mirror/glm
1
0
Fork 0
mirror of https://github.com/g-truc/glm.git synced 2024-11-30 03:44:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed matrix functions using SIMD code on unaligned types #518

Browse Source
This commit is contained in:
Christophe Riccio 2016-07-14 15:21:45 +02:00
parent b8b43e6a4d
commit 6e9b524be1
3 changed files with 51 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

76
glm/detail/func_matrix.inl
View File

@ -7,7 +7,7 @@
namespace glm{ namespace glm{
namespace detail namespace detail
{ {
template <template <typename, precision> class matType, typename T, precision P> template <template <typename, precision> class matType, typename T, precision P, bool Aligned>
struct compute_matrixCompMult struct compute_matrixCompMult
{ {
GLM_FUNC_QUALIFIER static matType<T, P> call(matType<T, P> const& x, matType<T, P> const& y) GLM_FUNC_QUALIFIER static matType<T, P> call(matType<T, P> const& x, matType<T, P> const& y)
@ -19,11 +19,11 @@ namespace detail
} }
}; };
template <template <class, precision> class matType, typename T, precision P> template <template <class, precision> class matType, typename T, precision P, bool Aligned>
struct compute_transpose{}; struct compute_transpose{};
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat2x2, T, P> struct compute_transpose<tmat2x2, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat2x2<T, P> call(tmat2x2<T, P> const & m) GLM_FUNC_QUALIFIER static tmat2x2<T, P> call(tmat2x2<T, P> const & m)
{ {
@ -36,8 +36,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat2x3, T, P> struct compute_transpose<tmat2x3, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat3x2<T, P> call(tmat2x3<T, P> const & m) GLM_FUNC_QUALIFIER static tmat3x2<T, P> call(tmat2x3<T, P> const & m)
{ {
@ -52,8 +52,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat2x4, T, P> struct compute_transpose<tmat2x4, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat4x2<T, P> call(tmat2x4<T, P> const & m) GLM_FUNC_QUALIFIER static tmat4x2<T, P> call(tmat2x4<T, P> const & m)
{ {
@ -70,8 +70,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat3x2, T, P> struct compute_transpose<tmat3x2, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat2x3<T, P> call(tmat3x2<T, P> const & m) GLM_FUNC_QUALIFIER static tmat2x3<T, P> call(tmat3x2<T, P> const & m)
{ {
@ -86,8 +86,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat3x3, T, P> struct compute_transpose<tmat3x3, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat3x3<T, P> call(tmat3x3<T, P> const & m) GLM_FUNC_QUALIFIER static tmat3x3<T, P> call(tmat3x3<T, P> const & m)
{ {
@ -107,8 +107,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat3x4, T, P> struct compute_transpose<tmat3x4, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat4x3<T, P> call(tmat3x4<T, P> const & m) GLM_FUNC_QUALIFIER static tmat4x3<T, P> call(tmat3x4<T, P> const & m)
{ {
@ -129,8 +129,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat4x2, T, P> struct compute_transpose<tmat4x2, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat2x4<T, P> call(tmat4x2<T, P> const & m) GLM_FUNC_QUALIFIER static tmat2x4<T, P> call(tmat4x2<T, P> const & m)
{ {
@ -147,8 +147,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat4x3, T, P> struct compute_transpose<tmat4x3, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat3x4<T, P> call(tmat4x3<T, P> const & m) GLM_FUNC_QUALIFIER static tmat3x4<T, P> call(tmat4x3<T, P> const & m)
{ {
@ -169,8 +169,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_transpose<tmat4x4, T, P> struct compute_transpose<tmat4x4, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat4x4<T, P> call(tmat4x4<T, P> const & m) GLM_FUNC_QUALIFIER static tmat4x4<T, P> call(tmat4x4<T, P> const & m)
{ {
@ -198,11 +198,11 @@ namespace detail
} }
}; };
template <template <typename, precision> class matType, typename T, precision P> template <template <typename, precision> class matType, typename T, precision P, bool Aligned>
struct compute_determinant{}; struct compute_determinant{};
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_determinant<tmat2x2, T, P> struct compute_determinant<tmat2x2, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static T call(tmat2x2<T, P> const & m) GLM_FUNC_QUALIFIER static T call(tmat2x2<T, P> const & m)
{ {
