- }
- };
-
- template<length_t L, typename T, typename U, qualifier Q, bool Aligned>
- struct compute_mix_scalar
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, U const& a)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<U>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'mix' only accept floating-point inputs for the interpolator a");
-
- return vec<L, T, Q>(vec<L, U, Q>(x) + a * vec<L, U, Q>(y - x));
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_mix_scalar<L, T, bool, Q, Aligned>
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, bool const& a)
- {
- return a ? y : x;
- }
- };
-
- template<typename T, typename U>
- struct compute_mix
- {
- GLM_FUNC_QUALIFIER static T call(T const& x, T const& y, U const& a)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<U>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'mix' only accept floating-point inputs for the interpolator a");
-
- return static_cast<T>(static_cast<U>(x) + a * static_cast<U>(y - x));
- }
- };
-
- template<typename T>
- struct compute_mix<T, bool>
- {
- GLM_FUNC_QUALIFIER static T call(T const& x, T const& y, bool const& a)
- {
- return a ? y : x;
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool isFloat, bool Aligned>
- struct compute_sign
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return vec<L, T, Q>(glm::lessThan(vec<L, T, Q>(0), x)) - vec<L, T, Q>(glm::lessThan(x, vec<L, T, Q>(0)));
- }
- };
-
-# if GLM_ARCH == GLM_ARCH_X86
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_sign<L, T, Q, false, Aligned>
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- T const Shift(static_cast<T>(sizeof(T) * 8 - 1));
- vec<L, T, Q> const y(vec<L, typename make_unsigned<T>::type, Q>(-x) >> typename make_unsigned<T>::type(Shift));
-
- return (x >> Shift) | y;
- }
- };
-# endif
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_floor
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return detail::functor1<L, T, T, Q>::call(std::floor, x);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_ceil
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return detail::functor1<L, T, T, Q>::call(std::ceil, x);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_fract
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return x - floor(x);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_trunc
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return detail::functor1<L, T, T, Q>::call(trunc, x);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_round
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
- {
- return detail::functor1<L, T, T, Q>::call(round, x);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_mod
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'mod' only accept floating-point inputs. Include <glm/gtc/integer.hpp> for integer inputs.");
- return a - b * floor(a / b);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_min_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
- {
- return detail::functor2<L, T, Q>::call(min, x, y);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_max_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
- {
- return detail::functor2<L, T, Q>::call(max, x, y);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_clamp_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal)
- {
- return min(max(x, minVal), maxVal);
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_step_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& edge, vec<L, T, Q> const& x)
- {
- return mix(vec<L, T, Q>(1), vec<L, T, Q>(0), glm::lessThan(x, edge));
- }
- };
-
- template<length_t L, typename T, qualifier Q, bool Aligned>
- struct compute_smoothstep_vector
- {
- GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& edge0, vec<L, T, Q> const& edge1, vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'step' only accept floating-point inputs");
