Fixed simd mat4 build

This commit is contained in:
Christophe Riccio 2010-12-16 10:32:43 +00:00
parent 866db89305
commit 4634ff3647
4 changed files with 72 additions and 82 deletions

View File

@ -11,9 +11,7 @@
#define GLM_DETAIL_INTRINSIC_MATRIX_INCLUDED #define GLM_DETAIL_INTRINSIC_MATRIX_INCLUDED
#include "../glm.hpp" #include "../glm.hpp"
#include "intrinsic_geometric.hpp"
#include <xmmintrin.h>
#include <emmintrin.h>
void _mm_add_ps(__m128 in1[4], __m128 in2[4], __m128 out[4]); void _mm_add_ps(__m128 in1[4], __m128 in2[4], __m128 out[4]);

View File

@ -16,9 +16,8 @@
// Dependency: // Dependency:
#include "../glm.hpp" #include "../glm.hpp"
#include "../core/intrinsic_matrix.hpp"
#include "../gtx/simd_vec4.hpp" #include "../gtx/simd_vec4.hpp"
#include <xmmintrin.h>
#include <emmintrin.h>
namespace glm namespace glm
{ {
@ -26,23 +25,18 @@ namespace glm
{ {
GLM_ALIGNED(struct, 16) fmat4x4SIMD GLM_ALIGNED(struct, 16) fmat4x4SIMD
{ {
static __m128 one; enum ctor{null};
enum no_init
{
NO_INIT
};
typedef float value_type; typedef float value_type;
typedef fvec4SIMD col_type; typedef fvec4SIMD col_type;
typedef fvec4SIMD row_type; typedef fvec4SIMD row_type;
typedef glm::sizeType size_type; typedef std::size_t size_type;
static size_type value_size(); static size_type value_size();
static size_type col_size(); static size_type col_size();
static size_type row_size(); static size_type row_size();
static bool is_matrix(); static bool is_matrix();
fvec4SIMD value[4]; fvec4SIMD Data[4];
////////////////////////////////////// //////////////////////////////////////
// Constructors // Constructors
@ -60,7 +54,7 @@ namespace glm
fvec4SIMD const & v2, fvec4SIMD const & v2,
fvec4SIMD const & v3); fvec4SIMD const & v3);
explicit fmat4x4SIMD( explicit fmat4x4SIMD(
tmat4x4 const & m); tmat4x4<float> const & m);
// Conversions // Conversions
//template <typename U> //template <typename U>
@ -129,7 +123,7 @@ namespace glm
//! GLM_GTX_simd_mat4 extension: SIMD implementation of vec4 type. //! GLM_GTX_simd_mat4 extension: SIMD implementation of vec4 type.
namespace simd_mat4 namespace simd_mat4
{ {
typedef detail::fmat4SIMD mat4SIMD; typedef detail::fmat4x4SIMD simd_mat4;
}//namespace simd_mat4 }//namespace simd_mat4
}//namespace gtx }//namespace gtx

