mirror/glm
1
0
Fork 0
mirror of https://github.com/g-truc/glm.git synced 2024-11-14 14:04:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

Revert "Add support for non aligned SIMD for vec4"

Browse Source 
This reverts commit 4e01e940de.
This commit is contained in:
Christophe 2023-12-20 21:33:19 +01:00
parent e516c13a06
commit 8bdcf22d72
12 changed files with 122 additions and 406 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
glm/detail/func_geometric.inl
View File

@ -59,13 +59,8 @@ namespace detail
{
GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
// VS 17.7.4 generates longer assembly (~20 instructions vs 11 instructions)
#if defined(_MSC_VER)
return a.x * b.x + a.y * b.y + a.z * b.z + a.w * b.w;
#else
vec<4, T, Q> tmp(a * b);
return (tmp.x + tmp.y) + (tmp.z + tmp.w);
#endif
vec<4, T, Q> tmp(a * b);
return (tmp.x + tmp.y) + (tmp.z + tmp.w);
}
};
@ -172,14 +167,14 @@ namespace detail
GLM_FUNC_QUALIFIER GLM_CONSTEXPR T dot(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
{
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'dot' accepts only floating-point inputs");
return detail::compute_dot<vec<L, T, Q>, T, detail::use_simd<Q>::value>::call(x, y);
return detail::compute_dot<vec<L, T, Q>, T, detail::is_aligned<Q>::value>::call(x, y);
}
// cross
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> cross(vec<3, T, Q> const& x, vec<3, T, Q> const& y)
{
return detail::compute_cross<T, Q, detail::use_simd<Q>::value>::call(x, y);
return detail::compute_cross<T, Q, detail::is_aligned<Q>::value>::call(x, y);
}
/*
// normalize