- vec<L, T, Q> const tmp(clamp((x - edge0) / (edge1 - edge0), static_cast<T>(0), static_cast<T>(1)));
- return tmp * tmp * (static_cast<T>(3) - static_cast<T>(2) * tmp);
- }
- };
-}//namespace detail
-
- template<typename genFIType>
- GLM_FUNC_QUALIFIER genFIType abs(genFIType x)
- {
- return detail::compute_abs<genFIType, std::numeric_limits<genFIType>::is_signed>::call(x);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> abs(vec<L, T, Q> const& x)
- {
- return detail::compute_abs_vector<L, T, Q, detail::is_aligned<Q>::value>::call(x);
- }
-
- // sign
- // fast and works for any type
- template<typename genFIType>
- GLM_FUNC_QUALIFIER genFIType sign(genFIType x)
- {
- GLM_STATIC_ASSERT(
- std::numeric_limits<genFIType>::is_iec559 || (std::numeric_limits<genFIType>::is_signed && std::numeric_limits<genFIType>::is_integer),
- "'sign' only accept signed inputs");
-
- return detail::compute_sign<1, genFIType, defaultp, std::numeric_limits<genFIType>::is_iec559, highp>::call(vec<1, genFIType>(x)).x;
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> sign(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(
- std::numeric_limits<T>::is_iec559 || (std::numeric_limits<T>::is_signed && std::numeric_limits<T>::is_integer),
- "'sign' only accept signed inputs");
-
- return detail::compute_sign<L, T, Q, std::numeric_limits<T>::is_iec559, detail::is_aligned<Q>::value>::call(x);
- }
-
- // floor
- using ::std::floor;
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> floor(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'floor' only accept floating-point inputs.");
- return detail::compute_floor<L, T, Q, detail::is_aligned<Q>::value>::call(x);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> trunc(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'trunc' only accept floating-point inputs");
- return detail::compute_trunc<L, T, Q, detail::is_aligned<Q>::value>::call(x);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> round(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'round' only accept floating-point inputs");
- return detail::compute_round<L, T, Q, detail::is_aligned<Q>::value>::call(x);
- }
-
-/*
- // roundEven
- template<typename genType>
- GLM_FUNC_QUALIFIER genType roundEven(genType const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'roundEven' only accept floating-point inputs");
-
- return genType(int(x + genType(int(x) % 2)));
- }
-*/
-
- // roundEven
- template<typename genType>
- GLM_FUNC_QUALIFIER genType roundEven(genType x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'roundEven' only accept floating-point inputs");
-
- int Integer = static_cast<int>(x);
- genType IntegerPart = static_cast<genType>(Integer);
- genType FractionalPart = fract(x);
-
- if(FractionalPart > static_cast<genType>(0.5) || FractionalPart < static_cast<genType>(0.5))
- {
- return round(x);
- }
- else if((Integer % 2) == 0)
- {
- return IntegerPart;
- }
- else if(x <= static_cast<genType>(0)) // Work around...
- {
- return IntegerPart - static_cast<genType>(1);
- }
- else
- {
- return IntegerPart + static_cast<genType>(1);
- }
- //else // Bug on MinGW 4.5.2
- //{
- // return mix(IntegerPart + genType(-1), IntegerPart + genType(1), x <= genType(0));
- //}
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> roundEven(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'roundEven' only accept floating-point inputs");
- return detail::functor1<L, T, T, Q>::call(roundEven, x);
- }
-
- // ceil
- using ::std::ceil;
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> ceil(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'ceil' only accept floating-point inputs");
- return detail::compute_ceil<L, T, Q, detail::is_aligned<Q>::value>::call(x);
- }
-
- // fract
- template<typename genType>
- GLM_FUNC_QUALIFIER genType fract(genType x)
- {
- return fract(vec<1, genType>(x)).x;