@ -210,8 +210,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_determinant<tmat3x3, T, P> struct compute_determinant<tmat3x3, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static T call(tmat3x3<T, P> const & m) GLM_FUNC_QUALIFIER static T call(tmat3x3<T, P> const & m)
{ {
@ -222,8 +222,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_determinant<tmat4x4, T, P> struct compute_determinant<tmat4x4, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static T call(tmat4x4<T, P> const & m) GLM_FUNC_QUALIFIER static T call(tmat4x4<T, P> const & m)
{ {
@ -246,11 +246,11 @@ namespace detail
} }
}; };
template <template <typename, precision> class matType, typename T, precision P> template <template <typename, precision> class matType, typename T, precision P, bool Aligned>
struct compute_inverse{}; struct compute_inverse{};
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_inverse<tmat2x2, T, P> struct compute_inverse<tmat2x2, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat2x2<T, P> call(tmat2x2<T, P> const& m) GLM_FUNC_QUALIFIER static tmat2x2<T, P> call(tmat2x2<T, P> const& m)
{ {
@ -268,8 +268,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_inverse<tmat3x3, T, P> struct compute_inverse<tmat3x3, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat3x3<T, P> call(tmat3x3<T, P> const& m) GLM_FUNC_QUALIFIER static tmat3x3<T, P> call(tmat3x3<T, P> const& m)
{ {
@ -293,8 +293,8 @@ namespace detail
} }
}; };
template <typename T, precision P> template <typename T, precision P, bool Aligned>
struct compute_inverse<tmat4x4, T, P> struct compute_inverse<tmat4x4, T, P, Aligned>
{ {
GLM_FUNC_QUALIFIER static tmat4x4<T, P> call(tmat4x4<T, P> const& m) GLM_FUNC_QUALIFIER static tmat4x4<T, P> call(tmat4x4<T, P> const& m)
{ {
@ -359,7 +359,7 @@ namespace detail
GLM_FUNC_QUALIFIER matType<T, P> matrixCompMult(matType<T, P> const & x, matType<T, P> const & y) GLM_FUNC_QUALIFIER matType<T, P> matrixCompMult(matType<T, P> const & x, matType<T, P> const & y)
{ {
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'matrixCompMult' only accept floating-point inputs"); GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'matrixCompMult' only accept floating-point inputs");
return detail::compute_matrixCompMult<matType, T, P>::call(x, y); return detail::compute_matrixCompMult<matType, T, P, detail::is_aligned<P>::value>::call(x, y);
} }
template<typename T, precision P, template <typename, precision> class vecTypeA, template <typename, precision> class vecTypeB> template<typename T, precision P, template <typename, precision> class vecTypeA, template <typename, precision> class vecTypeB>
@ -377,21 +377,21 @@ namespace detail
GLM_FUNC_QUALIFIER typename matType<T, P>::transpose_type transpose(matType<T, P> const & m) GLM_FUNC_QUALIFIER typename matType<T, P>::transpose_type transpose(matType<T, P> const & m)
{ {
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'transpose' only accept floating-point inputs"); GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'transpose' only accept floating-point inputs");
return detail::compute_transpose<matType, T, P>::call(m); return detail::compute_transpose<matType, T, P, detail::is_aligned<P>::value>::call(m);
} }
template <typename T, precision P, template <typename, precision> class matType> template <typename T, precision P, template <typename, precision> class matType>
GLM_FUNC_QUALIFIER T determinant(matType<T, P> const & m) GLM_FUNC_QUALIFIER T determinant(matType<T, P> const & m)
{ {
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'determinant' only accept floating-point inputs"); GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'determinant' only accept floating-point inputs");
return detail::compute_determinant<matType, T, P>::call(m); return detail::compute_determinant<matType, T, P, detail::is_aligned<P>::value>::call(m);
} }
template <typename T, precision P, template <typename, precision> class matType> template <typename T, precision P, template <typename, precision> class matType>
GLM_FUNC_QUALIFIER matType<T, P> inverse(matType<T, P> const & m) GLM_FUNC_QUALIFIER matType<T, P> inverse(matType<T, P> const & m)
{ {
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'inverse' only accept floating-point inputs"); GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'inverse' only accept floating-point inputs");
return detail::compute_inverse<matType, T, P>::call(m); return detail::compute_inverse<matType, T, P, detail::is_aligned<P>::value>::call(m);
} }
}//namespace glm }//namespace glm