View File

@ -15,10 +15,10 @@ namespace detail
inline fmat4x4SIMD::fmat4x4SIMD(float const & s) inline fmat4x4SIMD::fmat4x4SIMD(float const & s)
{ {
this->value[0] = fvec4SIMD(s, 0, 0, 0); this->Data[0] = fvec4SIMD(s, 0, 0, 0);
this->value[1] = fvec4SIMD(0, s, 0, 0); this->Data[1] = fvec4SIMD(0, s, 0, 0);
this->value[2] = fvec4SIMD(0, 0, s, 0); this->Data[2] = fvec4SIMD(0, 0, s, 0);
this->value[3] = fvec4SIMD(0, 0, 0, s); this->Data[3] = fvec4SIMD(0, 0, 0, s);
} }
inline fmat4x4SIMD::fmat4x4SIMD inline fmat4x4SIMD::fmat4x4SIMD
@ -29,10 +29,10 @@ namespace detail
float const & x3, float const & y3, float const & z3, float const & w3 float const & x3, float const & y3, float const & z3, float const & w3
) )
{ {
this->value[0] = fvec4SIMD(x0, y0, z0, w0); this->Data[0] = fvec4SIMD(x0, y0, z0, w0);
this->value[1] = fvec4SIMD(x1, y1, z1, w1); this->Data[1] = fvec4SIMD(x1, y1, z1, w1);
this->value[2] = fvec4SIMD(x2, y2, z2, w2); this->Data[2] = fvec4SIMD(x2, y2, z2, w2);
this->value[3] = fvec4SIMD(x3, y3, z3, w3); this->Data[3] = fvec4SIMD(x3, y3, z3, w3);
} }
inline fmat4x4SIMD::fmat4x4SIMD inline fmat4x4SIMD::fmat4x4SIMD
@ -43,21 +43,21 @@ namespace detail
fvec4SIMD const & v3 fvec4SIMD const & v3
) )
{ {
this->value[0] = v0; this->Data[0] = v0;
this->value[1] = v1; this->Data[1] = v1;
this->value[2] = v2; this->Data[2] = v2;
this->value[3] = v3; this->Data[3] = v3;
} }
inline fmat4x4SIMD::fmat4x4SIMD inline fmat4x4SIMD::fmat4x4SIMD
( (
tmat4x4 const & m tmat4x4<float> const & m
) )
{ {
this->value[0] = fvec4SIMD(m[0]); this->Data[0] = fvec4SIMD(m[0]);
this->value[1] = fvec4SIMD(m[1]); this->Data[1] = fvec4SIMD(m[1]);
this->value[2] = fvec4SIMD(m[2]); this->Data[2] = fvec4SIMD(m[2]);
this->value[3] = fvec4SIMD(m[3]); this->Data[3] = fvec4SIMD(m[3]);
} }
////////////////////////////////////// //////////////////////////////////////
@ -65,26 +65,26 @@ namespace detail
inline fvec4SIMD & fmat4x4SIMD::operator[] inline fvec4SIMD & fmat4x4SIMD::operator[]
( (
typename fmat4x4SIMD::size_type i fmat4x4SIMD::size_type i
) )
{ {
assert( assert(
i >= typename tmat4x4<valType>::size_type(0) && i >= fmat4x4SIMD::size_type(0) &&
i < tmat4x4<valType>::col_size()); i < fmat4x4SIMD::col_size());
return value[i]; return this->Data[i];
} }
inline fvec4SIMD const & fmat4x4SIMD::operator[] inline fvec4SIMD const & fmat4x4SIMD::operator[]
( (
typename fmat4x4SIMD::size_type i fmat4x4SIMD::size_type i
) const ) const
{ {
assert( assert(
i >= typename fmat4x4SIMD::size_type(0) && i >= fmat4x4SIMD::size_type(0) &&
i < fmat4x4SIMD::col_size()); i < fmat4x4SIMD::col_size());
return value[i]; return this->Data[i];
} }
////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////
@ -95,10 +95,10 @@ namespace detail
fmat4x4SIMD const & m fmat4x4SIMD const & m
) )
{ {
this->value[0].Data = m[0].Data; this->Data[0] = m[0];
this->value[1].Data = m[1].Data; this->Data[1] = m[1];
this->value[2].Data = m[2].Data; this->Data[2] = m[2];
this->value[3].Data = m[3].Data; this->Data[3] = m[3];
return *this; return *this;
} }
@ -107,10 +107,10 @@ namespace detail
