Eigen subtracting vector from matrix columns

Question

Matrix linesP0 is 3xN. I want to subtract it from vector planeP0 which is 3x1. Is there any smarter and faster way to do this?

Now I'm using a for loop. Sample code below:

MatrixXf temp(linesP0.rows(), linesP0.cols());
for (int i = 0; i < linesP0.cols(); i++)
{
    temp.col(i) = planeP0 - linesP0.block<3, 1>(0, i);
}

I tried to use colwise() but didn't work.

Bobseine

Answer 1

You can use .colwise() to do this, you just have to be a little creative.

Vector3d v      = Vector3d(1.0, 2.0, 3.0);
Matrix3d m      = Matrix3d::Random();
Matrix3d result = (-m).colwise() + v;
std::cout << result << std::endl;

Sample result:

v = [1 2 3]' (3x1)
m = [1 1 1; 2 2 2]' (3x2)
result = [0 1 2; -1 0 1]' (3x2)

Thematic

Jacob, can you please take a look at this similar problem: #42936444 – Tonsillitis

Sure, I posted an answer. – Thematic

Any short explanation why is it faster than a for loop? – Curler

@Curler Reasons I can conceive of: - Eigen will reliably vectorize, and more importantly, will vectorize in precisely the way that you want it to. Some vectorization patterns can't be discerned by the compiler, because it would have to know how your data looks at runtime - map-like operations make everyone feel warm and fuzzy deep inside, it can be easier to avoid off-by-one bugs - You want to do everything const – Thematic