155
glm/detail/qualifier.hpp
View File

@ -11,16 +11,11 @@ namespace glm
packed_mediump, ///< Typed data is tightly packed in memory and operations are executed with medium precision in term of ULPs for higher performance
packed_lowp, ///< Typed data is tightly packed in memory and operations are executed with low precision in term of ULPs to maximize performance
# if GLM_CONFIG_SIMD == GLM_ENABLE
unaligned_simd_highp, ///< Typed data is unaligned SIMD optimizations and operations are executed with high precision in term of ULPs
unaligned_simd_mediump, ///< Typed data is unaligned SIMD optimizations and operations are executed with high precision in term of ULPs for higher performance
unaligned_simd_lowp, // ///< Typed data is unaligned SIMD optimizations and operations are executed with high precision in term of ULPs to maximize performance
# endif
# if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
aligned_highp, ///< Typed data is aligned in memory allowing SIMD optimizations and operations are executed with high precision in term of ULPs
aligned_mediump, ///< Typed data is aligned in memory allowing SIMD optimizations and operations are executed with high precision in term of ULPs for higher performance
aligned_lowp, // ///< Typed data is aligned in memory allowing SIMD optimizations and operations are executed with high precision in term of ULPs to maximize performance
aligned = aligned_highp, ///< By default aligned qualifier is also high precision
# endif
highp = packed_highp, ///< By default highp qualifier is also packed
@ -28,24 +23,11 @@ namespace glm
lowp = packed_lowp, ///< By default lowp qualifier is also packed
packed = packed_highp, ///< By default packed qualifier is also high precision
# if GLM_CONFIG_SIMD == GLM_ENABLE
unaligned_simd = unaligned_simd_highp, ///< By default unaligned_simd qualifier is also high precision
# endif
# if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
aligned = aligned_highp, ///< By default aligned qualifier is also high precision
# endif
# if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE && defined(GLM_FORCE_DEFAULT_ALIGNED_GENTYPES)
defaultp = aligned_highp
# else
# if GLM_CONFIG_SIMD == GLM_ENABLE
defaultp = unaligned_simd_highp
# else
defaultp = highp
# endif
defaultp = highp
# endif
};
typedef qualifier precision;
@ -99,51 +81,7 @@ namespace detail
};
# endif
template<glm::qualifier P>
struct use_simd
{
static const bool value = false;
};
#if GLM_CONFIG_SIMD == GLM_ENABLE
template<>
struct use_simd<glm::unaligned_simd_lowp>
{
static const bool value = true;
};
template<>
struct use_simd<glm::unaligned_simd_mediump>
{
static const bool value = true;
};
template<>
struct use_simd<glm::unaligned_simd_highp>
{
static const bool value = true;
};
template<>
struct use_simd<glm::aligned_lowp>
{
static const bool value = true;
};
template<>
struct use_simd<glm::aligned_mediump>
{
static const bool value = true;
};
template<>
struct use_simd<glm::aligned_highp>
{
static const bool value = true;
};
#endif
template<length_t L, typename T, bool is_aligned, bool use_simd = true>
template<length_t L, typename T, bool is_aligned>
struct storage
{
typedef struct type {
@ -176,72 +114,24 @@ namespace detail
typedef glm_f32vec4 type;
};
template<>
struct storage<4, float, false, true>
{
typedef struct type{
float data[4];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
inline type(glm_f32vec4 v){_mm_storeu_ps(data, v);}
inline operator glm_f32vec4() const {return _mm_loadu_ps(data);}
} type;
};
template<>
struct storage<4, int, true>
{
typedef glm_i32vec4 type;
};
template<>
struct storage<4, int, false, true>
{
struct type
{
int data[4];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
type(glm_i32vec4 v) { _mm_storeu_si128((__m128i*)data, v); }
operator glm_i32vec4() const { return _mm_loadu_si128((__m128i*)data); }
};
};
template<>
struct storage<4, unsigned int, true>
{
typedef glm_u32vec4 type;
};
template<>
struct storage<4, unsigned int, false, true>
{
struct type
{
unsigned int data[4];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
type(glm_i32vec4 v) { _mm_storeu_si128((__m128i*)data, v); }
operator glm_i32vec4() const { return _mm_loadu_si128((__m128i*)data); }
};
};
template<>
struct storage<2, double, true>
{
typedef glm_f64vec2 type;
};
template<>
struct storage<2, double, false, true>
{
struct type
{
double data[2];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
type(glm_f64vec2 v) { _mm_storeu_pd(data, v); }
operator glm_f64vec2() const { return _mm_loadu_pd(data); }
};
};
template<>
struct storage<2, detail::int64, true>
{
@ -283,56 +173,17 @@ namespace detail
typedef glm_f32vec4 type;
};
template<>
struct storage<4, float, false, true>
{
typedef struct type {
float data[4];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
inline type(glm_f32vec4 v) { vst1q_f32(reinterpret_cast<float*>(data), v); }
inline operator glm_f32vec4() const { return vld1q_f32(reinterpret_cast<const float*>(data)); }
} type;
};
return ;
template<>
struct storage<4, int, true>
{
typedef glm_i32vec4 type;
};
template<>
struct storage<4, int, false, true>
{
struct type
{
int data[4];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
type(glm_i32vec4 v) { vst1q_u32(data, v); }
operator glm_i32vec4() const { return vld1q_u32(data); }
};
};
template<>
struct storage<4, unsigned int, true>
{
typedef glm_u32vec4 type;
};
template<>
struct storage<4, unsigned int, false, true>
{
struct type
{
unsigned int data[4];
GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR type() GLM_DEFAULT;
type(glm_i32vec4 v) { vst1q_u32(data, v); }
operator glm_i32vec4() const { return vld1q_u32(data); }
};
};
# endif
enum genTypeEnum

