thoehne/PhanesEngine
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fixed Matrix4 inverse abstraction.

Browse Source
This commit is contained in:
Thorben Höhne 2025-02-19 19:30:02 +01:00
parent 1fbbdf5f0a
commit 0da9755ae6
1 changed files with 19 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
Engine/Source/Runtime/Core/Math/SIMD/PhanesVectorMathSSE.hpp
View File

@ -26,6 +26,9 @@
// Common // // Common //
// ========== // // ========== //
#ifndef PHANES_VECTOR_MATH_SSE_HPP
#define PHANES_VECTOR_MATH_SSE_HPP
namespace Phanes::Core::Math::SIMD namespace Phanes::Core::Math::SIMD
{ {
Phanes::Core::Types::Vec4f32Reg vec4_cross_p(const Phanes::Core::Types::Vec4f32Reg v1, const Phanes::Core::Types::Vec4f32Reg v2) Phanes::Core::Types::Vec4f32Reg vec4_cross_p(const Phanes::Core::Types::Vec4f32Reg v1, const Phanes::Core::Types::Vec4f32Reg v2)
@ -48,10 +51,10 @@ namespace Phanes::Core::Math::SIMD
/// <returns>Sum stored in v[0:31].</returns> /// <returns>Sum stored in v[0:31].</returns>
Phanes::Core::Types::Vec4f32Reg vec4_hadd(const Phanes::Core::Types::Vec4f32Reg v) Phanes::Core::Types::Vec4f32Reg vec4_hadd(const Phanes::Core::Types::Vec4f32Reg v)
{ {
__m128 shufl = _mm_movehdup_ps(v); __m128 t = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 3, 0, 1));
__m128 sum = _mm_add_ps(v, shufl); t = _mm_add_ps(t, v);
shufl = _mm_movehl_ps(sum, sum); t = _mm_shuffle_ps(t, t, _MM_SHUFFLE(1, 0, 3, 2));
return _mm_add_ss(sum, shufl); return _mm_add_ps(t, v);
} }
/// <summary> /// <summary>
@ -61,12 +64,11 @@ namespace Phanes::Core::Math::SIMD
/// <returns>Sum of components.</returns> /// <returns>Sum of components.</returns>
float vec4_hadd_cvtf32(const Phanes::Core::Types::Vec4f32Reg v) float vec4_hadd_cvtf32(const Phanes::Core::Types::Vec4f32Reg v)
{ {
__m128 shufl = _mm_movehdup_ps(v); __m128 t = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 3, 0, 1));
__m128 sum = _mm_add_ps(v, shufl); t = _mm_add_ps(t, v);
shufl = _mm_movehl_ps(sum, sum); t = _mm_shuffle_ps(t, t, _MM_SHUFFLE(1, 0, 3, 2));
sum = _mm_add_ss(sum, shufl); t = _mm_add_ps(t, v);
return _mm_cvtss_f32(t);
return _mm_cvtss_f32(sum);
} }
/// <summary> /// <summary>
@ -586,12 +588,7 @@ namespace Phanes::Core::Math::Detail
pl.comp.data = _mm_set_ps(x, y, z, d); pl.comp.data = _mm_set_ps(x, y, z, d);
} }
static FORCEINLINE void map(Phanes::Core::Math::TPlane<float, true>& pl, const TVector3<float, true>& v1, const TVector3<float, true>& v2, const TVector3<float, true>& v3) // TODO: Create SSE constructor with 3 Points
{
TVector4<float, false> tmp;
}
}; };
@ -1281,7 +1278,7 @@ namespace Phanes::Core::Math::Detail
template<> template<>
struct compute_mat4_inv<float, false> struct compute_mat4_inv<float, true>
{ {
// From: GLM: https://github.com/g-truc/glm/blob/master/glm/simd/matrix.h (MIT License) // From: GLM: https://github.com/g-truc/glm/blob/master/glm/simd/matrix.h (MIT License)
static FORCEINLINE bool map(Phanes::Core::Math::TMatrix4<float, true>& r, const Phanes::Core::Math::TMatrix4<float, true>& m1) static FORCEINLINE bool map(Phanes::Core::Math::TMatrix4<float, true>& r, const Phanes::Core::Math::TMatrix4<float, true>& m1)
@ -1503,6 +1500,10 @@ namespace Phanes::Core::Math::Detail
r.c1.data = _mm_mul_ps(Inv1, Rcp0); r.c1.data = _mm_mul_ps(Inv1, Rcp0);
r.c2.data = _mm_mul_ps(Inv2, Rcp0); r.c2.data = _mm_mul_ps(Inv2, Rcp0);
r.c3.data = _mm_mul_ps(Inv3, Rcp0); r.c3.data = _mm_mul_ps(Inv3, Rcp0);
return true;
} }
}; };
} }
#endif