fmat4x4SIMD const & m fmat4x4SIMD const & m
) )
{ {
this->value[0].Data = _mm_add_ps(this->value[0].Data, m[0].Data); this->Data[0].Data = _mm_add_ps(this->Data[0].Data, m[0].Data);
this->value[1].Data = _mm_add_ps(this->value[1].Data, m[1].Data); this->Data[1].Data = _mm_add_ps(this->Data[1].Data, m[1].Data);
this->value[2].Data = _mm_add_ps(this->value[2].Data, m[2].Data); this->Data[2].Data = _mm_add_ps(this->Data[2].Data, m[2].Data);
this->value[3].Data = _mm_add_ps(this->value[3].Data, m[3].Data); this->Data[3].Data = _mm_add_ps(this->Data[3].Data, m[3].Data);
return *this; return *this;
} }
@ -119,10 +119,10 @@ namespace detail
fmat4x4SIMD const & m fmat4x4SIMD const & m
) )
{ {
this->value[0].Data = _mm_sub_ps(this->value[0].Data, m[0].Data); this->Data[0].Data = _mm_sub_ps(this->Data[0].Data, m[0].Data);
this->value[1].Data = _mm_sub_ps(this->value[1].Data, m[1].Data); this->Data[1].Data = _mm_sub_ps(this->Data[1].Data, m[1].Data);
this->value[2].Data = _mm_sub_ps(this->value[2].Data, m[2].Data); this->Data[2].Data = _mm_sub_ps(this->Data[2].Data, m[2].Data);
this->value[3].Data = _mm_sub_ps(this->value[3].Data, m[3].Data); this->Data[3].Data = _mm_sub_ps(this->Data[3].Data, m[3].Data);
return *this; return *this;
} }
@ -132,7 +132,7 @@ namespace detail
fmat4x4SIMD const & m fmat4x4SIMD const & m
) )
{ {
_mm_mul_ps(this->Data, m.Data, this->Data); _mm_mul_ps(&this->Data[0].Data, &m.Data[0].Data, &this->Data[0].Data);
return *this; return *this;
} }
@ -142,8 +142,8 @@ namespace detail
) )
{ {
__m128 Inv[4]; __m128 Inv[4];
_mm_inverse_ps(m.Data, Inv); _mm_inverse_ps(&this->Data[0].Data, Inv);
_mm_mul_ps(this->Data, Inv, this->Data); _mm_mul_ps(&this->Data[0].Data, Inv, &this->Data[0].Data);
return *this; return *this;
} }
@ -153,10 +153,10 @@ namespace detail
) )
{ {
__m128 Operand = _mm_set_ps1(s); __m128 Operand = _mm_set_ps1(s);
this->value[0].Data = _mm_add_ps(this->value[0].Data, Operand); this->Data[0].Data = _mm_add_ps(this->Data[0].Data, Operand);
this->value[1].Data = _mm_add_ps(this->value[1].Data, Operand); this->Data[1].Data = _mm_add_ps(this->Data[1].Data, Operand);
this->value[2].Data = _mm_add_ps(this->value[2].Data, Operand); this->Data[2].Data = _mm_add_ps(this->Data[2].Data, Operand);
this->value[3].Data = _mm_add_ps(this->value[3].Data, Operand); this->Data[3].Data = _mm_add_ps(this->Data[3].Data, Operand);
return *this; return *this;
} }
@ -166,10 +166,10 @@ namespace detail
) )
{ {
__m128 Operand = _mm_set_ps1(s); __m128 Operand = _mm_set_ps1(s);
this->value[0].Data = _mm_sub_ps(this->value[0].Data, Operand); this->Data[0].Data = _mm_sub_ps(this->Data[0].Data, Operand);
this->value[1].Data = _mm_sub_ps(this->value[1].Data, Operand); this->Data[1].Data = _mm_sub_ps(this->Data[1].Data, Operand);
this->value[2].Data = _mm_sub_ps(this->value[2].Data, Operand); this->Data[2].Data = _mm_sub_ps(this->Data[2].Data, Operand);
this->value[3].Data = _mm_sub_ps(this->value[3].Data, Operand); this->Data[3].Data = _mm_sub_ps(this->Data[3].Data, Operand);
return *this; return *this;
} }
