glm::mat4 * glm::mat4 yielding incorrect results? - c++

Question

For context, I've been learning OpenGL, and got to perspective projections. Spent a solid hour trying to figure out why my triangular pyramid wasn't showing up right, until I noticed one of my matrix multiplications (performed using operator* for glm::mat4) was giving me incorrect results. Tried multiplying two matrices with the same values, outside of that context, and sure enough, the result was still incorrect.
The two matrices I tried to multiply were:
glm::mat4 left, right;
left[0][0] = 2.09928f; left[0][1] = 0.0f; left[0][2] = 0.0f; left[0][3] = 0.0f;
left[1][0] = 0.0f; left[1][1] = 3.73205f; left[1][2] = 0.0f; left[1][3] = 0.0f;
left[2][0] = 0.0f; left[2][1] = 0.0f; left[2][2] = 1.0202f; left[2][3] = -2.0202f;
left[3][0] = 0.0f; left[3][1] = 0.0f; left[3][2] = 1.0f; left[3][3] = 0.0f;
right[0][0] = 0.999998f; right[0][1] = 0.0f; right[0][2] = -0.00174533f; right[0][3] = 0.0f;
right[1][0] = 0.0f; right[1][1] = 1.0f; right[1][2] = 0.0f; right[1][3] = 0.0f;
right[2][0] = 0.00174533f; right[2][1] = 0.0f; right[2][2] = 0.999998f; right[2][3] = 5.0f;
right[3][0] = 0.0f; right[3][1] = 0.0f; right[3][2] = 0.0f; right[3][3] = 1.0f;
Using left * right, I get:
2.09928, 0, -0.00178059, 0.00352592
0, 3.73205, 0, 0
0.00366393, 0, 6.0202, -2.0202
0, 0, 1, 0
which is incorrect. On the other hand, with this multiplication algorithm:
inline glm::mat4 Multiply(const glm::mat4& left, const glm::mat4 right)
{
glm::mat4 ret;
for (unsigned int i = 0; i < 4; i++) {
for (unsigned int j = 0; j < 4; j++) {
ret[i][j] = left[i][0] * right[0][j] +
left[i][1] * right[1][j] +
left[i][2] * right[2][j] +
left[i][3] * right[3][j];
}
}
return ret;
}
I get:
2.09928, 0, -0.00366393, 0
0, 3.73205, 0, 0
0.00178059, 0, 1.0202,3.08081
0.00174533, 0, 0.999998, 5
which is correct.
So, what am I doing wrong here?
Thanks in advance!

Answer 1

ret[i][j] = left[i][0] * right[0][j] +
left[i][1] * right[1][j] +
left[i][2] * right[2][j] +
left[i][3] * right[3][j];
GLM uses column-major matrices. As such, left[0] is the first column. You are doing a dot-product between the column of the lhs matrix and the rows of the rhs matrix. That's backwards; it's row * column.
What you are probably doing wrong is building your matrices incorrectly, in row-major fashion.