hello,
i'm trying to multiply a matrix with a huge amount of vectors.
the following code shows basically what i'm doing right now:

Code: Select all

bool CalculateVectors()
{
Matrix3f m;
Vector3f **A;Vector3f **B; Vector3f **C;
unsigned int Amount=1000000;
// make A,B and C 2D array's of vectors here ...
// set the values for the matrix and vectors here ...
   unsigned int i=0;unsigned int j=0;
   while(i<rows)
   {
      while(j<columns)
      {
      A[i][j]=m*B[i][j];
      C[i][j]=A[i][j].reverse().normalized();
      j++;
      }
   i++; j=0;
   }

return true;
}

this method works, and does what i want it to do, but i would love to see it go faster.
is there a way to multiply all vectors with the matrix at once, without looping through all the vectors individually? maybe i could get some performance boost there?
The amount of vectors could sometimes be 16 million, so i will probably still have to do it in multiple steps

sorry for my English

Hi,

yes you can store your (rows * columns) vectors into a 3 x (rows * columns) matrix and simply do:

Matrix<float, 3, Dynamic> A, B, C;
// ...
A = m * B;
C = A.colwise().reverse().normalized();

Then your "A[i][j]" become "A.col(i+j*rows)". You could wrap this into a function returning a Matrix<float, 3, Dynamic>::ColXpr object.