2
glm/detail/setup.hpp
View File

@ -72,7 +72,7 @@
#define GLM_LANG_CXXMS GLM_LANG_CXXMS_FLAG
#define GLM_LANG_CXXGNU GLM_LANG_CXXGNU_FLAG
#if defined(_MSC_EXTENSIONS)
#if (defined(_MSC_EXTENSIONS))
# define GLM_LANG_EXT GLM_LANG_CXXMS_FLAG
#elif ((GLM_COMPILER & (GLM_COMPILER_CLANG | GLM_COMPILER_GCC)) && (GLM_ARCH & GLM_ARCH_SIMD_BIT))
# define GLM_LANG_EXT GLM_LANG_CXXMS_FLAG

16
glm/detail/type_mat4x4.inl
View File

@ -629,15 +629,15 @@ namespace glm
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER mat<4, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
{
typename mat<4, 4, T, Q>::col_type const &SrcA0 = m1[0];
typename mat<4, 4, T, Q>::col_type const &SrcA1 = m1[1];
typename mat<4, 4, T, Q>::col_type const &SrcA2 = m1[2];
typename mat<4, 4, T, Q>::col_type const &SrcA3 = m1[3];
typename mat<4, 4, T, Q>::col_type const SrcA0 = m1[0];
typename mat<4, 4, T, Q>::col_type const SrcA1 = m1[1];
typename mat<4, 4, T, Q>::col_type const SrcA2 = m1[2];
typename mat<4, 4, T, Q>::col_type const SrcA3 = m1[3];
typename mat<4, 4, T, Q>::col_type const &SrcB0 = m2[0];
typename mat<4, 4, T, Q>::col_type const &SrcB1 = m2[1];
typename mat<4, 4, T, Q>::col_type const &SrcB2 = m2[2];
typename mat<4, 4, T, Q>::col_type const &SrcB3 = m2[3];
typename mat<4, 4, T, Q>::col_type const SrcB0 = m2[0];
typename mat<4, 4, T, Q>::col_type const SrcB1 = m2[1];
typename mat<4, 4, T, Q>::col_type const SrcB2 = m2[2];
typename mat<4, 4, T, Q>::col_type const SrcB3 = m2[3];
mat<4, 4, T, Q> Result;
Result[0] = SrcA0 * SrcB0[0] + SrcA1 * SrcB0[1] + SrcA2 * SrcB0[2] + SrcA3 * SrcB0[3];

2
glm/detail/type_vec4.hpp
View File

@ -50,7 +50,7 @@ namespace glm
struct { T r, g, b, a; };
struct { T s, t, p, q; };
typename detail::storage<4, T, detail::is_aligned<Q>::value, detail::use_simd<Q>::value>::type data;
typename detail::storage<4, T, detail::is_aligned<Q>::value>::type data;
# if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
GLM_SWIZZLE4_2_MEMBERS(T, Q, x, y, z, w)