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> fract(vec<L, T, Q> const& x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fract' only accept floating-point inputs");
- return detail::compute_fract<L, T, Q, detail::is_aligned<Q>::value>::call(x);
- }
-
- // mod
- template<typename genType>
- GLM_FUNC_QUALIFIER genType mod(genType x, genType y)
- {
-# if GLM_COMPILER & GLM_COMPILER_CUDA
- // Another Cuda compiler bug https://github.com/g-truc/glm/issues/530
- vec<1, genType, defaultp> Result(mod(vec<1, genType, defaultp>(x), y));
- return Result.x;
-# else
- return mod(vec<1, genType, defaultp>(x), y).x;
-# endif
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> mod(vec<L, T, Q> const& x, T y)
- {
- return detail::compute_mod<L, T, Q, detail::is_aligned<Q>::value>::call(x, vec<L, T, Q>(y));
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> mod(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
- {
- return detail::compute_mod<L, T, Q, detail::is_aligned<Q>::value>::call(x, y);
- }
-
- // modf
- template<typename genType>
- GLM_FUNC_QUALIFIER genType modf(genType x, genType & i)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'modf' only accept floating-point inputs");
- return std::modf(x, &i);
- }
-
- template<typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<1, T, Q> modf(vec<1, T, Q> const& x, vec<1, T, Q> & i)
- {
- return vec<1, T, Q>(
- modf(x.x, i.x));
- }
-
- template<typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<2, T, Q> modf(vec<2, T, Q> const& x, vec<2, T, Q> & i)
- {
- return vec<2, T, Q>(
- modf(x.x, i.x),
- modf(x.y, i.y));
- }
-
- template<typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<3, T, Q> modf(vec<3, T, Q> const& x, vec<3, T, Q> & i)
- {
- return vec<3, T, Q>(
- modf(x.x, i.x),
- modf(x.y, i.y),
- modf(x.z, i.z));
- }
-
- template<typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<4, T, Q> modf(vec<4, T, Q> const& x, vec<4, T, Q> & i)
- {
- return vec<4, T, Q>(
- modf(x.x, i.x),
- modf(x.y, i.y),
- modf(x.z, i.z),
- modf(x.w, i.w));
- }
-
- //// Only valid if (INT_MIN <= x-y <= INT_MAX)
- //// min(x,y)
- //r = y + ((x - y) & ((x - y) >> (sizeof(int) *
- //CHAR_BIT - 1)));
- //// max(x,y)
- //r = x - ((x - y) & ((x - y) >> (sizeof(int) *
- //CHAR_BIT - 1)));
-
- // min
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> min(vec<L, T, Q> const& a, T b)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'min' only accept floating-point or integer inputs");
- return detail::compute_min_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, vec<L, T, Q>(b));
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> min(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
- {
- return detail::compute_min_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, b);
- }
-
- // max
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> max(vec<L, T, Q> const& a, T b)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'max' only accept floating-point or integer inputs");
- return detail::compute_max_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, vec<L, T, Q>(b));
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> max(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
- {
- return detail::compute_max_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, b);
- }
-
- // clamp
- template<typename genType>
- GLM_FUNC_QUALIFIER genType clamp(genType x, genType minVal, genType maxVal)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || std::numeric_limits<genType>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'clamp' only accept floating-point or integer inputs");
- return min(max(x, minVal), maxVal);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> clamp(vec<L, T, Q> const& x, T minVal, T maxVal)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'clamp' only accept floating-point or integer inputs");
- return detail::compute_clamp_vector<L, T, Q, detail::is_aligned<Q>::value>::call(x, vec<L, T, Q>(minVal), vec<L, T, Q>(maxVal));
