matrixToQuaternion

[wing64]

Code: Select all

template <typename T> int sign(T val) {
    return (T(0) < val) - (val < T(0));
}
 
irr::core::quaternion m2q( const irr::core::matrix4& m ) 
{
    // Adapted from: http://www.euclideanspace.com/maths/geometry/rotations/conversions/matrixToQuaternion/index.htm
    irr::core::quaternion q;
    q.W = sqrtf( irr::core::max_( 0.0f, 1.0f + m[0] + m[5] + m[10] ) ) / 2.0f; 
    q.X = sqrtf( irr::core::max_( 0.0f, 1.0f + m[0] - m[5] - m[10] ) ) / 2.0f; 
    q.Y = sqrtf( irr::core::max_( 0.0f, 1.0f - m[0] + m[5] - m[10] ) ) / 2.0f; 
    q.Z = sqrtf( irr::core::max_( 0.0f, 1.0f - m[0] - m[5] + m[10] ) ) / 2.0f; 
    q.X *= sign( q.X * ( m[6] - m[9] ) );
    q.Y *= sign( q.Y * ( m[8] - m[2] ) );
    q.Z *= sign( q.Z * ( m[1] - m[4] ) );
    return q;
}

copy code from http://answers.unity3d.com/questions/11 ... ctor3.html

[chronologicaldot]

That might come in handy. Thanks for posting!

[CuteAlien]

Our irr::core::quaternion constructor does take a matrix (note that it got fixed in 1.8, so if you use an older Irrlicht version it didn't work well).

[wing64]

irr::core::quaternion constructor (1.8 trunk svn.4438) not work for my unreal psk/psa loader (no luck)

[CuteAlien]

Depending on how the matrix are described in psk/psa you might have to transpose them first before passing them to the quaternion.

[fmx]

I did tests on this shortly after 1.8 came out and I can confirm there is nothing wrong with irrlicht's Quaternion and Matrix calculations, so as CuteAlien said the problem is likely to be elsewhere in your pipeline