72
glm/detail/type_vec4.inl
View File

@ -113,7 +113,7 @@ namespace detail
{
GLM_FUNC_QUALIFIER GLM_CONSTEXPR static bool call(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
{
return !compute_vec4_equal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(v1, v2);
return !compute_vec4_equal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v1, v2);
}
};
@ -479,7 +479,7 @@ namespace detail
GLM_FUNC_QUALIFIER GLM_CONSTEXPR T& vec<4, T, Q>::operator[](typename vec<4, T, Q>::length_type i)
{
assert(i >= 0 && i < this->length());
switch (i)
switch(i)
{
default:
case 0:
@ -497,7 +497,7 @@ namespace detail
GLM_FUNC_QUALIFIER GLM_CONSTEXPR T const& vec<4, T, Q>::operator[](typename vec<4, T, Q>::length_type i) const
{
assert(i >= 0 && i < this->length());
switch (i)
switch(i)
{
default:
case 0:
@ -540,84 +540,84 @@ namespace detail
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator+=(U scalar)
{
return (*this = detail::compute_vec4_add<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_add<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator+=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_add<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
return (*this = detail::compute_vec4_add<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator+=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_add<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_add<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator-=(U scalar)
{
return (*this = detail::compute_vec4_sub<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator-=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_sub<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
return (*this = detail::compute_vec4_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator-=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_sub<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator*=(U scalar)
{
return (*this = detail::compute_vec4_mul<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_mul<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator*=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_mul<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
return (*this = detail::compute_vec4_mul<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator*=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_mul<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_mul<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator/=(U scalar)
{
return (*this = detail::compute_vec4_div<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_div<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator/=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_div<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
return (*this = detail::compute_vec4_div<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator/=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_div<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_div<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
// -- Increment and decrement operators --
@ -664,126 +664,126 @@ namespace detail
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator%=(U scalar)
{
return (*this = detail::compute_vec4_mod<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_mod<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator%=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_mod<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_mod<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator%=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_mod<T, Q, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_mod<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator&=(U scalar)
{
return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator&=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator&=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator|=(U scalar)
{
return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator|=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator|=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator^=(U scalar)
{
return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator^=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator^=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator<<=(U scalar)
{
return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator<<=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator<<=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator>>=(U scalar)
{
return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator>>=(vec<1, U, Q> const& v)
{
return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
template<typename T, qualifier Q>
template<typename U>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator>>=(vec<4, U, Q> const& v)
{
return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(*this, vec<4, T, Q>(v)));
return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
}
// -- Unary constant operators --
@ -1107,7 +1107,7 @@ namespace detail
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator~(vec<4, T, Q> const& v)
{
return detail::compute_vec4_bitwise_not<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(v);
return detail::compute_vec4_bitwise_not<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v);
}
// -- Boolean operators --
@ -1115,13 +1115,13 @@ namespace detail
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
{
return detail::compute_vec4_equal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(v1, v2);
return detail::compute_vec4_equal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v1, v2);
}
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
{
return detail::compute_vec4_nequal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::use_simd<Q>::value>::call(v1, v2);
return detail::compute_vec4_nequal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v1, v2);
}
template<qualifier Q>