@ -179,10 +179,10 @@ namespace detail
) )
{ {
__m128 Operand = _mm_set_ps1(s); __m128 Operand = _mm_set_ps1(s);
this->value[0].Data = _mm_mul_ps(this->value[0].Data, Operand); this->Data[0].Data = _mm_mul_ps(this->Data[0].Data, Operand);
this->value[1].Data = _mm_mul_ps(this->value[1].Data, Operand); this->Data[1].Data = _mm_mul_ps(this->Data[1].Data, Operand);
this->value[2].Data = _mm_mul_ps(this->value[2].Data, Operand); this->Data[2].Data = _mm_mul_ps(this->Data[2].Data, Operand);
this->value[3].Data = _mm_mul_ps(this->value[3].Data, Operand); this->Data[3].Data = _mm_mul_ps(this->Data[3].Data, Operand);
return *this; return *this;
} }
@ -191,29 +191,29 @@ namespace detail
float const & s float const & s
) )
{ {
__m128 Operand = _mm_div_ps(one, s)); __m128 Operand = _mm_div_ps(one, _mm_set_ps1(s));
this->value[0].Data = _mm_mul_ps(this->value[0].Data, Operand); this->Data[0].Data = _mm_mul_ps(this->Data[0].Data, Operand);
this->value[1].Data = _mm_mul_ps(this->value[1].Data, Operand); this->Data[1].Data = _mm_mul_ps(this->Data[1].Data, Operand);
this->value[2].Data = _mm_mul_ps(this->value[2].Data, Operand); this->Data[2].Data = _mm_mul_ps(this->Data[2].Data, Operand);
this->value[3].Data = _mm_mul_ps(this->value[3].Data, Operand); this->Data[3].Data = _mm_mul_ps(this->Data[3].Data, Operand);
return *this; return *this;
} }
inline fmat4x4SIMD & fmat4x4SIMD::operator++ () inline fmat4x4SIMD & fmat4x4SIMD::operator++ ()
{ {
this->value[0].Data = _mm_add_ps(this->value[0].Data, one); this->Data[0].Data = _mm_add_ps(this->Data[0].Data, one);
this->value[1].Data = _mm_add_ps(this->value[1].Data, one); this->Data[1].Data = _mm_add_ps(this->Data[1].Data, one);
this->value[2].Data = _mm_add_ps(this->value[2].Data, one); this->Data[2].Data = _mm_add_ps(this->Data[2].Data, one);
this->value[3].Data = _mm_add_ps(this->value[3].Data, one); this->Data[3].Data = _mm_add_ps(this->Data[3].Data, one);
return *this; return *this;
} }
inline fmat4x4SIMD & fmat4x4SIMD::operator-- () inline fmat4x4SIMD & fmat4x4SIMD::operator-- ()
{ {
this->value[0].Data = _mm_sub_ps(this->value[0].Data, one); this->Data[0].Data = _mm_sub_ps(this->Data[0].Data, one);
this->value[1].Data = _mm_sub_ps(this->value[1].Data, one); this->Data[1].Data = _mm_sub_ps(this->Data[1].Data, one);
this->value[2].Data = _mm_sub_ps(this->value[2].Data, one); this->Data[2].Data = _mm_sub_ps(this->Data[2].Data, one);
this->value[3].Data = _mm_sub_ps(this->value[3].Data, one); this->Data[3].Data = _mm_sub_ps(this->Data[3].Data, one);
return *this; return *this;
} }

View File

@ -156,9 +156,7 @@ namespace glm
template <comp a, comp b, comp c, comp d> template <comp a, comp b, comp c, comp d>
inline fvec4SIMD fvec4SIMD::swizzle() const inline fvec4SIMD fvec4SIMD::swizzle() const
{ {
int const Mask = ((int(d) << 6) | (int(c) << 4) | (int(b) << 2) | (int(a) << 0)); __m128 Data = _mm_shuffle_ps(this->Data, this->Data, ((int(d) << 6) | (int(c) << 4) | (int(b) << 2) | (int(a) << 0)));
__m128 Data = _mm_shuffle_ps(this->Data, this->Data, Mask);
return fvec4SIMD(Data); return fvec4SIMD(Data);
} }