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> clamp(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer || GLM_UNRESTRICTED_GENTYPE, "'clamp' only accept floating-point or integer inputs");
- return detail::compute_clamp_vector<L, T, Q, detail::is_aligned<Q>::value>::call(x, minVal, maxVal);
- }
-
- template<typename genTypeT, typename genTypeU>
- GLM_FUNC_QUALIFIER genTypeT mix(genTypeT x, genTypeT y, genTypeU a)
- {
- return detail::compute_mix<genTypeT, genTypeU>::call(x, y, a);
- }
-
- template<length_t L, typename T, typename U, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> mix(vec<L, T, Q> const& x, vec<L, T, Q> const& y, U a)
- {
- return detail::compute_mix_scalar<L, T, U, Q, detail::is_aligned<Q>::value>::call(x, y, a);
- }
-
- template<length_t L, typename T, typename U, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> mix(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, U, Q> const& a)
- {
- return detail::compute_mix_vector<L, T, U, Q, detail::is_aligned<Q>::value>::call(x, y, a);
- }
-
- // step
- template<typename genType>
- GLM_FUNC_QUALIFIER genType step(genType edge, genType x)
- {
- return mix(static_cast<genType>(1), static_cast<genType>(0), x < edge);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> step(T edge, vec<L, T, Q> const& x)
- {
- return detail::compute_step_vector<L, T, Q, detail::is_aligned<Q>::value>::call(vec<L, T, Q>(edge), x);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> step(vec<L, T, Q> const& edge, vec<L, T, Q> const& x)
- {
- return detail::compute_step_vector<L, T, Q, detail::is_aligned<Q>::value>::call(edge, x);
- }
-
- // smoothstep
- template<typename genType>
- GLM_FUNC_QUALIFIER genType smoothstep(genType edge0, genType edge1, genType x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'smoothstep' only accept floating-point inputs");
-
- genType const tmp(clamp((x - edge0) / (edge1 - edge0), genType(0), genType(1)));
- return tmp * tmp * (genType(3) - genType(2) * tmp);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> smoothstep(T edge0, T edge1, vec<L, T, Q> const& x)
- {
- return detail::compute_smoothstep_vector<L, T, Q, detail::is_aligned<Q>::value>::call(vec<L, T, Q>(edge0), vec<L, T, Q>(edge1), x);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> smoothstep(vec<L, T, Q> const& edge0, vec<L, T, Q> const& edge1, vec<L, T, Q> const& x)
- {
- return detail::compute_smoothstep_vector<L, T, Q, detail::is_aligned<Q>::value>::call(edge0, edge1, x);
- }
-
-# if GLM_HAS_CXX11_STL
- using std::isnan;
-# else
- template<typename genType>
- GLM_FUNC_QUALIFIER bool isnan(genType x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'isnan' only accept floating-point inputs");
-
-# if GLM_HAS_CXX11_STL
- return std::isnan(x);
-# elif GLM_COMPILER & GLM_COMPILER_VC
- return _isnan(x) != 0;
-# elif GLM_COMPILER & GLM_COMPILER_INTEL
-# if GLM_PLATFORM & GLM_PLATFORM_WINDOWS
- return _isnan(x) != 0;
-# else
- return ::isnan(x) != 0;
-# endif
-# elif (GLM_COMPILER & (GLM_COMPILER_GCC | GLM_COMPILER_CLANG)) && (GLM_PLATFORM & GLM_PLATFORM_ANDROID) && __cplusplus < 201103L
- return _isnan(x) != 0;
-# elif GLM_COMPILER & GLM_COMPILER_CUDA
- return ::isnan(x) != 0;
-# else
- return std::isnan(x);
-# endif
- }
-# endif
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, bool, Q> isnan(vec<L, T, Q> const& v)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'isnan' only accept floating-point inputs");
-
- vec<L, bool, Q> Result;
- for (length_t l = 0; l < v.length(); ++l)
- Result[l] = glm::isnan(v[l]);
- return Result;
- }
-
-# if GLM_HAS_CXX11_STL
- using std::isinf;
-# else
- template<typename genType>
- GLM_FUNC_QUALIFIER bool isinf(genType x)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'isinf' only accept floating-point inputs");
-
-# if GLM_HAS_CXX11_STL
- return std::isinf(x);
