include/foundation/PxVec3.h
File members: include/foundation/PxVec3.h
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions
// are met:
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
// * Neither the name of NVIDIA CORPORATION nor the names of its
// contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ''AS IS'' AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Copyright (c) 2008-2023 NVIDIA Corporation. All rights reserved.
// Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
// Copyright (c) 2001-2004 NovodeX AG. All rights reserved.
#ifndef PX_VEC3_H
#define PX_VEC3_H
#include "foundation/PxMath.h"
#if !PX_DOXYGEN
namespace physx
{
#endif
template<class Type>
class PxVec3T
{
public:
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T()
{
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T(PxZERO) : x(Type(0.0)), y(Type(0.0)), z(Type(0.0))
{
}
explicit PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T(Type a) : x(a), y(a), z(a)
{
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T(Type nx, Type ny, Type nz) : x(nx), y(ny), z(nz)
{
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T(const PxVec3T& v) : x(v.x), y(v.y), z(v.z)
{
}
// Operators
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T& operator=(const PxVec3T& p)
{
x = p.x;
y = p.y;
z = p.z;
return *this;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type& operator[](unsigned int index)
{
PX_ASSERT(index <= 2);
return reinterpret_cast<Type*>(this)[index];
}
PX_CUDA_CALLABLE PX_FORCE_INLINE const Type& operator[](unsigned int index) const
{
PX_ASSERT(index <= 2);
return reinterpret_cast<const Type*>(this)[index];
}
PX_CUDA_CALLABLE PX_FORCE_INLINE bool operator==(const PxVec3T& v) const
{
return x == v.x && y == v.y && z == v.z;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE bool operator!=(const PxVec3T& v) const
{
return x != v.x || y != v.y || z != v.z;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE bool isZero() const
{
return x == Type(0.0) && y == Type(0.0) && z == Type(0.0);
}
PX_CUDA_CALLABLE PX_INLINE bool isFinite() const
{
return PxIsFinite(x) && PxIsFinite(y) && PxIsFinite(z);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE bool isNormalized() const
{
const float unitTolerance = Type(1e-4); // PT: do we need a different epsilon for float & double?
return isFinite() && PxAbs(magnitude() - Type(1.0)) < unitTolerance;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type magnitudeSquared() const
{
return x * x + y * y + z * z;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type magnitude() const
{
return PxSqrt(magnitudeSquared());
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T operator-() const
{
return PxVec3T(-x, -y, -z);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T operator+(const PxVec3T& v) const
{
return PxVec3T(x + v.x, y + v.y, z + v.z);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T operator-(const PxVec3T& v) const
{
return PxVec3T(x - v.x, y - v.y, z - v.z);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T operator*(Type f) const
{
return PxVec3T(x * f, y * f, z * f);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T operator/(Type f) const
{
f = Type(1.0) / f;
return PxVec3T(x * f, y * f, z * f);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T& operator+=(const PxVec3T& v)
{
x += v.x;
y += v.y;
z += v.z;
return *this;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T& operator-=(const PxVec3T& v)
{
x -= v.x;
y -= v.y;
z -= v.z;
return *this;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T& operator*=(Type f)
{
x *= f;
y *= f;
z *= f;
return *this;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T& operator/=(Type f)
{
f = Type(1.0) / f;
x *= f;
y *= f;
z *= f;
return *this;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type dot(const PxVec3T& v) const
{
return x * v.x + y * v.y + z * v.z;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T cross(const PxVec3T& v) const
{
return PxVec3T(y * v.z - z * v.y, z * v.x - x * v.z, x * v.y - y * v.x);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T getNormalized() const
{
const Type m = magnitudeSquared();
return m > Type(0.0) ? *this * PxRecipSqrt(m) : PxVec3T(Type(0));
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type normalize()
{
const Type m = magnitude();
if(m > Type(0.0))
*this /= m;
return m;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type normalizeSafe()
{
const Type mag = magnitude();
if(mag < PX_NORMALIZATION_EPSILON) // PT: do we need a different epsilon for float & double?
return Type(0.0);
*this *= Type(1.0) / mag;
return mag;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type normalizeFast()
{
const Type mag = magnitude();
PX_ASSERT(mag >= PX_NORMALIZATION_EPSILON); // PT: do we need a different epsilon for float & double?
*this *= Type(1.0) / mag;
return mag;
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T multiply(const PxVec3T& a) const
{
return PxVec3T(x * a.x, y * a.y, z * a.z);
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T minimum(const PxVec3T& v) const
{
return PxVec3T(PxMin(x, v.x), PxMin(y, v.y), PxMin(z, v.z));
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type minElement() const
{
return PxMin(x, PxMin(y, z));
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T maximum(const PxVec3T& v) const
{
return PxVec3T(PxMax(x, v.x), PxMax(y, v.y), PxMax(z, v.z));
}
PX_CUDA_CALLABLE PX_FORCE_INLINE Type maxElement() const
{
return PxMax(x, PxMax(y, z));
}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3T abs() const
{
return PxVec3T(PxAbs(x), PxAbs(y), PxAbs(z));
}
Type x, y, z;
};
template<class Type>
PX_CUDA_CALLABLE static PX_FORCE_INLINE PxVec3T<Type> operator*(Type f, const PxVec3T<Type>& v)
{
return PxVec3T<Type>(f * v.x, f * v.y, f * v.z);
}
typedef PxVec3T<float> PxVec3;
typedef PxVec3T<double> PxVec3d;
class PxVec3Padded : public PxVec3
{
public:
PX_FORCE_INLINE PxVec3Padded() {}
PX_FORCE_INLINE ~PxVec3Padded() {}
PX_FORCE_INLINE PxVec3Padded(const PxVec3& p) : PxVec3(p) {}
PX_FORCE_INLINE PxVec3Padded(float f) : PxVec3(f) {}
PX_CUDA_CALLABLE PX_FORCE_INLINE PxVec3Padded& operator=(const PxVec3Padded& p)
{
x = p.x;
y = p.y;
z = p.z;
return *this;
}
PxU32 padding;
};
PX_COMPILE_TIME_ASSERT(sizeof(PxVec3Padded) == 16);
typedef PxVec3Padded PxVec3p;
#if !PX_DOXYGEN
} // namespace physx
#endif
#endif