242
glm/detail/type_vec4_simd.inl
View File

@ -51,7 +51,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_add<float, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
{
vec<4, float, Q> Result;
Result.data = _mm_add_ps(a.data, b.data);
@ -63,7 +63,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_add<double, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
{
vec<4, double, Q> Result;
Result.data = _mm256_add_pd(a.data, b.data);
@ -75,7 +75,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_sub<float, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
{
vec<4, float, Q> Result;
Result.data = _mm_sub_ps(a.data, b.data);
@ -87,7 +87,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_sub<double, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
{
vec<4, double, Q> Result;
Result.data = _mm256_sub_pd(a.data, b.data);
@ -99,7 +99,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_mul<float, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
{
vec<4, float, Q> Result;
Result.data = _mm_mul_ps(a.data, b.data);
@ -111,7 +111,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_mul<double, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
{
vec<4, double, Q> Result;
Result.data = _mm256_mul_pd(a.data, b.data);
@ -123,7 +123,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_div<float, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
{
vec<4, float, Q> Result;
Result.data = _mm_div_ps(a.data, b.data);
@ -135,7 +135,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_div<double, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
{
vec<4, double, Q> Result;
Result.data = _mm256_div_pd(a.data, b.data);
@ -147,7 +147,7 @@ namespace detail
template<>
struct compute_vec4_div<float, aligned_lowp, true>
{
GLM_FUNC_QUALIFIER static vec<4, float, aligned_lowp> call(vec<4, float, aligned_lowp> const& a, vec<4, float, aligned_lowp> const& b)
static vec<4, float, aligned_lowp> call(vec<4, float, aligned_lowp> const& a, vec<4, float, aligned_lowp> const& b)
{
vec<4, float, aligned_lowp> Result;
Result.data = _mm_mul_ps(a.data, _mm_rcp_ps(b.data));
@ -155,69 +155,36 @@ namespace detail
}
};
template<qualifier Q>
struct compute_vec4_and<int, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
{
vec<4, int, Q> Result;
Result.data = _mm_and_si128(a.data, b.data);
return Result;
}
};
template<qualifier Q>
template<typename T, qualifier Q>
struct compute_vec4_and<uint, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, int, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, uint, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm_and_si128(a.data, b.data);
return Result;
}
};
# if GLM_ARCH & GLM_ARCH_AVX2_BIT
template<qualifier Q>
struct compute_vec4_and<int64, Q, true, 64, true>
template<typename T, qualifier Q>
struct compute_vec4_and<T, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, int64, Q> call(vec<4, int64, Q> const& a, vec<4, int64, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int64, Q> Result;
Result.data = _mm256_and_si256(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_and<uint64, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint64, Q> call(vec<4, uint64, Q> const& a, vec<4, uint64, Q> const& b)
{
vec<4, uint64, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm256_and_si256(a.data, b.data);
return Result;
}
};
# endif
template<qualifier Q>
template<typename T, qualifier Q>
struct compute_vec4_or<int, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int, Q> Result;
Result.data = _mm_or_si128(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_or<uint, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
{
vec<4, uint, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm_or_si128(a.data, b.data);
return Result;
}
@ -235,134 +202,79 @@ namespace detail
}
};
template<qualifier Q>
template<typename T, qualifier Q>
struct compute_vec4_or<uint64, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint64, Q> call(vec<4, uint64, Q> const& a, vec<4, uint64, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, uint64, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm256_or_si256(a.data, b.data);
return Result;
}
};
# endif
template<qualifier Q>
template<typename T, qualifier Q>
struct compute_vec4_xor<int, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int, Q> Result;
Result.data = _mm_xor_si128(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_xor<uint, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
{
vec<4, uint, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm_xor_si128(a.data, b.data);
return Result;
}
};
# if GLM_ARCH & GLM_ARCH_AVX2_BIT
template<qualifier Q>
struct compute_vec4_xor<int64, Q, true, 64, true>
template<typename T, qualifier Q>
struct compute_vec4_xor<T, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, int64, Q> call(vec<4, int64, Q> const& a, vec<4, int64, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int64, Q> Result;
Result.data = _mm256_xor_si256(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_xor<uint64, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint64, Q> call(vec<4, uint64, Q> const& a, vec<4, uint64, Q> const& b)
{
vec<4, uint64, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm256_xor_si256(a.data, b.data);
return Result;
}
};
# endif
template<qualifier Q>
struct compute_vec4_shift_left<int, Q, true, 32, true>
template<typename T, qualifier Q>
struct compute_vec4_shift_left<T, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int, Q> Result;