-# elif GLM_COMPILER & (GLM_COMPILER_INTEL | GLM_COMPILER_VC)
-# if(GLM_PLATFORM & GLM_PLATFORM_WINDOWS)
- return _fpclass(x) == _FPCLASS_NINF || _fpclass(x) == _FPCLASS_PINF;
-# else
- return ::isinf(x);
-# endif
-# elif GLM_COMPILER & (GLM_COMPILER_GCC | GLM_COMPILER_CLANG)
-# if(GLM_PLATFORM & GLM_PLATFORM_ANDROID && __cplusplus < 201103L)
- return _isinf(x) != 0;
-# else
- return std::isinf(x);
-# endif
-# elif GLM_COMPILER & GLM_COMPILER_CUDA
- // http://developer.download.nvidia.com/compute/cuda/4_2/rel/toolkit/docs/online/group__CUDA__MATH__DOUBLE_g13431dd2b40b51f9139cbb7f50c18fab.html#g13431dd2b40b51f9139cbb7f50c18fab
- return ::isinf(double(x)) != 0;
-# else
- return std::isinf(x);
-# endif
- }
-# endif
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, bool, Q> isinf(vec<L, T, Q> const& v)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'isinf' only accept floating-point inputs");
-
- vec<L, bool, Q> Result;
- for (length_t l = 0; l < v.length(); ++l)
- Result[l] = glm::isinf(v[l]);
- return Result;
- }
-
- GLM_FUNC_QUALIFIER int floatBitsToInt(float const& v)
- {
- union
- {
- float in;
- int out;
- } u;
-
- u.in = v;
-
- return u.out;
- }
-
- template<length_t L, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, int, Q> floatBitsToInt(vec<L, float, Q> const& v)
- {
- return reinterpret_cast<vec<L, int, Q>&>(const_cast<vec<L, float, Q>&>(v));
- }
-
- GLM_FUNC_QUALIFIER uint floatBitsToUint(float const& v)
- {
- union
- {
- float in;
- uint out;
- } u;
-
- u.in = v;
-
- return u.out;
- }
-
- template<length_t L, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, uint, Q> floatBitsToUint(vec<L, float, Q> const& v)
- {
- return reinterpret_cast<vec<L, uint, Q>&>(const_cast<vec<L, float, Q>&>(v));
- }
-
- GLM_FUNC_QUALIFIER float intBitsToFloat(int const& v)
- {
- union
- {
- int in;
- float out;
- } u;
-
- u.in = v;
-
- return u.out;
- }
-
- template<length_t L, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, float, Q> intBitsToFloat(vec<L, int, Q> const& v)
- {
- return reinterpret_cast<vec<L, float, Q>&>(const_cast<vec<L, int, Q>&>(v));
- }
-
- GLM_FUNC_QUALIFIER float uintBitsToFloat(uint const& v)
- {
- union
- {
- uint in;
- float out;
- } u;
-
- u.in = v;
-
- return u.out;
- }
-
- template<length_t L, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, float, Q> uintBitsToFloat(vec<L, uint, Q> const& v)
- {
- return reinterpret_cast<vec<L, float, Q>&>(const_cast<vec<L, uint, Q>&>(v));
- }
-
- template<typename genType>
- GLM_FUNC_QUALIFIER genType fma(genType const& a, genType const& b, genType const& c)
- {
- return a * b + c;
- }
-
- template<typename genType>
- GLM_FUNC_QUALIFIER genType frexp(genType x, int& exp)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'frexp' only accept floating-point inputs");
-
- return std::frexp(x, &exp);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> frexp(vec<L, T, Q> const& v, vec<L, int, Q>& exp)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'frexp' only accept floating-point inputs");
-
- vec<L, T, Q> Result;
- for (length_t l = 0; l < v.length(); ++l)
- Result[l] = std::frexp(v[l], &exp[l]);
- return Result;
- }
-
- template<typename genType>
- GLM_FUNC_QUALIFIER genType ldexp(genType const& x, int const& exp)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'ldexp' only accept floating-point inputs");
-
- return std::ldexp(x, exp);
- }
-
- template<length_t L, typename T, qualifier Q>
- GLM_FUNC_QUALIFIER vec<L, T, Q> ldexp(vec<L, T, Q> const& v, vec<L, int, Q> const& exp)
- {
- GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_UNRESTRICTED_GENTYPE, "'ldexp' only accept floating-point inputs");
-
- vec<L, T, Q> Result;
- for (length_t l = 0; l < v.length(); ++l)
- Result[l] = std::ldexp(v[l], exp[l]);
- return Result;
- }
-}//namespace glm
-
-#if GLM_ARCH != GLM_ARCH_PURE && GLM_HAS_UNRESTRICTED_UNIONS
-# include "func_common_simd.inl"
-#endif