10
glm/detail/func_matrix_simd.inl
View File

@ -11,8 +11,10 @@ namespace glm{
namespace detail namespace detail
{ {
template <precision P> template <precision P>
struct compute_matrixCompMult<tmat4x4, float, P> struct compute_matrixCompMult<tmat4x4, float, P, true>
{ {
GLM_STATIC_ASSERT(detail::is_aligned<P>::value, "Specialization requires aligned");
GLM_FUNC_QUALIFIER static tmat4x4<float, P> call(tmat4x4<float, P> const & x, tmat4x4<float, P> const & y) GLM_FUNC_QUALIFIER static tmat4x4<float, P> call(tmat4x4<float, P> const & x, tmat4x4<float, P> const & y)
{ {
tmat4x4<float, P> result(uninitialize); tmat4x4<float, P> result(uninitialize);
@ -25,7 +27,7 @@ namespace detail
}; };
template <precision P> template <precision P>
struct compute_transpose<tmat4x4, float, P> struct compute_transpose<tmat4x4, float, P, true>
{ {
GLM_FUNC_QUALIFIER static tmat4x4<float, P> call(tmat4x4<float, P> const & m) GLM_FUNC_QUALIFIER static tmat4x4<float, P> call(tmat4x4<float, P> const & m)
{ {
@ -38,7 +40,7 @@ namespace detail
}; };
template <precision P> template <precision P>
struct compute_determinant<tmat4x4, float, P> struct compute_determinant<tmat4x4, float, P, true>
{ {
GLM_FUNC_QUALIFIER static float call(tmat4x4<float, P> const& m) GLM_FUNC_QUALIFIER static float call(tmat4x4<float, P> const& m)
{ {
@ -47,7 +49,7 @@ namespace detail
}; };
template <precision P> template <precision P>
struct compute_inverse<tmat4x4, float, P> struct compute_inverse<tmat4x4, float, P, true>
{ {
GLM_FUNC_QUALIFIER static tmat4x4<float, P> call(tmat4x4<float, P> const& m) GLM_FUNC_QUALIFIER static tmat4x4<float, P> call(tmat4x4<float, P> const& m)
{ {

7
test/core/core_func_geometric.cpp
View File

@ -87,6 +87,13 @@ namespace normalize
glm::vec3 Normalize1 = glm::normalize(glm::vec3(1, 0, 0)); glm::vec3 Normalize1 = glm::normalize(glm::vec3(1, 0, 0));
glm::vec3 Normalize2 = glm::normalize(glm::vec3(2, 0, 0)); glm::vec3 Normalize2 = glm::normalize(glm::vec3(2, 0, 0));
glm::vec3 Normalize3 = glm::normalize(glm::vec3(-0.6, 0.7, -0.5));
glm::vec3 ro = glm::vec3(glm::cos(5.f) * 3.f, 2.f, glm::sin(5.f) * 3.f);
glm::vec3 w = glm::normalize(glm::vec3(0, -0.2f, 0) - ro);
glm::vec3 u = glm::normalize(glm::cross(w, glm::vec3(0, 1, 0)));
glm::vec3 v = glm::cross(u, w);
int Error = 0; int Error = 0;
Error += glm::all(glm::lessThan(glm::abs(Normalize1 - glm::vec3(1, 0, 0)), glm::vec3(std::numeric_limits<float>::epsilon()))) ? 0 : 1; Error += glm::all(glm::lessThan(glm::abs(Normalize1 - glm::vec3(1, 0, 0)), glm::vec3(std::numeric_limits<float>::epsilon()))) ? 0 : 1;