Result.data = _mm_sll_epi32(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_shift_left<uint, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
{
vec<4, uint, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm_sll_epi32(a.data, b.data);
return Result;
}
};
# if GLM_ARCH & GLM_ARCH_AVX2_BIT
template<qualifier Q>
struct compute_vec4_shift_left<int64, Q, true, 64, true>
template<typename T, qualifier Q>
struct compute_vec4_shift_left<T, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, int64, Q> call(vec<4, int64, Q> const& a, vec<4, int64, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int64, Q> Result;
Result.data = _mm256_sll_epi64(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_shift_left<uint64, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint64, Q> call(vec<4, uint64, Q> const& a, vec<4, uint64, Q> const& b)
{
vec<4, uint64, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm256_sll_epi64(a.data, b.data);
return Result;
}
};
# endif
template<qualifier Q>
struct compute_vec4_shift_right<int, Q, true, 32, true>
template<typename T, qualifier Q>
struct compute_vec4_shift_right<T, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, int, Q> Result;
Result.data = _mm_srl_epi32(a.data, b.data);
return Result;
}
};
template<qualifier Q>
struct compute_vec4_shift_right<uint, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
{
vec<4, uint, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm_srl_epi32(a.data, b.data);
return Result;
}
};
# if GLM_ARCH & GLM_ARCH_AVX2_BIT
template<qualifier Q>
template<typename T, qualifier Q>
struct compute_vec4_shift_right<int64, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, int64, Q> call(vec<4, int64, Q> const& a, vec<4, int64, Q> const& b)
@ -373,58 +285,36 @@ namespace detail
}
};
template<qualifier Q>
struct compute_vec4_shift_right<uint64, Q, true, 64, true>
template<typename T, qualifier Q>
struct compute_vec4_shift_right<T, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint64, Q> call(vec<4, uint64, Q> const& a, vec<4, uint64, Q> const& b)
static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
{
vec<4, uint64, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm256_srl_epi64(a.data, b.data);
return Result;
}
};
# endif
template<qualifier Q>
struct compute_vec4_bitwise_not<int, Q, true, 32, true>
template<typename T, qualifier Q>
struct compute_vec4_bitwise_not<T, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v)
static vec<4, T, Q> call(vec<4, T, Q> const& v)
{
vec<4, int, Q> Result;
Result.data = _mm_xor_si128(v.data, _mm_set1_epi32(-1));
return Result;
}
};
template<qualifier Q>
struct compute_vec4_bitwise_not<uint, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v)
{
vec<4, uint, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm_xor_si128(v.data, _mm_set1_epi32(-1));
return Result;
}
};
# if GLM_ARCH & GLM_ARCH_AVX2_BIT
template<qualifier Q>
struct compute_vec4_bitwise_not<int64, Q, true, 64, true>
template<typename T, qualifier Q>
struct compute_vec4_bitwise_not<T, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, int64, Q> call(vec<4, int64, Q> const& v)
static vec<4, T, Q> call(vec<4, T, Q> const& v)
{
vec<4, int64, Q> Result;
Result.data = _mm256_xor_si256(v.data, _mm_set1_epi32(-1));
return Result;
}
};
template<qualifier Q>
struct compute_vec4_bitwise_not<uint64, Q, true, 64, true>
{
GLM_FUNC_QUALIFIER static vec<4, uint64, Q> call(vec<4, uint64, Q> const& v)
{
vec<4, uint64, Q> Result;
vec<4, T, Q> Result;
Result.data = _mm256_xor_si256(v.data, _mm_set1_epi32(-1));
return Result;
}
@ -434,7 +324,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_equal<float, Q, false, 32, true>
{
GLM_FUNC_QUALIFIER static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
{
return _mm_movemask_ps(_mm_cmpneq_ps(v1.data, v2.data)) == 0;
}
@ -444,7 +334,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_equal<int, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
{
//return _mm_movemask_epi8(_mm_cmpeq_epi32(v1.data, v2.data)) != 0;
__m128i neq = _mm_xor_si128(v1.data, v2.data);
@ -456,7 +346,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_nequal<float, Q, false, 32, true>
{
GLM_FUNC_QUALIFIER static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
{
return _mm_movemask_ps(_mm_cmpneq_ps(v1.data, v2.data)) != 0;
}
@ -466,7 +356,7 @@ namespace detail
template<qualifier Q>
struct compute_vec4_nequal<int, Q, true, 32, true>
{
GLM_FUNC_QUALIFIER static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
{
//return _mm_movemask_epi8(_mm_cmpneq_epi32(v1.data, v2.data)) != 0;
__m128i neq = _mm_xor_si128(v1.data, v2.data);
@ -491,21 +381,6 @@ namespace detail
data(_mm_set1_ps(_s))
{}
template<>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, unaligned_simd_lowp>::vec(float _s) :
data(_mm_set1_ps(_s))
{}
template<>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, unaligned_simd_mediump>::vec(float _s) :
data(_mm_set1_ps(_s))
{}
template<>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, unaligned_simd_highp>::vec(float _s) :
data(_mm_set1_ps(_s))
{}
# if GLM_ARCH & GLM_ARCH_AVX_BIT
template<>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, double, aligned_lowp>::vec(double _s) :
@ -922,6 +797,7 @@ namespace detail {
data(vcvtq_f32_u32(vec<4, uint, aligned_mediump>(_x, _y, _z, _w).data))
{}
template<>
template<>
GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(uint _x, uint _y, uint _z, uint _w) :

8
glm/gtx/dual_quaternion.hpp
View File

@ -109,11 +109,11 @@ namespace glm
template<typename T, qualifier Q>
GLM_FUNC_DECL vec<3, T, Q> operator*(vec<3, T, Q> const& v, tdualquat<T, Q> const& q);
template<typename T, qualifier Q, qualifier Q2>
GLM_FUNC_DECL vec<4, T, Q2> operator*(tdualquat<T, Q> const& q, vec<4, T, Q2> const& v);
template<typename T, qualifier Q>
GLM_FUNC_DECL vec<4, T, Q> operator*(tdualquat<T, Q> const& q, vec<4, T, Q> const& v);
template<typename T, qualifier Q, qualifier Q2>
GLM_FUNC_DECL vec<4, T, Q2> operator*(vec<4, T, Q2> const& v, tdualquat<T, Q> const& q);
template<typename T, qualifier Q>
GLM_FUNC_DECL vec<4, T, Q> operator*(vec<4, T, Q> const& v, tdualquat<T, Q> const& q);
template<typename T, qualifier Q>
GLM_FUNC_DECL tdualquat<T, Q> operator*(tdualquat<T, Q> const& q, T const& s);

10
glm/gtx/dual_quaternion.inl
View File

@ -169,14 +169,14 @@ namespace glm
return glm::inverse(q) * v;
}
template<typename T, qualifier Q, qualifier Q2>
GLM_FUNC_QUALIFIER vec<4, T, Q2> operator*(tdualquat<T, Q> const& q, vec<4, T, Q2> const& v)
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER vec<4, T, Q> operator*(tdualquat<T, Q> const& q, vec<4, T, Q> const& v)
{
return vec<4, T, Q2>(q * vec<3, T, Q>(v), v.w);
return vec<4, T, Q>(q * vec<3, T, Q>(v), v.w);
}
template<typename T, qualifier Q, qualifier Q2>
GLM_FUNC_QUALIFIER vec<4, T, Q2> operator*(vec<4, T, Q2> const& v, tdualquat<T, Q> const& q)
template<typename T, qualifier Q>
GLM_FUNC_QUALIFIER vec<4, T, Q> operator*(vec<4, T, Q> const& v, tdualquat<T, Q> const& q)
{
return glm::inverse(q) * v;
}

5
glm/simd/platform.h
View File

@ -180,11 +180,6 @@
// Visual C++
#elif defined(_MSC_VER)
# if INTPTR_MAX == INT64_MAX // 64bits compiler has always at least SSE2 support
# ifndef GLM_FORCE_INTRINSICS
# define GLM_FORCE_INTRINSICS
# endif
# endif
# if _MSC_VER >= 1920
# define GLM_COMPILER GLM_COMPILER_VC16
# elif _MSC_VER >= 1916

1
test/core/core_type_vec4.cpp
View File

@ -1,5 +1,4 @@
#define GLM_FORCE_SWIZZLE
#define GLM_FORCE_MESSAGES
#include <glm/gtc/constants.hpp>
#include <glm/gtc/vec1.hpp>
#include <glm/ext/scalar_relational.hpp>

2
test/gtx/gtx_hash.cpp
View File

@ -22,7 +22,7 @@ int test_compile()
std::unordered_map<glm::quat, int> map_quat;
Error += ++map_quat[glm::quat(0.0f, glm::vec3(0.0f))];
std::unordered_map<glm::dualquat, int> map_dualquat;
Error += ++map_dualquat[glm::dualquat(glm::quat(0.0f, glm::vec3(0.0f)), glm::vec3(0.0f))];
Error += ++map_dualquat[glm::dualquat(glm::vec3(0.0f))];
// Matrix types
std::unordered_map<glm::mat2x2, int> map_mat2x2;